2006 Book TheEngineeringOfSport6

The Engineering of Sport 6
Eckehard Fozzy Moritz and Steve Haake (Eds.)
The Engineering of Sport 6

Volume 1: Developments for Sports
~ Springer
Eckehard Fozzy Moritz Steve Haake
SportKreativWerkstatt GmbH Centre for Sport and Exercise Science
Herzogstraße 48 Collegiate Hall
D-80803 München Sheffield Hallam University
Germany Sheffield S10 28P
efm@SportKreativWerkstatt.de UK
www.SportKreativWerkstatt.de s.j.haake@shu.ac.uk
Library of Congress Control Numbcr: 2006927112
ISBN-10: 0-387-31773-2
ISBN-13: 978-0387-31773-l
Printed on acid-free paper.
© 2006 Springer Science+Business Media LLC, corrected publication 2021

All rights reserved. This work may not be translated or copied in whole or in part without the
written permission of the publisher (Springer Science+Business Media. LLC. 233 Spring Street,
New York, NY 10013. USA), except for brief excerpts in connection with reviews or scholarly
analysis. Use in connection with any form of information storage and retrieval, electronic
adaptation, computer software. or by similar or dissimilar methodology now known or hereafter
developed is forbidden.
The use in this publication of trade names. trademarks. service marks. and similar terms. even if
they are not identified as such. is not to be taken as an expression of opinion as to whether or
not they are subject to proprietary rights.
Printed in the United States of America. (EB)
9876 54321
springer.com
Preface
What you are holding in your hands is probably the best overview of activities in
sports engineering available at the time of printing; i.e. the state of the art in summer
2006. lt is the result of so many people's work to whom we are indebted that it is
difficult to name them: there are the authors, the scientific advisory board, the
scientific committee, the theme patrons, the publisher and printer, the advisors of
whatever kind - and, here we have to make an exception, there is Ingo and Amanda.
Nobody who has bccn part of thc production of this book could have done without
them, at the very least us: they handled issues you wouldn't even believe could turn up
with efficiency and charm. Thanks, Ingo Valtingoier; thanks, Amanda Staley.
In the accumulation of the contributions and the preparation of the proceedings
we encountered one development that we were very happy about: the sports
engineering community keeps growing - in the number or researchers and experts
involved, but also in the breadth of disciplines and institutions contributing. This
should definitely be interpreted as a positive development - even though in the
evaluation of contributions this lead to a number of intricate discussions. Is sports
engineering primarily science? Is it engineering? Is it science and engineering helping
sports? Some reviewers had differing views on that: if it is science, you need method,
data, and discussion; if it is engineering, you need method and an outcome with some
demonstrable usefulness, if it is an aide to sports then whatever has been done needs
demonstrable relevance. As a consequence, some contributions very weil done from
an engineering perspective have been tumed down by hardcore scientists, and vice
versa; in some cases we tried to intermediate, in others it may have been bad Juck for
the contributors. We think sports engineering will have to live with this variety of
perspectives and interests; it is rather the appeal of this field in the process of finding
itself. Openness combined with consistent reasoning will be needed to progress from
here; somewhere in-between academic traditions and Feyerabend's famous "Anything
goes".
As a quick glimpse behind the scene, besides the disciplinary quarrels skctched
above some "cultural" clashes could also not be avoided. One German reviewer put
his comments in a very direct way that was hard to bear for the British author; some
East Asian authors had a hard time in focusing their writing on the most interesting
results and were thus bluntly thrown out; some well-known members in one
community have seen their abstract tumed down by experts from another area who did
not know about the writer's fame ... these anecdotes point to just a couple of more
issues the sports engineering community will have get to grips with in the not too
distant future.
vi Preface
As the result of various influences in these proceedings you will find a number of
new topic areas indirectly related to but important to sports engineering. One area of
concem we like to especially highlight here is the topic of sustainability, which may
serve as an important yardstick for the future development of sports engineering and
hopefully other industrial activities. Furthermore, you will find contributions on
trends, cultural influences, human factors and on neural network modeling. Finally,
according to the special emphasis of this conference we were successful in seeking a
large number of papers in the area of innovation and design, including economic
perspectives and proposals for novel design approaches. To our regret, even though
we had tried hard we could get no contributions on industrial design - this area with so
much relevance to sports equipment apparently is still a step-child in our community.
In the assembly of these proceedings we have endeavored to realize some novel
approaches. First of all, we used "theme patrons" for different topic areas who not
only helped acquire contributions but were also asked to write a synopsis of the
contributions in "their" fields. This will hopefully increase the use value for readers,
who by just reading the synopses can have a basic idea about developments in certain
fields, and can then scan contributions on a much better knowledge basis. This is a
first step towards converting the proceedings into a sort of handbook which hopefully
will be taken up by future editors.
Then, as we tried to increase the relevance of sports engineering to sports, we
have asked authors to take special care to illustrate the respective relevance, and to put
their contribution into a sports-related category rather than a discipline-oriented
category. Therefore, one volume of these proceedings has been named "developments
for sports"; it is the biggest and could have even been bigger. The second volume is
termed "developments in disciplines", which consists mainly of contributions focusing
on modeling and measurements. A third volume has been named "developments for
innovation", a tribute to this special focus of this conference (being organized by a
center for innovation in sports ), and to the fact that we could accumulate an amazing
number of contributions in this field.
Finally, we hope that the reader will appreciate the outcome, and we'll be very
happy to receive comments of whatever kind, be it criticism, proposals for
improvement or grappa casks and flower arrangements.
Eckehard Fozzy Moritz

Stephen Haake
Editors
July 2006
Contents
1 Baseball
Synopsis 3
Alan M. Nathan
An Experimental lnvestigation of Baseball Bat Durability 5

Patrick J. Drane, James A. Sherwood. Rebecca H. Shaw
Bending Modes, Damping, and the Sensation of Sting in Baseball Bats 11

Daniel A. Russell
Experimental Investigations of the Relationship of Baseball Bat Properties 17

on Battered-Ball Performance
Rebecca H. Shaw, James A. Sherwood
The Effect of Spin on the Flight of a Baseball 23

Alan M. Nathan. Joe Hopkins. Lance Chong. Hank Kaczmarski
Rigid Wall Effects on Softball Coefficient ofRestitution Measurements 29

Lloyd Smith. Aaron /son
The Effect of Holding Methods on a Baseball Bat Performance Estimation 35

System
Hiroyuki Kagawa. Takeshi Yoneyama. Masaya Takahashi
2 Climbing - Instrumentation And Testing 0/ Equipment
Synopsis 43
Franz Konstantin Fuss
An Estimation ofthe Load Rate Imparted to a Climbing Anchor During Fall 45

Arrest
Dave Custer
Dynamics of Speed Climbing 51

Franz Konstantin Fuss. Günther Niegl
lnstrumented Climbing Holds and Dynamics of Sport Climbing 57

Franz Konstantin Fuss. Günther Niegl
Forces Generated in a Climbing Rope During a Fall 63

Andrew Phillips, Jeff Vogwell. Alan Bramley
viii Contents
Rock Climbing Belay Device Analysis, Experiments and Modelling 69

Liane! Manin, Matthieu Richard, Jean-Daniel Brabant, Marc Bissuel
3 Cycling
Synopsis 77
Martin Strangwood
Thermo-mechanical Modification Techniques for Structural Foams used in 79

Racing Bicycle Wheels.
Catherine Caton, Mike Jenkins, Martin Strangwood
The Effect of a Non Circular Chainring on Cycling Performance 85

Nicolas Horvais, Pierre Samozino, Frederique Hintzy
Dynamic Characteristics ofModem Mountain Bikes Rear Linkages 91

Angela Tempia, Aleksandar Subic, Ricardo M Pagliarella
An Ambient Intelligence System to Assist Team Training and Competition in 97

Cycling
Jngmar Fliege, Alexander Geraldy, Reinhard Gotzhein, Thomas Jaitner,
Thomas Kuhn, Christian Webei
Indoor-Simulation ofTeam Training in Cycling 103

Thomas Jaitner, Marcus Trapp, Dirk Niebhur, Jan Koch
A Bond Graph Model of a Full-Suspension Mountain Bicycle Rear Shock 109

Robin Redfield, Cory Sutela
Track Cycling: An Analytical Model 115

Richard Lukes, Matt Carre, Stephen Haake
Forces During Cycling After Total Knee Arthoplasty 121

Maximilian Mueller, Veit Senner, Markus Wimmer
A Study of Aerodynamic Drag and Thermal Efficiency of a Series of Bicycle 127

Heimets
Firoz Alam, Aleksandar Subic, Simon Watkins
4 Golf
Synopsis 135
Steve Mather
Contents ix
An Instrumented Grip Handle for Golf Clubs to Measure Forces and 137
Moments Exerted by Each Hand During Swing Motion
S. Koike, H. lida, H. Shiraki, M Ae
The Aerodynamic Influence of Dimple Design on Flying Golf Ball 143

T. Sajima, T. Yamaguchi, M Yabu, M Tsunoda
Experimental Verification of Trajectory Analysis of Golf Ball Under 149

Atmospheric Boundary Layer
Takeshi Naruo, Taketo Mizota
Validation of Accelerometers And Gyroscopes to Provide Real-Time 155

Kinematic Data for Golf Analysis
K. Fitzpatrick, R. Anderson
Investigation ofWrist Release During the Golf Swing by Using a Golf Swing 161
Robot
Yohei Hoshino, Yukinori Kobayashi, Soichiro Suzuki
Segmental Sequencing of Kinetic Energy in the Golf Swing 167

Brady C. Anderson, Jan C. Wright, Darren J. Stefanyshyn
5 Gymnastics
Synopsis 175
David G. Kerwin
Effect of Shoulder Compliance on Peak High Bar Forces During the Giant 177
Swing
Alison L. Sheets, Mont Hubbard
Effects of Horizontal Surface Compliance on Balance Strategies 183

Wendy Kimme!, Mont Hubbard
Predicting High Bar Forces in the Longswing 189

David Kerwin, Gareth frwin
Musculoskeletal Work in the Longswing on High Bar 195

Gareth lrwin, David G Kerwin
6 Lawn Sports
Synopsis 203
Matt Carre
x Contents
Quantification of the Cricket Bowling Delivery; a Study of Elite Players to 205

Gauge Variability and Controllability
Laura Justham, Andrew West, Andy Harland, Alex Cox
Ball Launch Characteristics for Elite Rugby Union Players 211

Christopher Holmes, Roy Jones, Andy Harland, Jon Petzing
A Novel Quantitative Method for the Determination of Wear in an Installed 217

Synthetic Turf System
Andrew McLeod, Jain James, Kim Blackburn, Gavin Wood
Multi-Optimization of Three Kicks in Rugby 223

Kazuya Seo, Osamu Kobayashi, Masahide Murakami
The Mechanical Behaviour of Cricket Soils During Preparation by Rolling 229

Peter Shipton, lain James, Alex Vickers
Studies on the Oblique Impact of a Cricket Ball on a Cricket Pitch 235

David James, Matt Carni, Stephen Haake
Test Devices for the Evaluation of Synthetic Turf Pitches for Field Hockey 241
Co/in Young, Paul Fleming, Neil Dixon
7 Skiing, Snowboarding and Ski Jumping
Synopsis 249
Veit Senner
Laboratory Device for Measuring the Friction Between Ski-Base Materials 251
and Ice or Snow
Mathieu Fauve, Lukas Bäurle, Hansueli Rhyner
Biomechanical Instrumentation of the Bergisel Jumping Hili in Innsbruck 257

and Exemplary Analysese
Kurt Schindelwig, Werner Nachbauer
Dynamic Properties ofMaterials for Alpine Skis 263

Christian Fischer, Mathieu Fauve, Etienne Combaz, Pierre-Etienne Bourban,
Veronique Michaud, Christopher J.G. Plummer, Hansueli Rhyner, Jan-
Anders E. Manson,
Calculation ofFriction and Reaction Forces During an Alpine World Cup 269
Downhill Race
M. Schiestl, P. Kaps, M. Mössner, W. Nachbauer
Contents xi
Measurement of Jumper's Body Motion in Ski Jumping 275

Yuji Ohgi, Kazuya Seo, Nobuyuki Hirai, Masahide Murakami
RETRACTED CHAPTER: Riding on Air: A New Theory for Lift Mechanics 281
of Downhill Skiing and Snowboarding
Qianhong Wu, Yesim Igci, Yiannis Andreopoulos, Sheldon Weinbaum
Subjective Evaluation ofthe Performance of Alpine Skisand Correlations 287

with Mechanical Ski Properties
Peter Federoif, Mirco Auer, Mathieu Fauve, Anton Lüthi, Hansueli Rhyner
Timing of Force Application and Joint Angles During a Long Ski Turn 293
Takeshi Yoneyama, Nathan Scott, Hiroyuki Kagawa
Effect of Bindings and Plates on Ski Mechanical Properties and Carving 299
Performance
Anton Lüthi, Peter Federoif, Mathieu Fauve, Hansueli Rhyner
Development of a Prototype that Measures the Coefficient of Friction 305

Between Skis and Snow
Paul Mil/er, Andy Hytjan, Matthew Weber, Mi/es Wheeler, Jack Zable, Andy
Walshe, Alan Ashley
8 Football
Synop~s 313
Matt Cami
An Investigation into the Link Between Soil Physical Conditions and the 315
Playing Quality of Winter Sports Pitch Rootzones
Marke Jennings-Temple, Peter Leeds-Harrison, Jain James
Measuring and Modelling the Goalkeeper's Diving Envelope in a Penalty 321

Kick
David G. Kerwin, Ken Bray
Flow Visualization on a Real Flight Non-spinning And Spinning Soccer Ball 327
Takeshi Asai, Kazuya Seo, Osamu Kobayashi, Reiko Sakashita
Gaze Point Analysis in Movement Prediction of Soccer Players By Image 333

Processing
Yuusuke Hiramatsu, Shigemichi Ohshima, Atsumi Ohtsuki
Traction Testing of Soccer Boots Under Game Relevant Loading Conditions 339
Thomas Grund, Veit Senner
xii Contcnts
Correlation between Support Foot Placement and Goal Accuracy for Instep 345
Kicks in the Soccer Field
Giuseppe Marcolin, Nicola Petrone, Claudia Robazza
Analysis of the Influence of Rubber lnfill on the Mechanical Performance of 351

Artificial Turf Surfaces for Soccer
Enrique Alcantara, David Rosa, Javier Gamez, Antonio Martinez, Maria
Comin, Maria lose Such, Pedro Vera, Jaime Prat
Soccer Ball Modal Analysis Using a Scanning Laser Doppler Vibrometer 357
(SLDV)
Jouni Ronkainen, Andy Harland
9 Tennis
Synopsis 365
Stuart Miller
Normal Impact ofHollow Balls on Fiat Surfaces 367

Yoshihisa Honda
Factors in Tennis Ball Wear 373

Carolyn Steele, Roy Jones, Paul Leaney
Measuring Ball Spin off a Tennis Racket 379

Simon Goodwill, Jamie Douglas, Stuart Miller, Stephen Haake
30 Player Testing in Tennis 385

Simon Choppin, Simon Goodwill, Stephen Haake
An Extended Study lnvestigating the Effects ofTennis Rackets with Active 391
Damping Technology on The Symptoms ofTennis Elbow
Robert Cottey, Johan Kotze, Herfried Lammer, Werner Zirngibl
10 Waier Sports
Synopsis 399
Jani Macari Pallis
Computational Fluid Dynamic Analysis of a Water Ski Jumper 401

John Hart, David Curtis, Stephen Haake
Feedback Systems in Rowing 407

Arnold Baca, Philipp Kornfeind, Maria Heller
Contents xiii
Biomechanical Analysis of Olympic Kayak Athletes During Indoor Paddling 413

Nicola Petrone, Andrea Isotti, Guglielmo Guerrini
So you think you know the ropes? White Water Rescue Ropes and 419
Techniques
Matt Barker
Computational Modelling of Surfboard Fins for Enhanced Performance 425

Dave Carswell, Nicholas Lavery, Steve Brown
Development of Swimming Prosthetic for Physically Disabled (Optimal 431

Design for One Side of Above-Elbow Amputation)
Keiko Yoneyama, Motomu Nakashima
Retraction Note to: Riding on Air: A New Theory Cl

for Lift Mechanics of Downhill Skiing and Snowboarding
Author Index 437
Subject Index 441

Contributors
Simon C. Adelmann
University of Birmingham, UK
Michiyoshi Ae
University ofTsukuba, Japan
UzomaAjoku
Loughborough University, UK
Shinichiro Akiyama
Toyota Motor Corporation, Japan
Firoz Alam
Royal Melbourne Institute ofTechnology
Pär-Anders Albinsson
Swedish Defence Research Agency, Sweden
Enrique Alcantara
Universitat Politecnica de Valencia, Spain
Brady C. Anderson
University ofCalgary, Canada
Lauren Anderson
xvi Contributors
Ross Anderson
University of Limerick, UK
Dennis Andersson
Swedish Defence Research Agency, Sweden
Yiannis Andreopoulos
The City College ofNew York, USA
Ali Ansarifar
Ayako Aoyama
Tokyo Institute ofTechnology
Takeshi Asai
Yamagata University, Japan
Andrew Ashcroft
University ofCambridge, UK
Alan Ashley
United States Ski Association, USA
MircoAuer
Swiss Federal Institute for Snow and Avalanche Research Davos, Switzerland
Andreas Avgerinos
Democritus University ofThrace, Greece
Arnold Baca
University ofVienna, Austria
Sarah Barher
University of Sheffield, UK
Franck Barbier
Universite de Valenciennes, France
Matthew R. Barker
Auckland University ofTechnology, New Zealand
Contributors xvii
Joseph Beck
United States Air Force Academy, USA
Nicolas Belluye
Decathlon, France
Alexey, Belyaev
Perm State Technical University, Russia
Göran Berg/und
Sandvik Material Technology, Sweden
Nils Betzler
Otto von Guericke University Magdeburg, Germany
Marc Bissuel
INSA Lyon, France
Kim Blackburn
Cranfield University, UK
Jane R. Blackford
University of Edinburgh, UK
Kim B. Blair
Massachusetts Institute ofTechnology, USA
Stephan Boerboom
Technische Universität München, Germany
Harald Böhm
Robert Bordas
Pierre-Etienne Bourban
Ecole Polytechnique Federale de Lausanne (EPFL), Switzerland
Jean-Daniel Brabant
INSA Lyon, France
xviii Contributors
Alan N. Bramley
University of Bath, UK
Ken Bray
Desmond Brown
University of Ulster, UK
Steve Brown
University of Wales Swansea, UK
Mark-Paul Buckingham
University ofEdinburgh, UK
JeremyBurn
Bristol University, UK
Mike P. Caine
Matt J. Carre
David J. Carswell
Catherine J. Caton
University ofBirmingham, UK
Chaochao Chen
Kochi University ofTechnology, Japan
lance Chong
University of Illinois, USA
Simon Choppin
Jeffrey J. Chu
Simbex, USA
Contributors xix
Steffen Clement
AUDI Sport, Germany
Etienne Combaz
Maria Comin
Alex Cork
James Cornish
Robert Cottey
HEAD Sport AG, Austria
Aimee C. Cubitt
Kieran F. Culligan
David Curtis
Sheffield Hallam University, UK
Dave Custer
Massachusetts Institute of Technology, USA
Tim Deans
Jeroen Dethmers
Universiteit Maastricht, Netherlands
Neil Dixon
Sharon J. Dixon
University of Exeter, UK
xx Contributors
Jamie Douglas
International Tennis Federation, UK
Patrick J. Drane
University of Massachusetts Lowell, USA
Melanie Dumm
Juan Vicente Dura

Co/in Eames
Markus Eckelt
University of Applied Sciences Technikum Wien, Austria
Jürgen Edelmann-Nusser
Frank Einwag
Klinik für Orthopädische Chirurgie und Unfallchirurgie Bamberg, Germany
Carl F. Ettlinger
Vermont Safety Research, USA
Paul Ewart
University of Waikato, New Zealand
Emanuela F aggiano
University of Padova, ltaly
Mathieu Fauve
Owen R. Fauvel
Contributors xxi
Peter Federolf
University of Salzburg, Austria
Monika Fikus
University of Bremen, Germany
Christian Fischer
Peter R. Fischer
University of Augsburg, Germany
Keith Fitzpatrick
University of Limerick, UK
Paul Fleming
Ingmar Fliege
Technical University Kaiserslautern
Matthieu Foissac
Decathlon, France
Kathryn Frank/in
University ofGlamorgan, UK
Philippe Freychat
Decathlon, France
Piergiuseppe Fumei

Nanyang Technological University, Singaporc
Javier Gamez
xxii Contributors
Nico Ganter
Paul Gebhard
Alexander Geraldy
Anton Gerrits
TNO, Netherlands
Alexandras Giannakis
CSEM - Swiss Center for Electronics and Microtechnology, Switzerland
Maria Giannousi
Paul J. Gibbs
Christophe Gillet
Juan Carlos Gonzales

Simon Goodwill
Philippe Gorce
Toulon University, France
Rae, Gordon
University of Glamorgan, UK
Reinhard Gotzhein
Richard M Greenwald
Simbex, USA
Contributors xxiii
Thomas Grund
Guglielmo Guerrini
ltalian Kayak Federation, ltaly
lose Maria Gutierrez

Stephen J. Haake
Christian Hainzlmaier
Nick Hamilton
Dang Chul Han

Seoul National University, Korea
R. Keith Hanna
Fluent Europe Ltd., UK
Andy R. Harland
John Hart
Thomas Härte/
Chemnitz University ofTechnology, Germany
Ulrich Hartmann
Andreas Hasenknopf
MLD, Germany
Dieter Heinrich
University Innsbruck, Austria
xxiv Contributors
Ben Heller
Maria Heller
Christian Henneke
SportKreativWerkstatt GmbH, Germany
Martin Herbert
F alk Hildebrand
Institute for Applied Training Science (IA T) Leipzig, Germany
Norbert Himmel
Institut für Verbundwerkstoffe GmbH, Germany
Frederique Hintzy
Laboratoire de Modelisation des Activites Sportives, France
Nobuyuki Hirai
Yuusuke Hiramatsu
Meijo University, Japan
Philip Hodgkins
Martin Hofmann
Frank Hoisl
Christopher E. Holmes
Yoshihisa Honda
Kinki University, Japan
Contributors xxv
Joe Hopkins
Western Michigan University, USA
Neil Hopkinson
Nicolas Horvais
Yohei Hoshino
Hokkaido University, Japan
Kenji Hosokawa
Chubu University, Japan
Mont Hubbard
University of Califomia, Davis, USA
Andrew Hytjan
University ofColorado at Boulder, USA
Yesim Jgci
Princeton University, USA
Hiroshi Jida
Polytechnic University Kagawa, Japan
Yoshio lnoue
Carl Johan Jrander

Sandvik Material Technology, Sweden
Jan lriberri Berrostegieta

Performance Enhancement Centre, Basque Govemment, Spain
Gareth Jrwin
University of Wales Cardiff, UK
Aaron lson
Cascade Engineering, USA
xxvi Contributors
Andrea lsotti
University of Padova, Italy
Koji lto
Japan Institute of Sport Sciences, Japan
Takuzo lwatsubo
Kansai University, Japan
Thomas Jaitner
Daniel A. James
Griffith University, Australia
David M James
lainJames
Mike J. Jenkins
Marke Jennings-Temple
Alexander W Jessiman
Simbex, USA
Tomohiko Jin
Toyota Motor Corporation, Japan
Robert J. Johnson
University ofVermont, USA
Clifton R. Johnston
RoyJones
Contributors xxvii
Andre Jordan
Laura Justham
Hank Kaczmarski
Hiroyuki Kagawa
Kanazawa University, Japan
Michael Kaiser
Institut für Verbundwerkstoffe GmbH, Germany
Nico Karnperrnan
TNO, Netherlands
Peter Kaps
Shozo Kawarnura
Toyohashi University ofTechnology, Japan
Jan C. Kenny
University ofUlster, UK
David G. Kerwin
Andreas Kiefrnann
Cheol Kirn
Kyungpook National University, Korea
Moo Sun Kirn

SunJin Kirn
xxviii Contributors
Wendy Kimme!
University of Califomia, Davis, USA
Efthimis Kioumourtzoglou
Bob Kirk
Sebastian Klee
/sabella Klöpfer
Karin Knall
Klaus Knall
Ted Knox
Wright Patterson Air Force Base, USA
Cheolwoong Ko
University of Iowa, USA
Osamu Kobayashi
Tokai University, Japan
Yukinori Kobayashi
Hokkaido University, Japan
Jan Koch
Hannes Kog/er
Fischer GmbH, Austria
Sekiya Koike
Contributors xxix
Philipp Kornfeind
Giorgos Kotrotsios
CSEM - Swiss Center for Electronics and Microtechnology, Switzerland
Johan Kotze
Christian Krämer
Maximilian Krinninger
Michael Krohn
Hochschule für Gestaltung und Kunst Zürich, Switzerland
Andreas Krüger
Thomas Kuhn
Herfried Lammer
Nicholas Lavery
Paul Leaney
Manryung Lee
Kyungin Women's College, Korea
Woo Il Lee
Peter Leeds-Harrison
xxx Contributors
Sebastien leteneur
Chris Lewis-Jones
Delcam plc, UK
Udo Lindemann
Daniel Low
Peter Lugner
Vienna University ofTechnology, Austria
Richard Lukes
Anton Lüthi
Reiner Lützeler
RWTH Aachen University, Germany
Jani Macari Pallis

Cislunar Aerospace Inc., USA
Liane! Manin
INSA Lyon, France
Graeme Manson
Jan-Anders E. Manson
Giuseppe Marcolin
Brett A. Marmo
University of Edinburgh, UK
Contributors xxxi
Antonio Martinez
Natividad Martinez
Tom Mase
Michigan State University, USA
Steve Mather
University of Nottingham, UK
Sean Maw
University of Calgary, Canada
Alex J. McCloy
University ofUlster, UK
Mark McHutchon
Andrew Mcleod
Hossain Md.Zahid
Kenneth Meijer
Daniel Memmert
University ofHeidelberg, Germany
Roberta Meneghello
Imke K. Meyer
University of Bremen, Germany
Michael Michailov
National Sports Academy, Bulgaria
xxxii Contributors
Veronique Michaud
Thomas Milani
Paul Miller
University ofColorado at Boulder, USA
Stuart Miller
International Tennis Federation, UK
Guillaume Mil/et
Universite Jean Monnet Saint-Etienne, France
Hirofumi Minamoto
Sean R. Mitehe!!
Chikara Miyaji
Yusuke Miyazaki
Tokyo Institute ofTechnology, Japan
Taketo Mizota
Fukuoka Institute ofTechnology, Japan
Stuart Mank
Ana Montaner
John Morgan
Eckehard Fozzy Moritz

SportKreativWerkstatt GmbH, Germany
Contributors xxxiii
Rhys Morris
Martin Mössner
Maximilian Müller
Masahide Murakami
University of Tsukuba, Japan
Werner Nachbauer
Daiki Nakajima
Kansai University, Japan
Motomu Nakashima
Takeshi Naruo
Mizuno Corporation, Japan
Alan M Nathan
Dirk Niebhur
Günther Niegl
Christian Nolte
University of Augsburg, Germany
Claudius Nowoisky
Wubbo Ockels
Delft University ofTechnology, Netherlands
xxxiv Contributors
Stephan Odenwald
Yuji Ohgi
Keio University, Japan
Shigemichi Ohshima
Atsumi Ohtsuki
Hiroki Okubo
National Defense Academy, Japan
Steve R. Otto
R&A Rules Limited, UK
Riccardo M Pagliarella
Royal Melboume Institute ofTechnology, Australia
Jürgen Perl
University ofMainz, Germany
Stephane Perrey
Universite de Montpellier, France
Christiane Peters
Nicola Petrone
Neil Pettican
Jon Petzing
Andrew Phillips
Contributors xxxv
John Plaga
Wright Patterson Air Force Base, USA
Christopher J. G. Plummer
Alexander Romanovich Podgaets

Delft University ofTechnology, Netherlands
Jaime Prat
Celine Puyaubreau
Decathlon, France
Franck Quaine
Universite Joseph Fourier Grenoble, France
lose Ramiro
Robin Redfield
Martin Reiche!
Hansueli Rhyner
Matthieu Richard
PETZL, France
Claudia Robazza
Bryan C. Roberts
Jonathan Roberts
xxxvi Contributors
Markus A. Rohde
University ofSiegen, Germany
Jouni A. Ronkainen
David Rosa
Steve Rothberg
Maxime Roux
Decathlon, France
Daniel Russell
Kettering University, USA
Anton Sabo
Takahiro Sajima
SRI Sports Limited, Japan
Reiko Sakashita
Kumamoto University, Japan
Toshiyuki Sakata
Pierre Samozino
Yu Sato
Nicholas Savage
Royal Melbourne Institute ofTechnology, Australia
Hans Savelberg
Contributors xxxvii
Michael Schiestl
David Schill
Kurt Schindelwig
Erin Schmidt
Heinz-Bodo Schmiedmayer
Vienna University ofTechnology, Austria
Alexander Schneider
Turn Till Burn GmbH, Switzerland
Isabelle Schöfjl
University of Erlangen-Nuremberg, Germany
Volker R. Schöfjl
Stefan Schönberger
Herwig Sehretter
HTM Tyrolia, Austria
Andreas Schweizer
Kantonsspital Aarau, Switzcrland
Carsten Schwiewagner
Nathan Scott
The University of Western Australia, Australia
Brian P. Seif
xxxviii Contributors
Terry Senior
Veit Senner
Kazuya Seo
Yamagata University, Japan
Sonali Shah
University of Illinois at Urbana-Champaign, USA
Rebecca H. Shaw
University of Massachusetts Lowell, USA
Jasper Shealy
Rochester Institute ofTechnology, USA
Alison L. Sheets
University ofCalifornia, Davis, USA
James A. Sherwood
University ofMassachusetts Lowell, USA
Kyoko Shibata
Jun Shimizu
Peter Shipton
Hitoshi Shiraki
Anton Shumihin
Gerard Sierksma
University ofGroningen, Netherlands
Contributors xxxix
Lloyd Smith
Washington State University, USA
Peter Spitzenpfeil
Technische Universität München, Gerrnany
Carolyn Steele
Darren J. Stefanyshyn
Gunnar Stevens
University of Siegen, Gerrnany
Victoria H. Stiles
Valeriy Stolbov
Martin Strangwood
University ofBirrningham, UK
Wolf Strecker
Martin Strehler
SportKreativWerkstatt GmbH, Gerrnany
Claude Stricker
AISTS - International Academy ofSports Science and Technology, Switzerland
William J. Strange
University of Cambridge, UK
Aleksandar Subic
Maria lose Such

xi Contributors
Cory Sutela
SRAM Corporation, USA
Soichiro Suzuki
Kitami Institute ofTechnology, Japan
Masaya Takahashi
Sumitomo Light Meta!, Japan
Hironuri Takihara
Ming Adin Tan

Nanyang Technological University, Singapore
Angela Tempia
Eva Tenan
Dominique Thevenin
Mark Timms
Hot Stix Technologies, USA
Daniel Toon
Marcus Trapp
Masaya Tsunoda
Sadayuki Ujihashi
Sandor Vajna
Contributors xli
Rafael Valero
AIJU, Technological Institute ofToys, Spain
Sergey Vasilenko
JSC Aviadvigatel - Perm Engine Company, Russia
Pedro Vera
Universitat Politecnica de Valencia, Spain.
Johan Verbeek
University of Waikato, New Zealand
Nicholas Vernadakis
Alex Vickers
Laurant Vigouroux
Universite Joseph Fourier Grenoble, France
Jejf Vogwell
Jörg F. Wagner
University Stuttgart, Germany
Klaus Wagner
David Walfisch
Eric S. Wallace
University of Ulster, UK
Tom Waller
Andy Walshe
United States Ski Association, USA
xlii Contributors
Simon Watkins
PekChee We
Christian Webei
Matthew Weber
University of Colorado at Boulder, USA
Sheldon Weinbaum
The City College ofNew York, USA
Andrew West
Cory West
Hot Stix Technologies, USA
Miles Wheeler
Josef Wiemeyer
Technische Universität Darmstadt Germany
Bart Wijers
Terra Sports Technology, Netherlands
Paul Willems
Simon Williams
University of Glamorgan, UK
Markus A. Wimmer
Rush University Medical Center Chicago, USA
Erich Wintermantel
Contributors xliii
Clive Wishart
Kerstin Witte
Gavin Wood
Jan C. Wright
TaylorMade-adidas GolfCompany, USA
Qianhong Wu
Villanova University, USA
Volker Wulf
University of Siegen, Germany
Bernd Wunder/ich
Masanori Yabu
Tetsuo Yamaguchi
Connie Yang
Keiko Yoneyama
Takeshi Yoneyama
Kanazawa University, Japan
Co/in Young
Allen Yuen
xliv Contributors
Jack Zahle
Michael F. Zäh
Eleni Zetou
Andreas Zimmermann
University of Siegen, Germany
Werner Zirngiebl
Praxisklinik für Orthopädie und Sportmedizin, München, Germany
1 Baseball
Synopsis of Current Developments: Baseball
Alan M. Nathan
University of Illinois at Urbana-Champaign, a-nathan@uiuc.edu
Introduction
Five papers specifically dealing with the science or engineering of baseball were
submitted to the conference. Four of these papers deal with topics associated with
issues related to the ball-bat collision. The fifth paper deals with the aerodynamics
of a baseball in flight. All of the submitted papers address issues of practical impor-
tance to the game of baseball. What follows is a brief summary of each paper, fol-
lowed by a synopsis of other activity in the field not reported at this conference.
Synopsis of Submitted Papers
The contribution of Smith and Ison investigates the effects of wall rigidity in meas-
urements ofthe ball coefficient ofrestitution (COR), one ofthe important parameters
affecting bat performance. Tue COR is measured by impacting the ball against a flat
rigid surface and is equal to the ratio of rebound to incident speed. The paper,
"Rigid Wall Effects on Softball Coefficient of Restitution Measurements," investi-
gates the effect of wall compliance on the measured COR by impacting softballs
against thin clamped plates of known thickness. The COR was generally found to
increase with decreasing plate thickness, in agreement with FEA simulations also
reported in the paper and with expectations based on a simple model of the trampo-
line effect. These results will be useful in setting specifications for the compliance
of surfaces used for COR measurements. Also reported in the paper is a potentially
novel technique for measuring the drag coefficient on a baseball in flight.
In the paper "Experimental Investigations of the Relationship of Baseball Bat
Properties on Batted-Ball Performance," Shaw and Sherwood investigate the effects
of barrel compression and bat moment of inertia (MOi) on bat performance, using
bats specifically designed to isolate one of these two properties. Tue experiment
involves static measurements of the barrel compression, modal analysis to determine
the frequency of the lowest hoop modes, and high-speed ball-bat collisions to meas-
ure the collision efficiency eA. They find that for the bats selected, the MOi con-
tributed more to eA than did the barrel compression. However, using a particular
4 Alan M. Nathan
swing speed fonnula, it was found that both properties contributed similarly to field
perfonnance.
Russell's contribution, "Bending Modes, Damping, and the Sensation of Sting in
Baseball Bats," uses modal analysis to detennine the frequency and damping rates of
the lowest two bending modes in youth baseball bats. He provides evidence that
damping reduces the sting feit in the top hand of the batter when the contact occurs
off the sweet spot and shows that a novel dynamic absorber tuned to damp the sec-
ond bending mode greatly reduces the sting. He concludes that the sting is mainly
due to the vibrations of the second bending modes, an important practical finding.
Drane, et al. in "An Experimental Investigation of Baseball Bat Durability" de-
scribe a new impact machine for use in the study ofbat durability. The machine uses
an air cannon to fire baseballs at high speed and high repetition rate at a bat which is
suspended vertically from a support system. Accelerometer and strain gage data are
used to investigate various methods for gripping the bat in the machine in order to
find the one that best replicates the gripping method used by players. The machine is
demonstrated to replicate the types of bat failures that is experienced in field use as
well as the gripping method used by players.
In the contribution "The Effect of Spin on the Flight of a Baseball," Nathan, et al.
report the results of new measurements of the lift on a spinning baseball in the range
of speeds and spins relevant for the game of baseball. The experimental technique
involves the use of high-speed motion capture cameras to track the trajectory of the
baseball and measure the spin. The lift coefficients detennined from the data help
resolve a discrepancy in the literature between two different parametrizations.
Synopsis of Related Activities
The relatively small number of baseball-related papers submitted to this conference

is not an accurate measure of the activity in the field. A variety of new investiga-
tions are in progress but not yet at the stage where they can be reported at ISEA2006.
These include the following studies: the effect of ball COR, dynamic stiffness, and
compression on bat perfonnance with the goal of nonnalizing bat performance to
these properties of the ball; oblique ball-bat collisions to leam about the ability of a
batter to put backspin on a batted ball; indirect metrics for baseball bat perfonnance
and their correlation with direct measures of performance; the effect of different grip
methods on baseball bat perfonnance; the visual characteristics observed by a batter
that are related to the spin axis of a pitched baseball; measurements of the effect of
backspin on the distance and optimum takeoff angle of a long fly ball; and refine-
ment of FEA models of hollow bats to include plasticity, with the goal of predicting
denting. With such a !arge level of research activity, we can expect great things at
the ISEA2008 conference.
An Experimental Investigation
of Baseball Bat Durability
Patrick J. Drane, James A. Sherwood and Rebecca H. Shaw

University ofMassachusetts Lowell, james_sherwood@uml.edu
Abstract. The service life of a baseball bat is a function of its durability. All wood bats crack,
and ash bats exhibit flaking of the barre! due to repeated impacts. In aluminum and composite
bats, repeated impacts can cause a change in the material properties, which in turn can lead to
dents and microcracks that ultimately coalesce to form macrocracks. A test machine for simu-
lating essentially any field condition for bat/ball impacts has been developed to study bat
durability. The system uses an air cannon capable of firing a baseball at speeds up to 180 mph
at a stationary bat which is supported in a grip that replicates a player's hands. This paper will
describe the system, present some supporting analysis of the gripping method, and present
results oftests from wood and aluminum bats.
1 Introduction
Baseball bat durability is a topic of importance to baseball players, teams and gov-
eming bodies and bat manufacturers. Players want a bat that is durable with respect
to reliable batted-ball performance, teams want durability with respect to controlling
operating expenses, manufacturers are interested in minimizing warranty claims, and
everyone is concemed about durability with respect to injuries.
What bat properties are important with respect to durability is one question that
engineers, players and fans of the game try to answer. For wood bats, the important
properties may be straightness ofthe grain, growth-ring density, mass density, color,
moisture content, drying method, or the particular forest that makes wood good for
baseball bats. For aluminum bats, the important properties may be the alloy, heat
treatment, forming process, and wall thickness variation. For composite bats, the
important properties may be the material choices for the resin and the fiber rein-
forcement, orientations of the respective layers, and wall thickness variation For
any bat, the durability probably depends on a combination of many of these factors
and, in addition, the bat profile.
Durability testing can be accomplished using players for field testing or batting-
cage studies or using a laboratory hitting machine. Field testing can take a long time
and can be very subjective. Batting-cage studies can be less time consuming, but
require a rotating supply of fresh hitters for every five hits. Of the currently avail-
able bat-performance testing machines, none is practical for doing durability testing
as it can take a long time to obtain the required number of impacts, and the gripping
6 Patrick J. Drane, James A. Sherwood and Rebecca H. Shaw
Fig. 1. Durability Test System
method may not exert forces and constraints on the bat similar to those experienced
in field use. The testing system described in this paper and operational at the
UMass-Lowell Baseball Research Center is capable of gripping the bat similar to
that of a player, getting a number of hits along the profile of the bat in a timely man-
ner, and providing data for quantitatively comparing the durability ofbaseball bats.
2 Testing System
The testing system used for performing the durability testing was developed by
Automated Design Corporation in Romeoville, Illinois in collaboration with the
Baseball Research Center. The system, shown in Fig. 1, operates with an automatic
loading air cannon that fires baseballs into the test chamber. The cannon is capable
of firing baseballs at speeds up to 180 mph. After the ball is fired, it is fed back into
the magazine by an elevator on the back of the chamber. The bat hangs vertically
inside the chamber and is able to pivot about an axis at the handle of the bat when
impacted. The automated air cannon for firing baseballs allows for getting repeat-
able impacts on a bat and for completing durability testing in a relatively short time.
Using the computer software, a range of hit locations and impact speeds can be pre-
scribed to run without operator intervention. The time between hits can be as short
as 5 seconds, thereby allowing for even the most durable bat to be tested to failure
within a few hours.
The bat is mounted in the chamber in such a way that the ball is fired at the
desired location on the bat, and the bat then rotates vertically from the impact. The
bat can be moved up or down between each hit to impact different locations along
the bat. The bat can also be rotated, referred to as clocking the bat, so that different
locations around the barre! can be impacted. Both clocking and moving the bat up or
down to different locations will lengthen the process, but are often weil worth the
extra time. Clocking is a critical component of testing composite and aluminum bats
as they are impacted on different sides in field use and the bat surfaces, especially
with aluminum, can be prone to denting. For the testing ofwood bats, the bat is not
rotated, because in field use the bat should always be impacted parallel to the grain
direction. The grip fixtures for mounting the bat into the machine are shown in
Fig. 2 (canister grip typically for aluminum and composite bats and roller grip typi-
cally for wood bats ).
An Experimental Investigation ofBaseball Bat Durability 7
Fig. 2. Canister Grip (left) and Roller Grip (right)
Table 1. Frequency response as a function of gripping method

Natural Frequency (Hz)
Gripping Me thod 1st Bending Mode 2nd Bending Mode
Free-Free 152 634
Hand-Held Loose 155 628
Hand-Held Tight 148 638
Roller-Grip Tight 222 703
Roller-Grip Loose 166 676
Canister Grip 168 704
3 Evaluation of the Gripping Method
One concem when investigating baseball bats in a simulated hitting scenario is en-
suring that the test replicates the field use of the baseball bat (Shaw and Sherwood
2006). For baseball bat durability testing, the impact speed and the method of grip-
ping the handle need to be realistic. The air cannon can easily duplicate the bat/ball
relative impact speed. To investigate how well the gripping methods in the durabil-
ity machine replicate a player's hands, several studies using strain-gages, acceler-
ometers, and different gripping strengths were preformed in the durability test sys-
tem and with players. These studies quantified how well the machine grip actually
replicates that of a player.
Two accelerometers placed along the length of a bat were used to determine the
first two bending natural frequencies. Experimental data were collected using five
different grip configurations for both fixtures shown in Fig. 2, a player's hands, and
free-free. The results are shown in Table 1. All of the accelerometer data for this
modal investigation were taken with the bat held stationary (i.e., not swinging).
The roller-grip-tight method shows a significant increase in the first natural
frequency when compared to the free-free and hand-held conditions. The roller-grip-
loose method (bottom roll er loosely touching bat handle) and the canister-grip
method both have natural frequencies only slightly higher than the person-held grips
for the first bending mode. The results for the second bending mode show slightly
more separation between the person-held bat and the machine-gripped bats. Just as
with the first bending mode, the roller grip more closely represents a player's hands
when the bottom set of rollers is left loose. The canister grip raises the natural fre-
quency of the second bending mode by about the same amount as the roller-grip
tight. The roller-grip-tight method and the canister-grip method both constrain the
deflection of the bat at two points on the handle, thus changing the effective length
of the bat and thereby changing the second bending mode. The roller-grip-loose
method allows some deflection within the lower pair of rollers, much like a player's
hands would. From these modal data, the roller-grip-loose method and the canister-
grip method are concluded to be good representations in the !ab of a player-held bat
in the field.
Strain gage data were collected for several hits from three college players using
the same bat, and the respective impact locations were marked after each hit. The
same bat was then loaded into the durability machine, and the bat was impacted at
the same locations as the field hits using each of the three gripping methods. The
impact speeds were varied to match the amplitudes ofthe peak strain and the shapes
of the responsc with those observed in the field-test data. Estimates for pitch speeds
and player swing speeds were used for comparison. Fig. 3 shows the strain-gage
response for one of the gages to a hit off live pitching for one of the eo liege players.
The strain response is also shown for a similar impact in each of the three gripping
methods used in the durability machine. In the strain response from the live hit,
there are clearly two modes present right after impact, and then the second mode
damps out leaving only the first bending mode in the response. A similar response is
seen in each of the grips in thc durability machine. However, there is less overall
damping prescnt in the durability-machine responses-implying a player's hands
absorb more of the vibration than the rubber rollers used in the machine grips. The
peak strain secn in a field hit where the impact occurred about 9 in. from the end of
the barrel was matched in each of the three machine grips with impacts of -100 mph.
The strain results for each of these hits are shown in Fig. 3. The shape of the strain
response is matched closely for all of the grips, but is best matched by the canister
grip for the aluminum bat used in this study.
The accelerometer and strain gage data support the conclusion that the machine
can be a good simulation of field-service conditions for durability testing in a lab
environment. Of the three grips tested, the roller-grip-loose method and the canistcr-
grip method are good representations of a player's hands. Based on experience, the
roller-grip-loose method is less time-consuming to load a bat than is the canister-grip
method. When testing wood bats, which can crack in the handle region and after
only a few hits, the roller-grip-loose method is the preferred gripping method in
comparison to the canister grip, because a bat can be loaded and unloaded relatively
quickly and a crack in the handle can be easily observed. When testing aluminum or
composite bats, which tend to fail in the barrel region of the bat, the canister grip is
the preferred gripping method as it allows the bat to be rotated between impacts as
the bat would be in the field.
An Experimental Investigation ofBaseball Bat Durability 9
Fig. 3. Strain gage measurements for four different grips during impacts
4 Durability Testing Methodologies
The baseball bat durability testing requires a protocol to ensure that the collected
data are comparable and sufficiently comprehensive for making conclusions with
respect to absolute and relative durability of the tested bats. An example routine that
would be programmed for testing an aluminum or composite baseball bat may in-
clude impacting different locations along the length of the bat as weil as different
locations around the barre! and varying the speed of the impact. The motors in the
durability machine can be controlled so that the testing could begin, for example,
with several impacts at the 6-in. location (measured from the end ofthe barre!), then
several impacts at the 4-in. location and then move to impact the bat several times at
the 8-in. location. The clocking device is often programmed to rotate the bat 1/4-,
1/4-, 1/4-, and 3/8-tum after each impact allowing eight consecutive hits to impact
the bat on eight evenly spaced locations around the barre!. These rotations are typi-
cally an important part of not unfairly causing premature denting of the surface.
Another aspect of the programming will adjust the firing velocity, by adjusting the
cannon pressure. The speed of impact, for example, can be adjusted to account for
the change in velocity of the swing speed of a bat as the impact location is adjusted
along the length of the bat. These example routine components allow for consider-
able flexibility when testing baseball bats, and depending on the routine, different
results will be attained.
Fig. 4. W ood bat broken in the durability machine with high-speed camera view
Fig. 5. W ood bats broken in two pieces
Fig. 6. Aluminum samples cracked and dented
5 Results
Baseball bats break many different ways as a result of field use, and Figs. 4, 5 and 6
show that similarly diverse results can be obtained from testing in the durability
testing system.
6 Conclusions
The durability testing system described in this paper and operational at the Baseball
Research Center is capable of replicating the types of failures that baseball bats ex-
perience during field use. The gripping methods used in the machine replicate a
player's grip.
References
Shaw, R. H. and Sherwood, J. A. (2006) Exploring the Crack ofthe Bat in the Lab: Perform-
ance and Durability", IMAC XXIV Conference Proceedings.
Bending Modes, Damping, and the Sensation
of Sting in Baseball Bats
Daniel A. Russell
Science and Mathematics Department, Kettering University, Flint, MI
drussell@kettering.edu
Abstract. The painful sensation of sting in thc top hand of a player holding a baseball or
softball bat may be a deterrcnt to enjoying the game, especially for young players. Several
mechanisms for rcducing the vibration of bending modes have been implemented in youth
baseball bats in order to reducc sting. One method of assessing thc effectiveness of thesc
mechanisms is to compare the damping rate they providc for the first two or threc bending
modes in a bat. Damping ratcs are comparcd for scveral wood, aluminum, composite, and
two-picce construction baseball bats, in addition to scveral bats with special damping control
mechanisms. Experimental evidcnce suggests that damping mcchanisms which reduce the
vibration of the sccond bending mode arc prefcrred by players. A novcl dynamic absorber in
the knob is shown to effectivcly rcduce thc vibration of the second bending mode and mini-
mize thc painful sting feit in thc top hand.
1 The Problem of Hand Sting
The problem of sting is often a deterrent to young players who are leaming how to
swing a baseball bat. When thcy do make contact with a pitched ball, young players
often hit the ball in the taper region or at the very end ofthe barre!. The painful sting
resulting from such poor impacts can be very frustrating, and can discourage young
players from continuing on in the sport. The prob lern of sting is not limited to young
players, however, and accomplished adult players will still occasionally hit the ball
badly resulting in painful sting in the hands. Discussions with players reveal that
impacts near the taper region in the bat often result in a sharp pain in the fleshy re-
gion between the thumb and forefinger of the top hand. This pain is significant
enough to sometimes cause bruising, and can persist for several days afterwards.
Aluminum bats tend to sting more than wood bats, and while the development of
specially designed padded batting gloves and special thick rubber grips on the han-
dles of aluminum bats has improved the sensation of feel somewhat, the problem of
sting still remains.
Scveral means of reducing vibration have been implemented. Because the prob-
lern of hand sting is more pronounced at the youth level, many of vibration reduction
mechanisms only appear in youth baseball bat models. These include inserting a
dynamic absorber (tuned-mass-damper) in the taper region ofthe barre), inserting an
elastomer plug into the knob in the handle, a two-piece construction in which the
12 Russell
First Bending Mode

Q)
"O
-~
C.
E
<(
----------- : Top
: Hand
"O
-~
cij
E
.. - +: - ..
'
.
'
0
z
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
Distance from Barrel End (inches)
Fig. 1. Mode shapes for the first two bending vibrational modes in a 32-inch youth
baseball bat. Barrel end is at the left and handle is at the right.
handle and barrel are separate pieces connected by a rubber joint, and the injection of
foam into the hollow handle.
2 Bending Modes and Hand Location
All baseball and softball bats exhibit a family of bending vibrations, similar to what
one would find in a free-free beam, with nodes (locations of zero displacement) on
the barrel and handle and antinodes (locations of maximum displacement) between
nodes. Most of the research published on the issue of sting has focused either on the
impact location relative to the nodes of the first two bending modes of vibration
(Noble and Walker 1994; Cross 2001) or on the frequencies ofthe lowest two bend-
ing modes (Noble, Walker and Ponte 1996).
Figure 1 shows the mode shapes for the first two bending modes for a typical 32-
inch youth baseball bat. The mode shapes were obtained by performing an experi-
mental modal analysis measurement. Two features of this graph are relevant to the
discussion of sting. First is the location of the nodes at the barrel end of the bat. The
first bending mode has a node approximately 7-inches from the end ofthe barrel, and
the second bending mode has a node approximately 5-inches from the barrel end.
An impact at a node will prevent the corresponding mode shape from contributing to
the resulting vibration of the bat. The region between 5-7 inches from the barrel end
is often referred to as the "sweet zone" due to the fact that impacts within this region
result in minimal vibration in the handle (Cross 1999; Cross 2001 ).
Of greater importance to the perception of sting is the location of the nodes and
anti-nodes at the handle end of the bat. The bottom hand is centered on a node for
the second bending mode and the fleshy part at the base of the bottom hand is at an
antinode for the first bending mode. This would suggest that, if sting is the result of
bending vibrations, the bottom hand should be more responsive to the first bending
mode but not affected much by the vibration of the second bending mode. The top
hand, meanwhile, is centered on a node for the first bending mode and the region
between the thumb and forefinger of the top hand is located at an antinode for the
second bending mode. This would suggest that the top hand is most responsive to
Bending Modes, Damping, and the Sensation of Sting in Baseball Bats 13
vibration of the second bending mode, and less to the vibration of the first bending
mode. Cross argued that the sting is the result of the impulse from the bat-ball colli-
sion traveling to the player's hands rather than the result of the bat vibrating in its
various free modes of vibration (Cross 1998). However, he does point out that the
impulse is indistinguishable from the vibration ofthe second bending mode.
3 Damping Rates for Bending Modes
lt has been shown (Brody 1986) that the natural frequencies of a baseball bat are not
significantly altered when the bat is gripped in the hands, thus allowing the hand-
held bat tobe modeled as a free-free bat (Nathan 2000). The hands do, however, add
a huge amount of damping so that the natural vibrations of the bat decay very
quickly. What is not known, however, is exactly how much damping the hands
provide nor how much damping is inherently present in the bat itself. There are very
little published data showing measured damping rates for the bending modes of
baseball bats. The data that do exist suggest that damping rates for aluminum bats
are roughly half those of wood bats (Collier 1992; Naruo and Sato 1998). There are
no data available for composite bats, nor for youth bats with vibration reduction
mechanisms. One ofthe aims ofthis paper is to provide some damping rate data.
The damping rate for a particular mode of vibration is one of the modal parame-
ters (mode shapes, frequencies, and damping) that may be determined by curve fit-
ting the frequency response functions (FRF) collected in a modal analysis experi-
ment (Gade, Herlufsen and Konstantin-Hansen 2002). The analytical function used
to perform the curve fitting assumes that the structure may be modeled as a 2nd order
time invariant system with an impulse response function ofthe form
L [R~-~)1e-akt sin(2efkt + ~)-_7,))],

n
hr_,.(t) = (1)
k=I
where hrsCt) is the impulse response at location r due to an excitation at location s,
and Rr,Ckl is the residue (mode shape) at location r due to excitation at location s for
mode k. Equation ( 1) indicates that the vibration resulting from an impulse is the
superposition of sinusoidal oscillations, each at their own natural frequency fk and
exponential damping rate crk. The quantity of interest in the present analysis is the
modal damping rate crk for the first two bending modes.
Most experimental modal analysis software packages report the modal damping
in terms of a non-dimensional critical damping ratio ~k, usually expressed as a per-
centage. The critical damping ratio is related to the modal frequency and modal
damping coefficient by (Formenti 1999)
q- (Yk (2)
k- ✓(YI + (2Jifk)2 .
In our laboratory we extract the damping rate by suspending a baseball bat vertically
from the knob using rubber bands. An accelerometer is attached to the knob, and the
bat is impacted with a force hammer at the barre! end. The Frequency Response
Function consisting of the ratio of acceleration/force is obtained using a two-channel
14 Russell
FFT analyzer and curve fitted to extract the critical damping ratio Sk- The damping
rate <h is determined from Eq.(2).
Damping rates for the first two bending modes of a sampling of youth baseball
bats of varying construction are shown in Table l. The data show that aluminum
bats have very little inherent damping. Wood and composite bats have similar
damping rates, both having damping rates that are approximately an order of magni-
tude greater than aluminum bats. The aluminum bats marked with '*' include a
vibration reduction mechanism which significantly increases the damping of either
the first and/or the second bending mode.
Table 1. Damping rates for wood, aluminum and composite youth baseball bats. Bats
marked with •*' include a vibration reduction mechanism.
First Bcnding Mode Second Bending Mode

Damping Damping
Frequency Damping Frequency Damping
Bat Type Rate Rate
/(Hz) Ratio~ /(Hz) Ratio~
c; (s-1) c; (s-1)
wood-ash 187 3.368c-3 3.96 691 5.009e-3 21.7
wood-ash 212 3.916c-3 5.22 663 l .209e-3 50.4
wood- maple 175 6.713e-3 7.38 580 4.278e-3 15.6
Aluminum 229 4.654e-4 0.67 763 7.844e-4 3.8
Aluminum 190 8.428e-4 1.01 690 1.019e-3 4.4
Aluminum 201 l.326e-3 1.67 780 8.224e-4 4.0
Aluminum * 163 l. l 12c-2 11.39 559 2.092e-2 73.5
Aluminum * 211 l.757e-2 23.30 752 2.2 l 3e-3 10.5
Aluminum * 197 6.439e-2 79.87 697 2.309e-3 10.1
Composite 168 3.966c-3 4.19 615 3.593e-3 13.9
Composite 105 6.433c-3 4.24 405 5.702e-3 14.5
Compositc 137 6.038e-3 5.20 529 6.837e-3 22.7
4 Evidence that Damping Reduces Sting
A preliminary correlation between damping and the perception of sting came

from an opportunity to test three youth baseball bats for a manufacturer. Bat A was
brand new (still in plastic wrapper) and served as a control while bats B and C had
been modified in an attempt to reduce sting, and had each been hit by 70 players.
The source providing the bats informed us that every single player preferred the
same bat because it feit better, but we were not told which bat was preferred. We
were asked to try to identify the preferred bat and explain why. Modal testing re-
vealed that all three bats had nearly identical bending and hoop frequencies. The
only difference between the bats was in the amount of damping for the first and
second bending modes. This was immediately apparent by gripping the bat barre!
lightly at the "sweet spot" and tapping the barrel. The vibration from bat C died out
immediately while bat B and the control bat A rang for several seconds. Measured
damping rates, shown in Table 2, show that the preferred bat (bat C) had roughly 6-8
Bcnding Modes, Damping, and the Sensation of Sting in Baseball Bats 15
times more damping for the first bending mode, and 20-30 times greater damping for
the second bending mode.
Table 2. Damping rates for three identical youth baseball bats, with bat C being pre-
ferred unanimously by 70 different players.
First Bending Mode Second Bending Mode

Frcquency Damping Frequency Damping
Bat
/(Hz) Rate cr (s- 1) /(Hz) Rate cr (s- 1)
A 173 0.58 643 2.8
B 172 0.74 641 3.6
C 167 4.82 623 82.4
A second correlation between sting and the damping rate of the second bending
mode is currently being investigated with the implementation of a novel dynamic
absorber (Albin 2004) into the knob of aluminum baseball and softball bats. This
vibration absorber may be tuned to reduce the vibration at a specific frequency by
adjusting the mass of the plug and/or the stiffness of the rubbcr support. The knob
with the absorber is )arger than a normal bat handle knob, and the combined mass of
the knob and absorber lowers the frequencies of the first two bending modes. Table 3
lists the damping rates for a 32-inch (81.3cm) youth senior league baseball bat with-
out the damper and with the damper tuned to the first and second bending modes.
When the vibration absorber is tuned to the frequency of either bending mode, the
amount of damping for that mode is huge, while the damping rate for thc other mode
is not significantly altered.
Table 3. Damping rates for a baseball bat with and without a dynamic absorbcr in thc
knob that has bcen tuned to thc first and sccond bending mode.
First Bending Mode Sccond Bending Mode

Frequency Damping Frcqucncy Damping
Bat
/(Hz) Rate cr (s- 1) /(Hz) Rate cr (s- 1)
No damper 162 1.75 582 3.3
With damper 1 146 124.6 547 8.5
With damper 2 142 1.52 573 182.0
Preliminary field tests, using bats with this absorber in the knob, indicate that the
painful sting in the top hand resulting from an impact near the taper region of the bat
can be greatly reduced by tuning the absorber to the second bending mode of vibra-
tion. In an attempt to further quantify the relative importance of the damping for the
first and second bending modes, we have instrumented a bat, with the tunable ab-
sorber in the knob, with strain gauges on the handle in ordcr to measure the force
under the hands during and following an impact with a ball. Adjusting the tuning of
the absorber allows variation of the damping rates of the first and second bending
modes, to compare how eithcr or both influence the perception of feel. This further
study was still in progress at the time this paper was submitted.
As a final demonstration of how increased damping might improve the feel of a
bat, Fig. 2 shows thc frequency response curve of the vibration amplitude at the
location of thc top hand for the baseball bat in Table 3. The dashed curve is for the
16 Russell
60
55 - - - Normal Bat
- With Damper in Knob
50
45 '
,,'
in 40 ,,'
~ ,,
,,
Q) 35 ,,
,,
(/)
30 1 32dB
C
;: ''
''
0
a.
(/)
25 ,: :' ''
Q)
a: 20
>, 15
u
C 10
(1)
:::,
5
O'"
~ 0
u..
·5
-10
·15
·20 ~-~~~~~~~~~~~~--~~~~~~~~-
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Frequency (Hz)
Fig. 2. Frequency response function for a baseball bat with a tuned-mass damper in the
knob. Tuning the damper to the frequency of the second bending mode effectively re-
moves that mode from the resulting vibration ofthe bat.
bat without the absorber, and the solid curve is for the bat with the damper inserted
and tuned to the second bending mode. The dynamic absorber reduces the vibration
amplitude of the second bending mode by approximately 32 dB, effectively remov-
ing this mode from the vibration of the bat following an impact with a ball away
from the sweet spot.
References
Albin, J. N. (2004) U.S. Patent No. 6,709,352. Washington, DC: U.S. Patent and Trademark
Office.
Collier, R. D. (1992) Material and structural dynamic properties of wood and wood composite
professional baseball bats. Proceedings 2 nd lnt. Congress on Recent Developments in Air
and Structure Borne Sound and Vibration, Aubum University, Aubum, AL, pp. 197-204.
Cross, R. (1998) The swcet spot of a baseball bat. Am. J. Phys. 66(9), 772-779.
Cross, R. (2001) Response to "Comment on 'The sweet spot of a baseball bat."' Am. J. Phys.
69(2), 231-232.
Gade, S., Herlufsen H. and Konstantin-Hansen, H. (2002) How to Determine the Modal Pa-
rameters of Simple Structures. Sound & Vib. 36(1 ), 72-73.
Formenti, D. (1999) Thc Relationship Bctween % ofCritical and Actual Damping in a Struc-
ture. Sound & Vib. 33(4), 14-18.
Nathan, A. (2000) Dynamics of the baseball-bat collision. Am. J. Phys. 68(11 ), 979-990.
Naruo, T. and Sato F. ( 1998) An experimental study of baseball bat performance. In: Haake,
S. (Ed.), Engineering of Sport - Design and Devefopment. Blackwell Pub., pp. 46-52.
Noble, L. and Walker, H. (1994) Baseball Bat Inertial and Vibrational Characteristics and
Discomfort Following Ball-Bat Impacts. J. Appl. Biomechanics. 10. 132-144.
Noble, L., Walker, H. and Ponte, M. (1996) The effect of softball bat vibration frequency on
annoyance ratings. Proceedings ofthe 14th International Symposium on Biomechanics in
Sport, Funchal, Portugal. 371-374.
Experimental Investigations of the Relationship
of Baseball Bat Properties on Batted-Ball Performance
Rebecca H. Shaw and James A. Sherwood

University of Massachusetts Lowell, becky@baseballrc.eng.uml.edu
Abstract. Laboratory tests are used to investigate the relationship between baseball bat
performance and two bat properties: moment of inertia (MOi) and barre! stiffness for
aluminum and composite bats. Each bat used in the current study is spccifically designed and
manufactured to isolate a particular property. Static tests, e.g. three-point bend and barre!
compression, are used to characterize the properties of each bat. The natural frequencies of
the bat are measured using modal techniques. Dynamic perforrnance testing is done using an
air cannon capable of throwing a baseball at collision specds equal to those seen in field play.
For the bats studied, variation in MOi contributed more to perforrnance in the !ab than did
barre! stiffness. However, the changes in predicted field perforrnance due to the two properties
were similar.
1 Introduction
The Ball Exit Speed Ratio (BESR) is the metric currently used to quantify the
performance of nonwood baseball bats. This research examines two bat properties,
barre! stiffness and mass moment of inertia (MOi), and their relationship to
performance. For this paper, performance is the speed of the ball as it leaves the bat-
ball collision as measured in the !ab and predicted in the field. Static nondestructive
tests are used to measure the barre) stiffness and handle stiffness of each bat before
performance testing. Modal tests are used to determine the first two bending natural
frequencies and the first hoop frequency. Each bat was manufactured to isolate a
singe property as closely as possible, e.g. all properties equal except for MOi.
Laboratory tests are used to determine the performance of each bat, and the results
are compared to predictions using the BESR. All testing is done using an air cannon
following the NCAA (2005) baseball bat certification protocol.
2 Background
One metric used to measure baseball bat performance is the BESR (Carroll 2000),
BESR=VR+0.5 (1)
VI
18 Rebecca H. Shaw and James A. Sherwood
. 1 changes m
Table 1. Th eoretJca . MOI
. 1ab an d fi1e ld batte d-ba11 speeds due to changes m
Relative Relative Relative
MOi Lab Lab Change in Field Field
MOi
(oz- BESR BES BES Swing BES BES
Class
in2) (mph) Diff. Speed (mph) Diff.
(mph) (mph) (mph)
Low 9000 0.701 93.4 -6.6 3.5 97.6 -2.4
Med 11000 0.750 100.0 0 0.0 100.0 0.0
High 13000 0.786 104.9 +4.9 -3.4 100.6 0.6
where V1 is the ball inbound speed and VR is the ball rebound speed for a collision
with a stationary bat. The BESR equation can also be written as,
BESR = l + 2e- µ* (2)
2(1 + µ*)
where e is the bat-ball coefficient of restitution (COR) and,
(3)
where I is the mass moment of inertia (MOi) measured about the axis of rotation
(6 inches from the knob end of the bat), mb is the mass of the ball and x is the
distance from the axis ofrotation to the impact location. Using Eqs. 1, 2 and 3, the
variation in BESR due to MOi can be calculated assuming the bat-ball COR
(coefficient of restitution) remains constant. If it is assumed that a medium MOi bat
has a BESR of 0. 750, then high and low MOi bats will have BESR values as shown in
Table 1. These calculations are all done assuming impact at the 6-in. location (as
measured from the tip ofthe barrel). The differences in performance as quantified by
the batted-ball speed (BBS) can be found using,
BBS = v(BESR-0.5)+ V(BESR+0.5) (4)
where v is the ball pitch speed (mph) and V is the bat swing speed (mph). For these
calculations, a 70-mph pitch speed and a 66-mph swing speed are assumed.
For bats in this MOi range, the high MOi bats are expected to hit ~5 mph faster
than the medium MOi bats and ~ 11.5 mph faster than the low MOi bats in the lab.
These calculations do not account for changes in player swing speed due to MOi.
Swing speed is known to be inversely proportional to MOi. Therefore, using one
swing speed for all bats in the lab is not a true representation of field performance. A
iatting cage study by Crisco and Greenwald (1999) analyzed the swing speeds of
1layers for different bats. The data from this study were analyzed by Nathan (2001,
003), and the relationship between bat swing speed and MOi was found to be
.2 x 10-3 mph/oz-in2, where swing speed is measured 6 inches from the end of the
1arrel and MOi is measured about the knob. The results for batted-ball speed
adjusted for swing speed are presented in the last three columns of Table l.
Adjusting for change in player swing speed brings the predicted field performance
(Field BBS) ofthe three bats much closer together than the Lab BBS values.
Investigations ofthe Relationship ofBaseball Bat Properties on Batted-Ball Performance 19
102
~
100
J ,,,, -- - -
98
~ 96 --
- - - e=O.S:O
94 - - - e:•0.629
92
8COO 10:00 12((0 14(00 16)00
MOi (az-itt)
Fig. 1. Batted-ball speed vs. MOi adjusting

for changes in swing speed
Eq. 2 can be combined with the swing speed model to determine the "ideal"
MOi for maximum batted-ball speed at a particular impact location. The results are
shown in Fig. 1. For these calculations, it is assumed that the MOi about the knob
varies linearly with MOi about the axis of rotation. lt is also assumed that the
average college player swings a medium MOi bat with a speed of 66 mph. These
calculations do not consider movement of the sweet spot due to changes in MOi.
This sweet spot vs. MOi relationship will be discussed later in this paper.
Figure 1 shows two different curves for BBS vs. MOi, one assuming e=0.529
and one assuming e=0.500. Both curves reach a peak at ~12600 oz-in2 indicating
that the ideal MOi is not dependant on e. This peak represents the ideal MOi for
maximum batted-ball speed for impacts at the 6-in. location.
An empirical model relating hoop frequency to softball bat performance was
developed by Russell (2004). The maximum efficiency was shown tobe at a hoop
frequency of just less than 1000 Hz. Because the bat-ball collision time is
approximately 0.001 s, a frequency of 1000 Hz would correspond to the barre!
moving in and out in harmony with the ball contacting the bat, thus minimizing the
collision energy lost to ball deformation. lt is assumed that the relationship between
baseball bat performance and hoop frequency would be similar to that of the softball
bat model. However, the maximum performance would occur at a slightly different
hoop frequency due to differences in the collision time between softball and baseball.
3 Results
3.1 Barrel Stiffness

Three bats were manufactured to have all properties equal except for the barre!
stiffness. Barrel stiffness was measured two ways, with a barre! compression test at
5 in. from the end of the barre! and with a hoop frequency measurement. The hoop
frequencies for the bats used in this study range from 2360 to 3950 Hz. Based on
Russell's observations, a dramatic change in performance due to hoop frequency is
not expected for these bats. However, the batted-ball performance should increase
slightly as hoop frequency decreases. Fora dramatic change due to hoop frequency,
it is expected the bats would need to be in the range of 1000 to 2000 Hz. The results
Table 2. Barre1 st1"ffness an d perfiormance results fior t hree baso 1 erent sfffn
t f d'ffi 1 esses
Lab
Barrel Sweet BBS e at the
Barrel Hoop
MO! Compression Spot at the sweet
Stiffness Freq.
(oz-in2) 5-in. Avg. Location sweet spot
Class (Hz)
(lbs) (in.) spot location
(mph)
Low 10106 723 2360 4.5 98.2 0.528
Med 10007 873 2670 4.0 96.6 0.513
High 10923 1239 3950 5.0 96.5 0.490
Table 3. Barre 1 stJ"ffness perfiormance result s ad.11usted fior d'ffi . MOI

1 erences m
Predicted
Hoop Sweet Spot e at the Field BBS
MO! BBS Diff
Freq. Location sweet spot at the
(oz-in2) (mph)
(Hz) (in.) location sweet spot
(mph)
2360 10000 4.5 0.528 97.8 1.3
2670 10000 4.0 0.513 96.5 0.0
3950 10000 5.0 0.490 93.2 -3.3
for these tests along with the experimental performance results are shown in Table 2.
All bats were tested per the 2005 NCAA Bat Certification Protocol.
The performance results follow the expected trend-as stiffness decreases BBS
increases. The stiffness measurements show the low- and medium-stiffness bats to
be much closer in stiffness than the medium- and high-stiffness bats. lt would
therefore be expected that the low- and medium-stiffness bats would have similar
performance results, and the high-stiffness bat would have a much lower
performance. However, the performance test results show the high- and medium-
stiffness bats to be very similar. The reason for the difference in performance from
the expected performance is the MO! difference between the bats. The low- and
medium-stiffness bats only differ in MO! by about 100 oz-in2, but the high-stiffness
bats are about 900 oz-in2 higher than the medium-stiffness bats. The increase in
MO! would cause an increase in performance. Thus, the high-stiffness bats are
higher performing than they would be if they had the same MO! as the other bats.
Using Eqs. 2 and 3, the e value in the BESR term can be backcalculated. The e term
represents the "bat-ball COR", or the performance due to all factors other than MO!.
lt is expected that the values for e at the sweet spot location will increase as the
stiffness decreases. These results are shown in Table 2. Looking at the results using
e shows a larger difference between the medium and high stiffness bats-as was
expected. In addition, the BBS values can be adjusted for MO! using Eqs. 2
through 4. Table 3 shows field-use BBS values for each bat adjusted to an MO! of
10000 oz-in 2 . As was expected, the low- and medium-stiffness bats have similar
performance values, and the high-stiffness bat is relatively low performing.
lnvestigations of the Relationship of Baseball Bat Properties on Batted-Ball Perfonnance 21
Table 4. Perfonnance results for bats with different MOi values

Barrel Lab BBS
Hoop Sweet
MOi Compression atthe e at the
Frequency Spot
(oz-in 2) 5-in. Avg. sweet spot sweet spot
(Hz) Location
(lbs) (mph)
9218 2470 821 5.0 91.9 0.503
9259 2500 843 5.0 92.8 0.510
10912 2530 808 4.5 100.9 0.528
11199 2560 762 4.5 100.7 0.518
12810 2920 827 4.0 105.7 0.519
12722 2910 800 4.0 104.6 0.511
Ta bl e 5. Battc d-baII spee d ca 1cu 1at10ns

. f,or the three bat mo de1s w1t. hd'ff,
1 erent MOi va ues
Sweet Avg. Relative Predicted
Avg. Predicted
MOi Spot Lab LabBBS Field BES
MOi Field BES
Class Location BES Diff. Diff.
(oz-in 2 ) (mph)
(in.) (mph) (mph) (mph)
Low 9239 5.0 92.3 -8.5 96.1 -4.5
Med. 11056 4.5 100.8 0.0 100.6 0.0
High 12766 4.0 105.1 4.3 101.0 0.4
3.2 Moment of Inertia

Six bats were manufactured with three different values for MOi, classified as low
(handle-loaded), medium (balanced), and high (end-loaded). The MOi values and
performance results are shown in Table 4. The range of barre! compression values
seen here is small compared to the range considered in Section 3.1.
The performance results in Table 4 show a clear increase in performance with an
increase in MOi. The BESR equation (Eq. 2) can be used to separate the MOi term
from the "bat-ball COR" term, e, as discussed previously. If each of these bats had
identical properties except for MOi, then each would have the same value for e. The
values for e at the sweet spot locations are presented in the last column of Table 4.
The values for e range from 0.503 to 0.528, and there does not appear to be any
correlation between MOi and e-indicating that the differences in e are due to
properties other than MOi. The variation in e seen here is similar to that between the
low- and high-stiffness bats discussed in Section 3.1. lt was seen earlier that for bats
with the same MOi a difference in e of 0.038 resulted in a difference in maximum
lab or field BES of 4.6 mph. As a result, a difference in e of 0.025 as seen here
would result in a difference of about 3 mph in BES. The difference in maximum lab
BES between the handle-loaded bats and the end-loaded bats is 12.8 mph. Therefore,
the variation in !ab BES due to the variation in e is small compared to the variation in
!ab BES due to MOi. The BES values of the two bats for each MOi class were
averaged, and the results are presented in Table 5.
Table 5 shows a batted-ball speed difference of 12.8 mph for a difference in
MOi of 3527 oz-in 2 . The predicted difference was 11.5 mph for an MOi difference
of 4000 oz-in2, assuming the sweet spot tobe at the 6-in. location. The balanced bats
had the highest values for e, which resulted in the balanced bats having a slightly
higher BBS than predicted-shifting them farther from the handle-loaded bats and
closer to the end-loaded bats. The total difference is also slightly increased due to
the fact that the sweet spot is not at the 6-in. location. Moving the sweet spot out to
the 4.5-in. location for the predicted calculations increases the difference due to MOi
slightly. For these bats, the sweet spot moves closer to the barre! end as MOi
increases. As discussed previously, changes in MOi have a direct effect on player
swing speed. Using the swing speed model developed by Nathan (2001, 2003), the
predicted field BBS values were calculated and are presented in the last two columns
ofTable 5.
Adjusting for player swing speed brings the performance of the low- and high-
MOi bats to within 5 mph of each other. The change in field performance due to
MOi (-5 mph) is similar to the change in performance due to barre! stiffness seen in
Section 3.1 (-4.6 mph) for the range ofproperties studied.
4 Conclusion
The equation for BESR states that baseball bat performance is dependent on two
factors: moment of inertia and bat-ball COR, or e. Experimental data from bats of
varying MOi correlate weil with results predicted using the BESR equation.
Increasing MOi increases performance in the !ab, but it also makes the bat more
difficult for a player to swing in the field. Variation in swing speed due to MOi is
currently not considered in the !ab test protocol. As a result, changes in MOi appear
to have a !arger effect on performance in the !ab than would be seen in the field.
Bat-ball COR is dependent on several factors, one of which is barre! stiffness. The
data from the bats ofvarying barre! stiffness show an increase in bat-ball COR with a
decrease in barre! stiffness. For the bats studied, predicted changes in field
performance due to MOi and barre! stiffness were similar.
References
Carroll, M.M. (2000) Assessment and regulation ofbaseball bat performance, Symposium on
Trend1· in the Application of Mathematics to Mechanics, edited by P.E. O'Donoghue and
J.N. Flavin (Elsevier, Amsterdam).
Crisco, J.J., Greenwald, R.M., Penna, L.H. (1999) Baseball Bat Performance: A Batting Cage
Study, Draft Report, July 14, 1999. http://www.nisss.org/BBSPEED6a.html.
Nathan, Alan M. (2001) Baseball and Bat Performance Standards, Presentation to the NCAA
Research Committee, June 13, 2001.
Nathan, Alan M. (2003), Characterizing the Performance of Baseball Bats, Am. J. Phys. ll,
pp. 134-143 (2003).
NCAA (2005) Baseball Bat Certification Protocol.
Russell, D.A. (2004) Hoop frcqucncy as a predictor ofperformance for softball bats, The
Engineering ofSport 5, Vol. 2, pp. 641-647 edited by M. Hubbard, R.D. Mehta, J.M.
Pallis.
The Effect of Spin on the Flight of a Baseball
Alan M. Nathan 1, Joe Hopkins2, Lance Chong 1, and Hank Kaczmarski 1
1 University of Illinois, a-nathan@uiuc.edu

2 Western Michigan University
Abstract. New mcasurements arc presented of the lift on a spinning baseball for speeds in the
range 50-110 mph and spins 1500-4500 rpm. The experiment utilizes a pitching machine to
project the baseball horizontally; a high-speed motion capture system to measure the initial
velocity and angular velocity and to track the trajectory over -5 m of flight; and a ruler to
measure the total distance traversed by the ball. The lift coefficients are extracted from the
data and comparcd to with previous measurcments or parametrizations.
1 Introduction to the Problem
In a recent paper, Sawicki et al. (Sawicki, Hubbard, and Stronge 2003) report a study
of the optimum bat-swing parameters that produce the maximum range on a batted
baseball. Using a model for the ball-bat collision and recent experimental data for
the lift and drag coefficients, they tracked the ball from collision to landing. For
given initial speed, angle, and spin of the pitched baseball, the bat swing angle and
undercut distance were varied to maximize the range. The study found the surprising
result that an optimally hit curveball travels some 12 ft. farther than an optimally hit
fastball, despite the higher pitched-ball speed of the fastball. The essential physics
underlying this result has to do the with the aerodynamic lift force on a baseball
projected with backspin. In general, a baseball will travel farther if it projected with
backspin. lt will also travel farther if is projected with higher speed. In general a
fastball will be hit with a higher speed. However, a curveball will be hit with !arger
backspin. The reason is that a curveball is incident with topspin and hence is already
spinning in the right direction to exit with backspin. A fastball is incident with back-
spin so the spin direction needs to reverse to exit with backspin. lt then becomes a
quantitative question as to which effect wins: the higher speed of the fastball or the
!arger backspin of the curveball. According to Sawicki et al., hereafter referred to as
SHS, the latter effect wins and the curveball travels farther.
The conclusion of SHS depends critically on the size of the lift force on a spin-
ning baseball. SHS used a particular model for the lift based largely on experimen-
tal data that will be reviewed in the next section. That model and the conclusions
that follow have been criticized by Adair (Adair 2005), who claims that SHS grossly
overestimate the effect of spin on the flight of a baseball. The goal of the present
paper is to resolve the disagreement between SHS and Adair, hereafter referred to as
RKA, by performing new measurements of the effect of spin on the flight of a base-
ball.
24 Alan M. Nathan et al.
2 Previous Determinations of Lift
When a spinning baseball travels through the atmosphere, it experiences the force of
gravity in addition to the aerodynamic forces of drag and lift, F0 and FL, Conven-
tionally the lift force is parametrized as FL= ½CLpAv 2 where Ais the cross sectional
area of the ball, v is the speed, p is the air density, and CL is the lift coefficient. In
the SHS parametrization, CL depends only on the spin parameter S=RmN and is a
rough fit to the data of Alaways (Alaways 1998; Alaways and Hubbard 2001) and
Watts and Ferrer (Watts and Ferrer 1987). Alaways used a motion capture technique
to determine CL for speeds up to approximately 75 mph (Re<l.6 x 105) and for
0.1 <S<0.5. Watts and Ferrer measured CL in a wind tunnel experiment for speeds up
to 37 mph (Re<0.8 x 10 5) and 0.4<S<l.0. RKA uses a parametrization (Adair 2002)
which follows from the argument that the lift on a spinning ball is related to the
difference in drag between two sides of the ball. This prescription leads to lift values
that are in good agreement with the Watts and Ferrer data and with SHS at low
speeds. RKA argues that extending this prescription to higher speeds leads to predic-
tions in reasonable agreement with observations from the game itself (Adair 2002),
such as the maximum break of a pitched curveball. lt is at the high speeds that are
typical of pitched or hit baseballs that the differences between SHS and RKA are
most pronounced. For example, at 100 mph and m= 1800 rpm, the SHS lift is about
three times !arger the RKA lift. This will have dramatic implications for the distance
traversed by a long fly ball.
3 Description of the Experiment
The essential technique is to project the ball approximately horizontally with back-
spin or topspin and to use a motion-capture system to measure the initial velocity and
angular velocity and track the trajectory over approximately five meters of flight.
Under such conditions, the vertical motion is particularly sensitive to the lift force,
which leads to a downward acceleration smaller or !arger than g when the ball has
backspin or topspin, respectively. Additional information is obtained by measuring
the total distance traversed by the baseball before hitting the floor, which is ap-
proximately 1.5 m above the initial height. The projection device was an ATEC
two-wheel pitching machine, which was operated in the range V=50- l l 0 mph and
m=l500-4500 rpm, corresponding to S=0.1-0.6 and Re=(l.1-2.4) x 10 5. A total of
22 pitches were analyzed, all in the "two-seam" orientation.
To measure the initial velocity and angular velocity of the ball, a motion analysis
system was used. The system, manufactured by Motion Analysis Corporation, con-
sisted of 10 Eagle-4 cameras operating at 700 frames per second and l /2000 second
shutter speed and the EVaRT4.0 reconstruction software. Each camera shines red
LED light onto the ball, attached to which is circular dot of retro-reflective tape. The
tape reflects the light back at the cameras, which then record the coordinates of the
dot in the CCD array. The reconstruction software determines the spatial coordinates
in a global coordinate system by using triangulation among the 10 cameras. The
cameras were positioned at various heights along a line approximately parallel to and
The Lift on a Spinning Baseball 25
six meters from the line-of-flight of the ball. In order to accomplish the triangula-
tion, the precise position and lens distortions of each camera must be determined
from an elaborate calibration scheme. The global coordinate system is defined by
positioning an L-shaped rod in the viewing volume of the 10 cameras simultane-
ously. The rod has four reflective dots located at precise relative locations to each
other. With these distances, the software determines a first approximation to the
location of each camera. These distances are further refined and the lens distortions
determined by waving a wand throughout the tracking volume of the cameras. The
wand has three reflective dots at known relative distances. Although the particular
calibration software is proprietary, it almost surely uses some variation of the direct
linear transformation technique described by Heikkila and Silven (Heikkila and
Silven 1997). A typical root-mean-square (rms) precision for tracking a single dot is
1.3 mm. Additional calibrations and consistency checks were performed. A plumb
line was used to establish that the y-axis of global coordinate system made an angle
of 0.16° with the vertical. The clock of the motion capture system was checked
against a precisely calibrated strobe light and found to be consistent within 0.5%. A
non-spinning baseball was tossed lightly in the tracking volume and the vertical
acceleration was measured tobe g to within 1.5%.
The setup for this experiment is similar in many respects to that used in the pio-
neering experiment of Alaways (Alaways 1998) but different is some key respects.
One difference concems the deployment of cameras. Alaways used two sets of mo-
tion capture cameras. One set tracked the ball over the first 1.2 m of flight and was
used to establish the initial conditions; a second set tracked the center-of-mass trajec-
tory over the last 4 m of flight, starting approximately 13 m downstream from the
initial position. This allowed a very long "lever arm" over which to determine the
acceleration but a short lever arm for determining the initial spin. In the present
setup, we used only one set of cameras distributed spatially so to track over the larg-
est distance possible, approximately 5 m. The same set of cameras determined both
the initial conditions and the acceleration. Tracking over a larger distance is useful
for measuring the spin at small S, since the angle through which the ball rotates over
distance D is proportional to SD. For all of the pitches analyzed in the current ex-
periment, the ball completed at least one complete revolution over the tracking re-
gion. We gained some redundant information by measuring the total distance R
traversed by the baseball while falling through a height of 1.5 m. The sec ond dif-
ference concems the deployment of reflective markers. Alaways utilized 4 dots on
the ball with precisely known relative positions. This allowed him to determine both
the initial orientation of the ball and the direction of the spin axis. We experienced
great difficulty in tracking more than one dot and therefore used only a single dot,
offset from the spin axis by approximately 15°. We were not able to measure the
spin axis but simply assumed it was constrained by the pitching machine to lie in the
horizontal plane, perpendicufar to the direction of motion. New measurements are
planned for early 2006 using multiple dots and those results will be reported at the
conference.
3000
1520
2000
1500
1000
1480
'<
E 3
-S 0 1460 3
N
-1000 1440
-2000 <v>=32 m/s 1420

ro=522 rad/s
-3000 1400
0.00 0.04 0.08 0.12 0.16
t (s)
Fig. 1. Trajectory data for one of the pitches projected. For clarity, the plot for the z coordi-
nate displays every other point. The inset shows the coordinate system. The ball is projected
in the +z direction with topspin. The curves are least-squares fits to the data.
A plot ofy(t) and z(t) for one ofthe pitches is shown in Fig. 1, where y and z are
the vertical and horizontal coordinates, respectively, of the reflective marker. Stan-
dard nonlinear least-squares fitting algorithms were used to fit these data to functions
of the form y(t) = Ycm(t) + Asin(rot+$) and z(t) = Zcm(t) ± Acos(rot+$) where cm refers
to the center ofmass coordinate ofthe ball. The center ofmass motion can be calcu-
lated by solving the equations of motion numerically, given the initial values of the
Ycm and Zcm positions and velocities and the lift and drag coefficients. The location of
the rotating dot with respect to the center of mass is characterized as an oscillation
with amplitude A, angular frequency ro, and initial phase $, with the +/- sign apply-
ing for backspin/topspin. Nine parameters were varied in the fit: the four initial
values, the three rotation parameters, and CL and C0 , both of were assumed to be
constant over the 5 m flight path. CL and C0 are primarily determined by the curva-
ture ofy(t) and z(t), respectively, where the latter is nearly completely masked by the
large linear term. The rms deviation of the fit from the data in Fig. 1 is 0.4 mm for
y(t) and 12.7 mm for z(t), which values are typical of the fits for the other pitches.
The inferred values of CL, the primary result of the experiment, are presented in Fig.
2. The error bars on the values are estimates based on the rms deviation of the data
from the fit and the calibration tests.
4 Results and Discussion
Our results for CL are shown in Fig. 2. Also shown are the SHS parametrizaton and
the earlier results upon which the SHS parametrization is based, including the data of
Alaways, Watts and Ferrer, and Briggs (Briggs 1959), as corrected by Hubbard
The Lift on a Spinning Baseball 27
• present
o Alaways 2-Seam : : '{)•
0.6 D Alaways 4-Seam ···········t···········O-··;;;.-•·······o·
o Watts & Ferrer : ,• :
0.5 .... ,;Briggs ........... [ ... a... ~~·····l···a· ........ .
:
·····SHS :,•• o :
J __ ,<,':oo[ .. --- j··- ---- ------

-~ 0 0 ;
0.4 i -
0. 3 ··············••: ••·····i:p,• • ,◊9 i :
··············!. ···············•·: .............. .
: ,j,o !• ! :
0.2 ·····~f·····
, ◊ /;,
6
········i·················:················i···············
: : :
0.1 ,' : r : t
,\ i : ' i
0.0 ~~~~~~~~~~~~~~~~
0.0 0.2 0.4 0.6 0.8 1.0
s
Fig. 2. Results for CL from the present and previous experiments, along with the parametri-
zation of SHS.
(Sawicki et al. 2005). In the region S<0.3, which is the region most relevant for long
fly balls, our data are in excellent agreement with SHS and the earlier data. In fact,
only for S>0.5 do our results start to deviate from SHS, which is constrained primar-
ily by the high-S data of Watts and Ferrer, most of which were taken at low Re,
below 0.6 x 105 (v<25 mph). Interestingly, the present data near S=0.5 agree with
Alaways and are below the general trend of the Watts and Ferrer data. Perhaps this
is an indication of a dependence of CL on Re, for a fixed S. On the other band, in the
range S=0.15-0.30 the values of CL measured in the present experiment are tightly
clustered around 0.20, despite a variation in Re in the range (1.1-2.4) x 105 • This
seems to suggest only a weak dependence of CL on v, at least in that regime. For
v=lO0 mph and co=1800 rpm (S=0.154), the present result for CL is in the range
0.20-0.22, the SHS parameterization is 0.18, and the RKA prescription is considera-
bly smaller at 0.06. We conclude that the present data support SHS and do not sup-
port RKA.
5 Summary and Conclusions
An experiment has been performed utilizing high-speed motion capture to determine

the effect of spin on the trajectory of a baseball. From these data, CL is extracted
with good precision over the range 50<v<l 10 mph and 0.1 <S<0.6. These values are
in excellent agreement with the SHS parameterization for S<0.3 but falls below SHS
for larger S. Our results are not consistent with the RKA parameterization for v>70
mph.
References
Adair, R. K. (2002) The Physics of Baseball, 3rd Edition. HarperCollins, New York.
Adair, R. K. (2005) Comment on "How to hit home runs: Optimum baseball bat swing pa-
rameters for maximum range trajectories." Am. J. Phys. 73, 184-185.
Alaways, L. W. (1998) Aerodynamics ofthe Curve Ball: An Investigation ofthe Effects of
Angular Velocity on Baseball Trajectories. Ph.D. thesis, University of Califomia, Davis.
Alaways, L. W., Mish, S. P., and Hubbard, M. (2001) Identification ofrelease conditions and
aerodynamic forces inpitched-baseball trajectories. J. Appl. Biomech. 17, 63-76.
Alaways, L. W. and Hubbard, M. (2001) Experimentaldetermination ofbaseball spin and lift.
J. Sports Sei. 19, 349-358.
Briggs, L. J. ( 1959) Effects of spin and speed on the lateral deflection (curve) of a baseball and
the Magnus effect for smooth spheres. Am. J. Phys. 27, 589-596.
Heikkila, J. and Silven, 0. (1997) A four-step camera calibration procedure with implicit
image correction. Proc. CVPR '97, IEEE, 1997, 1106-1112.
Sawicki, G. S., Hubbard, M. and Stronge, W. (2003) How to hit home runs: Optimum base-
ball bat swing parameters for maximum range trajectories. Am. J. Phys. 71, 1152-1162;
Am. J. Phys. 73, 185-189.
Watts, R. G. and Ferrer, R. (1987) The lateral force on a spinning sphere: Aerodynamics of
curve ball. Am. J. Phys. 55, 40-44.
Rigid Wall Effects on Softball Coefficient of Restitution
Measurements
Lloyd Smith 1 and Aaron Json 2
1 Washington State University, lvsmith@wsu.edu

Casecade Engine
Abstract The coefficient of restitution (COR) is often measured by impacting a ball against a rigid
wall. The ball COR is used to regulate performance in many sports. In the case of ball impact
sports, such as baseball and softball, the ball COR also contributes to bat performance.
Understanding the accuracy of the ball COR has received greater scrutiny as bat performance tests
have been refined and performance limits lowered. The effect of the idealized rigid wall surface
was considered in this study by impacting softballs against flat surfaces of controlled stiffness. The
ball COR was generally observed to decrease as the impact surface stiffness increased. An
exception to this observation was found for the thinnest impact plate (6 mm) which provided the
lowest COR value in this study. Only impact surfaces ofrelatively !arge compliance were observed
to affect the measured ball COR. The affect of the distance of the rigid wall from the speed
measurement due to drag was also considered. The common distance of 450 mm was shown to
report a ball COR that is approximately 0.005 lower than would be obtained by measuring the ball
speed at the impact surface.
1 Introduction
Thc cocfficicnt of rcstitution (COR) of two colliding objccts is defincd as the ratio of
their relative speed after and before impact, respectively (Beer and Johnston, 1997). The
ball COR, independent of the object being impacted, is often desired. To this end balls
are impacted against a flat rigid wall. Accordingly, the ball COR is taken as the ratio of
the ball speed after to before impact. The rigid wall ball COR is used by a number of
goveming associations to certify balls and bats for softball and baseball. Little has been
done, however, to define or quantify a rigid wall. ASTM 1887 dcfines a rigid wall "as
cinder block or concrete, minimally 20-cm (8-in.) thick." lt is unclear whether this
definition is sufficient or unnecessarily conservative.
The motivation to understand the effect of a rigid wall on ball COR is two-fold.
First, obtaining uniform ball COR measurements from laboratory round-robin studies has
been surprisingly difficult. lt is possible that differences in respective laboratory rigid
wall impact surfaccs contributc to this problem. Sccond, most ball and bat manufacturcrs
design their products near the allowed performance limits. Reducing the variation in ball
COR measurements will allow manufacturers to design their products closer to the
allowed limit.
30 Lloyd Smith and Aaron Ison
The effect of impact surface stiffness will be examined by impacting plates

supported near. their edges as a function of plate thickness. Plate displaccment and ball
COR will be examined as a function ofthickness.
2 Experiment
Softballs nominally 300 mm ( 12 in) in circumference were projected toward a flat impact
surface at 27 m/s (60 mph). Thc balls werc dclivcrcd using an air cannon as dcscribed in
ASTM 2219 striking the impact surface without rotation. The balls were impacted
sequentially on the four surfaces with the widest distance between the stitches.
The impact plate was nominally 300 mm (12 in) in diameter. Six impact plates were
studied ranging in thickncss from 13 to 44 mm (0.5 to 1.75 in) in 6 mm (0.25 in)
increments. The impact plate was attached to a support structure through 12 equally
spaced 19 mm (0.75 in) bolts arranged on a 240 mm (9.5 in) diameter circle. Twelve
spacers measuring 32 mm ( 1.25 in) in diameter were placed between the impact plate and
support structure to allow free motion of the center of the impact plate. A 25 mm (1 in)
diameter hole was drilled through the support structure which allowed a non-contact
laser vibrometer (Polytec OFV-511 Fiber Interferometer with a Polytec OFV-5000
Controller) to measure thc surfacc motion ofthc impact plate as depicted in Fig. 1.
3 Numerical Simulation
The effect of impact plate stiffness was studied numerically using a finite element
simulation. The model consisted of nearly 30,000 elements and is shown in Fig. 2. The
ball was meshed using 8 noded brick elements while the plate was meshed using 4 noded
platc clcments. Symmetry conditions were applied to the sectioned edges of the model.
The round holes in the plate represent the edges ofthe spacers which were constrained in
the direction of the impcnding ball motion. The ball was modeled as a viscoelastic
material whose properties were tuned to match the nominal ball density, COR, and
stiffness as described elsewhere (Shenoy, Smith, Axtell 2001 ).
~-'6
SUPPORT
SPACERS
STRUCTURE
SPEED
VIBROMETER \
~
~
IMPACT PLATE
Fig. 1. Schcmatic of impact plate stiffness study.

Rigid Wall Effects on Softball Coefficient of Restitution Measurements 31
4 Results
The average COR was found for 6 softballs impacted 6 times against each of the impact
plates, as shown in Fig. 3. (The experiment was repeated with 24 balls with similar
results.) The error bars represent the standard deviation of the mean for each thickness.
The COR was observed to decrease with increasing plate thickness. This trend is
analogous to the so-call trampoline effect from which thin walled hollow bats derive
their perforrnance. The balls were also impacted against a reinforced concrete wall,
labeled "Fixed" in the figure. The results suggest that wall compliance must be relatively
!arge to affect the measured ball COR.
The results from the FEA model are also presented in Fig. 3, and show good
agreement with experiment. The plate clamping condition was observed to have a !arge
effect on the numerical results. Pinned supports at the outside edge of the plate, for
instance, produced a much !arger change in ball COR with plate thickness than was
obscrved cxpcrimcntally; whilc clamped supports at the fastencr holes rcsulted in a
smaller change in ball COR with plate thickness than was observed experimentally. For
the results presented in Fig. 3, the plate was pinned at the hole circle radius since
mechanical fasteners are not able to produce a true clamped boundary condition.
The model also considered a 6 mm (0.25 in.) thick plate. The ball COR with this
plate decreased sharply (0.291 ). The FEA model of the plate was linear elastic and did
not consider yielding, which would have occurred for a 6 mm plate. The reason for the
low COR using the 6 mm thick plate is likely related to its high compliance and low
natural frequency (-300 Hz). The contact time between the ball and plate is of order 1
ms. The 6 mm thick plate continucd to dcform while the ball was recoiling. This is an
example of an elastic mechanism of impact energy loss, and shows that decreasing rigid
wall compliance may not always reduce the measured ball COR.
The vibrometer was used to measure the displacement of the impact plates.
Unfortunately the results of these measurements did provide a reliable description of the
plate motion. The displacement rate of the thinner plates exceeded the capacity of the
vibrometer, while the motion of the millbase, used to support the impact plates, affected
the displacement of the thicker plates. Although the millbase motion appeared to be
relatively small during the ball contact (-0.01 mm), its contribution was difficult to
Fig. 2. Mesh ofthe simulated ball-plate impact.

32 Lloyd Smith and Aaron lson
separate from the plate motion. The results of thc numerical simulation were used,
thcrcforc, to dcscribc thc cffect of rigid wall motion on COR.
Ball COR is shown as a function of the plate displacement in Fig. 4. The ball COR
appears relatively constant for plate displacements below 0.1 mm. A peak COR occurred
for plate displacements near 0.3 mm before decreasing sharply with ]arger plate
displacements (i.e. the 6 mm thick plate). lt should bc noted that displacement magnitude
alone may not be sufficient to describe rigid wall compliance. Plate speed may also play
a roll if the ball contact duration changes appreciably with plate compliance. For the case
at hand the contact duration was relatively constant resulting in comparable speed and
displacement trends with COR, where the spccd rangcd from 6 to 0.04 m/s.
The effect of air resistance is oftcn neglected when measuring the ball COR. The
short flight distance ofthe ball and a misconception that the effect of drag approximately
canccls in the COR calculation have been cited to discount drag effects. For a non-
rotating sphere (which is the case with the air cannon used for this work) travcling at a
speed, V, the drag force, F, may be found from (White 1986)
(1)
where CD, p, and A are the drag coefficient, air density, and sphere cross sectional area,
respectively. Over the range of speeds common to softball (below 110 mph), C0 falls
betwccn 0.5 and 0.3 (Adair, 2002). Thc ratio ofthe measured speeds, v/ v;, may be found
as
0.475
0.470
- oExperiment
aFEA
0.465
rf:
0::
0
(.)
0.460
äi
CD
0.455
rf~
rf
- - rf- - rii
'% rI
0.450
0.445
13 19 25 32 38 44 Fixed
Plate thickness (mm)
Fig. 3. Average ball COR from 6 softballs.

Rigid Wall Effects on Softball Coefficient ofRestitution Measurements 33
v = COR)e-k)
0 (2)
V;
where COR0 is measured at the impact plate and
k= pCDA X
m (3)
where m is the ball mass andx is the distance from the speed sensor to the impact plate.
Six balls were impacted against a rigid impact plate at 60 mph six times each. The
speed sensors were moved in 125 mm (5 in) increments, where the distance to the impact
plate was taken as the center of the speed sensors. The ratio of the after to before impact
speeds are presented in Fig. 5 as a function of sensor distance from the impact plate.
While the relatively srnall range of sensor distances considered here have a small effect
on the ball COR, the effect is measurable and outside the range of experimental scatter.
For the case of softball and baseball standardized COR measurements, the center of the
sensors are 450 mm from the impact plate. This implies that the actual ball COR is on
average 0.005 higher than the commonly measured value. For comparison Eq. (2) is also
plotted in Fig. 5, which appears to support Cd= 0.5.
SSummary
Tue foregoing has considered the effect of impact surface compliance and distance in
measuring the ball COR. The reinforced concrete wall commonly used as a rigid surface
provided a similar COR measure as suspended plates which much larger compliances.
0::
0
(.)
0.25 +---~--~~--~--- --!

0.001 0.01 0.1 10
Plate Displacement (mm)
Fig. 4. Ball COR as a function ofthe impact plate displacement from the numerical
simulation.
34 Lloyd Smith and Aaron Ison
The results of this work suggest that the current practice of using reinforced concrete
provides a good approximation of a rigid impact surface and that more compliant impact
surfaces can also provide comparable COR values. The distance between the speed
sensors and the impact surface was shown to have a small, but measurable effect on the
measured ball COR. This distance should be controlled and reported when measuring
ball COR.
0.460 ~ - - - - - - - - - - - - ~
0.455
·"~ 0.450
:,
E,
>
0.445
o Experiment
----Cd= 0.3
--Cd=0.5
0.440 + - - - - - , - - - - - - - r - - - - - - l
0 500 1000 1500
Distance from rigid wall (mm)
Fig. 5. Ball COR as a function of speed sensor distance from the impact plate.
6 Acknowledgements
The authors gratefully acknowledge the American Softball Association and the Sports
Science Laboratory at Washington State University whose support made this work
possible. The authors also acknowledge the assistance ofNicholas Smith and Ryan Smith
whose meticulous experiments greatly benefited this work.
7 References
Adair, R. K., (2002) The Physics of Baseball, 3rd Ed., Perennial, New York.
Beer, F. P., Johnston, E. R. (1997) Vector Mechanics for Engineers, 6th Ed., McGraw-
Hill.
Shenoy, M. M., Smith, L. V., Axtell, J. T., (2001) Performance Assessment ofWood, Metal
and Composite Baseball Bats, Composite Structures, 52:397-404.
White, F. M. (1986) Fluid Mechanics, 2nd Ed., McGraw-Hill.
The Effect of Holding Methods on a Baseball Bat
Performance Estimation System
Hiroyuki Kagawa 1, Takeshi Yoneyama 1, and Masaya Takahashi 2
1Kanazawa University, kagawa@t.kanazawa-u.ac.jp

2Sumitomo Light Meta!
Abstract. To obtain a value for the coefficient of restitution or of the trampoline effect for a
given bascball bat it is typical to hold the bat in a clamping fixturc or robot during simulated
hits. This fixture and the structurc or mechanism that connects the bat to earth form a dynamic
system in which the bat is only one part. Ifwe wish to measure some property ofthe bat alone,
therefore, it is important to understand thc influencc of the clamp method and boundary
conditions it imposes. To investigatc this matter we have developed an evaluation system for
the coefficient of restitution. Bats were held in three different ways while standard baseballs
were fired at them using a pitching machine. In this paper wc will describe the evaluation
method and give examples of the results obtained on some mctal bats. We conclude that the
fixing method does affect the mcasurcd value of the coefficient of restitution, indicating that
caution is needed whcn comparing the rcsults from different test fixtures.
1 Introduction
The coefficient of restitution (COR) is often measured by researchers who are

studying the performance of baseball bats (for example Tanaka et al. 1992,
Nishikawa et al. 1994, Naruo et al. 1997, and Sherwood et al. 2000). These
researchers have made many valuable contributions to the development and science
ofbaseball bats. Nathan 2004 and Hongo et al. 2004 introduced the term "trampoline
effect" to describe the relationship between stiffness and COR in baseball bats and
golf clubs. The position and significance of the "sweet spot" was thoroughly
explored by Takagi 1983. However, so far as we know, the effect of the clamp
method ofthe bat on the measured COR has not yet been investigated.
In this study we used three types of bat clamp:
1) suspended on soft supports so that the bat was nearly "free";
2) clamped but allowed to rotate about a fixed axis (fixed pivot); and
3) rigidly clamped.
In actual batting we expect the effective "clamp" condition to be somewhere
between the "free" and "rigid" conditions because the human batter will usually hold
the bat very firmly but his or her body is still a fairly flexible mechanism. The real
"clamp" condition, from the point of view of dynamics, will also probably be
different for each batter, batting style and so on. We thus considered that it would be
interesting to find out how much the clamp condition affects the apparent COR. This
paper describes our methods and some results obtained with meta! bats.
36 Hiroyuki Kagawa
2 Measuring Equipment
Our apparatus consisted of a pitching machine, a ball velocity measuring frame and a
bat fixture, arranged as shown in Fig. 1 (a).
The pitching machine was a commercially available general purpose machine
with two rotors. lt was able to pitch both hard and rubber balls. The machine had the
ability to drive the rotors at different rates to produce various amounts of ball spin:
however in this work we set the rotor speeds to be equal i.e. to produce little or no
spin. The maximum pitching velocity of the machine was about 180 km/hr (50 m/s ).
The ball velocity measuring frame was a pair of vertical plates with eleven laser
sensors. The plates faced each other and the ball passed between them. The sensors
were aligned radially about the bat center to detect the ball after impact, even if its
path was up to 15° above or below the horizontal, see Fig. 1 (b). The ball velocity
was estimated from the displacement of the laser beams and the time between the
beam interruption events.
The bat was held in one of three fixtures, as described above (Fig. 2).
Additional laser beams were placed near the bat to record its velocity. In the case of
the free clamp type, six vertical laser beams were used; in the case of the fixed-pivot
clamp type, only one beam was used but it could record both edges of the
interruption event of an optical shielding plate fixed on the pivot axis.
Batting events were filmed using a high-speed camera capturing 20 000 frames
per second. The high-speed camera was also used to verify the laser velocity
measurement system, to measure the velocity of the bat after the impact, and to
record the contact time and elastic behaviour of the ball.
Using this equipment described above, the pitching velocity v 0 and the hitting
velocity v 1 were measured during experiments. Additionally, in the case of the
method shown in Fig.2 (a) the translational velocity vb and the rotational velocity ffit,
of the bat around the center of mass G were measured. In the case of holding method
shown in Fig.2 (b) the rotational velocity ffii of the bat around the grip point A was
measured.
~ -□ &
•(htM l~IJ~"I
fr••
(a) Simplified general top view of apparatus. (b) Alignment oflaser sensors ( side view)
Fig. 1 Measuring equipment.
The Effect ofHolding Methods on a Baseball Bat Performance Estimation System 37
mf mf
i,o
v,
i
v,
7( P• •B G• 7( p,. •9 a.
R
R
.~ 52.:1
(a) Suspended on soft supports (b) Clamped but allowed to (c) Rigidly clamped
rotate about a fixed axis
Fig. 2 Holding methods for bats
3 Dynamic Relations
Let us define:
M Mass of the bat R Distance between hitting point P and grip point A
m Mass of the ball Rb Distance . between hitting point P and the bat
I Moment of inertia of the bat center of mass G
The value of COR is then:
(a) Suspended on soft supports
(1)
(b) Clamped but allowed to rotate about a fixed axis

COR= VI +R())., (2)
Vo
where R=_!!!L_ (3)

m(v0 +v1 )
(c) Rigidly clamped
(4)
COR=2
Vo
In the case of (b) not only the COR but also the hitting position can be
evaluated from the velocities alone.
4 Experiments
The properties of the bat and balls used in this study are summarized in Table 1. The
grip position (or pivot position, for case (b) experiments) was chosen as 152mm
from the grip end according to ASTM 1998. The bat was very slightly under the
regulation weight for Japanese high-school use (0.9kg) because we removed the grip
tape.
5 Results
Let us first consider clamp condition (b ): rigid pivot type. The value of the distance
from A to P, i.e. R, as estimated using only the measured velocities, is compared to
38 Hiroyuki Kagawa
Table 1. Summary of the properties of

the bat and balls used. ~
.., 700 . . . - - - - - - - - - - - - -
Baseball bat .s
I
0
C.
Material Aluminium
600
Entire length, L (mm) 842
1l
Weight, M (kg) 0.899 ~ so□
Moment of inertia, I (kg m2) 0.180 ..,
~
0
Distance between pivot and -~c.
377 400
center of mass, AG (mm) C
~
Distance betwccn pivot and

531
!~
center of percussion, AB (mm) .c 300
Hard ball
Diameter, dh (mm) 71
200 300 400 so□ 600 700
Weight, mh (kg) 0.150
Distance between pivot and batting point by actual
Rubber ball rneasurement (mm)
Diameter, d, (mm) 71
Fig. 3 Estimation of batting point for
Weight, m, (kg) 0.136
the case ofrigid pivot type clamping
the actual value in Fig. 3. The solid line in the figure is the diagonal; the dotted lines
are an error band of ±20mm. lt will be seen that the data points are generally within
that error band. lt seems that it is possible to estimate the hitting position using only
the three velocities v0, v 1, and co 1•
Fig. 4 shows CORs for hard ball impacts. The solid line shows the position of
the center of percussion B, and the dotted line shows the position of the center of
mass G. The same pitching velocity setting was used within each clamp experiment
type, but a different average velocity setting was used for each clamp type. For
clamp method (a) ("free") the pitching velocity was about 172 km/hr; for method (b)
(rigid pivot) it was 168 km/hr; and for method (c) ("rigid") it was 150 km/hr. There
was some variation in the actual pitching velocity, caused by variability in the
pitching machine. According to our previous experiments (Nasu et al. 2005), a higher
pitching velocity produces a lower observed COR. So it is expected that the order of
the COR data for the methods would not be changed by the variations in the pitch
velocity.
The forces on the bat when it was held in clamp type (c) were at times very high.
When the hitting point P was near the center of mass G, we had some cases where the
clamp bolts were broken by the impact. Our first experimental bat was actually bent
near the grip by one impact at the hitting end (the top).
Fig. 4 shows that the COR for clamp types (a) and (b) increased from the grip to
the top, and that it was generally a little higher for clamp type (a). However for
impacts near the center of percussion the clamp method made little difference to the
measured COR. This may be understood from the expected forces during an impact.
In the case of clamp method (b ), the rigid pivot type, we expect a force at the pivot
that is increased according to the distance between the hitting point P and the center
of percussion B. When the point P is inside B, the direction of the force at the pivot
becomes backward. On the other hand, the direction becomes forward when P is
outside B.
If there is a large force at the pivot, it bows the bat and excites slow vibrations.
The Effect of Holding Methods on a Baseball Bat Perfonnance Estimation System 39
We will call this type of bat deformation "reflection". We also expect deformation of
the cross-section of the bat, which we will call "hoop" type deformation; however
we won 't report about that here.
Fig. 4 also shows that the COR for clamp type (c) is considerably lower than for
the other clamp types. We suspect that this is due to several high-speed-impact
phenomena including the mode of vibration excited in the bat, deformation and
energy loss in the bat clamp system, and damping in the bat material. All the clamp
types allow reflection deformation, but in the rigid clamp case the magnitude of the
oscillations after impact was observed to be the greatest. So we expect that it would
be more useful than the other clamp types to investigate the effect of bat deformation
"reflection".
Qm- cf nm;,G Qm-cf pIDH,in,B

0-7
C
01,
·B • Suspended on soft supports, (a)
·B
VI
~
.....
0
....C
0 .S '
. •c
~
ll
~ ce
-• Clamped but allowed to rotate
about a fixed axis, (b)
.111 0~ ••-:c Rigidly clamped , (c)

~QJ
u
0
03 Iili1
qs \
02
200 300 400 500 600 700
D :ist.ance betw een pilot and battiig pont rm m )
Fig. 4. Distribution of the coefficient of restitution along the longitudinal axis of the bat
Ottr cf nm;,G Onr cf pIDH,in,B
0-7
C
:B
·B
VI
01,
0 .s
~,-t.,,,,,. J~
• Rubber ball (about 84km/h)
Hard ball (about l 32km/h)

.....~
....C
0
.111 0~
',o
~QJ ooo
0
u 03
80
02
200 300 400 500 600 700
D :ist.ance betw een pilot and batti,g poi1t rm m )
Fig. 5. Distribution ofthe coefficient ofrestitution along the longitudinal axis ofthe bat in the
case ofthe fixed pivot clamp type.
40 Hiroyuki Kagawa
Fig. 5 shows some additional results obtained using a different bat and both
hard and rubber balls. There is a significant effect in the case of a very elastic ball
like a rubber one because the deformation is so extreme. The high-speed video
footage for the rubber ball impacts showed that the ball became almost completely
flattened and wrapped around the bat for a brief instant. If we look only at the trend
within each set of results, it is interesting that the distribution of COR along the
length of the bat, as measured using a rubber ball, was nearly constant - unlike the
measurement done with the harder ball. The bat's characteristics were not visible
because the elasticity of the ball was so dominant.
5 Conclusions
An evaluation system for the coefficient of restitution of baseball bats with three
kinds of clamp methods has been developed.
( 1) In the case of clamping by suspension on soft supports, that is, "free", the
coefficient of restitution for impacts with a hard baseball was high. The main
deformation in the impact is local change in the circular cross-section of the bat
(hoop deformation).
(2) In the case of the fixed-pivot clamp type, the coefficient of restitution matched
the "free" clamp case but only for impacts near the center of percussion. We found
that it was possible to estimate the hitting position from only three velocity
measurements.
(3) The rigid clamp type produces very high forces on the bat, for a given impact
velocity, so the maximum velocity must be restricted to prevent damage to the bat.
Also the observed COR for this clamp type tends tobe rather low.
References
ASTM (1998), Standard Test Method for Measuring Baseball Bat Performance
Factorl, Fl881, pp.1-4.
Koening, K. et al. (2004), Engineering of Sport 5, Vol.2, pp.87-93.
Hongo, T. et al. (2004), Prepr. of Jpn. Soc. Mech. Eng., No.04-26, pp.162-167. (in
Japanese)
Nishikawa, et al. (1994), Prepr. of Jpn. Soc. Mech. Eng., No.940-59, pp.21-25. (in
Japanese)
Naruo, T. et al. ( 1997), Proc. 5th Japan Int. SAMPE Symposium, pp 1311-1316.
Nasu et al. (2005), Prepr. of Jpn. Soc. Mech. Eng., No.05-16, pp.9-13. (in Japanese)
Nathan, A.M. (2003), Am.J. Phys., Vol.71, pp.134-143.
Nathan, A.M., et al. (2004), Engineering ofSport 5, Vol.2, pp.38-44.
Sherwood, J.A., et al.(2000), Engineering of Sport 3, pp.377-387.
Takagi, R. (1983), Mechanics of Sports, Koudansha. (in Japanese)
Tanaka, et al. (1992), J. Jpn.Soc.Mech.Eng., Vol.64, No.623, pp.59-64. (in Japanese)
2 Climbing - Instrumentation and Testing of Equipment
Synopsis of Climbing - Instrumentation and Testing of
Equipment
Division ofBioengineering, School ofChemical and Biomedical Engineering, and SPERT

(Sports Engineering Research Team), BioMedical Engineering Research Centre, Nanyang
Technological University, Singapore, mfkfuss@ntu.edu.sg
Design, development, and testing of sports equipment and measurement devices are
classical engineering domains, and try to lead an athlete to greater heights. However,
unlike other sports disciplines, climbing equipment does not target at gaining addi-
tional centiseconds or centimeters, which are rather an equipment factor than directly
related to the athlete's performance. In climbing, equipment concerns safety and
prevention of injuries as weil as overuse syndromes. Thus, equipment becomes an
important legal issue, once more people are involved and interacting, such as lead
climber and belayer.
Instrumentation of climbing equipment is crucial for understanding and quantify-
ing the mechanical parameters of climbing, which, in turn, is an invaluable tool for
measuring the climber's performance and training success, and offers advice to
climber in terms of equipment selection and prevention of injuries.
The climbing sport, unfortunately not an Olympic discipline so far, is represented
by UIAA (International Mountaineering and Climbing Federation, Union Interna-
tionale des Associations d' Alpinisme ). UIAA comprises a special Safety Committee,
which deals with equipment related safety issues, especially with safety standards.
Climbing equipment encompasses belaying devices such as ropes, anchors (screws,
cams), karabiners, helmets, harnesses, crampons, ice-anchors, ice tools, slings, shoes,
and grip-enhancing means. Yet, as mentioned above, these kinds of equipment do
not increase the climber's performance: a better grip due to higher friction does not
necessarily affect the performance. lt rather depends on the climber's experience
whether to consider all grip parameters involved, such as surface, material, size,
shape, sweat, fatigue and weight distribution among all four limbs.
The current research and development in the field of equipment testing moved
from simple material and structural testing to more sophisticated testing approaches,
such as:
1) changes of material and structural properties of ropes after consecutive falls,
influence of knots, rope length, and fall factor (fall height over rope length);
2) impact of load rate and jerk on anchors and anchoring media like rock faces
and ice walls
3) development of safety standards for belaying devices (brakes)
44 Konstantin Fuss
The results, and the advice for the climber, are quite clear: 1) consecutive falls
increase the stiffness of a rope; 2) use low modulus ropes for ice climbing is pre-
ferred; 3) belaying devices with a !arge breaking coefficient stop a fall more effi-
ciently but also more uncomfortably.
These studies are highly essential for the climber, as they provide either practical
advice for the climber, or compare and rank different brands of equipment available
in the market.
The current research and development in the field of instrumented climbing
equipment moved from posturographic laboratory studies to more dynamic meas-
urements during training and competition, thereby providing a performance indica-
tor. Usually and conventionally, climbing performance is measured with difficulty
scales (UIAA, YDS, French, UK, etc.) either red-point (to lead climb a familiar route
without falling or hanging on the rope) or on-sight (to lead a climb without falling or
hanging on the rope on the first attempt without any prior information about the
route), as weil as the World Cup Ranking for top-climbers. Yet, mechanical parame-
ters of climbing proved to provide a more sophisticated kind of measurement, by
differentiating between climbing technique and style, experience, and performance
related to a single move. The instrumented climbing equipment consists of wall,
holds, and force sensors, placed in between, allowing measurement of contact time,
forces, impulse, smoothness factor, and friction on a single handhold. Moreover, the
force vectors changing with time can be visualised on the hold though vector dia-
grams. These parameters correlate weil with World Cup Rank and training success.
Furthermore, they offer advice to the climber in terms of injury prevention.
Future equipment developments expected are safety standards and test reports for
all kinds of safety gear, recommendation for the climber which equipment to prefer,
and to raise the technical standard for these kinds of equipment. Future instrumenta-
tion developments will concentratc on fully instrumented climbing walls for pcr-
formance analysis, by equipping each hold with a 6-DOF transducer; such a wall is
currently in development.
lt is recommended to set up centralised and independent test labs, preferably in
collaboration with UIAA, which perform standardised testing of new brands, or
newly developed equipment and publishes the main results on the Internet, by com-
paring them to other existing equipment. This will ensure the availability of data to
all climbers.
An Estimation of the Load Rate Imparted to a Climbing
Anchor During Fall Arrest
Dave Custer
Massachusetts Institute ofTechnology, USA, custer@mit.edu
Abstract. The strength of ice anchors is, in part, a function of the load rate; increasing the jerk
of the decelerating climber reduces anchor strength in brittle ice. To provide a rule of thumb
estimate of the load rate imparted to a climbing anchor during fall arrest, a simple, algebraic
formula for load rate is derived. As a first approximation, the load rate is proportional to both
the rope modulus and the square root of the fall factor and is inversely proportional to the
square root of the length of rope between the belayer and the falling climber. Load rate con-
siderations suggest that ice climbers should use low modulus ropes and avoid high fall factor
falls close to the anchor.
1 Introduction
Load rate, the time derivative of force, is unimportant in most climbing situations
because the components of the safety system are not affected by changes in load rate
over the range of values that can occur during fall arrest. For example, metal equip-
ment is expected tobe slightly stronger at higher loading rates (Newby 1985). Fur-
ther, the strength of granite, a rock favored by climbers, also exhibits little change
over several orders of magnitude changes in strain rate, and some rock types increase
in strength with increased strain rate (Lockner 1995). In contrast, the strength of ice
anchors decreases with increasing load rate. While the metal ice screw itself is unaf-
fected, a two order of magnitude increase in strain rate roughly halves the ice
strength (Gold 1977). Tests on ice anchors show a similar halving of strength with a:
two order of magnitude increase in loading rate (Blair, Custer, Alziati, Ben-
nett 2004). This phenomenon is complicated by the fact that ductile ice exhibits an
increase in strength with increased strain rates. Further complications are the paucity
of data and conflict in existing data that pinpoints the brittle/ductile transition at
temperatures above -40°C, temperatures at which most ice climbing occurs (Ara-
kawa, Maeno 1997). Nonetheless, the documented decrease in ice and ice anchor
strength at higher load rates suggests that climbers need to understand and control
the loading rate of ice anchors. To this end, a (very) simple model of a climbing fall
is used to derive an algebraic formula to estimate the load rate imparted to a climbing
anchor during fall arrest, and this simple model is compared to more realistic models
and to existing experimental data.
46 Dave Custer
2 Development of a Simple Estimate of Load Rate
The geometry of the roped climbing game determines the load rate that results from
a fall. Typically, the belayer is anchored at a stance, and the climber progresses up.
The rope connecting the two is clipped into intermediate anchors to reduce the dis-
tance the climber would drop in the event of a fall. Should a fall occur, the potential
energy of the climber is converted into spring energy in the rope and into heat energy
in the rope and at the belayer and the top anchor (Fig. 1). An estimate ofthe resulting
load rate can be developed by equating the potential energy of the falling climber
with the spring and heat energy stored in the rope at füll extension. Because the
resulting model is better suited for spread sheet or finite element analysis than for
rule of thumb estimation by climbers, a very simple model (VSM) is developed in
which the potential energy of the falling climber is converted only into spring en-
ergy. The VSM is then compared to more complicated models that take the various
non-spring energies into account. Additionally, the VSM is compared to empirical
fall data.
________ j_ _________ _
Fig. 1. Where energy goes in the roped climbing game. Figure la shows the roped climbing
game with a belayer anchored at a stance and the climber above an intermediate anchor. In the
event of a fall, the climber falls a distance h, and the rope, length /, will stretch a distance y.
Figure 1b shows the situation after the fall when the rope is maximally stretched. The potential
energy of the falling climber has been converted to heat energy at the top anchor (cf), spring
and damping energy imparted to the rope (rsd), and energy lost at the belayer (bb). The inset
in Fig. 1b shows the capstan effect of the rope loaded over the carabiner at the top anchor.
Typically, the tension ofthe rope section between the anchor and the climber is approximately
twice that ofthe rope between the belayer and the anchor (Pavier 1998; McMillan 2003).
2.1 The Very Simple Model (VSM) of Rope Behavior

A simple understanding of rope behavior can be derived by modeling a rope as a
spring and by equating the potential energy of the falling climber with kinetic energy
as the rope engages; in turn, the kinetic energy is also equated with the energy stored
An Estimation ofthe Load Rate Imparted to a Climbing Anchor During Fall Arrest 47
in the spring at its maximum extension. Figures 2a-2c show "snapshots" of the fal-
ling climber along with the associated quantities and approximations; Fig. 2d pro-
vides an overview of the resulting differential equation, its solutions, and the result-
ing VSM estimate for load rate. Load rate is proportional to the rope modulus, M, the
product of the Young's modulus of the rope material and the rope's cross sectional
area (thus, the rope's spring constant, k, is equal to Mil); load rate is proportional to
the square root of the fall factor, F, the ratio of fall height to rope length, h/l; and
load rate is inversely proportional to the square root of the rope length, /. This reduc-
tion of load rate with increasing rope length may be counterintuitive because every-
day experience suggests that langer falls have more severe consequences. Another
counterintuitive result is that the load rate is independent of the climber mass.
Simple mass/spring oscillation:

f =ky=-my
Boundary conditions:
mgh=½m.Y!ax =½ky!u
Solutions:
y=Wsin(mt)
j; = .figFi cos(mt)
.
y=- ✓2gFM . ( )
--smmt
m
... M µgF ()
y = --;; f-11- cos OJt
Load rate:
l_<!td = M W cos(mt)
Fig. 2a. Fig. 2b. Fig. 2c. Fig.2d. rdi7 Vt
Fig. 2. The derivation ofthe very simple model (VSM) ofjerk during climbing fall arrest.
Figure 2a shows the climber just before falling; all the climber's energy is gravitational poten-
tial energy, mgh. Figure 2b shows the climber just as the rope begins to pull taught; all the
energy is kinetic, ½my2. Figure 2c shows the climber at the maximum rope extension; the fall
energy has all been converted to energy stored in the rope spring, ½ky2 (Wexler 1950). Figure
2d develops the equations goveming the VSM, culminating in an expression for the load rate.
2.2 Comparison of the VSM to Models that Include Friction Losses and
the Potential Energy Produced by Rope Stretch
The model developed in 2.1 is quite simple and deserves to be compared to models
that include the potential energy due to rope stretch, friction at the carabiner, the
damping effect of the rope, and the effects of belayer behavior and energy absorbing
systems (EAS).
To include the gravitational potential energy of the climber due to rope stretch, y,
the development in Fig. 2 needs one modification, the inclusion of an additional
potential energy term in the boundary conditions such that the potential energy of
both the fall and the stretch equals the energy stored in the spring at maximum exten-
sion: mgh+mgymax=½k/max· The solution to this quadratic yields the Wexler Equa-
48 Dave Custer
tion, Eq. 1, for the tension in the rope. The differential equation itself remains un-
changed, so the frequency of oscillation is also unchanged (Wexler 1950).
The force predicted by Eq. 1 is )arger than that of the VSM because the addition
of the potential energy associated with the spring stretch results in an increase in
spring energy and thus force. Equation I predicts load rates slightly (- I 0%) high er
than the VSM predictions, but the Ioad rate remains roughly proportional to the rope
modulus, M, and the square root ofthe fall factor over the rope length, (F/f) 112 •
(1)
Another observation drawn from Eq. I is that the MF!mg term is the product of
the energy component proportional to y2 and the energy component that is independ-
ent of y. Thus, Eq. I can be used to estimate forces resulting from situations in which
energy is lost due to friction, damping in the rope, and other energy absorbing phe-
nomena such as belayer behavior or EAS.
The inclusion of a carabiner at the top anchor has two related effects on the load
transmitted to the anchor. The friction at the top anchor affects the load rate by effec-
tively reducing the length of the rope. The extreme cases can be shown by inspec-
tion: no friction loss makes the füll rope length available to absorb energy, and com-
plete "binding" at the top carabiner reduces the length of rope available to absorb
energy to half the fall height. In typical climbing situations, the ratio of rope tensions
on either side ofthe anchor carabiner, y, is about 2: I (McMillan 2004). Energy is lost
to friction proportional both to the difference of tensions in the rope on either side of
the top carabiner and to the stretch of the rope section between the belayer and the
top anchor. Equation 2 shows thc cffective rope length as a function of rope length,
fall factor, and the ratio of rope tensions across the anchor carabiner. This reduced
effective rope length increases the load rate, but does not alter the proportionality of
load rate to M or (Fl[)'ll_
_(i-n
(rrective - l 2 F+y
1
(2)
A spring-dashpot model of the rope has been developed that closely matches em-
pirical data (Pavier 1998). The inclusion of a dashpot in the rope model reduces the
maximum force and increases the time it takes to reach the maximum force. Using
the values of rope moduli and damping coefficient from Pavier, estimates of the
effect of damping on the load rate can be derived by estimating the ratio of energy
absorbed by damping to energy absorbed in the spring, which is about 1: 1. Thus, the
MF!mg term in Eq. 1 is halved, and the maximum force is reduced by a factor ofthe
square root of 2; a 1: I ratio of spring to damping energy produces a 45° phase lag
that increases rise time by about 50%. While rope damping reduces the load rate, it
does not effect the dependence of load rate on M or (Ff[) 112 •
An Estimation of the Load Rate Imparted to a Climbing Anchor During Fall Arrest 49
As with rope damping, any energy lost at the belay, by pulling the belayer up-
wards or by pulling rope through the belay device, can be deducted from the MF/mg
term in Eq. 1. Empirical evidence suggests that belayer behavior can increase the
maximum force by at most 15% and decrease the maximum force by as much as
30%. The effect of belayer behavior on the time to maximum force is less well docu-
mented, though variations of about 15% can be seen in experimental data (Zan-
toni 2005; Richard 2003). The effects of belayer behavior on the load rate are diffi-
cult to estimate or model.
The deployment of EAS can also yield force reductions of up to 30%, and time to
maximum force is increased by the duration of EAS deployment, which can be on
the order ofthe rise time without the EAS (Richard 2003).
3 Testing the VSM
In this section, the VSM is tested on the fall data available in Helmut Mägdefrau's
PhD thesis (Mägdefrau 1989). The data comprise two subsets large enough to at-
tempt statistical correlation. The single rope subset contains data for 29 falls on an 11
mm (UIAA "single") rope; the halfrope subset contains data for 14 falls on a pair of
8.8 mm (UIAA "half') ropes. The belay device used for both subsets is an antz-
bremse. Force, rope length, fall height, climber mass, duration of loading, and the
rope slip through the belay device are recorded. The researchers used themselves as
falling climbers, so high fall factor falls are not represented, nor is there any control
for belay technique, as evidenced by the variation in rope slippage through the belay
device. The VSM model predicts that within subsets the load rate should be propor-
tional to (F/[) 112 , and that the load rate between the subsets should differ proportion-
ally to the difference in rope moduli. The scatter plots of load rate vs. (F/[) 112 are
shown in Fig. 3. Indeed, the load rates differ between subsets by about 30%, which
would be expected if the rope moduli are proportional to the cross sectional area of
the half rope pair and the single rope. The correlation coefficient for the single rope
data is 0.2, and hypothesis testing suggests there is little certainty of correlation.
When the high slip data points are removed, the correlation is 0.6; hypothesis testing
gives 90% confidence that there is a correlation. The half rope data have a correla-
tion coefficient of 0. 7 and a 98% confidence of correlation.
Similar correlation analyses were calculated from the Wexler and Pavier models
using the fall scenarios from the Mägdefrau single rope data. Additionally, ±15%
noise was added to both force and time results from the Wexler model to simulate
belayer behavior. The W exler and Pavier models show correlation coefficients of
0.99 and 0.93 respectively, with almost certain correlation. The noisy Wexler calcu-
lations give a correlation coefficient of0.77.
4 Conclusion
The VSM provides a rule of thumb estimate for the jerk and the resulting loading
rate, which is approximately proportional to both the rope modulus and the square
root of the fall factor and inversely proportion
50 Dave Custer
nal to the square root of rope length. This relationship changes very little when more
complicated models that include rope damping and carabiner friction are invoked.
The correlation is greatly reduced by belayer behavior. Climbers can control the
loading rate through choice of low modulus ropes, judicious anchor placement in
proximity to the belay anchor to reduce the fall factor, and use of EAS.
0.30
◊
~0.25 ◊ ◊◊-'O◊~~ .:.◊ ♦ ..

.... V
◊«>•i ♦ ♦
10.20
◊◊ ♦♦
[0.15
~0.10
er
rn 0.05 ◊ single rope
• half rope pair
0.00
0.00 10.00 20.00 30.00 40.00 50.00
Load rate (kN/s)
Fig. 3. Scatter plot of load rate vs. square root of (F/[) for Mägdefrau data.
References
Arakawa, M., Maeno, N., (1997), Mechanical strength of polycrystaline ice under uni axial
compression, Cold Regions Science & Technology, 26, 215-229.
Blair, K., Custer, D., Alziati, S., and Bennett, W., (2004) The effect of load rate, placement
angle, and ice type on ice screw failure, The Engineering ofSport 5, Volume 2 (ed. M.
Hubbard, R.D.Mehta, and J.M. Pallis), International Sports Engineering Association, Shef-
field, UK, 283-289.
Gold, L., (1977) Engineering Properties ofFresh-Water lce, Journal ofG!aciology, 19, 81,
197-211.
Lockner, D., (1995) Rock Failure, Rock Physics and Phase Relations A Handbook of Physical
Constants, http://www.agu.org/reference/rock/l l_lockner.pdf, Dec. 17, 2005.
Mägdefrau, H., ( 1989) Die Belastung des menschlichen Körpers beim Sturz ins Seil und deren
Folgen, dissertation, Ludwig-Maximilians University, translated by David LiaBraaten,
1994.
McMillan, N., (2003) How Strong Does Your Climbing Gear Need to Be?, British Mountain-
eering Council Technical Commillee, Note 04/03.
Newby, J., (1985), ASM Handbook, Vol 8, ASM International, Metals Park, Ohio.
Pavier, M. ( 1998) Experimetal and theoretical simulations of climbing falls, Sports Engineer-
ing, 1, 79-91.
Richard, M, (2003) Modelization ofthe fall in climbing and mountaineering, Annex 21, 2003
U/AA Safety Commission Meeting Minutes.
Wexler, A. ( 1950) The theory ofbelaying, American Alpine Journal, 7, 379-405.
Zantoni, C, (2005), Presentation, 2005 UIAA Safety Commission Meeting, Chamonix, France.
Dynamics of Speed Climbing
Franz Konstantin Fuss 1 and Günther Niegi2
1 Nanyang Technological University, Bioengineering, SG, mfkfuss@ntu.edu.sg

2 University of Vienna, Anthropology, AT, a9203213@unet.univie.ac.at
Abstract. This paper aims to provide the first insight into the dynamics of speed climbing. For
this purpose, two jug-type climbing holds were manufactured and instrumented with 3D force
transducers. At a speed climbing route, three climbers with different experience levels per-
formed four runs at three different speeds (slow, medium, maximal possible speed). The 36
experiments served to analysc the following parameters: contact time at the hold, magnitude of
contact forces and of the shock spike at initial contact, and the climbing velocity between the
two holds. In speed climbing, the higher the climbing speed, the higher are the finger reaction
forces at the hold, thc highcr is the shock spike at initial contact, and the shorter is the contact
time. As speed is a factor of experience, the contact time and thc magnitude of forces also
correlate with the experience.
1 Introduction
As far as lead climbing and bouldering, two of the three sub-disciplines of sport
climbing, are concemed, very few literature sources are available. Most of these
sources deal mainly with posturographic studies on experimental climbing walls
( Quaine et al. 1997 ab, 1999, Testa et al. 1999, 2003, N oe et al. 2001) and kinematic
studies (Werner et al. 2000, Bursnall and Messenger, 2000).
Yet, to date, there is no literature source available on the biomechanics of speed
climbing, the third crown discipline of sport climbing. Due to this fact, it was highly
challenging to provide insight into the biomechanics of speed climbing for the first
time. In speed climbing, i.e., vertical running up the wall, top speed climbers reach a
velocity of approximately 2 m/s.
lt was the aim of this study, to investigate speed climbing with instrumented
climbing holds and to evaluate which mechanical parameters correlate with the
climbing speed.
2 Methods
Two sloper holds were designed and manufactured for this study (Fig. I ). Each of
the two holds was connected to a Kistler force sensor (type 9327, Kistler, Winter-
52 Franz Konstantin Fuss and Günther Niegl
thur, Switzerland), placed between the hold and the aluminium frame (Fig. 1). The
holds were moulded from a mixture of epoxy resin and sand. The data sampling
frequency during the experiment was 500 Hz. The hardware equipment consisted of
Kistler amplifiers (DAQ-5040A23), an A/D board (National Instruments) and a
laptop. We collected the data with DASYLab 5.6 by Datalog. The 3D piezoelectric
transducers retumed the forces according to the 3 axes of the co-ordinate system.
The co-ordinate system was: x - out of the wall, towards the climber; y - upwards
(climbing direction); z - to the left.
The experiment was carried out at Climb Adventure Pte Ltd Climbing Gym, Sin-
gapore. Two factors were considered for the experiment: the experience of climbers
(3 climbers with different experience levels, beginner intermediate, and top speed
climber), and the speed of ascend (slow, intermediate, and maximal speed). Each
climber performed the runs at the 3 different speeds four times, resulting in a total of
36 experiments.
The climbing wall adapted for the speed climbing experiment is shown in Figure
2. The two instrumented holds of a jug type were placed in the middle of the route
such that they will be contacted by the right hand.
The mechanical parameters calculated from the raw data were: contact time,
forces in y- and x-direction (mean and maximum, normalized to body weight), im-
pulse of y-force, ratio of the x-axis to y-axis forces, magnitude of shock spikes, vec-
tor diagrams, and climbing speed in the middle of the route. The latter is defined by
the distance between the two instrumented holds ( 1.2 m) over the time taken by the
climber to get from the initial contact of the 1st instrumented hold to the initial con-
tact of the 2 nd instrumented hold.
All the statistical calculations were carried out in SPSS 11.0 (Statistical Package
for Social Sciences). The level of significance was set at p<0.05.
3 Results
Figure 1 shows the typical curves of the resultant forces on the two instrumented
handholds. The shapes of these two graphs are basically identical. At the start of the
contact on the handhold, there is a force shock spike, followed by a steep increase in
forces to a first peak. The force then decreases to a valley and subsequently up to a
second peak, which has a lower maximum force. After the second peak, the contact
forces decrease sharply. The most striking result was the existence of shock spikes
(Fig. 2) at the initial hold contact. These spikes are apparent in both vertical and
horizontal directions and come from the impact of the hand, when it hits the hold.
-
..
I'.,..
j
1-
..
r ,
Fig. 1. Climbing wall for speed climbing experiment (left side; the instrumented jug holds are
in position 3 and 5); jug hold and force sensor (middle ); resultant force - time graphs of the
two instrumented holds (right side; climbing speed approximately 0.73 m/s)
... 0
Shodr.Spakt
!
J
-·-·u 12
Fig. 2. Shock spikes in x- and y-direction (right side; forces in absolute values); correlation
between contact time at the Ist (a) and 2nd (b) hold and the climbing velocity (right).
The contact time at the hold and the climbing velocity correlate highly. Figure 2
shows the relationship between contact time (t) at the two instrumented holds and the
climbing velocity (v). The regressions are non-linear and follow a power fit: hold 1: t
= 0.953 v- 1109, hold 2: t = 1.135 v-0-887 . The correlation coefficients for the 1st
(lower) and 2nd (upper) hold are 0.999 and 0.936 respectively.
The next significant effect of velocity is on the mean and maximum resultant
forces (norrnalised to body weight) of the two handholds. The p-values range from p
< 0.001 to 0.002. The various normalised resultant forces are plotted against velocity
(Fig. 3). A logarithmic fit provided the best correlatiori with coefficients between
0.825 and 0.92. The force - velocity relationship can also be seen in the vector dia-
grams (Fig. 4), which display the resultant force vectors with time on the hold.
The resultant shock spikes and the shock spikes in the x-axis direction are also
significantly influenced by the velocity of ascend, at p < 0.001. This can be seen
from Figure 3, where the shock spikes are plotted against the velocity. The shock
spikes in the direction of the x-axis are negative because the direction of the shock
spikes is into the wall. The correlation was ev.aluated with a linear fit. The correla-
tion coefficients are between 0.8 and 0.884. This relationship shows that the higher
the speed of ascend, the higher are the shock spikes.
Experience of the climber has a significant effect on maximal velocity of ascend
(p<0.001). The more experienced a climber is, the faster he can ascend the route. The
top velocities are (most experienced climber first): 1.44 m/s, 1.00 m/s, and 0.68 m/s
The contact time on the handhold is also significantly influenced by the experi-
ence ofthe climber, p<0.001 for first handhold and p<0.003 for the second handhold.
The more experienced a climber is, the shorter is the contact time on the handhold, as
the contact time decreases with increasing speed. The shortest contact times are
(most experienced climber first): 0.71 s, 1.33 s and 1.61 s.
The forces on the handholds are also significantly influenced by the experience
of the climbers. These include the mean and maximum forces along the x-axis, y-
axis and also their resultant. The p-values range from p<0.001 to p<0.006. The more
experienced a climber is, the higher are the mean resultant forces of both holds for
the fastest ascend trail (most experienced climber first): 0.46lxBW (body weight),
0.390xBW and 0.378xBW.
--...
D D
•
....
li: -
0 j 0
J„
.·.,. ..-.. .
C
d" • •
·-
d
. ..
o • ,. •• •'-'
..
... ..
-- ...
...
_,_..lu u
Fig. 3. Resultant forces (normalized to body weight) vs. velocity (left side; a = maximal value
at 1SI hold, b = maximal value at 2nd hold, c = mean value at I SI hold, d = mean value at 2nd
hold); shock spikes vs. velocity (right side; e = resultant shock spike at 2nd hold, f = resultant
shock spike at I st hold, g = shock spike in x-direction at I st hold, h = shock spike in x-direction
at 2nd hold)
a b C
Fig. 4. Vector diagrams at different speeds (a: fast, 0.3-0.7 m/s, b: medium, 0.7-1.1 m/s, c:
low, 1.1-1.5 m/s); the resultant vectors point upwards and towards the wall (side view); the
vertical bar on the left side corresponds to a force vector of 500 N.
4 Discussion
Speed climbing can be compared to horizontal level walking and running: the higher
the velocity, the shorter is the contact time (duration of stance phase in walking and
running), and the higher are the forces (ground reaction forces in walking and run-
ning). Furthermore, a shock spike is also present in speed climbing. The spike at the
fingers has two components, one in y-direction (upward), the other in x-direction
(towards the wall), which coincide in time. This stands in contrast to walking, where
the vertical shock spike at the heel occurs later then the horizontal claw-back spike.
lt is evident that the maximal speed of a climber depends on his experience and
training level. As a short contact time, high finger reaction forces, and high shock
spikes correlate with high velocity, it is not further astonishing that these parameters
also correlate with experience. At this point, a striking difference becomes apparent:
in lead climbers, the better a climber, the smaller are the contact forces at the hold'
(Fuss and Niegl, 2006); in speed climbers, the opposite is true, i.e., the better the
climber and thus the faster, the higher are the contact forces. In lead climbers, the
magnitude of these forces depends on how economical the climbing style is; in speed
climbers, the force is a function of speed. Y et, these contact forces do not linearly
increase with speed: at higher speeds, the force-velocity function flattens out and
increases at a smaller rate than the velocity does (Fig. 3a). This fact does not substan-
tially increase the risk of overuse syndromes and finger injuries, as big velocity dif-
ferences at higher speeds result into only small force differences. The shock spike,
however, increases linearly with speed. Shock spikes are generally know for causing
overuse syndromes when experienced at a high frequency over a longer period (e.g.,
running: chronic knee injuries due to running on hard ground with inadequate shoes;
navicular disease in shod horses with a steep fetlock angle). Speed climbing, how-
ever, is applied only for a couple of seconds (top athletes climb 20 min about 13 s),
hence the risk of overuse syndromes in individual climbs is small and depends on the
training frequency.
The contact time at the first instrumented hold correlates perfectly with the veloc-
ity by following a power function (line "a" in Fig. 2). The contact time at the second
instrumented hold displays the same correlation although the data points are more
scattered (line "b" in Fig. 2). The reason for this is that the climbing speed between
the two instrumented holds was calculated from the distance in between and the time
between the two contacts. As the first contact provides the thrust for covering the
distancc between the two holds, it is the first contact time which shows a higher
correlation.
5 Conclusion
In speed climbing, the higher the climbing speed, the higher are the finger reaction
forces at the hold, the higher is the shock spike at initial contact, and the shorter is
the contact time.
Speed climbers should be aware of the fact that big velocity differences at higher
speeds result into small finger force differences, which in turn does not substantially
increasc the risk of injuries. The shock spike at initial contact, however, increases
linearly with speed and bears the risk of overuse syndromes, depending on the train-
ing frequency.
References
Fuss, F. K. and Niegl, G. (2006) Instrumented Climbing Holds and Dynamics ofSport Climb-
ing. In: E.F. Moritz and S. Haake (Eds), The Engineering ofSport 6. Springer, Munich.
Noe, F., Quain, F. and Martin, L. (2001) lnfluence of steep gradient supporting walls in rock
climbing: biomechanical analysis. Gait Posture, 13, 86-94.
Quaine, F. and Martin, L. (1999) A biomechanical study ofequilibrium in sport rock climbing.
Gait Posture, 10, 233-239.
Quaine, F., Martin, L. and Blanchi, J.-P. ( 1997a) Effect of a leg movement the organisation of
the forces at the holds in a climbing position: 3-D kinetic analysis. Hum. Mov. Sei., 16,
337-346.
Quaine, F., Martin, L. and Blanchi, J.-P. ( 1997b) The effect ofbody position and number of
supports on wall reaction forces in rock climbing. J. Appl. Biomech., 13, 14-23.
Testa, M., Martin, L. and Debu, B. ( 1999) Effects of the type of holds and movement ampli-
tude on postural control associated with a climbing task. J. Gait Posture, 9, 57-64.
Testa, M., Martin, L. and Debu, B. (2003) 3D analysis ofposturo-kinetic coordination associ-
ated with a climbing task in children and teenagers. Neurosci. Lett., 336, 45~9.
Bursnall J. and Messenger N. (2000) Analysis of climbing technique using the ProReflex 3d
motion analysis system. In: N. Messenger, W. Patterson and D. Brook (Eds.) The Science
of Climbing and Mountaineering. Human Kinetics Software, Champaign, IL, CD-ROM.
Werner 1., Gebert W. and Kauer B. (2000). Three dimensional analysis ofrock climbing tech-
niques. In: N. Messenger, W. Patterson and D. Brook (Eds.) The Science of Climbing and
Mountaineering. Human Kinetics Software, Champaign, IL, CD-ROM.
Instrumented Climbing Holds and Dynamics of Sport
Climbing
Franz Konstantin Fuss 1 and Günther Niegl2
1 Nanyang Technological University, Bioengineering, SG, mtkfuss@ntu.edu.sg

2 University ofVienna, Anthropology, AT, a9203213@unet.univie.ac.at
Abstract. This paper gives an overview of usage of instrumented climbing holds (equipped
with force sensors) for assessing the performance ofa climber, be it during training or compe-
tition, as weil as assessment of training progress and success, gripping techniques, and value
of equipment (e.g., chalk). The experience of the climber is expressed by a short contact time,
a lower force at the handhold, a low impulse, and high friction. During training, the contact
time, mean force, maximal force, and impulse decrease whereas the mean and maximal fric-
tion and the smoothness factor increase. There is no difference between dry hand and liquid-
chalked hands, between dry and wet hands, and between powder-chalked hand on a clean
surface and dry hand on a messy surface. However, powder chalk is far better than liquid
chalk or a dry hand, and on messy surfaces, a dry hand is better than a powder-chalked hand.
The higher a climber jumps during double-dynoing, and the more fatigued he is during boul-
dering, the lower are the finger reaction forces and thus the injury risk.
1 Introduction
The research group at University of Grenoble, France (Quaine, Martin and Blanchi,
1997a, 1997b; Quaine and Martin, 1999; Noe, Quaine and Martin, 2001; Testa, Mar-
tin and Debu, 1999, 2003) has used instrumented climbing holds for posturographic
studies on experimental laboratory climbing walls. These used sterile standard condi-
tions, equipped with strain-gauged holds for analysis of the weight shift, and load
distribution, from 4-limb contact to 3-limb contact when moving upward. In addi-
tion, the overall acceleration, and after subsequent integrations, the velocity as well
as displacement of the centre of mass of the body are obtained.
Quaine, Vigouroux and Martin (2003) investigated individual forces applied by
the fingers in different crimp positions. The data for the finger tip forces were col-
lected in a seated posture (and not during climbing) in which the test subjects exerted
a horizontal force on the force sensors in three different finger postures: extended
crimp, opened crimp and closed crimp without the usage ofthe thumb.
Li, Margetts and Fowler (2001) calculated the static friction coefficient between
the hand and different rock surfaces by measuring the normal and tangential forces
with strain gauges and detecting slippage with a potentiometer. Non-chalk contact
58 Franz Konstantin Fuss and Günther Nicgl
between hand and rock produced a friction coefficient of 3 whereas chalk reduced
the coefficient to 2.5. Consequently, Li et al. (2001) considered 'chalking' a myth.
The goal of the present studies was to apply instrumented climbing holds to real
sport climbing situations (competition and training; difficult climbing and boulder-
ing) rather than to laboratory conditions, as well as to reinvestigate the influence of
chalking. Eight different studies have been carried out so far, related to (1) mechani-
cal parameters of climbing during competition, (2) influence of experience on the
mechanical parameters of climbing during competition, (3) training influence on the
mechanics of climbing, (4) friction between hand and climbing hold under different
conditions, (5) mechanics of the double dyno technique, (6) mechanics of the pinch
grip, (7) finger load distribution during climbing, and (8) fatigue influence on the
mechanical parameters of climbing.
2 Methods
For all instrumented holds used, the forces were measured with from 1-4 Kistler
force sensors per hold (type 9317 and 9327, Kistler, Winterthur, Switzerland), placed
between the wall and one handhold. The data sampling frequency was 100-200 Hz.
The hardware equipment consisted of Kistler amplifiers (DAQ-5040A23), an A/D
board (National Instruments) and a laptop. The data was collected with DASYLab
5.6 (Datalog). The 3D piezoelectric transducers retumed the forces in the 3 axes of
the co-ordinate system. The co-ordinate system was: x out of the wall, towards the
climber; y upwards (climbing direction); z to the left. The usage oftwo force sensors
in the y-direction allowed calculation of the moment about the z-axis, and the centre
ofpressure (COP, origin ofthe force vector) in the xy-plane on the hold surface. For
this task, the surface of the hold was digitised in the xy-plane and a 1st , 3rd , 9'\ or 10 th
order polynomial regression function was determined with respect to the original
sensor co-ordinate system, depending on the nature ofthe curvature.
Thus, the y-coordinate ofthe COP (Ycop) is a function of XcoP•
YCOP = f (XuJP) (1)
The moment equilibrium, calculated from Fx and Fr (extemal forces applied to

the handhold, calculated from the sensor forces measured), M1 (the moment about
the z-axis), YcoP and Xcop, becomes
Mz-Fx Ycop + FrXcoP = 0 (2)

After substituting Eq. ( 1) into Eq. (2), the resulting equation was solved for XcoP
with Matlab (MathWorks). In the case of more than one root, the selection of the
correct root was based on a continuous COP movement. The vector diagrams to-
gether with the handhold surface were reconstructed in AutoCAD 2000 (Autodesk).
The vectors (Fig. 2) thereby were positioned on the hold, yet they were directed
towards the hand, comparable to standard vector diagrams in clinical gait analysis
(Pedotti or butterfly diagrams).
The mechanical parameters of handhold-hand interaction, calculated subse-
quently with respect to time were: 1) contact time, 2) force in y-direction (vertical
lnstrumented Climbing Holds and Dynamics of Sport Climbing 59
force), 3) force in x-direction (sagittal force), 4) resultant force, 5) angle of the resul-
tant force (in the xy-plane), 6) position of the resultant force (COP), 7) impulse of
the y-force, 8) instantaneous friction (µ;), 9) smoothness factor (SF), and l 0) vector
diagram (2D in the xy-plane). The instantaneous friction (µ;) is the tangential force
divided by the nonnal force, by converting x- and y-forces into nonnal and tangen-
tial components according to the position ofthe COP and the slope at the COP (first
derivative of Eq. 1). The instantaneous friction (µ;) changes with time and is not
related to the static or dynamic friction coefficient, as during climbing, µ; occurs
usually before impending slippage. The dimensionless smoothness factor (SF) is the
body weight divided by the mean of the absolute difference (in N) between the y-
force-time graph and a parabolic curve of the same impulse. The parabolic curve
serves as a model for the ideal force time graph of the handhold contact. The closer
the y-force-time graph approaches a parabolic curve, the smoother is the force appli-
cation, and the higher the SF. The parabolic force(F)-time(t) graph was calculated
from Eq. 3, in which T is the total handhold-hand contact time, and J is the impulse
ofthe actual y-force-time graph.
F = (t/T - (t!D 2 ) (6 JID (3)
Fig. 1. Different instrumented holds (see below)
The events and instrumented holds at the 8 different studies were:

(1) Mechanical parameters of climbing during competition: National Climbing
Championship of Singapore, men 's semifinal (n=30), at the National Rock Climbing
Centre, SAFRA Yishun Country Club, on 17 August 2002; jug type hold (Fig. 1-a).
(2) Influence of experience on the mechanical parameters of climbing during
competition: UIAA-ICC Sport Climbing World Cup, ladies' quarter-final (n=23; all
top-6 climbers and 15 of the top-25 climbers of the then current world rank !ist),
Singapore EXPO, 24 August 2002; curved sloper (Fig. 1-b,c ), at a 15° overhang.
(3) Training influcnce on the mechanics of climbing: training session at National
Rock Climbing Centre, Singapore, 17 mm deep edge (Fig. 1-d,e), self-manufactured.
(4) Friction between hand and climbing hold under different conditions: Nanyang
Technological University, Singapore; self-manufactured horizontal hold with surface
of an artificial climbing hold and under different conditions (see results)
(5) Mechanics of the double dyno technique: training session at National Rock
Climbing Centre, Singapore; self-manufacturedjug type (Fig. 1-f).
(6) Mechanics ofthe pinch grip: training session at National Rock Climbing Cen-
tre, Singapore; self-manufactured CNC-machined pinch grip type (Fig. 1-g,h,i) with
4 force sensors for independent measurement of finger and thumb forces.
(7) Finger load distribution during climbing: training session at Climb Adventure
Pte Ltd Gym, Keppel Towers, Singapore; self-manufactured CNC-machined edge
(Fig. 1-j,k) with 4 force sensors for independent measurement of all 4 fingers.
(8) Fatigue influence on the mechanical parameters of climbing: bouldering ses-
sion at Climb Adventure Pte Ltd Gym, Keppel Towers, Singapore; 4 self-
manufactured edges, 15 mm - 50 mm deep (Fig. 1-1).
3 Results
( 1) Mechanical parameters of climbing during competition: the contact time was

7.93 ± 2.89 s (mean ± standard dev), the mean resultant force 117.0 ± 30.6 N, the
impulse 897 ± 296 Ns, the mean friction 0.316 ± 0.129, the maximal friction 0.608 ±
0.210, and the smoothness factor was 24.6 ± 10.4. The force applied during the set-
up phase is still low, about 30 Non average (Fig. 2a,b). The beginning ofthe crank
phase is marked by a sudden rise of the force, which reaches 163 N on average. The
force during the lock-off phase is 62 N on average, and the friction is far higher
(0.34) than during the first two phases (0.05).
(2) Influence of experience on the mechanical parameters of climbing during
competition: the contact time was 8.24±5.03 s, the mean resultant force 100.0±30.1
N, the impulse 786±435Ns, the mean friction 0.401±0.189, the maximal friction
0.626±0.225, and the smoothness factor was 18.1±8.0. Generally, the experience of a
climber is expressed by a short contact time, a lower force at the handhold, a low
impulse, and a high friction (greatcr inclination of forcc vectors; Fig. 2c,d). The
better the climber is, the higher is the rate of reaching a high friction. The climbers
were divided into three groups, according to the world rank )ist after updating the
World Cup results of Singapore. The percentage of the climbers per group, produc-
ing a mean friction > 0.5 was: group 1 (rank <30): 42% >0.5; group 2 (rank 30-60):
33% >0.5; group 3 (rank >60): 20% >0.5. Thus, it appears that better climbers can
produce a higher friction, even if there is no need to do so. A high positive correla-
tion could be found between smoothness factor and friction (r = 0.916).
(3) Training influence on the mechanics of climbing: the climbers had to climb a
route eight times. The mechanical parameters were tested as to a significant differ-
ence (p ::=: 0.05) between first and last 2, 3, and 4 climbs. The contact time, mean
force, maximal force, and impulse decreased significantly from first to last 2, 3, and
4 climbs. The mean and maximal friction and the smoothness factor increased sig-
nificantly from first to last 4 climbs. Thus, the mechanical parameters behaved as
expected and the training effect could be proven.
(4) Friction between hand and climbing hold under different conditions ( static
friction coefficient in parentheses): 1) dry hand (0.722 ± 0.087), 2) wet hand (0.675 ±
0.164), 3) hand covered with powder chalk (chalk bag, MgCO 3 ; 0.958 ± 0.145), 4)
hand covered with liquid chalk and left to dry ('Pure Grip' liquid chalk, by Beal,
Vienne, France; 0.763 ± 0.155), 5) dry hand on messy surface (covered with liquid
Instrumented Climbing Holds and Dynamics of Sport Climbing 61
chalk; 0.925 ± 0.072), 6) hand covered with powder chalk on messy surface (0.650 ±
0.089). The moving COP indicated the onset of slippage; the static friction coeffi-
cient was detennined on the graphs displaying friction (tangential force over nonnal
force) versus displacement of the COP; the static friction coefficient is marked by a
sharp bend of the graph. The results revealed that there is no significant difference
between dry hand and liquid-chalked hand, between dry and wet hand and between
powder-chalked hand on clean surface and dry hand on messy surface. However,
powder chalk is far better than liquid chalk or a dry hand, and on messy (polluted
with chalk) surfaces, a dry hand is better than a powder-chalked hand.
(5) Mechanics of the double dyno technique: from the forces at the Iower hand-
hold and the two footholds, the take-off velocity and the jumping height were calcu-
lated. The results showed that in unsuccessful jumps, the take-off velocity is too low.
In successful jumps the climber jumps higher than required and Junges for the upper
hold far before the dead point. Furthennore, the closer to the dead point the climber
latches the upper handhold, the higher is thc peak reaction force at the fingers. The
advice for the climber is to jump higher than necessary, and to grab the handhold as
early as possible (before the dead point). This results into a high success rate and a
minimal finger injury risk.
(6) Mechanics ofthe pinch grip: in the pinch grip, the force ofthe thumb is about
1/5 - 1/4 ofthe finger force (Fig. 2e,f). The friction at the thumb is 2-3 times higher
than that of the fingers. Friction and forces on fingers and thumb increase with the
wall inclination; the finger forces increase more than the thumb forces do. Finger
mean rcsultant force (in multiples ofthe body weight) at 5° and 30° wall inclination:
0.275±0.047 and 0.395±0.038. Thumb force at 5° and 30°: 0.067±0.012 and
0.077±0.016. Finger friction at 5° and 30°: 0.222±0.066 and 0.376±0.062. Thumb
friction at 5° and 30°: 0.460±0.099 and 0.657±0.232.
(7) Finger Ioad distribution during climbing: we compared three different grip
types: closed crimp, open crimp, and open-hand grip. The greater force was applied
by the index finger in the closed crimp whereas the middle finger applied the signifi-
cant force in comparison with the index and the ring finger for both open crimp and
open-hand grip. The Iatter 2 have a similar force distribution shape, but the closed
crimp has a descending distribution with the index having maximum force to the
Iittle finger with the minimum force. The force distribution of the closed crimp cor-
relates to the previous works. The normal mean force (downward vertical force) of
the 4 fingers, index, middle, ring, and Iittle, during the closed crimp was 34.1 %,
28.8%, 22.4%, 14.6% respectively, during the open crimp 23.6%, 30.1 %, 25.6%,
20.8%, and during the open hand grip 24.0%, 30.3%, 26.8%, 18.8%. The distribution
ofthe horizontal mean force (out ofthe wall towards the climber) was: closed crimp
- 42.8%, 24.0%, 21.6%, 11.6%, open crimp - 28.0%, 30.6%, 24.5%, 16.1 %, open
hand grip - 25.1 %, 34.2%, 25.5%, 15.1 %.
(8) Fatigue influence on the mechanical parameters of climbing: finger reaction
forces were measured with instrumented holds during climbing a difficult bouldering
route. The fatigue level of the participants was assessed according to the Borg scale.
The mechanical parameters were correlated to the fatigue lcvel. Once fatigue sets in,
the force at thc holds drops more than 10% between two groups of Borg< 5.5 and
Borg> 5.5. This indicates that a fatiguing climber exert less force on the handholds,
which, in turn, leads to a lower injury risk during fatigue.
2 2 2
a b
Fig. 2. Different veetor diagrams; a, b: National Championship (l=set-up, 2=erank, 3=loek-

off); e, d: World Cup (e is more experieneed than d); e, f: pineh grip at 5° and 30° wall.
4 Discussion and Conclusion
Instrumented climbing holds provide a useful tool for assessing the performance of a
climber, be it during training or competition, as weil as training progress and suc-
cess, gripping techniques, equipment, and result in feedback advice for climbers,
e.g., lower injury risk while fatiguing, and higher risk for perfect dead-pointing.
Specifically, the results ofLi et al. (2001) could not be verified: chalk is not a 'myth'
as it provides higher friction than a dry hand. However, on messy ('chalked') sur-
faces, a dry hand is significantly better than a powder-chalked hand. The latter fact is
important for competitions, as this indicates a clear disadvantage of climbers who
start first, even if the route setters clean the holds during the competition. A hold
cannot be cleaned perfectly by a brush and gets messy faster than a washed hold.
References
Li, F. X., Margetts, S. and Fowler, I. (2001) Use of'ehalk' in roek climbing: sine qua non or
myth? J. Sports Sei., 19, 427-432.
Noe, F., Quain, F. and Martin, L. (2001) Influenee of steep gradient supporting walls in roek
climbing: biomeehanieal analysis. Gait Posture, 13, 86-94.
Quaine, F. and Martin, L. (1999) A biomeehanieal study of equilibrium in sport roek climbing.
Gait Posture, 10, 233-239.
Quaine, F., Martin, L. and Blanehi, J.-P. (1997a) Effeet of a leg movement the organisation of
the forees at the holds in a elimbing position. Hum. Mov. Sei., 16, 337-346.
Quaine, F., Martin, L. and Blanehi, J.-P. (1997b) The effeet ofbody position and number of
supports on wall reaetion forees in roek climbing. J. Appl. Biomeeh., 13, 14-23.
Quaine, F., Vigouroux, L. and Martin, L. (2003) Effeet of simulated roek climbing finger
postures on foree sharing among the fingers. Clin. Biomeeh. 18, 385-388.
Testa, M., Martin, L. and Debfi, B. (1999) Effeets of the type of holds and movement ampli-
tude on postural eontrol assoeiated with a climbing task. J. Gait Posture, 9, 57-64.
Testa, M., Martin, L. and Debfi, B. (2003) 3D analysis ofposturo-kinetie eoordination assoei-
ated with a climbing task in ehildren and teenagers. Neurosei. Lett., 336, 45-49.
Forces Generated in a Climbing Rope During a Fall
Andrew Phillips 1, Jeff Vogwell2, Alan Bramley3

1University ofBath, UK, Dcpartment ofMechanical Engineering.
2 UniversityofBath, UK, Departmcnt ofMechanical Engineering. J.Vogwell@bath.ac.uk
3University of Bath, UK, Departmcnt of Mechanical Engineering.A.N.Bramley@bath.ac.uk
Abstract. The use of micro-protection is key to the development of the spart of rock climbing
as hardcr, blanker rock faces arc attempted. Howcvcr in many situations thc force of a fall will
be severc cnough to injure a climber or exceed thc strength of this cquipment and so an im-
proved understanding of the factors affecting maximum impact force and subsequent minimi-
sation of this force is essential for the safc use of micro-protection. A laboratory scale rig has
bccn designed to mcasure the impact forces gcnerated during simulated simple climbing falls.
Thc results show an increase in force with subsequent falls on the same rope due to irreversi-
ble damage, however this effect becomes saturated after a certain number of falls. A simple
analysis using a linear ropc stiffness is describcd and its predictions compared with thc ex-
perimental rcsults. The theorctical force equation is gcnerally found to bc valid.
1 Introduction
Rock climbing as a sport has been in existence for over 150 years during which time
there have been major improvements in the safety equipment used. From rudimen-
tary beginnings modern equipment is now based on sound scientific principles and
has greatly reduced the risks involved (Smith 1998). Standards and popularity have
risen phenomenally over the last 50 years and for the average climber the sport is
now considered relatively safe. For those pushing the limits however the risks are
still significant and a thorough understanding of the capabilities and limitations of
their equipment is essential. The principles of traditional rock climbing involve the
ascent of a rock face using only natural features and technical climbing ability.
Safety equipment is used to guard against the consequences of a slip or fall rather
than a direct mechanical aid. The behaviour of the rope is therefore crucial and this
paper is concemed with the validation of a simple analysis for the rope forces gener-
ated in a fall, the effect of the knot used to attach the rope and the effect of repeated
falls and associated rest periods on the magnitude of these forces. An earlier theo-
retical analysis of this situation has been provided (Pavier 1998)
2 Analysis
The forces generated in the climbing rope during a simple fall can be determined by
considering the conversion of the potential energy of the falling climber into strain
energy of the extending rope. A simple analysis can be derived by assuming linear
elasticity of the rope which in reality will be visco-elastic. The model used is shown
in Fig 1. A climber runs out a length of rope / above the second (the belayer) who is
64 Andrew Phillips, JeffVogwell, Alan Bramley
at a distance h/2 when a fall occurs. After falling a distance h the rope starts to pro-
vide a restriction. This conversion of potential energy into strain energy is given by
Eq. 1.
ß[FORE FALi. AFTER FALL
Leader
Sccond
Fig.1. Geometry of simple climbing fall
EAö 2 (1)
mgh+mgö=--
21
where Eis the tensile elastic modulus of the rope and A its cross sectional area.
The force F generated in the rope is given by Hooke's law as
(2)
Solving Eq. l for ö gives t5 = mgl

EA
[! + l + 2AEh]
mgl
(3)
F=mg [ I+ l +2AEh]
Substituting for ö from Eq.2 gives --
mgl (4)
Thus for a climber of weight mg using a rope of stiffness EA/1, the maximum theo-
retical tension generated in the rope is dependent on the ratio h/1. This ratio, often
referred to as the "fall factor" by climbers enables an assessment of the integrity of
the safety chain in any given situation.
3 Experiments
A simple rig was constructed to enable drop weight test to be performed on a sus-
pended rope up to a maximum drop height of 790 mm. Instrumentation was provided
to measure the forces generated in the rope. Static tests were used to determine the
nominal stiffness values of the rope The rope used throughout the tests was 9 mm
Edelris Rocky Half. The typical behaviour of the rope during a drop test at maximum
length is shown in Fig. 2. The weight used was a 55 kg mass and this corresponds to
the recorded force (-0.5 kN) as the mass is suspended from the load measuring de-
vice. The maximum force recorded of 3 kN corresponds to about six times the static
weight.
...,1\---""-
2 3 4 5
im ( )
Fig.2. Typical force time relation during a drop test with a rope length of 855 mm, a drop
height of 790 mm and a falling mass of 55 kg.
The effects of various rest periods on maximum impact force generated was investi-
gated with the rope knotted to the mass carrier and using the mass of 55 kg. An ini-
tial drop test was carried out on three new rope samples from the maximum height of
790mm and total length of 855mm. In each case the mass was then left to hang stati-
cally on the rope for 5 minutes before being raised to the release position. For the
first rope sample a second drop was carried out after a rest period of 5 minutes, for
the second sample the second drop occurred after a rest period of 30 minutes and for
the third sample the rest period was 60 hours. At the beginning of each rest period
the knot was loosened and reset to its original dimensions.
The results are shown in Fig.3. Each of the initial drops occurred under identical
conditions so these results have been averaged. Despite minor variation, rest period
appears to have little effect on the impact force of a subsequent drop or fall under
these conditions. These conditions are clearly sufficient to cause permanent damage
and the resultant increased impact force, being greater than mg is significant. This
supports recommendations that ropes be disposed of if subject to fall factors of I or

above. Clearly, however, the increase in stiffness due to the tightening of the knot
during a fall is much less than that occurring as a result of permanent damage.
4 5
4 ,-
- verage a.x1mum
~
lmpaet orce of
35
z 3
... - ... . First Drop
~ 25 ■ lmp et Force of
i 2
econd Drop
Q lmpac1 Force of
second dr p with
untigh1ened kn01
05
0 . min Omin 60hr
--
Re. t Period
Fig. 3. Effect ofrcst pcriod on maximum impact forcc .
The effect of rapid consecutive drops on maximum impact force was investigated.
The rope attached to the mass carrier with a knot and using the füll mass of 55 kg, an
initial drop test was carried out on one new rope sample of total length of 855mm
from the maximum height of 790mm and. This remained under a static load for 5
minutes before being raised to the release position. After a 5 minute rest period a
second drop was carried out. Maintaining the same intermediate static load and rest
period a total of six consecutive drops were carried out. A total rope length of
855mm was again used. This series oftests was then repeated but excluding the knot.
The results are shown in Fig 4. The results suggest that the increasing stiffness from
repeating drops appears to saturate. lt is also apparent that the increased force re-
duces the safety margin available for the climber.
6r-------------------------,
-+- Including Knol

.....,.. Excluding Knol
0 .____ _....__ __.__ _......,_ ___..___ _....__ _.......__ _...,

0 2 4 6 7
Drop No.
Fig. 4. Variation of impact forcc with consecutivc drop tests
The effect of varying the Fall Factor was investigated used a rope attached but ex-
cluding a knot and a falling mass of 55kg. Using the füll rope length of 855mm, 5
new samples were tested each with a different drop height from 790mm to 190mm in
steps of 150mm corresponding to a reduction in fall factor from 0.92 to 0.22. Each
sample is only subject to one test. Using the linear tensile modulus (hence stiffness)
value calculated from the earlier static tests Eq. 4 can be evaluated to give predicted
impact forces for a range of fall factors. This is shown in Fig.5 along with the actual
experimental results obtained by varying drop height with a constant rope length. As
can be seen there is good correlation between the experimental and theoretical data
in terms of the shape of the curve the experimental data is at a lower force than the
theoretical. This suggests two things; Eq.4 is potentially valid for this situation, but
the empirical stiffness value is possibly inaccurate resulting in the higher theoretical
values. Altematively friction at the sliding interfaces reduces the maximum kinetic
energy of the mass.
4 ~--~--~--~---~--~
3.5
z 2.5
c 2
1!
&: 1.5
♦ Experimental Data
- Theoretical data based on

0.5 empirical linear stiffness value
0 ~-----~--~~--~--~
0.00 0.20 0.40 0.60 0.80 100
Fall Factor, h/1
Fig. 5. Variation offorce with fall factor
The variation of rope length at a constant fall factor was investigated using a rope
attached without a knot and a füll mass of 55kg. The rope length varied from 855mm
to 255mm in steps of 150mm with a new sample being used for each test.
3.5
♦
♦
♦ ♦ ♦
♦ Experimental Data
- Theoretical data
0.5 based on empirical
linear stiffness value
o~--~--~----'---~~---'
0 200 400 600 800 1000
Rope Lenth (mm)
Fig. 6. Effect ofrope length with a constant Fall Factor on the force generated in the rope
4 Conclusions
An initial impact force of approximately 3kN corresponding to a fall factor of 0.92,

is sufficient to cause permanent damage and a resultant increase in stiffness. This
does not reduce within the maximum rest period considered of 60 hours and results
in a significant increase in the force during a subsequent drop of approximately 25%.
Eq. 4 is generally valid within the context of these tests and for un-knotted rope
samples stiffness may be represented linearly, ignoring the initial less steep section
of the force-extension graph. The inclusion of the knot (as is the case in practice)
reduces impact force due to reduced stiffness. The stiffness of a knotted rope sample
is directly related to the percentage of the overall rope length occupied by the knot.
Consequently at short rope lengths force is not constant for a given fall factor (h/[).
As rope length increases the knot has a lesser effect on stiffness. These results may
be of particular relevance to climbers attempting rock climbs requiring the use of
micro-protection, aiming to minimise the impact force in order to maximise their
safety margin. The inclusion of one or more additional knots close to the hamess
attachment may be of benefit although this will obviously dictate that the climber
must always be at least a minimum rope length away from the nearest runner as the
knots will not pass through the karabiner. Furthermore; these results show that, with
respect to the rope, for those climbers operating at lower standards (who are always
using protection devices with the maximum strength rating and with the experience
to identify good placements), there are no consequences of a fall that will signifi-
cantly compromise their safety margin in a subsequent fall.
Acknowledgements
The authors acknowledge the University of Bath for the provision of facilities for
this work.
References
Pavier, M.J. (1998). Failure of climbing ropes resulting from multiple leader falls.
Proceedings ofthe 2nd International Conference on Engoneering of Sport (ed. S J
Haake) Blackwell Science, Oxford.
Smith, R. ( 1998). The development of equipment to reduce risk in rock climbing.
Journal ofSports Engineering. 1,27-39.
Rock Climbing Belay Device Analysis, Experiments
and Modeling
Lionel Manin 1, Matthieu Richard2, Jean-Daniel Brabant', and Marc Bissuel 1

1 INSA Lyon, France, lionel.manin@insa-lyon.fr
2 PETZL, France
Abstract. Amazingly no standard exists for belay devices used in rock or ice climbing. This
paper presents an analysis performed on several belay devices that permits their comparison in
terms of characteristics and efficiency. The belay device characterization consists in testing
the brakes in the case of a given fall arrest in a testing room. Here, a belay device is character-
ized by its braking coefficient, the rope slip through the brake till the arrest of the fall and the
impact load on the last runner. The braking coefficient is the ratio between the rope tensions
on the tight and slack sides ofthe belay device. A device called "virtual hand" has been devel-
oped, it'tries to reproduce the belayer hand action on the rope during the fall arrest, and it also
enables the measurement of the load on the brake and the control of the rope tension on the
belayer hand side. The fall arrest intensity or brutality is evaluated from the measurements of
the rope slip in the brake and the load on the last runner. The analysis of the three measured
characteristics enables the comparison of the six belay devices tested. lt appears that these
three quantities are related, indeed the !arger the braking coefficient the lower the rope slip and
the higher the impact load. The energy of the fall is absorbed over a shorter time period for
belay device having a high braking coefficient. Therefore, an efficient braking gives way to a
high impact force on the climber and a more brutal arrest of the fall. A basic model for the
brake has been implemented in an already developed climber fall arrest model. Comparison of
experimental and numerical results is made and is satisfactorily.
1 Introduction
Rock climbing material has considerably developed this two last decades. Many
technological innovations have been created for karabiners and belay device or
brake, they aim at getting products that are lighter, safer, more ergonomic and effi-
cient, and that also respect international standards (Smith 1996; Blackford and May-
cock 2001). Surprisingly, no standard exists for belay device (also called brake or
descender), it is only the manufacturer know-how and the climber experience that are
used as reference or base for new development and design.
In order to be able to compare several types of brake and to build a database for
the brake characteristics, an experimental and numerical study has been conducted.
The objectives are to characterize and to classify the brakes. This study is part of a
larger project that aims at simulating the climber fall arrest dynamics taking into
account fall parameters, rope route, anchor points and material characteristics
70 Manin, Richard, Brabant, Bissuel
(Manin, Mahfoudh, Jauffres and Richard 2005). In this project, a numerical model
has been developed and the brake characteristics are needed for the fall arrest simula-
tion. To accomplish this work, Petzl manufacturer of rock climbing material and the
French university INSA-Lyon have collaborated.
Here the purpose is not to discuss about belaying techniques and safe use of be-
lay devices, but to describe a characterization method that can give way to a classifi-
cation of the brakes. Previous works have studied the factors affecting the various
belaying techniques (Schad 2000; Smith 1996; Zanantoni 2000): use of twin or half
ropes, slip ratio, type of device, position and weight of the belayer, length of slipping
rope, amount of friction along the rope. The major aim is the evaluation of the im-
pact load on the last runner and the rope slip depending on the belay device used.
These two quantities give indications on the brutality of the fall arrest and on the
brake efficiency. Several models have been developed and they perform satisfacto-
rily (Bedogni 2002; Pavier 1998; Manin et al. 2005), however few work has been
done on descenders.
2 Belay Device Experimental Analysis
The presence of a large amount of friction in practically all real falls is a great advan-
tage in mountaineering, but it conceals a number of basic factors which are funda-
mental for the analysis of the dynamic belay process (Zanantoni 2000), and it does
not permit to see the contribution of the belay device in the fall arrest. Moreover, the
behaviors ofbelayers when stopping a fall are very different, and even with the same
operator repeatability is difficult to obtain. Therefore, in order to test in the same
way all the belay devices considered, it was necessary to have a fall easily repro-
ducible with a fixed configuration and a minimum number ofperturbing factors. The
tests have been realized in Petz! test room equipped with a fall tower and a specific
device called "virtual hand" to which the brake is attached. The virtual hand plays
the role of the second and insures an almost identical braking action for each fall
arrest.
2.1 Test Rig Description

The fall configuration used to test the belay devices is very simple (Fig. 1): only one
karabiner (the last runner) in the rope route except at the retum anchor. The "virtual
hand" (bottom right of Fig. 1) has been developed to reproduce the belayer hand
action on the rope during the fall arrest, and it also enables the measurement of the
load on the brake and the rope tension on the belayer hand side. The "virtual hand"
uses a torque limiter coupled with a barrel on which the rope is coiled to apply an
almost constant retaining load on the rope as the belayer band does. The torque lim-
iter has been adjusted so that it resists to a rope tension smaller or equal to 160 N,
over that limit the barrel rotates and the rope slips through the descender. The value
of 160N has been determined from the measurement of the maximum load applied to
a rope a hand could retain. A torquemeter placed between the torque limiter and the
barrel permits to measure the torque on the barrel and therefore the rope tension on
Rock Climbing Belaying Device Analysis, Experiments and Modeling 71
the slack side of the belay device. Load sensors are placed at the last runner and at
the brake. All sensors are connected to a computer via a data acquisition board.
,/ M
7.3 kN
23
device s!de L 1 621 m
L2 15m H,
643m
15m
07m
0.12
M 80
v,nual hand flXed
/////II////////////////////////
Fig. 1. Fall configuration parameters, fall tower and virtual hand
2.2 Belay Device Measured Characteristics and Results

Three characteristics have been measured for the six brakes tested: the impact load
on the last runner, the braking coefficient and the final rope slip in the brake. The
braking coefficient ßf is defined as the ratio between the rope tensions on the tight
span (T1) and the slack span (Fmax_hand) ofthe brake:
ßt T1 (1)
Fmax_hand
Figure 2 shows the evolution of all the measured quantities versus time during a
fall arrest done by the virtual band for two different brakes. T2 is the rope tension on
the mass side. lt can be seen that the type ofbelay device used has great influence on
force magnitudes and displacements. For the two cases of Fig.2, differences of lkN
for the impact force and 1.5m for the displacement are observed. This points out,
regarding the climber potential injury, that the force supported by the climber and the
rope slip through the brake during a fall arrest can be very different depending on the
belay device used. The arrest of a given fall can therefore be progressive or brutal.
The braking coefficient is not constant during a fall arrest (Fig. 3), it increases
during the slowing of the fall until the start of the rope slip in the belay device and
then it remains almost constant until fall is stopped (about 1.3s) and then slightly
decreases. lt is this constant value that is considered. The measured belay device
characteristics are summarized in Table 1, they correspond to mean values obtained

from at least three tests. A good reproducibility of the test was observed.
Reverso
ATC XP
~000 4000 7
3
5
e1. 2
4 i "] -
: ~
•u •u
-; 2000 ~
•
=::u 2000. j
3-ä_.
i:5
l
1500
3 il.
III
.
i5
is
...
1500
2 2
1000 1000
500 500
0 0 0 - ~ -' - - - - ~ - - - -
!5 00 05 1 o I Cs) 1 s 2.0 25 00 05 , 0 I (S) 1 5 20
-- Brake karab,nor load v,r1ua1 hard for~ - LaSlnnierload
_ _ . · T2 -massdi"'lacem~ L-n -~==
Fig. 2. Load and displacement comparison for two types ofbelay device
B 1a,in!! Rev 0 Tube Tube :\.l>

de,ice
PETZL PITZL .\T . .\T . p ZL 1-.:nol
name
La,-t runner
2500 2800 3300 3950 2800 2 50
load (!
Br:ik
eo fficicnl ·'- 5.5 II ~.5 6
P,
Rop ·llp
U6 1. 6 0. 6 0.3-4 1. 1.35
(m)
Table 1. Brake measured characteristics
The brake coefficient is related to the magnitude of the impact force at the last
runner, they increase together. On the contrary, a high brake coefficient corresponds
Rcick Climbing Belaying Device Analysis, Experiments and Modeling 73
to a small rope slip in the descender. Regarding the efficiency of stopping a fall,
brakes with large braking coefficient perform better as they stop the fall quicker than
brakes with smaller braking coefficient. However, the energy of the fall being ab-
sorbed on a shorter time, the brutality ofthe fall arrest is higher. Note that the values
presented are for a given fall in a test room with a fixed belayer role played by the
virtual hand.
G
14
- Fig1.,-e8
- - reverso
12 · · ·ATC
--pirana
• • Cabestan knot
~ 1: 1
ATCXP
o l
05 07 09 15 17 19
11 t(s) 1 3
Fig. 3. Braking coefficients evolution during a fall arrest
3 Brake Modeling and Comparison with Experiments
A basic modeling of the belay device has been implemented in our model (Manin et
al. 2005). Here the brake is considered as a force multiplier (Bedogni 2002). lt helps
retaining a rope tension by amplifying the load applied by the belayer's hand on the
rope. Slip occurs at the belay if the tension on the tight side of the brake divided by
the braking coefficient ßJ is equal or larger to a critical value (Pavier 1998). The
critical value corresponds to the maximum rope tension a hand can hold (160N for
results presented on Fig.4), it can be adjusted in the model.
Figure 4 shows simulation results (thin line) compared to measurements (hold
line) obtained for two different brakes used for the arrest of the same fall. The
agreement between calculations and measurements is good, it is better for the mass
displacement and the virtual hand force than for the last runner and brake loads
where the calculated and measured shapes ofthe load evolution differ. Some discon-
tinuities are observed for the calculated last runner load at the start of the rope ten-
sioning phase. lt is due to the time incremental numerical scheme used, that point is
being improved.
___
j~
___________,
]~
l ..;
~ ·,,.
: ,..__
i
i „
! :,~
: ---
. ......
Fig. 4. Simulation results compared to measurements for Figure 8 and Reverso Brakes
4 Conclusion
A method for characterizing belay devices has been presented, it permits to compare
several designs from the determination of three parameters. Impact load, rope slip in
the brake an braking coefficient observed for each brake are related and their analy-
sis enables the estimation of the brutality of a fall arrest. Regarding the comfort of
the climber, a descender with a large braking coefficient will stop his fall efficiently
but uncomfortably.
References
Bedogni, V. (2002) Computer mathematical models in belaying techniques, Report Jtalian

Alpine Club.
Blackford, J., Maycock, E. (2001) Mountaineering Equipment- Ropes Harnesses, Karabiners
and Anchors, Materials WorldVol. 9 no. 8, pp. 8-12 Special European Supplement.
Manin, L., Mahfoudh, J., Jauffres, D., Richard, M. (2005) Modeling the climber fall arrest
dynamics, 5th International Conference on Multibody Systems, Nonlinear Dynamics, and
Contra!, ASME, Long Beach, USA, Sept. .
Pavier, M . ( 1996) Derivation of a rope behavior model for the analysis of forces developed
during a rock climbing leader fall, The Engineering of Sport, Haake S., pp.271-279.
Pavier, M. (1998) Experimental and theoretical simulations of climbing falls, Sports Engineer-
ing, pp. 79-91.
Schad, R. (2000) Analysis of climbing accident, Accident Analysis & Prevention, Volume 32,
Issue 3, pp. 391-396.
Smith, R. ( 1996), The development of protection systems for rock climbing, The Engineering
ofSport, Haake S., 1996, pp229-238.
Zanantoni, C. (2000), Analysis ofbelaying techniques: A typical UIAA Activity, World
Mountaineering and Climbing, The Journal ofthe UIAA, no. 3, pp. 7-11.
3 Cycling
Synopsis of Current Developments: Cycling
Martin Strangwood
Sports Materials Research Group, The University of Birmingham, Departmcnt of Metallurgy

and Materials, m.strangwood@bham.ac.uk
Scope of the sport
Cycling, as a sport, encompasses a range of types from the reproducible conditions

prevailing in indoor velodrome track cycling through road racing to the very variable
conditions experienced in mountain or all-terrain biking. All of these aspects of the
sport are represented by the papers presented in this section of the conference. In-
deed, by covering the performance of helmets the paper by Alam et al. is also rele-
vant to everyday commuting by bicycle. The papers presented also cover a number
of aspects involved in circular development from design through performance to
improved design, with a strong concentration on sensor development and analysis to
optimise performance and training.
Equipment design - mechanics and sensors
The need to maximise mechanical efficiency in cycling equipment has long been an
area of research and innovation and this continues in the papers presented here,
which cover all three disciplines noted above.
One method of increasing the specific stiffness of a bicycle and its components
comes from the use ofimproved ('advanced') materials in traditional and more inno-
vative designs. In recent years titanium- magnesium- and aluminium-based (e.g. Al-
Li) alloys have been introduced to the frames of road bikes, whilst track bikes have
made increasing use of carbon-fibre composites. The effective use of these new, and
generally less forgiving, materials requires greater understanding of the operating
conditions of the equipment and hence more optimized design and manufacture. The
paper by Caton et al. is an example of this where a composite-skinned foam sand-
wich structure, which has been used for many years for disc wheels, has been modi-
fied in its manufacture to reduce weight and maintain stiffness and strength. The use
of more specialized materials, such as polymer and meta! matrix composites and
high strength steels is likely to see greater use of composite structures such as these
in giving a mix of properties.
Whilst the design of road bikes is fairly traditional, innovations can occur in compo-
nents, which are exemplified by Horvais et al. who have assessed the force evolution
78 Martin Strangwood
from non-circular chainring and related this to triathlete performance. This is a good
example of linking theoretical mechanical advantages with what can actually be
achieved by athletes using the equipment. At the same time any reaction forces must
be within acceptable limits for the athlete's physiology and, whilst Hervais et al.
concentrate on optimizing performance, Müller et al. address the issue of measuring
reaction forces for cycling after total knee replacement. Research using these meth-
ods and in this area could be expanded to physiotherapy for cyclists after injury and
an optimization of their retum to füll fitness levels, minimizing the risk of aggravat-
ing the injury.
Greater design innovation is associated with mountain biking and this is shown
by the various suspension designs studied by Tempia et al.. The dynamic response of
suspension systems is crucial to both performance and athlete 'feel' and this requires
careful monitoring. Tampia et al. have modeled the suspension travel, whilst Red-
field and Sutela have modelled and dynamically measured the response of a rear
shock to give its force / velocity characteristics. Greater development of systems to
dynamically measure the response of cycling equipment in use will be seen in future
and should lead to greater athlete / equipment optimization when used to verify mod-
els on readily accessible code such as Matlab and Mathematica.
Sensors and training
As well as sensor development for the mechanical response of the equipment, then
increased use of sensors is being seen in assessing the performance of athletes in
competition and during training. The two submissions by Jaitner and co-workers
describe the development of sensing systems and protocols to allow coaches to opti-
mize the training of individuals in a group environment. Validation of these ap-
proaches should open up this approach to a range of other sports and is an interesting
development.
Aerodynamics and safety
In common with other vehicles, aerodynamics is an important parameter influencing

performance and 'feel'. The relatively controlled atmosphere of indoor track cycling
lends itself to both measurement and modelling of aerodynamic parameters, which
form part of an optimization pro gram for cycling [Lukes et al.] along with power
output and frictional forces. The verification of this approach should allow forms of
cycling in more varied conditions, e.g. road racing, to be modelled and linked to
variations in equipment, such as those noted above.
Finally, and most importantly, it is necessary for sport to be safe for the partici-
pants and this is covered by Alam et al.s contribution on cycling helmets, which
would be pertinent to all forms of cycling. The need to balance comfort, i.e. cooling,
with performance, i.e. low drag, is a common problem in many sports and one where
quantitative study is needed along with determination ofthe athlete's perception.
Thermo-mechanical Modification Techniques for
Structural Foams used in Racing Bicycle Wheels
Catherine Caton, Mike Jenkins and Martin Strangwood
Sports Materials Research Group, The Univcrsity of Birmingham, Department of Metallurgy

and Materials, m.strangwood@bham.ac.uk
Abstract. The effects of a range of thermo-mechanical treatments on closed cell polyme-

thacrylimide foam have been quantified in terms of surface roughncss, closed cell dimensions
at the surface and in the bulk, and cell wall thickncss. These have been carried out for a range
of specimen sizes and have lcd to thc determination of optimal conditions to balance expan-
sion of closed sub-surface cells and collapse of cell walls for the open surfacc open cells for
smallcr foam samples, where the modification is constrained to the surface regions and leaves
the bulk compressive and shear properties of the foam unaffectcd. Surface modification in this
manncr reduccs thc uptakc of rcsin by thc foam during manufacture of sandwich beams and
discs. Application of the surface modification to !arger samples resulted in exccssive thermal
gradients and bulk porc formation highlighting thc need for cdgc constraint.
1 Introduction
Composite sandwich disc wheels are frequently used in track racing where low
weight and low aerodynamic drag are required whilst providing a wheel of high
stiffness and strength. Disc wheel structures comprise a high shear strength, low
density polymer core, surrounded by high modulus carbon fibre composite skins.
Production of sandwich beams uses of an adhesive resin layer to produce a coherent
bond between skin and foam core. Due to the cellular structure of the foam, this
leads to a significant amount of resin uptake into the core, so increasing the mass and
inertia of the structure, important factors in wheel design and performance.
Previous work [Caton et al., 2004] investigated a thermo-mechanical surface
modification process to prevent resin uptake into a polymethacrylimide (PMI) closed
cellular core. Using a matrix oftime, temperature and pressure conditions, lcm3 cube
specimens were heat treated in air on a hot plate to determine optimum thermal
treatment conditions of 212 °C for 5 minutes under a pressure of 491 Pa, which did
not affect the bulk material, but reduced resin ingress by 14% when manufactured as
carbon fibre/PMI sandwich beams. Mechanically stiffness was found to be the same
as in beams with untreated cores and all failed through the foam in the near interfa-
80 Cathcrine Caton et al.
cial region showing no reduction in the interfacial bond with surface modification,
but this needs to be investigated for )arger structures.
2 Experimental Procedure
Scaling up was achieved in two stages; initially 'small' (sm) foam samples (40mm 2
surface area, 7mm in depth) of Rohacell PMI 51 IG density (Pc) of 52 kg/m 3 were
treated thermo-mechanically on one face in air on a hotplate at temperatures between
185 and 240 °C; 1159 - 1465 Pa pressure and for times between 1 and 3 minutes.
51 IG [Roehm, 2004] softens at around l 80°C. Top surface and centre temperature
profiles for the 200°C and 240°C, 1465 Pa pressure and 2 minutes schedules were
recorded using Type-K thermocouples. 'Large' (1) sized specimen, 125 x 262.5 mm 2
surface area, (7mm depth, 51 IG), were thermo-mechanically treated on both faces
simultaneously using two heated Al platens in an oven for the same conditions as
above. An edge-constrained specimen, using a steel ring and the sample was circular
(15 mm radius and 10 mm thick) was also investigated (185 °C, 1465 Pa, 2 minutes).
Surface-modified sm samples were impregnated with ACG VT A260 PK 13 epoxy
adhesive under vacuum (pressure under 20 mm Hg) following a eure cycle of
0.5°C/min heating ramp rate, 2 hour dwell at 100 °C and 2°C/min cooling rate.
Modified surfaces were gold coated and observed in a Jeol 6060 scanning elec-
tron microscope (SEM) operating at 20 kV. Bulk and ingress samples were sectioned
normal to the top face using a slow speed diamond saw to minimise damage, gold
coated (bulk) or cold mounted in Epofix (ingress), ground to 500 grit and polished to
a 1 µm finish, and examined by SEM (bulk) or optically (ingress). Axiovision 4.0
image analysis software was used to characterise cell and pore areas, lengths and
breadths along with cell wall thicknesses. Resin ingress depths were measured di-
rectly from an optical microscope fitted with KS300 image analysis software, whilst
sectioned surface images were measured to give Ra roughness values.
3 Results And Discussion
The sm samples resulted in a significant temperature variation through thickness,

Fig. 1 (a), which was not seen in the I samples (heated from both sides), Fig. 1 (b).
This results in distortion (buckling) of the foam with strains limited at low tempera-
ture; increasing through-thickness direction as temperature increased. Table 1, which
indicates that greater (gas) expansion ofthe cells, without permeation occurs.
The use of dual-sided heating for the I samples reduced the curvature, but at the
expense of )arge central pore formation, Fig. 2. The size of the pores was !arger at
lower temperatures, Table 1, which corresponded to greater strains suggesting that,
at higher temperatures compression of the sample through thickness combated the
gas expansion. Pore formation was associated with audible signals from cracking
along the centre-line and this occurred within the first 30 seconds of the heat treat-
Thermo-mechanical Modification Techniques for Structural Foams 81
ment, which corresponds to the initial steep temperature rise noted in thermal pro-
files, Fig. 1 (b ). The steel ring constraint eliminated the central pores at 185°C .
•I "'
1:
F•--•- - -1
---
0 - - "- " - - - ..
(a} (b}
Fig. 1. Thermal profiles for (a) single-sided and (b) dual-sided heating.
Table 1. Mean sample and pore dimensions for modified PMI 511G (1465 Pa).
Temperature Time Height Width Length No. ofMean area

(min) (mm) (mm) (mm) pores (mm2)
185 (!) 1 8.4 130.6 278.5 10 80.2
185 (1) 2 8.2 130.5 275.0 5 70.6
190 (1) 1 7.5 136.0 285.6 7 34.4
230 (1) 2 7.2 127.6 269.5 16 3.6
240 (1) 2 7.1 111.5 240.0 5 1.7
200 (sm) 7.1 40.1 39.8
240 (sm) 8.1 41.08 41.01
Fig. 2. Central pores after dual-sided heating of PMI 511G foam (185 °C / 1 min./ 1465 Pa).
The roughness of thermo-mechanically treated surfaces varied with time, tempera-

ture and pressure, but not in any simple, monotonic manner, Fig. 3. Lower treatment
temperatures, e.g. Fig. 3 (a), generally show the lowest Ra values at low applied
pressure increasing again at the intermediate pressure level before being reduced at
the highest pressure level. This order is followed over the time range studied, but an
initially high Ra is reduced from 1 to 2 minutes exposure, but then increases up to 3
minutes exposure. With increasing temperature then Ra increased to maximum val-
ues at 210°C except for high er pressures and longer times, when the Ra maximum is
82 Catherine Caton et al.
shifted to 220°C. At 230°C and above the Ra values showed a decrease, Fig. 3 (b ),
which was largely time-independent at low pressure but varied at higher pressures.
The heat-treated sm sample surfaces showed collapse of the original surface cell
walls causing a reduction in Ra values. Heating of the sub-surface cell walls softens
them whilst the gas inside expands leading to an increase in cell volume and bulging
at the surface tending to increase Ra values, Fig. 4 (a). The variation in Ra with pres-
sure and time at low temperatures would be consistent with high local stresses on the
surface (open) cell walls giving rise to rapid plastic collapse. This would increase the
contact area reducing the local stresses with time, coupled with increased heat trans-
fer causing later temperature rises and gas expansion so that Ra increases again .
...
...
.
,
, ...
-·
/
....'"' j•M
J·• 1::
.
,..________.======
[:m" .. . , · ~
-· " .
a (b)
Fig. 3. Ra variation for heat-treated Sill PMI 5 lIG foaill saillples at (a) 200°C and (b) 240°C.
(a) (b)
Fig. 4. Sill foaill saillple surfaces heat treated at (a) 200°C and (b) 240°C (2 illins / 1159 Pa)
Increased treatment temperature reduces the failure stress of the surface cell walls
leading to more rapid collapse of the surface cell walls, but also easier flattening of
the tops of the expanding sub-surface cells so that the surface is flatter overall with
evidence of greater elevated temperature material flow, Fig. 4 (b). A cyclic variation
of surface roughness may be expected with temperature/time increase as the greater
expansion of the enclosed gas in sub-surface cells increases Ra until cell wall
strength is reduced and rupture occurs to increase R. further, only to be followed by
cell wall collapse reducing Ra and leading to bulging of a lower layer of closed cells.
Thermo-mechanical Modification Techniques for Structural Foams 83
In the un-heat-treated condition the open surface cells tended to fill with resin
from any adhesive layer resulting in ingress to an average depth of 339 µm (between
282 and 410 µm [Caton et al., 2004]). In the sm samples, closure ofthe surface cells
has resulted in a much more unifonn ingress depth, which was temperature-
dependent, but largely time- and pressure-independent. At 210 °C, all three pressures
and times gave resin ingress between 148 and 188 µm, i.e. a 50 % reduction.
Single-sided heat treatment of sm samples gave satisfactory surface modification
and resin ingress reduction, but bowing of the sheets resulted in local variations in
pressure and temperature and inhomogeneous behaviour. Dual-sided heat treatment
of the 1 samples overcomes the bowing problem, but introduces central pores. Pores
were !arger for lower temperature heat treatments, Table 1, when significant expan-
sion of the sample took place; increasing the holding time at the lowest temperature
reduced the pore size, but did not reduce the overall expansion of the samples. In-
creasing the treatment temperature reduced both the pore size and sample expansion,
with the highest temperature resulting in overall shrinkage, but pores did not re-heal.
Individual pores showed size variation, which may result from size and gas content
variations in the original foam; this still needs confinnation. Pore imaging by SEM
shows both spherical and elliptical (higher temperature) expansion of cells. Examina-
tion of the ruptured cell walls associated with the central pores reveals the presence
ofbuckling and shear fracture, Fig. 5 (a), akin to that seen in room temperature shear
testing of the unmodified foam, Fig. 5 (b ). These features would be consistent with
shear loading of the cell walls to failure before they have been heated sufficiently to
defonn plastically; consistent with early pore fonnation.
(a) (b)
Fig. 5. SEM imaging ofshear failure surface of(a) a porc; and (b) an untreated core.
Detennination of cell sizes for the variously heat treated PMI 51 IG foam samples
indicates a !arge scatter, as expected from foamed material, but, importantly samples
showing !arge pores have !arger cells to the edge of the specimen, but this is reversed
(smaller edge cells) for samples showing limited pore fonnation.
The variation in cell structure in the bulk and at the surface above is consistent
with collapse of the open surface cell walls initially as the high local stresses com-
bined with elevated temperature cause plastic flow. Heating of sub-surface and lower
cells follows softening their walls with intemal gas expansion and wall thinning, in
extreme cases leading to cell wall rupture. Poor foam thennal conductivity causes the
84 Catherine Caton et al.
centre to be cooler than the surfaces, which means that the cell walls are still brittle
whilst any heated surface is expanding so that the gas in the centre cells is being
compressed shearing the vertical (cool) cell walls. Dual-sided heating increases
shearing to expansion from above and below and can exceed the fracture stress of the
walls to cause the failure features noted in Fig. 5. Lower treatment temperatures
result in longer times for cell wall heating and plasticity so that pore expansion can
occur to the !arge dimensions noted in Table 1. Once formed, pores open up on
unloading to give expansion of the sample in the thickness direction, Fig. 2. Hence,
extended time at l 85°C would not be expected to reduce the pore size suggesting
that the trend shown in Table I may represent statistical variations in the starting
material. Increasing treatment temperature increases thermal gradients resulting in
greater thermal flux and so a faster temperature rise in the centre so that the cell
walls are more plastic and can accommodate increased gas pressure without fracture.
On unloading the still intact cell walls will relax and contract so that the overall
sample dimensions show less through thickness expansion, Table 1.
Thermally activated gas permeation and diffusion should also occur. This is eas-
ier for higher cell wall temperatures, so that, for lower temperatures, higher intemal
closed cell gas pressures occur, further enhancing cell wall fracture. Intact cells will
be !arger at lower temperatures due to higher intemal pressure without permeation
relief. The use of a steel ring for edge constraint may not have prevented gas loss,
but have accelerated it by providing a high conductivity path to heat the edge cells to
reduce the stresses caused by gas expansion in the un-constrained I samples. The use
of a non-conductive constraint needs to be carried out to confirm this. Thus indicate
controlled heating rates balancing of expansion and heat transfer may be needed.
4 Conclusions and Further Work
Thermo-mechanical treatments leading to surface modification of PMI 51 IG such

that resin ingress is decreased by 50% have been identified. The behaviour of the
foam is based on a balance of heat transfer, thermal softening, gas expansion and
diffusion, and plastic flow, which has not allowed scaling up ofthe treatment without
pore formation. Identification of a working hypothesis for pore formation has indi-
cated that greater control of thermal transfer is likely to be a successful approach.
References
Caton, C.J., Jcnkins, M.J., Strangwood, M., (2004), The Engineering of Sport 5, Eds
M.Hubbard, R.D.Mehta & J.M.Pallis, ISEA, Shcffield, pp 227-233.
Gibson L.J., (1984), Optimization of stiffness in sandwich beams with rigid foam corcs, Mate-
rials Science and Engineering, 67 125-135.
Rochm (2004) www.roehm.com
The Effect of a Non-Circular Chainring on Cycling
Performance
Nicolas Horvais, Pierre Samozino, Frederique Hintzy
Laboratoire de Modelisation des Activites Sportives, Bourget du Lac, France,

nicolas.horvais@etu.univ-savoie.fr
Abstract. The aim of this work was to analyse the effect of a non-circular chainring
during sub and supra-maximal cycling conditions on physiological, mechanical and
muscular data. Results showed that the use of the non-circular chainring was
beneficial during top and bottom dead centres by decreasing the effective force for a
same extemal force for sub-maximal condition and by increasing the crank angular
velocity for supra-maximal condition. However, this non-circular chainring was
without effect during the pedal downstroke.
1 Introduction
Cycling propulsion only result from forces applied to the pedals. The study of pedal
forces application could help to understand the cycling performance since Ericson
and Nisell (1988) showed that the effective force applied to the pedal (tangential
force) was maximal during the downstroke at 90° and minimal at the top and bottom
dead centres. This effective force was transmitted to the chain via the crank and the
chainring. Thus, an improvement performance could then be realized by modifica-
tion on chainring. Traditionally, chainring were circular. Recently, a non-circular
chainring called Osymetric chainring (OC) has been manufactured on the no constant
forcc application principle throughout the pedal crank revolution (Fig. 2). Indeed, the
OC radius described a sinusoid curve similar to the effective force evolution during
the pedal crank (Fig. 1), i.e. )arge radius when the effective force was maximal and
inversely. Theoretically, OC would allow facilitating the foot path around the top
and bottom dead centres by a lower radius than a standard circular chainring (CC). In
order to keep the same development, it was necessary to increase the radius chain-
ring in another zone. In order to minimize the negative effect of this radius increase,
this one has been placed where the effective force applied to the pedals is the most
important (i.e. at 90°). To our knowledge, only one study tried to compare the oxy-
gen consumption between OC with CC in sub-maximal condition (Rate!, Duche,
Hautier, Williams and Bedu 2004) and did not show a significant. lt could be ex-
plained by a no muscular pattem modification. However, the muscular activity was
not measured.
86 Nicolas Horvais, Pierre Samozino, Frederique Hintzy
90 180 270 360

crank Mgle !")
Figure 1: Radius of a circular chainring (CC)

Figure 2: Geometry ofthe
and an Osymetric chainring (OC) for 52 teeth
Osymetric chainring.
chainrings during a pedal revolution
The aim of our study was then to compare and analyse mechanical, physiological
and muscular data obtained during endurance and anaerobic fatiguing Wingate tests
using OC and CC.
2 Materials and methods
2. 1 Subjects
12 male triathletes volunteered to participate in this study. Their age was 32.0 ± 6.7
years, height was 179 .3 ± 5 .2 cm, weight was 77 .5 ± 10.2 kg and fat mass percentage
was 16.8 ± 4.6 %.
2.2 Experiment protocol
Each subject participated in randomized two sessions (OC and CC sessions) sepa-
rated of 24h recovery. During the two sessions, subjects realised:
- a 8 min sub-maximal test at of 100 and 200 W (4 min per power output in an as-
cending order) at a constant pedalling rate fixed at 80 rpm.
- a supra-maximal test corresponding of a 30 s Wingate test against a friction load of
0.834 N.kg- 1 body mass.
The starting position was standardized with the right foot placed at 45°. At the signal
given by the experimenter, subjects were vigorously encouraged to sprint maximally
(i.e. from zero to maximal velocity) and to maintain this maximal effort during 30 s.
During each test, subjects had to stay seated on the saddle. The handlebar and the
saddle height were adjusted to height's subjects and remained constant for all tests
and sessions. Feet of subjects were not fixed on the pedals in order to avoid a traction
phase during the upstroke.
2.3 Material and data analysis
Mechanical data analysis
A typical friction-loaded cycle ergometer was used for this study (Monark 818E,
Stockholm, Sweden), specifically equipped with strain gauge (Interface MFG type;
The effect of a non circular chainring on cycling perforrnance 87
Scottsdale, Az, USA) for the friction force measurement and with optical encoder
(Hengstler type RIS IPSO, Aldingen, Germany) for the flywheel measurement, (Ar-
sac, Belli and Lacour 1996). Flywheel angular displacement and brake belt force
were sampled at 50 Hz. The cycle ergometer has been modified by the addition of a
52 teeth CC and a 52 teeth OC.
Instantaneous extemal force (Fe,t in N) produced by subjects was determined as the
sum of frictional force (measured by the strain gauge) and inertia force (dependent
on the acceleration of the flywheel) (Arsac et al. 1996). Moreover, the effective force
(Ferrcc in N) applied to the pedals was calculated from the instantaneous extemal
force (Cavanagh and Sanderson 1986).
During the sub-maximal test, instantaneous maximal and minimal extemal and effec-
tive force (F ext-max, Fext-min, Feffcc-max and Fcffcc-min), pedalling rate and power output
were recorded over each pedal downstroke, which was limited between the two
minimal values of instantaneous power output corresponding to top and bottom dead
centres, and were analysed between OC and CC. During the 30 s Wingate test, power
output was averaged during the first 5 s (P mean 0 - 5 s), the last 5 s (P mcan 25 - 30 s)
and during the entire test (P mcan 0 - 30 s ). The rate of decrease of power output during
the test, i.e. the fatigue index, was obtained by the ratio between the Pmcan 25 - 30 s
and the Pmean 0 - 5 S.
Physiological data analysis
Oxygen consumption (V; 0 2) and carbon dioxide production (V; C0 2) were meas-
ured breath to breath with an automatic gas analyser system (Cosmed K4b2, Rome,
ltaly). Both 0 2 and C0 2 fraction were calibrated with know reference gas mixture
(room air and a standard certified commercial gas preparation). The expired gas
volume tlow rate calibration was performed by mean of 31 syringe (Hans Rudolph).
During the sub-maximal test, V; 0 2 and V; C0 2 have been measured continuously
and averaged over the last 30 s of each power output. Gross efficiency (GE) was
obtained by the ratio of mechanical energy to energy expenditure.
Muscular data analysis
Electromyographic surface signal (EMG) was collected at 1000 Hz during all tests
using an eight ways system (Mega-ME3000P8, Mega Electronics, Finland). Bipolar
electrodes were placed on the skin (spaces of 2.5 cm) over the gluteus maximus
(Gma,), the biceps femoris (BF), the rectus femoris (RF), the vastus lateralis (VL), the
tibialis anterior (TA) and the gastrocnemius (GAS) muscles of the right lower limb
for both tests. The skin at each electrode site was shaved and cleaned with an alco-
hol-ether mixture. The EMG signals were amplified (x 600) and filtered and the
treatment was carried out thanks to the software Megawin (Mega Electronics Ltd,
Finland). The raw EMG data were full-wave rectified and smoothed using a 50 ms
moving averaging windows. From the raw EMG, averaged EMG (aEMG) has been
chosen for analysis.
88 Nicolas Horvais, Pierre Samozino, Frederique Hintzy
Ouring the sub-maximal test, aEMG was averaged during the last I Opedalling cycles
of each power output. aEMG was normalized to the maximum value observed during
one burst across the Wingate test for each individual muscle (%). The averaged total
EMG (aEMGT in%) was averaged from the 6 muscles studied. Ouring the 30 s Win-
gate test, aEMG was averaged during the first 5 s and the last 5 s and during all the
30 s of the test. aEMG was normalized (i) to the maximum value observed during
one burst across the Wingate test for each individual muscle and (ii) to the power
output realised (%.W- 1). The averaged total EMG (aEMGT in (%.W- 1) was calculated
with the 6 muscles studied. The rate of decrease of aEMG during the test, i.e. the
fatigue index was obtained by the ratio between the aEMG of the last 5 s and the
aEMG ofthe first 5 s.
Statistics
Results were presented as mean ± standard deviation (SO) values. Comparison of

data between CC and OC were realised using a Wilcoxon test. The limit for statisti-
cal significance was set at P < 0.05.
3 Results
Conceming the sub-maximal test, data were presented in table 1. No significant

difference appeared in all data between OC and CC except for Fcrrec-max in both pow-
ers output and for F cffec-min at 100 W (F errec-max with OC significantly higher than CC
and Ferrec-min OC were significantly lower on CC). Conceming the Wingate test, data
were presented in table 2. There was no significant difference in both mechanical
and muscular studied data between the use of OC and CC.
100W 200W
Submaximaltest
CC oc CC oc
Phys1olog1cal
GE 13 9 ± 1 7 135±09 15 4 ± 1 8 14 8 ± 1 4
data
.•
fotma• 227±37 235±16 382±39 380±32
Mechan1cal Fo,-mon 42 ±14 37 ± 14 104±32 95±25
data (N) F,H..:•m•• 248 0 ± 40 4
409±153
2779±192
371±139
4171±422
1139±351
• 4500±382
938±251
F,u.c-nwi
EMGr 14 1 ± 7 1 139±49 12 0 ± 6 0 11 7 ±5 D
Gmax 92±51 99±76 100±56 126±109
BF 76±53 9 7 ±77 104±98 106±87
aEMG (%) VL 11 3 ± 7 D 12 5 ± 8 2 9 1±55 98±47
RF 235±100 21 5 ± 11 6 19 5 ± 9 2 189±103
GAS 18 1 t 11 8 22 1 ± 21 1 12 2 ± 9 2 137±122
TA 142±123 9 1±6 7 99±93 6 3± 3 6
Table 1: Values (mean ± SO) of data measured during the sub-maximal condi-
tion in two conditions: circular chainring (CC) and Osymetric chainring (OC). *:
significantly difference between CC and OC (P < 0.05).
0 - 5 second 25 • 30 second 0 - II second Fat1~ue Index
Wrngate test
CC oc CC oc CC oc CC oc
Mechanical
p-~ 85±11 7 9 ± 15 52±07 47±08 67±07 61± 1 1 387±84 397±83
data k '
aEMGr 013±003 014±003 019±003 021±003 016±003 018±003 534 ± 185 50 0 ± 24 0
Gmax 014±004 016±003 021 ±DOS 023±003 017±004 018±002 621±415 521±325
aEMG BF 014±003 016±003 019±005 020±004 017±003 018±003 37 3 ± 19 6 284±198
(¾W')
RF 0 11 ±003 012±004 020±004 022±004 016±003 018±003 803±380 929±533
VL 013±004 0 14 ±003 020±004 022±003 D 16±003 018±002 660±388 666±371
GAS 014±003 0 16±003 016±003 018±003 016±003 018±003 36 8 ± 31 2 274±229
TA 013±003 013±003 016±003 018±005 016±003 017±003 378±231 32 3 ±23 3
Table 2: Values (mean ± SO) of data measured during the Wingate test with two
chainrings: circular chainring (CC) and Osymetric chainring (OC).
The effect of a non circular chainring on cycling performance 89
4 Discussion
Sub-maximal condition did not show significant difference on efficiency when using
OC versus CC. This result ~as in line with Rate! et al. (2004) who explained that the
physiological demand associated to the moving lower limbs at the top and bottom
dead centres and the production of higher force for a !arger part of the downstroke
was insignificant in relation to the total physiological demand. Theoretically, OC
would first permit to facilitate the foot path at the top and bottom dead centres. This
theoretical assumption has been verified in the present study by the analysis of the
effective force applied to the pedals. Indeed, Ferrcc-min (i.e. the effective force at top
and bottom dead centres) was significantly lower with the use of OC versus CC
(table 1). Moreover, with a smaller Fcrrec-min on OC versus CC, Fext-min (i.e. the exter-
nal force developed to the driving wheel) was identica\ between OC and CC. These
results confirmed a facilitated foot path during the top and bottom dead centres.
Since OC facilitated the foot path during the top and bottom dead centres, it would
bc expected to observe a lower muscular activity (level and / or hurst duration) per-
mitting the top and bottom dead centres foot path as RF and BF. In the present study,
analysis of RF and BF level and burst duration did not show significant difference
between the use of OC and CC. Two hypotheses could be proposed. First, this result
could be due to the bi-articular function of both muscles who participated also to the
pcdal downstroke for RF and to the pedal upstroke for BF; and secondly, this result
could be due to important SD, them even due to important inter-individual differ-
ences on muscular activity. The second theoretical assumption was the minimizing
effect of the radius increased during the downstroke since the effective force applied
to the pedal at 90° was important. Results of table I showed that Fext-max was identical
between OC and CC. This result was logical since power output and pedalling rate
were similar and fixed for both OC and CC. However, to produce the same Fext-max
with OC and CC, subjects should have produced more Feffcc-max with OC than CC.
This result was significant at 100 W and a tendency appeared at 200 W. Since OC
constrained subjects to produce more Fem,c-max, it would be Iogical to find more mus-
cular activity on muscles permitting the pedal downstroke as GAS, VL and Gmax· The
muscles analysis permitting the pedal downstroke did not show significant difference
bctween the use of OC and CC. Consequently, the cycling muscular pattem of sub-
jects had not been modified with OC. This result could be positive since there was an
increase of Feffec-max without increasc of the muscular activity responsible of the Fcffec-
max• Thus, sub-maximal results seemed shown that the use of OC was beneficial
during the top and bottom dead centres but without effect during the downstroke.
Identical result on GE could be due to this result since, like suggested by Rate! et al.
(2004), the physiological demand associated to the moving lower limb at the top and
bottom dead centres was insignificant in relation to the total physiological demand.
Nevertheless, an increase of power output would have allowed a significant differ-
ence on GE since the physiological demand associated to the moving lower limb at
the top and bottom dead centres increase with power output. Similar results on GE
could also come by a Jack of adaptation to OC. lt would be interesting to realise a
leaming period of OC use.
90 Nicolas Horvais, Pierre Samozino, Frcderiquc Hintzy
Conceming the Wingate test, there was no significant difference between the use of
OC and CC on mechanical and muscular parameters studied. An increase of Pmcan
could have appeared if subjects would have succeeded to increase their pedalling
rate. Indeed, the friction force being fixed, the only solution to increase power output
was to increase pedalling rate. Therefore, it would seem that subjects did not succeed
to increase their pedalling rate. Yet, according to sub-maximal condition results, the
OC use would lead to an increase of crank angular velocity during the top and bot-
tom dead centres foot path if the same effective force was applied to the pedal. lt
would mean that the crank angular velocity during the pedal downstroke was lower
with the use of OC than CC. Unfortunately, the method used in this study did not
permit to verify theses hypotheses since velocity was measured at the flywheel and
that it evolution was influenced by the inertia. However, the no significant difference
between muscular OC and CC data permitted to give an answer to the non improve-
ment of Pmcan during the Wingate test. lt would seem that the muscular force applied
on the pedal with OC was not sufficient to keep a crank angular velocity permitting
to produce higher power output. lt would seem therefore that the use of OC would be
beneficial for subjects having an important muscular capacity. lt was the result of the
study of Hintzy, Belli, Rouillon, Grappe (2000) where sprint specialist subjects had
improved their maximal power output with OC vs. CC while increasing their pedal-
ling rate for sprints of 8 s.
In conclusion, results of this study indicated that theoretical benefits brought by OC
did not significantly improve performance whatever the cycling condition for sub-
jects studied. This result could partly be due to a lack of adaptation to OC design.
References
Arsac, LM. Bclli, A. Lacour, J .R. ( 1996) Musclc function during bricf maximal exercisc:
accuratc measurements on a friction-loaded cycle crgometer. Eur. J. Appl. Physiol. 74,
100-106.
Cavanagh, P.R. Sanderson, J.S. (1986) The biomechanics of cycling: Studies ofthc pedaling
mcchanics of clitc pursuit riders. In: Burke, E.R. (Eds.), Science of cycling. Human Kinet-
ics, Champaign, Illinois, pp. 90-122.
Ericson, M.O. Nisell, R. ( 1988) Efficiency of pcdal forces during ergomctcr cycling. lnt. J.
Sports Med. 9, 118-122.
Hintzy, F. Bclli, A. Rouillon, J.D. Grappe F. (2000) Effects ofnon circular chainwccl on
force-velocity rclationship during sprinting on a cycle crgometer (in French). Sei. Mot. 40,
42-47.
Rate!, S. Duchl\ P. Hautier C.A. Williams C.A. Bedu, M. (2004) Physiological rcsponses
during cycling with "harrnonic" and circular chainrings. Eur. J. Appl. Physiol. 91, 100-
104.
Dynamic Characteristics of Modern Mountain Bikes
Rear Linkages
Angelo Tempia, Aleksandar Subic, Riccardo M. Pagliarella

angelo.tempia@rmit.edu.au
Abstract. Recent years have seen a tremendous development of full-suspended mountain

bikes, especially for downhill and free-ride market. These mountain bikes strongly rely on the
ability of the rear suspension shock absorber to manage extremely !arge forces and wheel
travel. Wheel travel is generally achieved through the geometry of the suspension linkage;
whereby forces are counterbalanced by the shock hydraulic and spring. The Sports Engineer-
ing Research Group (SERG) at the School of Aerospace, Mechanical and Manufacturing
Engineering, RMIT University has investigated the dynamic characteristics of different rear
linkages suspension geometry. The kinematic analysis has been performed using different
programs. A customised Matlab™ code has been developed to analyse in detail the kinematic
of the linkages, a MSC.Adams™ simulation has been conducted to better understand the
kinematic and kinetic behaviour. After-market shock absorbers are becoming more and more
popular but how their performance is influenced by the rear linkage geometries is often under-
estimated or not even considered. The results of this research aim to investigate and compare
the dynamic characteristics of rear linkage of modern full-suspended mountain bikes.
1 lntroduction
Mountain biking is a relatively new sport that can be seen as an evolution of cyclo-
cross. This sport came popular in the early 1980's in California and has continued to
grow in popularity to a point where it was introduced as an Olympic sport in 1996.
This exposure has driven bicycle manufacturers to develop an ever increasing num-
ber of model claming increasing performances.
The use of suspension on off-ride bicycles has proliferated in the past few years.
Although such a proliferation may be partially attributed to fashion (Sasaki, 2001 ),
bicycle suspension has technical merits and interests. Bicycles equipped with sus-
pension result in increased rider comfort, enhanced wheel contact and control, and
they also allow riders to manage the high jumps and the heavy landings that have
made this sport popular among youngsters and sportive from other disciplines.
In this paper a Downhill and a Cross Country rear suspension configurations of
two of the most popular mountain bikes in Australia are analysed in their kinematic
bchaviour.
92 Angelo Tempia et al.
2 Mountain bike Rear Suspension
The primary design consideration for suspension systems has to be inserted in the
context of the bicycles use. Cross Country mountain bikes need a suspension system
that responds to bumps but does not respond to rider induced forces, so that the input
of the rider is not wasted into modifying the geometry of the rear end of the bike. If
the suspension responds to the rider's forces, energy is used to compress the suspen-
sion instead of propel the rider. Free-Ride and Downhill mountain bikes, instead,
require a huge amount of travel in order to manage high drops and big jumps typical
of these disciplines, but the interaction between pedalling and the rear suspension it
is not of a prior importance.
The most important design parameters of a mountain bike rear suspension system
include the various torques and forces that are imposed on the suspension kinematic
by the rider and the terrain. These forces can be classified in 3 main categories. Pedal
induced forces compel all current rear suspension configurations to either compress
or extend the rear geometry. They act through drive chain and wheel drive load. A
design that causes the rear suspension to move under pedal induced forces, wastes
some of the rider input power into either heat energy (shock compression) or forcing
the mass of the rider to lift (negative sag) interacting with reaction to bumps.
Braking induced forces may cause a jacking force on the rear linkage. These
forces can cause the suspension to lose its effectiveness under heavy braking loads.
The easiest way to overcome this problem is to use floating brakes; however, this
solution increases both weight and cost
Terrain induced forces are the responsible in the first instance for the adoption of
suspension on a mountain bike. These forces are received through the contact be-
tween the ground and the rear wheel, and are transmitted to the shock absorber
through the suspension linkages.
Many typical rear suspension designs have the rear wheel follow an arc-like path
when a bump is encountered. This forces the wheel to be displaced in a forward and
upward direction. As a result of this, there occurs the undesirable situation of the
wheel compressing uphill when climbing instead of the wheel following the curva-
ture of the terrain. This increases the bump shock force transmitted to the sprung
mass of the bicycle because the wheel is not moving perpendicular to the bump.
The perfect kinetic solution would be a suspension that is able to respond only to
terrain induced forces and not to pedal and braking induced forces.
2.1 Centre of Curvature

Being a series of linkages bond together and forced to move around fixed points,
motion of suspensions can be analysed using the instantaneous centre of rotation.
This point identifies the point the linkages are moving around. This point, being
centre of rotation, can be seen as the point where if a force is applied it produces no
moment.
The suspension mechanism is needed in order to be able to create some sort of
suspension ratio, but since its centre of rotation moves with the movement of the
linkages it becomes difficult to understand the wheel path. The Centre of Curvature
Dynamic Charactcristics ofModem Mountain Bikes Rear Linkages 93
lncroulng
Otat1slng
~
0
LL
Disp acement
Fig. 1. Typical suspension motion ratio
instead is the parameter that can easily identify the wheel path. This point can be
calculated as analysing two consecutive different instants of motion.
3 Rear Suspension Geometry
The final aim for a rcar suspension varies from context to context; some times it is
required to increased comfort, other times it is required to increase the contact be-
tween the tyre and the ground, other times it is used to absorb a great amount of
energy. For these reason nowadays it is possible to have many different configura-
tion for a rear suspension design, but they all can be divided in major categories:
• Mono-shock and Cantilever Beam Designs
• Four Bar Linkage Designs
• Unified Rear Triangle (URT) Designs
• McPherson Strut (Macstrut) Designs
• Lawwill Linkage Designs
• Bottom Bracket Pivot Designs
• GT I-Drive
3.1 Suspension Motion Ratio

The motion ratio of a rear suspension is critical to proper suspension operation. Cur-
rent bicycle design incorporate motion rations in the rangc from rapidly rising to
rapidly falling.
Rapidly rising rates can cause the suspension to be too soft and active in the ini-
tial part of the wheel travel, mainly causing a pedalling energy waste; this can also
create prob lern with the travel of the suspension that might bccome no compliant.
A falling rate also involves negative aspects. A falling rate suspension is initially
stiff and gets softer as the suspension travel along; this might cause again the suspen-
sion to become non complaint.
All rear suspension geometries can achieve almost any suspension rates, Fig. l.
The rate of a bicycle suspension is composed of the intemal rate of the shock and the
rate inhcrent in the suspension geomctry, hence suspension rate is to be considered
94 Angelo Tempia et al.
Fig. 2. Specialized StumpJumper Expert Fig. 3. Kona Stab Deluxe
when design or choosing the appropriate shock absorbers, which generally have
different intemal rates. Pairing a falling rate frame with a linear coil shock or an
extremely rising rate frame with an air shock might not have acceptable results.
The accepted suspension component used on most production bicycles is either
coil spring and oiled damper combination or an air spring shock absorber. Coil
springs tend to have more linear rates, while air springs tend to have rising rates. All
frames may be fitted with a range of shocks, which these days generally have one of
two lengths and standard mounts.
The contribution to rate from suspension geometry is determined by the way in
which the shock mounts, front and rear wheel axles, and main triangle move relative
to one another. The front wheel axle establishes frame orientation to the ground but
generally may be neglected, since bottom brackets are almost universally at the same
distance from the ground without rider.
4 Analyses
The rear suspensions of a Specialized StumpJumper Expert and a Kona Stab Deluxe
have been analysed. The StumpJumper design features a four bar linkages with a
very short upper linkage; the Stab Deluxe instead has a Lawwill Linkage Designs
with the wheel attached to the lower linkage. An analysis of suspension ratios and
wheel paths was conducted using a Matlab™ code customised for each particular
suspension.
The four bar linkages of the Specialized showed an unexpected very lightly de-
creasing linear behaviour. This characteristic suits weil the original air shock ab-
sorber, but it is not the best solution for a coil over shock. The path of the rear the
wheel followed a trajectory very close to a circumference, due to the fact that the
centre of curvature did not move much during the whole motion of the suspension.
Tue wheel path is widely curved and it is running slightly up and back, as it is be-
lieved that this configuration is the best solution, Fig 4-6.
The Lawwill Linkage design of the Kona showed an again unexpected very
lightly decreasing linear behaviour. Even though it some reckon that a linear behav-
iour is the best way to cope with drops, the Kona suspension characteristic, together
with the original coil over shock absorber, results in a slightly decreasing suspension
rate, that is not ideal for Downhill mountain bikes. In this case the path of the rear
Dynamic Characteristics ofModern Mountain Bikes Rear Linkages 95
C:
.Q
.i
"O
N
Y directlon
. . . . .
Wheel travel
~ .
Fig. 4. Specialized StumpJumper Expert Centre Fig. 5. Specialized StumpJumper Expert suspen-
oflnstant Rotation (circle) and wheel path (<lots) sion ratio
E
,g•
8,
0
B
C •
1
,:j
N
Ooo~
i1
o•
. • ·"- ..
vvneel travel •
Y dlrectlon
Fig. 6. Specialized StumpJumper Expert Centre Fig. 7. Kona Stab Deluxe Centre oflnstant
of curvature distance from pivoting point Rotation (circle) and wheel path (<lots)
the wheel follows a circumference, due to the fact that the centre of curvature coin-
cides with one of the pivoting points of the linkages. The wheel path is widely
curved and it is running slightly up and back, as it is believed that this configuration
is the best solution. Fig 7-8
5 Conclusions
A bicycle suspension has many input forces, but in order to limit the analysis we
only considered the kinematic implication of the movement of the rear linkages. This
approach simulates sudden compression by the ground either through wheel contact
with an obstacle such as a rock or from the impact of a drop-off. In general, it is
believed that a widely curved rear axle path running slightly up and back is the best
solution (Sasaki 2001 ). In the case of an obstacle, the bump force will be up and
back relative to the frame, so the initial tangent should be up and back. The direction
of the force will turn more vertical as the bike clears objects of "ride-able" size, so a
96 Angela Tempia et al.
• .----.---"""T""-----.---.----r-----.
" • •
Wheel travel
• .. "'
Fig. 5. Kona Stab Deluxe suspension ratio
widely curving path tuming slightly upward should be ideal. This configuration also
allows the wheel base to increase and the wheel to follow the asperities. Tight curves
are generally inferior for shock absorption.
Having a widely curved rear axle path is sometimes difficult to achieve with a
linkage with a fixed pivoting point; 4-bars linkages represent a better solution that
often allows the tight curve deficiency to be mitigated to some degree by having the
path tangent tilting backward through all or most of travel
Even though a 4-bar linkage solution allows for more freedom in the design, it is
difficult, if not impossible to clearly identify what suspension solution represents the
optimum. Clearly for cross country competition a kinematic that minimises the nega-
tive drag and the energy lost during pedalling, would be the best choice; this can be
achieved with a configuration whose centre of instantaneous rotation is not moving
to much (as it can be seen for the Specialized suspension). In downhill, it would be
preferable a high travel suspension. When the ultimate design target is travel, it is
sometimes impossible to obtain a restricted space frame fro the centre of instantane-
ous rotation, as it can be seen for the Kona configuration. Rising rates benefit short
travel designs, since this will allow better initial compliance, while reducing the
probability of hard bottom-outs; a linear suspension ratio will offer the smoothest,
most consistent compliance in the event of a drop-off. Both mountain bike object of
the research showed good design geometry even though the Lawwill Linkage high-
lighted a slightly decreasing suspension ratio, leaving some doubts regarding the
optimal choice of the shock absorber that can be associated with the suspension. lt
would be interesting to have the possibility to modify the anchor point for the shock
absorber on the linkages in order to vary the suspension ratio.
References
Ola, H. (2000) MTB Suspension Tuning and Technology
http://www.math.chalmers.se/-olahe/Bike/index.html.
Herath, P. (2004) Design and Analysis of a Mountain Bike Rear Suspension System. Me-
chanical Engineering Thesis, RMIT.
Sasaki, K. (2001) A Bicycle Rear Suspension Analysis Method.
http://www.mtbcomprador.com/content/category/3/67 /l 05/.
An Ambient Intelligence System to Assist Team Training
and Competition in Cycling
Ingmar Fliege, Alexander Geraldy, Reinhard Gotzhein, Thomas Jaitner,

Thomas Kuhn, Christian Webel
Technical University Kaiserslautern, jaitner@rhrk.uni-kl.de
Abstract. Teamwork in cycling plays an important role, not only during competition to push a
cyclist, but also in training to maximize individual training effects. In this paper, we present a
fully operational prototype of the Assisted Bicycle Trainer, a distributed ambient intelligence
system to enhance outdoor group training of cyclists. The prototype is designed to run on
different hardware platforms and communication technologies, in particular, embedded PC
communicating via WLAN and Bluetooth, and light-weight micro controllers using ZigBee
for inter-bicycle communication. A focus of the paper is on the tailored communication solu-
tions and different broadcast schemes.
1 lntroduction
Ambient Intelligence (Ami) is a vision where we will be surrounded by unobtrusive

electronic devices, sensitive and responsive to people and objects, seamlessly em-
bedded into the environment, and interconnected through wireless ad-hoc networks ·
[ 1]. Ami systems will provide ubiquitous services that enhance human capabilities
and the quality of life. lt is expected that Ami systems will have an impact on all
spheres of our life, including professional work, leisure activities, public health,
transportation, communication, and sports.
In this contribution, we present a specialized Ami system to support outdoor bi-
cycle group training. Even though cycling is primarily known as an individual sport,
teams play an important role in training and competition. In particular, the team time
trial is an outstanding event in a major competition like the Tour de France. Even
more, road cycling in groups is common in training. In typical team training, a group
of cyclists covers a distance of up to 200 km, with a varying road profile. For best
training effects, each cyclist should ride with an individual exercise intensity that
depends on various conditions such as physical capabilities and skills of the cyclist,
speed, road incline, head wind and temperature.
In Section 2, we describe the Assisted Bicycle Trainer (ABn, a distributed Ami
system that enhances the outdoor training of a group of cyclists, and survey the im-
plementation of our fully operational prototype. Tailored communication solutions of
the ABT are addressed in Section 3. We conclude with abrief outlook in Section 4.
98 I. Fliege, A. Geraldy, R. Gotzhein, T. Jaitner, T. Kuhn, C. Webei
2 The Assisted Bicycle Trainer
The objective ofthe Assisted Bicycle Trainer (ABT) is to improve the training effects
such that each cyclist is as close to his individual exercise intensity as possible. To
achieve this objective, the ABT dynamically collects status data of each cyclist, and
displays a summary of these data to the human trainer accompanying the group of
cyclists by car. Based on this information, the trainer may adjust training parameters,
for instance, by ordering the group to change speed, or by ordering a particular cy-
clist to take thc lead, exposing him to the headwind, while all others can exploit the
slipstream and thus need less pcdal power. Orders of the trainer are shown on small
displays attached to each bicycle. The ABT is a self-organizing system, supporting,
in particular, dynamic group formation and mobility. Communication among cyclists
and human trainer is performed via a wireless ad-hoc network.
Currently, the Assisted Bicycle Trainersupports two outdoor training modes: in-
dividual training, where every cyclist can monitor the recorded sensor data during
the training session, and group training, where the training of every cyclist is con-
trolled by software and directed by the human trainer. In a typical training session, a
group of cyclists covers a !arge distance. While the speed ofthe whole group must be
the same to avoid partitioning, the position ofthe individual cyclists within the group
can be used to control the power that every cyclist requires to hold the group speed.
The Assisted Bicycle Trainer measures and collects various data of each cyclist and
uses this information to calculate the group speed and the position of every cyclist
within the group to enhance the outcome of the training. Depending on the availabil-
ity of sensors, a variety of data may be used for controlling the training. Examples
are the road profile, the slope, the wind speed and the power output of each cyclist.
Our Assisted Bicycle Trainer consists of three main components:
• ABT application. The distributed ABT application has been designed for
supporting outdoor bicycle training with a varying number of cyclists. lt
consists of a cyclist application running on every bicycle system, and of a
trainer application that runs on the traincr laptop. The cyclist application in-
teracts with the cyclist by a combination of a graphical user interface (GUI)
and voice output.
• Communication middleware. The communication middleware controls inter-
bicyclc communication for the acquisition of sensor data and GUI manage-
ment as weil as the intra-bicycle communication. The middleware is divided
into an application specific and into a generic part. The generic part handles
communication tasks that appear in evcry distributed application, such as
media access and multi-hop routing. The application-specific middleware
provides services that have been tailored for the ABT application. These scr-
vices include sensor data acquisition and distribution as weil as determining
the current number of cyclists and detection of field partitioning. The com-
munication middleware is running on all nodes of the ABT system.
• Hardware platform. Currently, there exist two hardware platforms for the as-
sisted bicycle trainer. The first platform is an embedded-PC solution using
WLAN for inter-bicycle communication and Bluetooth for intra-bicycle
communication. The sccond platform is a small, ultra low power solution
An Ambient Intelligence System to Assist Team Training and Competition in Cycling 99
that is based on MicaZ motes [3], manufactured by Crossbow Technologies.

The MicaZ motes mainly consist of a micro controller, and communicate us-
ing ZigBee, a low power wireless technology.
Fig. 2 Crossbow MicaZ mote

(OEM edition)
Fig. 1 The Wireless LAN solution for the

Assisted Bicycle Trainer
Figure l shows the embedded-PC solution, installed on a bicycle. The embedded

PC, WLAN stick, Bluetooth adapter, a pulse rate receiver, and batteries are mounted
on the carrier. A PDA showing the current driver status (e.g., pulse rate, actual
speed) and the trainer's orders (e.g., required speed, position changes) is attached to
the handle bar. The cyclist carries a transmitter belt that detects the pulse rate and
sends it to the pulse rate receiver. The trainer system (not shown in Figure 1) is in-
stalled on a laptop, with a sophisticated graphical interface to monitor and direct the
training. Intra-bicycle communication between PDA and embedded PC is by Blue-
tooth, wireless LAN is used for inter-bicycle data exchange.
Figure 2 shows a MicaZ mote, which has the size of a stamp, and which replaces
the embedded PC and the wireless LAN communication of the embedded-PC plat-
form. In comparison to the embedded-PC platform, the micro controller solution is
preferable due to its low weight and low power consumption.
To validate the technical functioning of the embedded-PC solution, we have run
several outdoor training sessions with one cyclist and a trainer accompanying the
cyclist in a car. Sensor values were communicated via WLAN up to a distance of
l20m. On the trainer laptop, the heart rate values of the cyclist were recorded. A
sample taken during a short training run is shown in Fig. 3.
100 I. Fliege, A. Geraldy, R. Gotzhein, T. Jaitner, T. Kuhn, C. Webel
Training aesslon
"" ~--------~
""'
...
Time (H<Ond•I
....
Fig. 3 Training session Fig. 4 Partitioning of a group of cyclists
3 Communication Middleware
The main task ofthe communication middleware is the provision ofinter-bicycle and
intra-bicycle communication. W e have developed a middleware that serves as a
hardware abstraction layer to the application, and that can be adapted to the specific
needs of different configurations. When developing distributed systems based on
wireless ad-hoc networks, routing across multiple hops is of particular importance.
The ABT requires a broadcast service - every message should be received by every
node including the trainer laptop. Here, we can distinguish two different types of
broadcasts:
• Local broadcasts are transmitted by a node and received by every node that
is within range of the transmitter. In this case, no routing protocol is re-
quired. Local broadcast can be used in scenarios where the range of the
transmission technology is high enough to reach all other nodes.
• Global broadcasts are transmitted through the entire network, regardless of
the transmitter's range, as long as the network is not partitioned. If some
nodes are not within range ofthe transmitter, this requires multi-hop routing.
Depending on the implementation of the routing protocol, the reliability and
the overhead ofthe global broadcast may vary.
For inter-bicycle communication of the ABT, we have tailored two communica-
tion systems, supporting local broadcast and global broadcast, respectively. Global
broadcast is realized by NXP/MPR [5], which uses a selective flooding strategy to
save bandwidth. Both protocols have been specified with SDL [2], a formal, stan-
dardized Ianguage for the design of distributed systems and protocols.
An Ambient Intelligence System to Assist Team Training and Competition in Cycling 101
To compare protocol performance, we have performed extensive simulations, us-

ing ns+SDL [4], our network simulator for SDL systems. One simulation scenario
consisted of a field of 20 cyclists and a trainer. At simulation times 150 and 550, a
group of cyclists separates from the field, resulting in the partitioning of the net-
works when using local broadcasts. Figure 4 gives an idea of this situation. Here, the
remaining field consists of cyclists 1, 2, and 3, with cyclist 4 of the separated group
still being within their communication range. Thus, when using local broadcast, the
other nodes can not communicate with the trainer, who stays close to the main field.
This situation is improved by NXP/MPR, where messages are forwarded across
multiple hops.
25-----------------
,. '
" 15 *:+---~---;---L-----+------"-------!
'g '
.,
C:
:;;
'iii
.
·;;:
Ö 10 :
5 :
- global broadcast
•· ·-· ·• local broadcast
200 400 600 800 1000

Time(s)
Fig. 5 Comparison of Iocal and global broadcast performance
Simulation results for the scenario are shown in Figure 5. With local broadcast,
network connectivity goes down when a group of cyclists separates from the field
until the field is reunited. With global broadcast based on NXP/MPR, there is füll
network connectivity most of the time. However, there are short periods of time
during which some nodes can not be reached, which is due to node mobility, frame
collisions, and signal interference.
Whether local or global broadcasts are used should be transparent for the ABT
application. In the case of a communication system that supports •local broadcasts
only, the ABT application is notified about the cyclists currently within reach. This
would also be the case for a communication system using global broadcasts when
two groups of cyclists are too far away from each other for the communication sys-
tem to establish a connection between these two groups.
102 1. Fliege, A. Geraldy, R. Gotzhein, T. Jaitner, T. Kuhn, C. Webei
4 Conclusion
In this paper, we have presented a fully operational prototype of the Assisted Bicycle
Trainer (ABT), a distributed ambient intelligence system to enhance outdoor group
training of cyclists, and have demonstrated its technical functioning. In particular, we
have described the tailored communication solution ofthe ABT, supporting different
types of broadcasts, and the advantages and drawbacks of these broadcast schemes.
Future work includes the incorporation of additional sensors into the ABT, the con-
ception and implementation of sophisticated control algorithms for optimizing train-
ing effects, and the application of the ABT during regular group training sessions.
References
E. Aarts, R. Harwig, M. Schuunnans. Ambient intelligence. The invisible future: the seamless
integration of technology into everyday life, Mc-Graw Hili, 2002.
International Telecommunications Union. Specijication and Description Language (SDL).
ITU-T Recommendation Z.I00, August 2002.
Crossbow. Micaz wireless measurement system. http://www.xbow.com/Products/
Product_pdf_ filcs/Wireless _pdf/MICAz _ Datasheet.pdf.
T. Kuhn, A. Geraldy, R. Gotzhein, and F. Rothländer. ns+SDL - The Network Simulator for
SDL Systems. In A. Prinz, R. Reed, and J. Reed, editors, SOL 2005, Lecture Notes in
Computer Science (LNCS) 3530, pages 103-116. Springer, 2005.
I. Fliege, A. Geraldy: NXPIMPR - An Optimized Ad-Hoc Flooding Algorithm, Technical
Report 343/05, Computer Science Department, University of Kaiserslautern, Gennany,
2005
Indoor-Simulation of Team Training in Cycling
Thomas Jaitner, Marcus Trapp, Dirk Niebuhr, Jan Koch
TU Kaiserslautern, jaitner@rhrk.uni-kl.de
Abstract. For the single cyclist performance parameters such as power, specd, or heart rate
can be monitored during training and competition. Although cycling is primarily a single
sport, riding in groups is a very common in training. An ambient intelligence system has been
developed for the training of a group of cyclists. The objective of this system is to improve
tcam training such that each cyclist is as close to his individual exercise intensity as possible.
Besides physiological and biomechanical data, subjective sensations are also considered. The
focus of this papcr is on feedback training. Based on the comparison of dynamically collected
status data and set values, the feedback training system adjusts training parameters, for in-
stance by advising thc group to change the order or the formation, to increase or decrease the
speed, or to split the group. In a final vcrsion, the system should run undcr outdoor conditions.
As an intcrmediate stcp, a prototype for indoor training was established.
1 Introduction
Cycling in groups is a very common method in training especially for long lasting
training sessions (Gregor and Conconi, 2000). For best training effects, each cyclist
of a group should ride with his prcdetennined exercise intensity that at least will
slightly differ between athletes due to individual physical capabilities and skills,.
While the speed for all cyclists must be the same, the power output depends on the
position within the group. Because of the head wind the power output of the leading
cyclist is up to 36% higher than the power output of subsequent cyclists (Neumann,
2000). In consequence, cardiocirculatory and metabolic effort of subsequent cyclists
will be lower. To improve team training the cyclists might regularly change posi-
tions, adjust the speed of the whole group or arrange their positions according to
individual differences in exercise intensities.
Besides physiological and biomechanical parameters the subjective sensations are
considered as reliable and highly relevant indicator to detennine the appropriate
exercise intensity (Gregor and Conconi, 2000). However, subject's sensations are
nonnally not monitored by technological measurement systems in cycling. A promis-
ing approach for the evaluation of physical exertion during training is offered by the
RPE scale (Borg, 1998)
An ambient intelligence system has been developed for the training of a group of
cyclists (Litz et al., 2004 ). The objective of this system is ( 1) to improve training for
104 Thomas Jaitner, Marcus Trapp, Dirk Niebuhr, Jan Koch
the single athlete considering physiological and biomechanical data as weil as sub-
jective sensations and (2) to improve team training such that each cyclist is as close
to his individual predetermined exercise intensity as possible. The system consists of
three major parts: data acquisition, communication between multiple bicycles, moni-
toring training data and feedback training.
The focus of this paper is on feedback training. In a final version, the system should
run under outdoor conditions. As an intermediate step, a prototype for indoor train-
ing was established that consists of 4 bicycles and 4 ergometers. The ergometer data
are processed by a simulator, which calculates the power each cyclist must generate
for a given speed considering head wind, lee and road profile. Simulated power,
speed, cadence as weil as the heart rate are transmitted to each cyclist.
2 Prototype of the Bicycle Trainer
2.1 Hardware Setting and Software Architecture

Four bicycles were mounted on ergometers (Tacx Tl680 FLOW) and equipped each
with a Sony Vaio U-71 micro laptop with a touch screen as user interface. All micro
laptops are connected among each other using Wi-Fi technology forming an ad-hoc
network. The control software running on this micro laptop uses the current speed,
the current cadence, the current pedal power, and the current heart rate of each cy-
clist as sensory input data. This can be seen in Fig. 1 exemplarily for only two bicy-
cles. The sensors are connected wireless or altematively by wire to a proprietary
sensor board which delivers the sensor values to our control software using the
RS232 serial interface (Wahl, 2004). A standard pulse belt was used as pulse sensor.
Processing of real sensor values as weil as conversion of control variables of the
brake current as result of the force information from the simulation is done on an-
other control board. They are connected via CAN-Bus. This separation between two
different boards is done since the system can be easily used outdoor just by removing
the simulator board circuits, the simulator and the ergometer.
The hardware setting (Fig. I) is supported by a software system that allows vari-
ous flexible hardware configurations without changing or replacing the software
system. Therefore, a Service Oriented Architecture was chosen (Bartelt et al., 2005)
where services are identified only by their software interfaces enabling the easy use
of different hardware. By service we refer to the software representation of devices
(e.g. sensors) as weil as to software units (e.g. single training control functionality)
which are executed on a device with computation capacity. This is necessary since it
cannot be assumed, for instance, that all cyclists use the same sensors or the same set
of sensors. Additionally, by using this software architecture, the number of cyclist
that can addend group training is not bounded above or below. The architecture is
based on a Configuration Service, where each service registers together with differ-
ent configuration sets of needed services, which allows more flexibility at runtime.
Thus, it is possible to select another configuration at runtime- e.g. introduce a new
sensor, switch from single training to group training, and so on - without changing
the software or even restarting the system. For example, the single training control
service registers at the configuration service stating that it needs a pulse sensor ser-
Indoor-Sirrmlation ofTeam Training in Cycling 105
Simulator
Fig. 1: Hardware architecture of the indoor prototype
vice, a power sensor service and many services more in order to be run. When it
notifies the Configuration service, that it wants to be run, the best set of services is
determined by the Configuration Service, based on quality of service descriptions
and contextual information. After determining the best required services, they are
handed over to the single training control service which can then start working.
2.3 Simulation of head wind and road profile

The power generated by the cyclist can only be gathered indirectly with knowledge
of speed and the imposed force or moment of crank torque, respectively. The power
is derived from the torque and the speed while the torque is calculated by the force
and the wheel radius. The moment of torque which is the force normalized to the
radius is assumed to be approximately equal at crank and rear wheel. Hence the
power can be determined by the adjustable brake ofthe ergometer.
Fora more realistic training in the indoor environment, the properties of a dynamic
world were simulated by a simplified model (Palm, 2005). The brake forces of each
ergometer were adjusted considering dynamic friction, wind resistance, and slope
dependent lift forces:
(2)
The single components ofthe brake force therefore are determined as follows (Eq.3):
106 Thomas Jaitner, Marcus Trapp, Dirk Nicbuhr, Jan Koch
(3)
Fw=½·p·cw· A(v+vw )2
with mass of cyclist and bicycle m, acceleration of gravity g, slope a at current posi-
tion, coefficient of friction cR, density of air p, air drag coefficient cw, front surface A,
wind velocity in counter direction v •. Air drag coefficient as weil as front surface
were approximated according to Gressmann (2003).
The exploitation of the slipstream by cyclists on subsequent positions is consid-
ered by a linear reduction of the wind resistance depending on the velocity and the
distance to previous cyclists. This approximation is based on calculations by Gress-
mann (2003) and Neumann (2000).
3. Feedback training control
3.1 Single training

An individual training plan serves as initial input for the feedback control of the
single cyclist. The exercise intensity is described by target power,· corresponding
heart rate and cadence for given time intervals. The primary task of the system is to
control the cyclist's heart rate by adjusting the target power. When a cyclist's heart
rate is above the upper bound of the tolerance range for a certain time the preset
power will decrease. This procedure will be repeated until the cyclists heart rate will
remain in the range of tolerance. Additionally, a limit for the decrease of power has
been defined, which cannot be underrun by the feedback control. The system reacts
in the opposite way every time the cyclist stays 30 seconds below the lower bound of
the heart rate corridor. All control parameters can be adjusted individually. The ini-
tial settings are a range of ±3 beats for the heart rate, a time period of 30 s and a
decrease rate of 10%.
3.2 Influence of Subjective Sensations

During the whole training the system asks the cyclist regularly to enter the actual
rating of perceived exertion as feedback about the current subjective sensations using
the RPE scale (Borg, 1998). On a scale from 6-20, values between 10 and 14 are
presumed as adequate exercise intensity, whereas lower or higher values are consid-
ered as non-adequate exercise intensity. The subjective sensations influence the
feedback control in the way that the tolerance range ofthe heart rate will be modified
by low as weil as high rating.
3.3 Group training

The group training control is based on the single training control as described above.
Two or more cyclist can form a training group. During the training the group control
Indoor-Simulation ofTeam Training in Cycling 107
maximizes the training effect of every single cyclist while keeping the group to-
gether. Therefore, the optimal group speed is calculated by minimizing the sum of
differences for all cyclists between the target values of their initial training plans and
the new target values of the group. Moreover, a formation is calculated which deter-
mines the position of each cyclist. If the cyclist's sensor values are not within the
tolerance corridor, the following means are taken for optimization in the given order:
l. Changing positions within the group
2. Changing the formation of group
3.Adjusting speed
4. Splitting the group
All data gathered during the single or group training is stored persistently and can be
used for evaluations afterwards.
4 Evaluation of the indoor simulation prototype
So far, several training sessions were run successfully. Fig. 2 shows exemplarily the
heart rates of two cyclists during a simulated team training. The training was started
with preset leading intervals of 90 seconds and individually determined heart rate
boundaries. Grey bars indicate the Ieading intervals of cyclist l (solid line). Leading
intervals of cyclist 2 (dotted lines) are marked in white.
180
170
IIO
~
C
,: 150
•• .: \ : .
:'.,
,:
1<0
." ·•.
.
130
120
00
,.,,
Fig. 2: Heart rates of two cyclists during a simulated team training

In the third leading interval of cyclist 2, the heart rate exceeds the tolerance range.
Therefore, the cyclist are instructed to change position, before the preset time target
is reached. Due to the feedback control mechanism, all subsequent le~ding intervals
of this cyclist are reduces. In Fig. 1, this can be observed by varying amplitudes of
the white bars. Cyclist l is able to maintain the preset time targets up to the third last
108 Thomas Jaitner, Marcus Trapp, Dirk Niebuhr, Jan Koch
leading interval. A change of positions is initiated after 60 seconds due to a trans-

gression of the individual heart rate boundary. In this case, the intervention of the
feedback control mechanism was accompanied by a higher rating of the subjective
sensation, which results in a decrease ofthe upper heart rate boundary.
5 Conclusion
According to the preliminary results, the indoor simulator as weil as the implemented
feedback control system seem to be an effective aid for cycling training. The estab-
lished communication structure is reliable under indoor conditions. By a set of train-
ing experiments tolerance ranges for cyclists of different levels of performance are
determined to optimize the feedback control algorithms. While the system should run
under outdoor conditions, parallel and future work will focus on the integration of
different sensors (e.g. force sensors at the crank, GPS) as weil as on the optimization
of the communication structure.
Acknowledgements
This work was supported by the Research Center Ambient Intelligence of the Uni-
versity of Kaiserslautern.
References
Bartelt, C., Fischer, T., Niebuhr, D., Rausch, A., Seid!, F., Trapp, M. (2005) Dynamic Integra-
tion ofHeterogeneous Mobile Devices. In: Proceedings ofthe Workshop in Design and
Evolution of Autonomie Application Software (DEAS), ICSE 2005, St. Louis
Borg, G. (1998) Borg's Perceived Exertion and Pain Scales. Human Kinetics, Champaign
Gregor, R.J., Conconi, F. (2000) Raad Cycling. Blackwell Science,Oxford
Gressmann, M. (2003) Fahrradphysik und Biomechanik. Delius Klasing, Bielefeld
Litz, L.. Wehn, N. and Schuermann, 8. (2004) Research Center "Ambient Intelligence" at the
University of Kaiserslautern. VDE Kongress 2004, 1, 19-24, VDE, Berlin
Neumann, G. (2000) Physiologische Grundlagen des Radsports. Dt. Zeitschrift f.
Sportmedizin, 5, 169-175
Palm, S. (2005) Realisierung eines Simulators für den Fahrraddemonstrator BicMon.
Unpublished work, University of Kaiserslautern.
Wahl, T. (2004) Dokumentation der programmierten Anwendung für das Sensor-Board des
BicMon-Demonstrators. Unpublished work, University of Kaiserslautern.
A Bond Graph Model of a Full-Suspension Mountain
Bicycle Rear Shock
Robin Redfield 1 and Cory Sutela2
1 Department ofEngineering Mechanics, United States Air Force Academy, Colorado

Springs, CO, USA, rob.redfield@usafa.af.mil
SRAM Corporation, Colorado Springs, CO, USA
Abstract. As the spart of mountain biking matures, equipment continually evolves to afford
better biking performance, enjoyment, and safety. In the arena of suspension systems, moun-
tain bikes have moved from rigid suspensions with large, knobby tires to front fork suspen-
sions, and finally füll suspensions. Suspensions have gone from elastomeric compliance to air
and coil springs with adjustable travel. Damping has progressed from fixed to adjustable
rebound, compression, and lockout. The current trend is to add force or frequency depcndent
damping to minimizc response of a suspension from pedal input. A bond graph model of a
mountain bike rear shock is developed incorporating adjustable rcbound/low-speed compres-
sion, high-specd compression, and adjustable, compression damping initiation. An air shock
with a nitrogen chargc is modeled with velocity across the shock as input. The dynamic equa-
tions that come from a bond graph model are simulated to predict key responscs. Experimen-
tal response of the modeled shock is acquired subject to periodic velocity inputs. The experi-
mental responsc is used to tune the design paramcters of the modcl and for validation. Future
use of the model is to bettcr undcrstand thc physics and pcrformance of the mountain bike
shock and to relate performance to the rcquircments of expert mountain bikers.
1 Introduction
The popularity of füll suspension mountain bicycles continues to increase as new
designs allow users to ride increasing technical terrain with more control. Today's
all-purpose trail riding frames are designed to perform reliably in harsh downhill
conditions without adding unnecessary weight that would compromise climbing
performance. Suspension elements (forks and rear shocks) must provide isolation
from !arge impacts, attenuation of high frequency bumps, and dissipation of kinetic
energy, while adding the least weight possible. A relatively new development in
mountain bike forks and rear shocks is a regressive damping configuration: the so
called "stable platform" or compression initiation control (cc) system. The cc system
provides increased efficiency during climbing and hard pedaling by reducing or
preventing suspension movement. In the ideal embodiment of this system, the sus-
pension system becomes fully active immediately upon encountering an impact force
greater than some threshold size.
Analyzing the behavior of cc damping systems is challenging because there is a
wide range of conditions relating to force, position, and velocity to which the shock
is subjected. The most popular approach to assessing rear shock performance is the
110 Robin Redfield and Cory Sutela
test ride - a qualitative technique which necessarily contains a degree of subjectivity.

Shock dynamometer curves (plots of force with displacement and force with veloc-
ity) are also used, but these are complex and difficult to interpret.
Previous research relating to mountain bicycle suspension performance includes
both experimental and theoretical analyses of complete bicycles (Sutela (2004) refer-
ences a selection of research examples). Redfield (2005a and b) used a bond
graph/conservation of energy approach to model a front only and füll suspension
bicycle coupled to a rider to predict system performance for a bike/rider exposed to
changes in terrain. Until now, a detailed model ofmountain bike rear shock behavior
has not been described.
The goals of this research are to reduce the design cycle time for rear shocks by:
Creating a mathematical model of a modern bicycle rear shock, including considera-

tion of a cc system. The model must reflect the physical behavior of the system so
that once it is validated it can be used to guide the performance optimization that is
possible by adjusting geometry and operating pressures.
Measuring the physical performance of the modeled shock. The experimental results
can be used to validate the model.
Adjusting model parameters to identify which physical parameters inside the shock
most strongly influence its performance.
2 Methods
2.1 Rear Shock Model

Mountainbike rear shocks control relative motion between the sprung mass (main
frame, mj) and the unsprung rear triangle (rt). Fig. 1 is a schematic of a shock as-
sembly that incorporates a damper body that moves over a piston containing oil flow
ports. These ports are essentially fitted with 2 check valves (one user tuned) and a
user adjustable orifice. One check valve is factory set for high speed compression
relief; the other is set for compression initiation control, the shock force necessary to
initiate compression. The orifice controls rebound and low-speed compression damp-
ing. The piston is supported by the air pressure, PA, and is contained also by the seal
head and shock body.
The shock also contains a sealed volume of pressurized nitrogen, N2 , which acts
on the floating piston (fp) volume to maintain positive pressure within the damping
oil chambers at all times, preventing oil vaporization during shock movement.
Yr 1 and Vmf represent the axial components of the main frame and rear triangle
velocities. The damper body is divided into 2 chambers (1 and 2) which are both
completely füll of oil. The size of chambers 1 and 2 depends on the position of the
piston within the damper body. The piston itself is hollow and contains a re-
bound/low-speed compression port and high speed compression relief on one side,
and a cc valve on the other. The ports are modeled representing a flow proportional
to the square-root ofthe oil pressure drop, while the high speed reliefvalve limits the
Mountain Bike Rear Shock Modeling, Testing, and Evaluation 111
maximum pressure that can be generated in chamber 2 during the compress10n

stroke.
Seal head Pi ton Damper body
Air
Oil
Vn (t)
Shock body 1p Floa mg piston
Fig. 1. Rear shock schematic
During compression, PA increases as VA is reduced, oil flows into the piston

through the cc valve (Qcc flows right to left). Qrc also flows left as oil moves out of
the piston through the parallel flow paths of the low speed compression port and the
high speed relief valve (if the maximum pressure is reached in the piston chamber).
Vi increases as V2 and VN2 decrease. The pressurized nitrogen chamber is designed
to prevent a vacuum from developing in Chamber 2 (and 1), but the model does
allow a vacuum to develop in chamber 1 under high compression speeds.
During the rebound stroke, the motion of the damper body over the piston is
reversed, and the speed is controlled by the oil flow rate through the rebound port
(Qrc now flows to the right which is defined as positive). The rebound force is pro-
portional to the square of Qrc•
Fig. 2a shows the calculated rebound, and low and high speed compression flow
rates (Qrc) through the piston resulting from various values of P 1 - Pp. Rebound
flows are positive, while low & high speed compression flows are negative. Fig 2b
shows the calculated value of Qcc for various pressure differences. This is flow
through the compression initiation control valve. In this case a PP - P2 of -4 atm. is
required before the cc valves opens, allowing significant relative motion between the
damper body and the piston.
The important dynamic effects of the shock are captured by a bond graph model
(Kamopp et. al., 2006), depicted in Fig. 3. The primary model input (flow source) is
the relative velocity of the ends of the shock. C elements represent energy storage
due to fluid compression, and R elements describe dissipation losses due to fluid
flow through the various orifices. The energetic elements of the model are described
clockwise from the flow source (Sr), Energy is: stored in the air chamber, stored as
pressurized oil in chamber 1, dissipated due to compression and rebound oil flow,
dissipated through the cc valve, and stored in the nitrogen chamber. cc spring repre-
sents the stiffness (kcc) of the cc valve spring. A "top-out" spring cushions füll re-
bound but is not engaged in the testing protocol of this paper. The rest of the struc-
ture of this bond graph captures the kinetics, and kinematics of the shock.
1 12 Robin Redfield and Cory Sutela
w 100 Cl) 100

. . . ., 50 ........
eu o
-so
M
E
u
50
Oi-----...--------+-----1
.. -100
-so
~ -150
u -100
O -200 .,,,. &-150
-10 -5 0 5 10 -6 -4 -2 0 2
P1 Pp, a m Pp-P 2 , am
Fig. 2. Orc and 0cc versus pressure drop
Fig. 3. Rear shock bond graph
A coupled set of nonlinear differential equations was derived from the bond
graph in accordance with Kamopp et al. (2006). These were simultaneously solved
using the commercial computational package Mathematica to calculate the force
experienced at thc ends of the shock as a result of a prescribed input velocity profile.
2.2 Experimental Setup

A shock of the type described in the model was subjected to a prescribed input veloc-
ity sine wave of the mf eyelet (rt eyelet fixed), using an MTS 24.11 dual servo-valve
hydraulic load. The actuator contains an integral LVDT and is fitted with a cali-
brated Interface 1210 ACK-5K-B load cell (measuring the mf eyelet force). The
shock was subjected to displacement sine waves of varying frequency, with peak-to-
peak stroke length of 25mm. The MTS damper test software (V3.5.3) produces
Mountain Bike Rear Shock Modeling, Testing, and Evaluation 113
Force vs. Displacement and Force vs. Velocity curves. Results are given for lHz
and 5Hz sine waves, and in each case the graph represents the 2nd of 3 sequential sine
waves at a given frequency.
3 Results
Figure 4 shows the eyelet force predicted by the shock model under 3 distinct
damper settings in addition to one cycle of test data, all in response to a typical 1Hz
sinusoidal displacement input. Geometrie data and initial fluid pressures are taken
from the tested rear shock.
- 75 - 50 - 25 0 25 50 75
Velocity, mm / s
Fig. 4. Applied shock force versus relative velocity
The shape of the output curve can be manipulated by changing model parameters
which relate to physical design parameters within the shock: The peak to peak range
of forces at zero velocity was adjusted by modifying the initial pressure and volume
in the N 2 chamber; the aspect and skew of the graph were adjusted by changing the
product of damping orifice discharge coefficient and area (i.e. CDA for the orifices in
the piston rc valve), and the curve offset near zero velocity was adjusted by modify-
ing the initiation pressure for cc valve actuation.
In Fig. 4 the test data is the dotted line and the tuned model prediction is the
lighter of the thick curves. TDC and BDC are the top and bottom of the compression
stroke. The compression portion of the prediction curve fits the data better than re-
bound. Tested rebound magnitudes are lower than predicted during the first half of
the rebound stroke. The thin, solid curve demonstrates halving the compression ini-
tiation pressure. A lower initiation pressure allows compression flow sooner and
thus the compression side of the curve is offset upward and less than the tuned curve.
Rebound is not affected by this change. The thick and darkest curve is for a 20%
decrease in rebound orifice CctA value. Rebound forces are higher for all rebound but
more so as velocity increases.
After tuning the model at lHz, it was applied to the 5Hz case as shown in Fig. 5.
The dotted curve is still the test data. There is good agreement between the test data
114 Robin Redfield and Cory Sutela
and the model in this case with the most notable discrepancy occurring at the com-
pression initiation transitions near zero shock velocity.
z........
~ 0J,-------+----+---7"-i---jC_-__.
u
g - 10001------6---==--+--"7"!;_+---
~
---4
- 2000~~:±::::::=i_ _ L_ _j
- 400 - 200 0 200 400
Velocity, mm / s
Fig. 5. 5 Hz test data and model prediction
5 Conclusions
A model, based on fluid pressure and flow through orifices in a production mountain
bicycle shock, was created using the bond graph technique. The model was cali-
brated at an operating frequency of lHz by adjusting model parameters that relate to
design choices that could be physically implemented in the shock. The calibrated
model was applied to a 5Hz operating frequency, demonstrating good agreement
with measured data. Deviations from the test data might be related to friction inside
the shock body, inertia of the fluid, or the thermodynamic conditions during com-
pression ofthe N 2 chamber, although these were not investigated in this work.
The model can be applied by design engineers to identify the physical parameters
which most significantly affect the measured performance of the shock, and how
these are manifested in the damper curves. This will ultimately result in reduced
design cycle time as new shocks are developed.
References
Sutela, C. (2004), Measurement of suspension efficiency in mountain bicycles during hill

climbing, The Engineering of Sport 5 - Procedings of the International Sports Engineer-
ing Association (JSEA), Vol. 1, pp. 487-493.
Kamopp, D., Margolis, D., and Rosenberg, R. (2006), System Dynamics; Modeling and Simu-
lation of Mechatronic Systems. 4th ed., Wiley InterScience, New York.
Redfield Robin (2005), "Large Motion mountain biking dynamics," Vehicle System Dynamics,
Vol. 43, No. 12, pp. 845-865.
Redfield, Robin C. (2005), "Planar, Large Excursion Bond Graph Model for Full Suspension
Mountain Biking," Proceedings of the ASME Dynamic Systems and Contra! Division-
2005, ASME International Mechanical Engineering Congress and Exposition, 2005.
Track Cycling: An Analytical Model
Richard Lukes 1, Matt Carre 1 and Stephen Haake 2
1 Sports Engineering Research Group, University of Sheffield, r.lukes@sheffield.ac.uk

2 Sports Engineering, CSES, Sheffield Hallam University
Abstract. This paper presents an analytical model of track cycling the purpose of which is to
provide a tool that allows subtle changes to be made to the bike, rider or environment and a
corresponding change in performance realised. The model has been derived specifically for
track cycling, and considers the implications of riding in a velodrome. Various inputs are
required by the model, such as; rider power, atmospheric conditions, tyre properties, velo-
drome geometry, aerodynamic properties and bike and rider characteristics. A fundamental
principle of the model is that the centre of mass travels a shorter distance in the bends than the
wheels. An application is demonstrated by examining Chris Boardman's 4 km individual
pursuit world record ride. The predicted completion time shows excellent agreement with the
record, however assumptions regarding atmospheric conditions and equipment dictate that
further validation is necessary. Examining the output demonstrates three fundamental princi-
ples of track cycling; (1) aerodynamic resistance is highly dominant, (2) the bike accelerates in
the bends and decelerates in the straights and (3) the rolling resistance increases in the bends.
A graphical-user-interface is to be produced for the model providing coaches and researchers
with an accessible and practical investigative tool.
1 Introduction
In the sport of cycling margins of victory are often extremely small, particularly in
track cycling. As a result equipment selection in track cycling is always of para-
mount importance. The choice of equipment is often govemed by an educated reck-
oning that one piece of equipment will improve performance over another. The
actual performance difference of selecting one piece of equipment over another is
not often realised. This paper presents an analytical model for track cycling which
provides a tool that allows the difference in equipment to be quantified in terms of
overall performance.
Analytical models of cycling are by no means a new concept. A number of au-
thors have produced equations of motion for the purpose of examining particular
aspects of cycling (van Ingen Schenau 1988; Olds, Norton, Lowe, Olive, Reay and
Ly 1995; Bassett, Kyle, Passfield, Broker and Burke 1999). Where this analytical
model aims to further previous understanding is in the derivation of a model specifi-
cally for track cycling.
116 Richard Lukes, Matt Carre and Stephen Haake
2 Model Formulation
The model is too lengthy to show the entire formulation, however the fundamental
principles will be presented. The basis of the model is that there are two states of
riding; one in the straights and one in the bends. The forces that dictate the rider's
motion are ho\ n in Fig. 1.
~
Fig. 1. The dominant forces on a bike
FD is the drag force, FA is the acceleration force, FR is the rolling resistance, FN is

the contact force, Fw is the weight and Fr is the transferred force . Resolving these
forces in the direction of motion gives the goveming equation,
(1)
In the straights the rider is travelling in a straight line and therefore not tipping,
leading to the simple free-body diagram shown in Fig. 2(a). In the bends the change
in velocity results in an angular acceleration and the addition of a centripetal force
(Fig. 2(b )). To maintain balance the bike tips, moving the centre of mass in towards
the track centre. The shift of mass is an important principle in the analytical model
as it is assumed the work of the rider is used to propel the centre of mass around the
track. In the bends the centre of mass travels less distance than the base of the
wheel. Therefore the movement of the centre of mass effectively cuts the comer,
re ulting in acc leration at the ba e of the wheel for a con tant p , er input.
Fw
(a) Straights (b) Bends
Fig. 2. The forces on the bike in the straight (a) andin thc bend (b)
An Analytical Model for Track Cycling 117
Angle ßis the banking angle of the track, a is the tipping angle of the bike, FF is
the lateral friction force at the wheel, Fc is the centripetal force, rw is the radius of
the wheel-to-track contact point and rm is the radius of the centre of mass. To con-
tinue the model derivation the forces shall be examined individually.
2.1 Transferred Force

The transferred force, Fr, is the force propelling the rider along. In this model Fr
accounts for the intemal losses within the bike, occurring mostly due to chain ineffi-
ciency and frame and component deflection and is given by,
(2)
where, T/ is the bike efficiency, Pis the rider power and v is the bike velocity.
2.2 Rolling Resistance

The rolling resistance, FR, is proportional to normal contact force and rolling re-
sistance coefficient, µR,
(3)
The term in the bracket in Eq. 3 is the normal contact force accounting for its in-
crease in the bends. The term Cs a coefficient that accounts for the increased rolling
resistance due to 'scrubbing'. Scrubbing occurs when the bike is not perpendicular
to the track and the rider has to steer the front wheel into the slope. This small angle
of steering causes the front wheel to skid or 'scrub' as weil as roll along the surface
leading to increased rolling resistance, as demonstrated in Fig. 3. The scrubbing
coefficient, C, has been determined using data reported by Kyle (2003) and is de-
1 1
p ndcnt upon thc angle of the bike r lativ t the tra k.
Direction of motion 1s identlcal In both cases
Rolling
Reslstance
foll in
R I g
es stance
Difference in rolling resistance due to 'scrubbing'
(a) Fiat track: Rolling

t (b) Banked track: Scrubbing
Fig. 3. Comparing flat and banked tracks illustrating scrubbing

2.3 Aerodynamic Drag
Aerodynamic drag is the major contributor to the cyclist's resistance. The aerody-
namic drag force, FD, is calculated using,
(4)
where CD is the drag coefficient, A is the frontal area and p is the air density. Drag
coefficients and frontal areas have been obtained from CFD simulations.
2.4 Completed Equation
Putting the terms from Eqs. 2-4 back in Eq. 1 gives the governing equation for
the model, Eq. 5.
ma = ( ry~)-( CJAl pv 2 ) - [ ( m :>osa+ mgsina JµRC, l (5)
3 Using the Model
3.1 Solution Procedure
The model uses the following procedure at a finite number of time steps to provide a
solution.
START
From power profile and governing equation

calculate distance travelled by centre of mass
Resolve forces in the bends to determine
tipping angle and radius of centre of mass
Use ratio ofradii (wheels to centre ofmass) to
calculate the distance travelled by the wheels
Does distance travelled
'--------1 exceed event distance?
y
STOP
Fig. 4. The solution procedurc
The model has been created for a 1 km time trail or 4 km individual pursuit. Ge-
neric power profiles for each event have been created, however the model allows the
use of power profiles extracted from an SRM™ power measuring crank. Once all
the input parameters for the specific event have been selected the model can calcu-
late the forces throughout the event, tipping angles and split times.
An Analytical Model for Track Cycling 119
3.2 Example Application

To demonstrate the model a 4 km pursuit shall be examined. In 1996 Chris
Boardman set the world record for this event, covering the distance in 4 m 11.1 s.
The model has been used to replicate this performance. A power profile for this
event has been constructed using the work of Broker, Kyle and Burke ( 1999), who
stated that Chris Boardman averaged 520 W over the entire event and 474 W once
cycling at a constant speed. The drag coefficient area (CoA) has been taken from
Hili ( 1993), who examined the same rider on a slightly earlier version of the bike in a
wind tunnel. The frontal area has been calculated based on the rider's height and
weight using the method presented in Bassett et al. ( 1999), giving a C 0 of 0.52.
Values for rolling resistance, bike efficiency and climate conditions have been se-
lected to mimic the actual conditions as closely as possible. Using these conditions
the model predicted a completion time of 4 minutes 10.7 s, which is a good approxi-
mation of the actual event. However it must be stated that this does not serve as a
strict validation due to the approximate nature ofthe input parameters.
The output of forces and velocity are shown in Fig. 5, illustrating some interest-
ing aspects. Firstly, the dominance of aerodynamic resistance is highly apparent.
Secondly, the fundamental concept that the movement of the centre of mass acceler-
ates the bike in the comer is demonstrated by examining the velocity profile. In the
model, the bike's velocity in the bends increases by approximately 0.4 m/s, a sensa-
tion feit by experienced riders. Thirdly, the rolling resistance noticeably increases in
the bends due to the far greater contact force as the bike tums.
so r
1
1
40
1
1
30
V"'j
~
...0~ 20
LL 10
,-- - ~ - --------, ---➔ -5
60 120 180 240 300

Time (s)
Fig. 5. The force and vclocity output from the model
The model can also be used for comparative analyses. Hili (1993) also reported
the Co,4 of a conventional track bike with a rear disc wheel, giving a Co of 0.62.
Using the model to compare the conventional track bike to the previous LotusSport
bike shows that Chris Boardman would have completed the 4 km event 13.9 s
slower, highlighting the efficiency ofthe LotusSport bike and the use ofthe model.
4 Discussion
Continuation of this analysis will involve some improvements to the model, mostly
involving a more detailed representation of the track geometry. In addition to modi-
fications, validation will be carried out whereby SRM power data is used as the input
and actual split times are compared to the model output. Once these steps are com-
plete a user interface will be created, providing coaches and researchers with an
accessible and practical tool for future investigations.
One such further investigation would be comparing the use of !arge and small
sprockets. Large sprockets are more efficient than small ones (Burgess 1998), how-
ever are disadvantaged in terms of weight and aerodynamics. Once sufficient data
has been obtained the model would provide an ideal tool for this and similar com-
parative analyses.
5 Conclusions
The fundamental principles of an analytical model for track cycling have been pre-
sented. lt has been demonstrated that this model can be used to scrutinise various
track cycling events. The output from the model has been shown to give the forces
and split times for the event. The application of such a model as both a research and
training tool can be greatly beneficial to give a fuller understanding of the event in
search of a performance advantage, which is of course the ultimate aim.
6 Acknowledgements
Many thanks to Dr Simon Goodwill for offering time to code the model.
References
Bassett, D.R.J., Kyle, C.R., Passfield, L., Broker, J.P. and Burke, E.R. (1999) Comparing
cycling world hour records, 1967-1996: modelling with empirical data. Med. and Sei. in
Sportsand Exercise, 31 ( 11 ), 1665-1676.
Broker, J.P., Kylc, C.R. and Burkc, E.R. (1999). Racing cyclist power requirements in the
4000 m individual and team pursuits. Mcd. and Sei. in Sportsand Excrcisc, 31(11 ), 1677-
1685.
Burgess, S. C. ( 1998). lmproving cycling pcrformancc with !arge sprockets. Sports Engineer-
ing, 1, 107-113.
Hili, R.D. ( 1993) Design and development ofthe LotusSport pursuit bicycle. Proceedings of
thc Institution ofMechanical Engineers, Part D: Journal of Automobile Engineering,
207(4), 285-294.
Kyle, C.R. (2003). Selecting Cycling Equipment. In: E.R. Burkc (Ed.), High-Tech Cycling:
The science ofridingfaster. Human Kinctics, Colorado, pp. 1-48.
Olds, T.S., Norton, K.I., Lowe, E.L.A., Olive, S., Reay, F. and Ly, S. (1995) Modelling road-
cycling pcrformancc. Journal ofapplied physiology, 78(4), 1596-1611.
van lngen Schenau, G.J. (1988) Cyclc power: a predictive model. Endeavour, 12(1), 44-47.
Forces During Cycling After Total Knee Arthoplasty
Maximilian Müller 1, Veit Senner 1, Markus Wimmer2

1 Technische Universität München, Germany, mueller@sp.tum.de
2 Rush University Medical Center Chicago, USA
Abstract. Recreational sport activity after total knee replacement (TKR) is of growing interest
to patients. Therefore, the purpose of this study was to evaluate dynamic loads acting on artifi-
cial and normal knees during cycling. A force measurement system has been developed and
was installed to evaluate the external loads on the pedals of a stationary bicycle. To analyze
the data, different evaluation algorithms were programmed to calculate pedal force levels of
the subjects during cycling.
1 Introduction
Recreational sport activity after total knee replacement (TKR) is of growing interest
to patients. In particular, cycling is one of those hobbies with up to 50% of TKR
patients riding a bicycle during leisure times (Kuster et al. 2000). The specific con-
tact mechanics of the tibiofemoral joint is well known for walking; however, there is
limited information for cycling activities (in particular for TKR joints). Thus, the
purpose of this study was to evaluate knee kinematics and kinetics acting on artificial
and normal knees during cycling.
2 Methods
2.1 Overview
The determination of knee loads in cycling using an inverse dynamics approach
requires the contact forces at the pedals as well as precise data on the joint kinemat-
ics (Fig. 1). This presentation focuses on the design of a force measurement system
and on the determination of the contact forces at the pedals as well as on the meas-
urement of knee kinematics.
External Loads Interna!

+ D Inverse Dynamics D Loading
Kinematics
Fig. 1. Calculation of internal loading by inverse dynamics

122 Maximilian Müller
The following model was considered on how to obtain knee loads (intemal load-
ing) from the contact forces at the pedals later on and therefore served as a basis for
the design of the measurement system in terms of the required load components and
reference frames. In a system of segments inverse dynamics is usually started at the
point next to the one of interest conceming the loads. For the present study this is the
contact point at the pedal (Fig. 2). The reactive knee forces can be calculated by
Newton mechanics which is cut to a 2-dimensional case for the example.
m .a = Ll = Fcontact + Fg + Fknee (1)

Fkncc is the wanted knee force. To obtain reactive muscle moments, the principle
of conservation of angular momentum is used as follows
0 ·~ = LM = fkncc X fknce + Mmuscle + fcontact X fcontact + Mcontact (2)
The left side of the equation is known from the kinematics analysis and while all
other parts of the right side are given from equation ( 1) or from the anthropometric
data Mmuscle can be calculated by solving the equation.
hip
femoral ~
knee-
F,=
joint
shank ~ fg
foot Fc.ontac!
Fig. 2. Model of leg segments for inverse dynamics
2.2 Force Measurement System

The force measurement system for the pedal loads was integrated into the crank
shafts and is based on strain gauge technology. For all load types Wheatstone full-
bridge arrangements were used to minimize the influence of temperature changes
and cross-sensitivity. The cells were designed to measure the three forces in space on
both pedals. The magnitude of loads was adjusted to expected values of elderly peo-
ple on stationary bicycles. Therefore, the maximum pedal force was set to a load of
500 N. The axial forces were read out directly by strain gauge bridges, the radial and
lateral forces were determined by separate bridges determining bending moments
(taking the Jever arms into account). The data transmission from the cranks to the
acquisition equipment as weil as the power supply of the load cells were realized by
collector rings mounted to the cranks and the bicycle frame respectively. Miniature
Power and Forces During Cycling After Total Knee Arthoplasty 123
voltage amplifiers were put directly onto the load cells to avoid deterioration of the
signals during the transmission (Fig. 3).
Fig. 3. Load cell with on board voltage amplifiers
2.3 Kinematics Analysis

3D-motion data of the lower extremity segments were obtained by a video based
camera system with passive markers placed on the skin. Data were recorded at 120
Hz with four cameras (Qualysis, Sweden). As skin movement relative to the underly-
ing bone is a primary factor limiting the resolution of detailed joint movement using
skin-based systems, in this study, the cyclist's legs were equipped with an advanced
marker set to conduct the so-called 'Point Cluster Technique' (Andriacchi et al.
1998). Twenty-one markers, distributed on shank and thigh, reduce the influence of
non-rigid body motion artefacts during human motion testing. Two additional mark-
ers were placed on the cranks and on the pedals to obtain the dynamic crank angle
that is essential to relate the force data to the motion data.
2.4 Post-Processing
Calibration matrix - The load cells are designed to measure strain by a generalized
load Q which is applied to the cell. During calibration the load cell's voltage signals
were related to actual applied loads. This was accomplished by applying known
static loads in various combinations. Cross sensitivity between the different sensor
signals was diminished by displaying the signals from cross sensitivity along with
the main signal from the applied load type. In order to obtain a valid linear regres-
sion, seven different loads of every load type were applied. Finally, the linear coeffi-
cients were determined, relating the applied load Q to the voltage response V of the
load cell as follows
V=a* Q (3)
Presuming that every load type is represented by the signals of a separate strain
gauge füll bridge a calibration matrix [C] was generated which related every single
generalized load to a voltage signal by means of the linear coefficient. [V] and [Q]
are column vectors and the equation is given by
[V] = [C] x [Q] (4)
The value of a calibration matrix grounds in the fact that it can be inverted to
produce a sensitivity matrix [Cinv ]. The latter can be used to determine unknown
loads. Equation (4) may be rewritten as
[Cinv] X [V] = [Q] (5)
This step was performed for every set of the 120 data sets total by a MatLab®
software routine.
Calculation of forces - The loads initially are seen by the load cells at the posi-
tion of the strain gauges, namely the measurement plane M (Fig. 4 ). In order to cal-
culate the forces acting on the bicycle pedals they have to be translated to the point
of load incidence P by cutting the accordant part free and setting the mechanical
equilibrium equations with the corresponding geometric parameters. Furthermore,
the loads currently are displayed in the local coordinate system of the tuming crank
and have to be transformed to a fixed coordinate system in order to facilitate inter-
pretation of the data. This was accomplished by accounting the dynamic crank angle
with a transformation matrix for the loads.
Fig. 4. Translation of loads to the point of load incidence at the pedal (P)
2.5 Setup of the System

Figure 5 shows the setup of the load cell at the crank of the stationary bicycle that
was used for the study.
Fig. 5. Setup of the load ccll at the crank

Power and Forces During Cycling After Total Knce Arthoplasty 125
Figure 6 shows the overall setup of the testing equipment with the stationary bicycle,
the camera system and the data acquisition computer prepared for the study.
Fig. 6. Overall sctup of the testing equipment
2.6 Study Design

Thirty subjects with an average age of 57.4 yrs. were included into the study. 10
subjects (av. age 64,5yrs.) had a total knee prosthesis with a constrained tibio-
femoral articulation (i.e. intemal-extemal and frontal plane rotations were limited
with this design); 10 subjects (av. age 53.lyrs.) had a total knee prosthesis with simi-
lar design but a mobile bearing allowing unconstrained intemal-extemal rotation; 10
subjects (av. age 54.5yrs.) had a normal (natural) knee joint and served as age-
matched controls. Patients were tested at 2 levels of resistance with 5 seconds of
acquisition time each. Before every testing, the patients would undergo a personal
survey and examination conducted by the medical doctor assigned to the project.
In the following the specific results of a normal subject are exemplarily presented.
The range of flexion du ring cycling was 41 °-126°. With increasing flexion angle the
epicondylar axis of the femur rotated 37° intemally and translated 11mm posterior
(based on a fixed tibia reference frame). The forces measured at the contact between
foot and pedal were analyzed in vertical (left foot: max. -1 0N/min.-l 20N), horizontal
(I00N/ 35N) and lateral (30N/-15N) direction (displayed in a fixed reference frame)
to comparc them with the anterior-posterior translation and intemal/extemal rotation
ofthe knee (Fig. 7.).
As has been reported by others, the forces during biking were relatively low;
however, the knee joint underwent considerable intemal/extemal rotation and ante-
rior-posterior translation. Total knee arthroplasty designs are typically optimized for
walking and not for deep flexion maneuvers. Therefore, despite the overall low
forces, generated movements at the articulation can still be troublesome causing
wear and/or increased constraint forces at the tibia plateau and fixation.
.,..,...,..,..,.1,..,.......,
1„ i---------'I!~ ~
!u f----,------,c-~ ~.,,.----:,:--~
Pedal farca Left lkMi
1•
--.......
Fig. 7. AP-translation and intemal rotation (upper row), pedal forces (below) of anormal
4 Conclusions
To our knowledge this is one of the first studies who aim to compare force genera-
tion in patients after total knee arthroplasty. The measurement system was validated
and the study with TKR patients is finished. Now, the data has to be processed to
obtain information about effects of TKR on cycling activity. In addition to the pedal
loads, however, we also acquired the contact forces at the saddle and at the handlebar
to preserve the possibility to build and validate a forward simulation model of the
whole cyclist which is the content of a future project.
References
Andriacchi, T., et al. (1998) A Point Cluster Method for In Vivo Motion Analysis: Applied to
a Study ofKnee Kinematics. J Biomech Eng. 120(6):743-9.
Kuster, M., et al. (2000) Knee endoprosthesis: sports orthopedics possibilities and limitations.
Orthopäde 29(8):739-45.
Raskulinecz, (1980) Design and Construction of an Anthropometrie Dummy of the Alpine
Skier. Case Western Reserve University, Ohio.
Wimmer, M.A. (1999) Wear ofthe Polyethylene Component created by Rolling Motion ofthe
Artificial Knee Joint. Shaker, Aachen.
A Study of Aerodynamic Drag and Thermal Efficiency
of a Series of Bicycle Heimets
Firoz Alam, Aleksandar Subic and Simon Watkins
RMIT University, firoz.alam@rmit.edu.au
Abstract. The primary objective of a helmet is to provide head protcction during fall or acci-
dent, however, thermal comfort and aerodynamic efficiency are becoming important design
criteria. Heimet with venting generally increases thermal comfort but decreases aerodynamic
efficiency. Thereforc, an optimal design for helmet is very important in order to satisfy both
acrodynamic and thermal efficiency. The primary objective of this work is to study the aero-
dynamic efficiency and thermal comfort of a series of current production helmets available in
Australia. Aerodynamic drag and thermal comfort was measured under a range wind specds,
yaw and pitch angles and compared.
1 Introduction
Bicycle helmets are mandatory for recreation_al or professional bicycle riders in many
countries including Australia. Although the primary objective of a helmet is to pro-
vide head protection during fall or accident, thermal comfort and aerodynamic effi-
ciency are becoming important design criteria (Alam et al. 2005, Bruhwiler 2003,
Reid and Wang 2000). Most bicycle helmets are made of foam that holds up heat,
generated by the rider's head during cycling. Humidity and high ambient tempera-
ture make the situation worse as trapped heat causes significant discomfort (sweat-
ing, stickiness etc). Heimets with venting can minimise this problem. However,
venting generally increases aerodynamic drag. Therefore, an optimal design for hel-
met is very important in order to satisfy both aerodynamic and thermal efficiency.
The primary objective of this work is to study the aerodynamic efficiency and ther-
mal comfort of a series of current production helmets available in Australia for rec-
reational users. Each helmet was tested for their aerodynamic efficiency and heat
dissipation characteristics under a range wind speeds, yaw and pitch angles in the
RMIT University Industrial Wind Tunnel. Descriptions about RMIT Industrial Wind
Tunnel and other equipment were given in Section 2. Heimets were ranked according
to their aerodynamic efficiency and thermal comfort.
128 Firoz Alam, Aleksandar Subic and Simon Watkins
2 Test Procedures and Heimet Descriptions

The aerodynamic efficiency in terms of drag and heat dissipation characteristics for
five helmets were experimentally measured under a range of speeds (20, 30, 40, 50
and 60 km/h wind speeds), yaw angles (0, ±30°, ±60° and ±90°) and pitch angles (90,
60, 30 and 0 from horizontal axis). The aerodynamic drag was measured using a six
component force sensor. The thermal efficiency in terms of heat dissipation (tem-
perature drop) was measured using a heat pad on the dummy head. Seven thermo
couples were attached with the heat pad located around the head under the helmet.
An instrumented dummy head with the heat pad, thermo couples and helmet is
shown in Figure 3. All five helmets are different in terms ofventing holes and struc-
tural geometry. All five helmets were new and manufactured by Rosebank Australia.
These helmets are: Blast, Mamba, Nitro, Summit and Vert (see Fig 1). In order to
understand the effects of venting, the Vert helmet was modified (venting blanked
oft) and tested twice as standard and modified configurations, see Fig lf.
a) Nitro Heimet b) Mamba helmet c) Summit helmet
d) Vert helmet f) V ert helmet modified

e) Blast
(simply blanked oft)
Fig. 1: A bird's eye viewofall helmets
Fig. 2: Experimental setup in the Fig. 3: Experimental setup for thermal

test section with a dummy head testing (dummy head, heat pad and helmet)
A study of aerodynamic drag and thermal efficiency of a series of bicycle helmets 129
Each helmet was tested with and without the head assembly (dummy head and
mounting device) for all speeds, yaw angles and pitch angles; and the aerodynamic
forces due to helmets were determined. All forces were converted to their non-
dimensional parameters and only the drag force coefficient (Cd) is presented in this
work. Thermal efficiency was measured by heating up the heat pad at 60°C which
was selected arbitrarily. A thermostat was used to keep the temperature constant on
the heat pad. Seven thermocouples were attached on the heat pad under the helmet to
monitor the temperature drop around the head. Two thermocouples were attached at
the front of the head, two on each side (left & right), two on the rear of the head and
one thermocouple at the centre of the head in order to obtain a comprehensive tem-
perature distribution (see Fig 3). The temperature drop was monitored for 5 minutes
at all wind speeds. The temperature readings from all seven thermocouples were
averaged and presented in this paper. The results for zero yaw angles are presented
here. The drag coefficient as a function of Reynolds numbers and pitch angles and
the average temperature drop for the Mamba, Nitro and Vert helmets are shown in
Figs. 4-9. The results for other helmets are not shown here. The drag coefficient (Cd)
is relatively independent of Reynolds numbers for all helmets at 90° and 60° pitch
angles except at very low speeds. However, minor Reynolds number dependency for
all helmets was noted at other pitch angles (30° and 0°). The pitch angle was meas-
ured from the horizontal. The airflow at 30° pitch angles becomes complex due to the
interaction of the flow separation from the local venting and fitting strips. The high-
est aerodynamic drag was found at 0° pitch angles for all helmets except the Vert
helmet with and without venting. With an increase of pitch angles, the projected
frontal area of the helmet reduces and the airflow becomes more streamlined. The
V ert helmet has minimum venting and it produces relatively lower aerodynamic drag
compared to other helmets. No significant variation in drag coefficient of the un-
modified and modified Vert helmets was noted. The Vert helmet generates the low-
est drag at over 30 km/h and 90° pitch angles. On the other hand, the Nitro and
Summit helmets produce higher drag at the same speeds and pitch angles. The pitch
angle has virtually no impact on the drag coefficient for the Vert helmet (see Fig 6).
i =:::::::::::
Cd Variation with Renolds numbers (Mamba hetmet, 0 yaw, various Cd Variation With Reynolds numbers (Nitro Heimet, 0 Yaw, Various
Pitch angles)
püohaaglu)
1
1.20 ---···--- ··---·- -------· --MM'ioaOPilch,OY<>N 1.20 - - - ---- --
1,.,,-
B 1.00 ----- ---- -~ -
'.:. 1.oor~=="===-~===..====;;-
~ 0.80 - - - - - - - - - - - - --------·------·-----
/J 0.60 ~ - - - - - - - - - - - - - - - j
~ 0.60 ---------------~~-
l 0.40 ~~~-----. -
m
5 0.40 r-==: ..:.:..---·
•
0.20 t---- - - - - - - - - - - - - - - - j
020+---------------
0.00+----~------~------,
100000 150000 200000 250000 300000 350000 0.00+------~-------r----"
100000 150000 200000 250000 300000
Reynolds numbers (Re)
Reynolds numbers (Re)
Fig. 4: Drag coefficient as a function of Fig. 5: Drag coefficient as a function of

Reynolds number and pitch angles Reynolds number and pitch angles (Nitro
(Mamba helmet, 0° Yaw) helmet, 0° Yaw)
130 Firoz Alam, Aleksandar Subic and Simon Watkins
The temperature drop is evident for all helmets with the increase of Reynolds num-
ber. The pitch angle has negligible effects on temperature drops for the Blast,
Mamba, Nitro and modified Vert helmets. However, a significant variation in tem-
perature drops due to pitch angles is noted for the unmodified Vert helmet. Addition-
ally, a small variation in temperature drops was noted between 90° and other pitch
angles for the Summit helmet. The lowest temperature drop was noted for the modi-
fied helmet as expected due to no venting (not shown here). The highest temperature
drop was evident for the Mamba helmet. Although the Vert helmet was aerodynami-
cally more efficient, it is the worst performer in terms of thermal efficiency. On the
other hand, the Nitro and Summit had relatively higher aerodynamic drag at 90° pitch
angle and generate significant temperature drops. However, the Mamba helmet is the
optimal helmet for its aerodynamic and thermal efficiency as it produces the highest
temperature drops at all pitch angles and low aerodynamic drag (second lowest
drag). The number ofventing, venting geometry and venting location play an impor-
tant role in temperature drops. However, they also can generate more aerodynamic
drag.
~===
Cd variation with Raynolds numbers (Vart heim et, 0 Yaw, Various Average Terfl)erature Variation (~rrba. 0 Yaw, various Rtch
Pltch angle,) angles) ~~~
'-"~---------<J-:::::::::::1
1.20+------------ VerlOPlt<:11,0Y:/M·I
70 ~ - - - - - - - - - - - - < - - + - 9 0 Atch
---a--- 60 Atc h
_,._30Atch
B 1.oo+-----------------1 ~OAtch
1::: ---------------------j
! 0.40 t;;;~i~~§~~~;;;§~~;;;;;;;~
0.20 .---·--------
0.00
100000 150000 200000 250000 300000 50000 100000 150000 200000 250000
Reynolds numbers (Ra) Reynolds Nurrber
Fig. 6: Drag coefficient as a function of Fig. 7: Temperature variation with Rey-

Reynolds number and pitch angles (Vert nolds numbers as a function of pitch
helmet, 0° Yaw) angles (0° yaw), Mamba helmet
Average Te~rature (Nitro, O Yaw, various Atch angles) Terrperature Variation (Ver! heln-et, O Yaw, various Atch angles)
70 --+-90 Atch
70 --+- 90 Aich
---so Atch __,._ 60 Aich
--
1
60
1
......,_ 30 Atch 60 ' \ _...,._30Alch
50
~ ~ o Atch
50~
---¾--0Atch
~
~
t \~
~
!
40
"""'----------- t 40
-
30
- ~~
,.,_E
30 ~ -
~
20 20
0 50000 100000 150000 200000 250000 0 50000 100000 150000 200000 250000
Reynolds t-l,rrber Reynolds N„mt>er
Fig. 8: Temperature variation with Rey- Fig. 9: Temperature variation with Reynolds
nolds numbers as a function of pitch numbers as a function of pitch angles (0°
angles (0° yaw), Nitro helmet yaw), Vert helmet
A study of aerodynamic drag and thermal efficiency of a series of bicycle helmets 131
4 Conclusions
The drag coefficient is relatively independent of Reynolds numbers at high speeds

under 90° and 60° pitch angles for all helmets.
The Yert helmet produces the lowest aerodynamic drag but worst performer for heat
dissipation.
The Nitro and Summit generate higher aerodynamic drag but perform weil in heat
dissipation.
The pitch angles have significant effects on aerodynamic drag, however less effects
on heat dissipation.
The Mamba is the optimal helmet for its aerodynamic and thermal efficiency.
The number of venting, venting geometry and venting location on the helmet play
the key role for the thermal and aerodynamic efficiency.
5 Acknowledgments
The authors are very grateful to Anthony Resta and Rhys Solomons for their assis-
tance with the testing and data analysis. We are also indebted to Rosebank Australia
for providing the helmets.
References
Alam, F., Watkins, S. and Subic, A. (2005) Aerodynamic efficiency and thermal comfort of
bicycle helmets, Proc. of the 6th International Conference on Mechanical Engineering
(ICME2005), TH-32 (1-6), ISBN 984-32-2846-4, 28-30 Dccember, Dhaka, Bangladesh.
Bruhwiler, P. A. (2003) Heatcd, perspiring manikin headform for the mcasurement ofhead-
gear ventilation charactcristics, Meas. Sei. Techno!. 14: 2 I 7-227.
Reid, J. and Wang, E. L. (2000) A system for quantifying the cooling effectiveness ofbicycle
hclmcts, J Biomech Eng. 122 (4): 475-460
4 Golf
Synopsis of Current Developments: Golf
Steve Mather
University of Nottingham, UK, stephen.mather@nottingham.ac.uk
The main challenge in the golf industry is to design equipment which can be used by
the 95% of golfers who do not play the game weil. This is where the market lies.
Unfortunately, the trend appears to be to design clubs for the professional and
better amateur golfers and then allow for small changes in mass distribution or shaft
flexibility to aid those with poor swings.
The first step in the building block process of redesigning clubs must be to under-
stand the nature of all swings, particularly the major differences between the good
and poor swing. Some ofthe papers in this category ofthe proceedings offer building
blocks for researchers to use in studying the golf swing, and are therefore very useful
additions to the current situation.
We must not lose sight of the fact that the elite end of the sport is still the shop
window for golf clubs. lt is interesting to note from one of the papers that even with
first class golfers, the variations from swing to swing can be marked. Perhaps these
athletes can take advantage of some of the latest technology.
There is no doubt that much of the distance gained and accuracy attained in the
game of golf comes from the technology applied to the design of the golf ball. The
effect of ball construction on performance was first brought to public attention in
papers at golf conferences 10 years ago. Since then, the market has made huge
strides in applying the technology to marketable balls. However there is still room
for matching the ball to the club design, or perhaps more relevantly, changing the
club design to match the technology offered by the ball. Much data is therefore
needed on the spin generation mechanisms and the aerodynamics of the flight of
spinning spheres. One paper in this category deals with the variation of Lift and
Drag coefficients during the fight of the ball.
An Instrumented Grip Handle for Golf Clubs to Measure
Forces and Moments Exerted by Each Hand During
Swing Motion
S. Koike 1, H. Iida2, H. Shiraki 1 and M. Ae 1

1 Institute ofHcalth and Sports Sciences, University ofTsukuba, lbaraki, Japan,
koike@taiiku.tsukuba.ac.jp
2 Machine and Control Department, Polytechnic University, Kanagawa, Japan
Abstract. An instrumented grip handle was designed to simultaneously measure the forces and
moments exerted by each hand on the handle during golf swing. Eleven pairs of strain gages
were attached on the surface of an aluminum bar inserted under separated grip covers. The
device was calibrated under static conditions and revealed good agreement betwecn applied and
calculated loads. The output ofthe sensors was converted into forces and moments by resolving
static equilibrium equations. A profcssional golf player participated in this study and performed
golf swings with several clubs. Reflective-markers on the body segment endpoints and on the
clubs were captured by VICON motion system with 8 cameras operating at 250Hz. The results
obtained in this study were: ( 1) long axial load of shaft affected the sensor coefficients of the
device, and (2) intemal forces and moments that do not cause the motion of the club were
observed in the swings.
1 Introduction
In golf swing, the upper limbs and the golf club form a closed multi-segment loop.
Due to the kinetic redundancy caused by the closed loop, it is impossible to determine
the forces and moments exerted by each hand on the grip handle ofthe club with only
visual information. The only way to obtain such information is to measure it directly.
Previous research on the kinetics of golf swing have investigated single hand models
(Milne and Davis, 1992; Sprigings and Neal, 2000). The purpose of this study was to
propose an instrumented grip handle for golf clubs capable of obtaining kinetic
information of each hand during swing motion.
138 Sekiya Koike, Hiroshi Iida, Hitoshi Shiraki and Michiyoshi Ae
2 Methods
2.1 Structure of the Club

Figure 1 shows the fully assembled club and the definition of the club coordinate
system (oc-XcYcZc)- Two spherical markers with negligible-mass shafts were attached
on the club for the purpose ofmeasuring the orientation ofthe moving club. Figure 2
shows the structure of the instrumented grip handle. The grip handle consists of an
aluminum alloy bar and two sets of aluminum covers. Eleven pairs of strain gages
were attached on the bar to calculate: 1) the torsional moment acting around the grip
axis between the hands; 2) the bending moments; and 3) the tensile and compressive
axial forces.
Sphere markers for measurcment

of position and orientation ---C:::::::::
i
system
Connector]
Head side grip Instrurnented
Butt side grip grip hao<lle
Fig. 1. The fully assembled golf club.
~$ Head side-~p cover ;:nnector
Butt side grip cover ;,

i".'• ;luminum
alloy rod
P ,
Gages a_(bending moments)
~::::::i
Gages b (axial force)
Gag es d (:::::gc moments)

' Gages e (axial force)
Gag es ~ (bending moments)
' : Gages g (tovr~._i·b···n···al

moments)
!1
!'.
'·
//,
Fig. 2. The structure of the instrumented grip handle.
2.2 Device Calibration

Each strain gage was calibrated under various static load conditions with an aluminum
An Instrumented Grip Handle for Golf Clubs 139
calibration rod attached to the grip handle by a connector. For loading and unloading
conditions, calibration was carried out under long axial tensile load exerted by string 6
and two-axial transversal forces exerted by strings 1-4 as shown in Figure 3(a). The
transversal forces were measured by the load-cell 1. For the purpose of studying
various long axial tensile load conditions, which are due to mainly the centrifugal
forces that result from the rotational movement of the club, calibration of the strain
gages for bending moments was completed using several axial-load conditions. The
tensile loads were applied by hanging weights with a string attached to the end of the
calibration rod and passed through a simple pulley system. The value was detected
with load-cell 2 and was changed with the positioning of the weights on a rail as
shown in Figure 3(b).
2.3 Calculation of Forces and Moments

Although the forces and moments exerted by each hand on the club are distributed all
over the handle area, these loads were represented by vectors, assuming that they act at
the fixed points of the hands in swing motion. The inertial forces of the handle pieces
were neglected.
From equations under static force/moment equilibrium conditions (Koike et al,
2004), the forces and moments are derived as follows:
Nhead Ta I I [(r h,head - rsJ X] [er h,butt - rsa) X

Nbutt =A-1 Tc I I [crhhead -r JX] [er h,butt - rsc) X]
5
A= (1)
Fhead Td 0 I 0 [(rh,butt - rsd)X]
Fbutt Tf 0 I 0 [er h,butt - rsf) X]
where vectors F; and N; indicate the vectors of force and moment for the lh location,
the subscripts head and butt stand for the values at the head-side and butt-side hands,
and the vectors rh,head and rh.butt represent the position vectors of the fixed points of the
head-side and butt-side hands, and the vectors 1°k and rsk indicate moment and position
vectors at the l<h sensors, respectively.
The instrumented grip handle
(a). The calibration device. (b). The loading device.

Fig. 3. The setup for the calibration.
0%-50%
(a). The swinging club coordinate 50% - 100%

system. (b). Forward swing motion.
Fig. 4. Stick diagrams of the coordinate system and the forward swing motion.
2.4 Swing Motion Analysis

The forces and moments exerted by each hand were expressed according to the
swinging coordinate system using a plane named "swing plane". The plane was
approximately calculated so as to involve the head and butt positions ofthe club from
the moment the head of club was highest to impact with the ball. Figure 4(a) shows
stick diagrams of the swing plane and the orientation of the swinging club coordinate
system. .
A professional golf player volunteered to participate in this study. The player
performed swings with driver, fifth iron and sand wedge clubs. Markers on the body
segment endpoints and on the clubs were captured by VICON motion system with 8
cameras operating at 250Hz. A personal computer was used to store the strain gage
signals, which were amplified by dynamic strain amplifiers. The sampling frequency
of the data collection was 500Hz. The forward swing motion was defined from the
top of swing to impact. Figure 4(b) shows examples of stick pictures of standardized
forward swing motion.
3 Results
Figure 5(a) shows the plots between sensor output voltages and applied y-axial
moments with respect to the sensors 'tr„ and 'tr,Y under various loading conditions.
From the calibration results of the sensors for bending moments, the relationships
showed very good linearity and little crosstalk in terms of the moments acting about
other axes at each loading condition. Since the tensile loads affected the relationship,
the gradients of linearity were calculated by using interpolating functions with respect
to the tensile loads as shown in Fig. 5(b).
Figure 6 shows an example of the curve pattems of actual and calculated values of
the forces applied at Xswc-axis under axial load of about 150N. The curves show good
agreement.
Figures 7 (a), (b) and (c) show the forces at Xswc-axis, Yswc -axis and the moments
around Zswc axis in the forward swing with fifth iron club. Solid and dashed lines
represent the components ofhead- and butt-side hand, respectively. These values were
expressed in the swinging club coordinate system, and were normalized to forward
An Instrumented Grip Handle for Golf Clubs 141
swing phase. The duration of the phase was 0.3 sec.
□a
"
~ □2
" "'~
E
i!i
J!l .. □ J.
□
~8 -0,1,
C
OJ
5
E L
0
~ -□ c
" -□ a
-□~~--~--~--~~
2D □ 1 □□ 2 □□ 3 □□
Sensorvo:tage IY] Lang axBlbads (N]
(a). Calibration plots for a sensor. (b ). Coefficients under various loading

conditions.
Fig. 5. Effects of long axial loads on sensor calibration.
30
- fy <catul3ted)
--~ fy Cacb.JaD
20
..,
i::i
OJ
10
~
0
"-
-10
0 2 6 8 10
Tin e [sJ
Fig. 6. Curve pattems of actual and calculated forces.
The X8w0 -axial forces of the hands showed approximate reverse pattems. The couple
force that accelerate the angular motion of the club decreased gradually until 25% of
the motion normalized time, increased until 70% time, and then decreased rapidly
toward impact. Large intemal forces were observed between the hands along Yswc-axis
until 70% and the resultant force along the axis increased toward the impact reaching
!arge positive values.
Although the moment of the butt side hand around Z8w0 -axis showed negative
values during the forward swing phase, the moment of the head side hand showed
positive values until 80% of the motion normalized time. Since the magnitudes of the
forces of the head side hand were greater than those of the butt side hand until 80%
time, the resultant moment acted as an angular accelerative torque around the axis up
to 80% time.
4. Discussion
Since the sensors were located in the grip handle attached to the club shaft with the
connector, it was possible to obtain kinetic information of each upper limb during golf
swing with several types of clubs.

As is the case with an instrumented baseball bat (Koike at el, 2004 ), the output
voltages of the strain gages for the bending moments were affected by the long axial
forces, which were mainly caused by centrifugal forces and tensile forces. Because of
that, it was necessary to calibrate the sensor device under various relatively high axial
forces.
In hitting motions with the hands, infinite combinations of forces and moments can
be exerted by each hand to realize the same motion ofthe hitting tool due to the kinetic
redundancy of the closed segments loop. From the results obtained in this study,
forces and moments of head side and of the butt side hands canceled each other out.
Hence, it is necessary to obtain kinetic information of each hand by using the device to
realize complete understanding of golf swing mechanism.
5. Conclusions
An instrumented grip handle capable of measuring forces and moments exerted by

each hand was proposed to achieve a kinetic analysis of each hand in golf swing. The
merit of the grip handle type device is that kinetic information about various types of
clubs can be obtained by exchanging shafts via the connector. The ability to quantify
acting forces and moments of each hand in golf swing has the potential to ( 1)
understand the mechanics of swings and to (2) evaluate technique skills of increasing
head speed for golfers.
---· Butt-söe hand 20

Resulantforce , ,,
,, .. _......,/
~ 200 ,../ '
i!i -----------
-100 -100 _ _ _ _ _.,,

-200 ,-----............. ,
-200
-3 □□~-----~ -20~-----~
□ so 100 so 100 □ so 100
Narm alized ti:n e 1/.:1 Norm alized ti:n e li::1 Norm alized ti:n e [%]
(a). The forces inxswc· (b). The forces alongyswc• (c).The moments aroundzswc•
Fig. 7. The forces and moments expressed by the swung club coordinate system.
References
S. Koike, H. lida, T. Kawamura, N. Fujii and M. Ae. (2004) An instrumented bat for
simultaneous measurement of forces and moments exerted by the hands during batting. The
Engineering of Sport 5, 2, 194-199.
R.D.Milnc and J.P.Davis. (1992) The Role of the shaft in the golf swing, Journal of
Biomechnics, Vol.25, No.9, 975-983.
E.J.Sprigings and R.J.Neal. (2000) An insight into the importance ofwrist torquc in driving the
golfball: A simulation study, Journal of Applied Biomechanics, Vol.16, 356-366.
The Aerodynamic Influence of Dimple Design on Flying
Golf Ball
T.Sajima 1, T.Yamaguchi 1, M.Yabu 1 and M.Tsunoda2
1SRI Sports Ltd., Hyogo, Japan

2SRI Research and Development Ltd., Hyogo, Japan
Abstract: The dimples on the golf ball surface affect the aerodynamic force while flying.
They are designed for this effect with various sizes, depths, contours and shapes; those are
combined and distributed on the surface. In this study, the aerodynamic influence of the
dimple design on drag and lift force is investigated by analyzing the result of the indoor and
outdoor ball launcher tests that measure CD, CL, trajectory and flying distance.
1 Introduction
The flying performance of a golf ball is changed by the dimples on the surface of golf ball. In
other words, the drag force and lift force are changed by size, depth, contour and shape of
dimple. In this study, the dimple depths are selected for the analysis and the relationship
between the depth and CD (drag coefficient) and CL (lift coefficient) while golfball is flying
is investigated with the numerical analysis for the indoor launcher test and the outdoor
launcher test. Aoki, Oike and Nonaka (2002) and Ting (2005) clarified change of CD and CL
that the rotating ball with different dimple depth. However, the level of depth is not suitable
for the recent trend of golf ball and numerical analysis and experiment is not performed in the
condition of the drivcr hitting of average golfcr. Then dimple depth in our study is designed
within the depth of rcccnt golf balls. Carry distance and trajectory are measured with the
outdoor test under avcrage golfcrs' initial launch condition.
2 Design of test balls
Specifications of the dimples are shown in Fig. 1 and Table 1. And the pattern of dimples is
shown in Table 2. This study adopts same dimple pattern and different dimple depth for each
sample. The difference of depth makes difference of the total dimple volume and the
curvature of cross-section on the surface of samples. Depth d is distance from a dimple outline
plane to the central decpcst part and volume V is capacity ofthe part below the dimple outline
plane. Diameter D of dimple is 4.07mm and number of dimples N are 300. The dimple pattcrn
on golf ball surface is completed by connection of two hemispheres shown in Table 2. Test
balls are made of ionomer rcsin cover and rubber core by the injection molding. The diameter
of the balls is 42.7mm and the cover thickness is 2.3mm. However, as several pins for core
holding and air vent is arranged around the pole of cavity for injection molding, twelve
144 Takahiro Sajima
dimples made by these pins are added to above 300 dimples. Finally, 3 l 2dimples are arranged
on all the balls. Though golf balls are usually painted after injection molding, in this study,
samples are prepared without painting. The dimcnsion of these real balls fit in CAD model for
the numerical analysis.
Fig.1. Sectional-vicw of Dimplc
SampE D in p:e Num ber Dm eter Deptt, Radi.Js of VolJm e

Curvatl.Jre
'm R 'u
' C "'
(m m l (m m l (m m l (mm'>
a 300 407 OJ.69 12ll 1J.O
' OJ.50
b 2 3.93 130 0-91
,f 10 306 OJ.63 73 Ob□
' a a 300 407 OJ.78 11;;> 1J.6
b 2 3.93 OJ.50 130 DR1
,f 10 306 DJ.63 73 Ob□
'b a 300 407 OJ.89 110 123
b 2 3.93 DJ.50 130 0-91
,f 10 306 OJ.63 73 Ob□
Table 1. Specification of Dimples
D in pla Latiude Lonqi!J.Jde

<dea> <dea>
a 60 00 120 240 360 480 600 720 840 960 1080
1200 1320 1440 1560 1680 1800 1920 2040 2160 2280
2400 2520 2640 2760 28110 3000 3120 3240 3360 3480
160 296 42ll 1016 114ll 1736 186ll 2456 258ll 3176 330ll
170 70 650 790 1370 1510 2090 2230 2810 2950 3530
200 18.S 53.S 90.S 125.S 162.S 197.S 234.S 269.S 306.S 341.S
25.S 00 360 720 1080 1440 1800 2160 2520 2880 3240
300 120 600 840 1320 1560 2040 2280 2760 3000 3480
310 250 470 970 1190 1690 1910 2410 2630 3130 3350
380 00 360 720 1080 1440 1800 2160 2520 2880 3240
412 142 578 862 1298 1582 2018 2302 2738 3022 3458
47.S 27.S 44 .S 99.S 116.S 171.S 188.S 243.S 260.S 315.S 332.S
50.S 00 720 1440 2160 2880
568 17.S 54.S 89.5 126.S 161.S 198.S 233.S 270.S 305.S 342.S
630 360 1080 1800 2520 3240
64J. 00 720 1440 2160 2880
74.S 360 1080 1800 2520 3240
b 900 00
C 738 00 720 1440 2160 2880
Table 2. Coordinates of Dimples (hemisphcre)

The Aerodynamic Influence ofDimple Design on Flying Golf Ball 145
3 The analysis of the initial launch condition for golfers
3.1 Condition for analysis

CD and CL of flying golf ball are changing from moment to moment and these variations
make a big influence on distance and trajectory. In particular, it is important to know CD and
CL at the initial launch condition for langer distance. Then CD, CL analysis is performed
under initial launch condition of average golfers. Ball speed (equal to flow velocity given in
simulation) and Back spin rate are set up with 58m/s and 2400rpm by the result of our
research.
3.2 Results
CAD models those have a difference of dimple depth are designed by SolidWorks 2003. The
design ofthese models is similar to current golfballs on sale. Numerical analysis is performed
by the fluid analysis software FLUENT using DES model and making 6,000,000 cells in mesh.
This analysis shows that as dimple depth becomes deeper, CD becomes greater and CL
becomes smaller. The indoor test to know CD and CL is carried out using ITR that was
developed by USGA . The technical structure and procedure is described in "The Indoor Test
Range (ITR) Technical Description and Operation Manual Revision 2.0" edited by United
States Golf Association Research and Test Center. Figure 2 shows comparison between
numerical analysis and experimental result at indoor test range . CD and CL of the numerical
analysis are average value between t=0.003 and t=0.01285 (t: analyzing time). Subscript (S)
shows the result of simulation and subscript (1) shows the result of indoor test. Analysis
shows as dimple depth becomes deeper, CD becomes greater and CL becomes smaller.
Though the value difference of CD is about 0.02, the tendency is similar to numerical analysis.
From CD analysis, shallower dimple will be more advantageous than deeper dimple for flying
distance. Nevertheless, it is important to consider the balance of CD and CL for flying
distance as CL tendency towards dimple depth is opposite to CD.
028
···········•
..,,...J 026
..........- , .
., •••••~···--&--~·
A
u 024
.µ
C:
,III
022
~QJ
0 -----CD CD
u 020 --<>--CLCD ~
.~
.,,
...J
C:
0J.8
· · ·A· · · CD ('S)
· · ·El· · · CL(';)
"' - -'----'----~----..,-.:
"'~ 0J.6 o~------'-'-'~--ee>--~--_-_----~-'--:..:-.:.:·-~-:__:--~·. ·0
A -----e
0J.4
0J.65 0J.70 0J. 75 0J.80 0J.85 0 J.90 0J.95
depth d Cm m)
Fig.2. Comparison between numerical analysis and experimental result at Indoor Test Range
146 Takahiro Sajima
4 Indoor and outdoor launcher test
4.1 Indoor launcher test

The detail of USGA for overall distance and symmetry test is described in "ACTUAL
LAUNCH CONDITIONS OVERALL DISTANCE AND SYMMETRY TEST PROCEDURE
(PHASE 2) Revision l" edited by United States Golf Association and Royal and Ancient Golf
Club of St. Andrews. Figure 3 shows conditions of the test. These conditions are set within the
bounds of ball speed and back spin rate on flying ball hit with driver club. CD and CL under
these conditions are measured by ITR and CD, CL while flying are identified by software
(USGA ITR DATA ANALYSIS ver.2.0.0). CD and CL on flying of three samples are
identified by this procedure under the initial condition of typical Japanese average golfer
(Initial ball speed: 58rn/s, Back spin rate: 2400rpm, Launch angle: 12deg). Figure 4 and 5
shows the result of CD and CL simulation. As dimple depth becomes deeper, CD becomes
greater from 0yd to about 80yd after test ball fired. But the tendency becomes opposite after
that. As dimple depth becomes deeper, CL becomes smaller in all flying.
3400
♦
~ 2900 - --------♦
♦- •
~ ♦ ♦
! 2400 ♦ ♦
~
♦
•
m 1900
♦
♦ • ♦ •
1400
20 30 40 50 60 70 BO 90
BanSpeedtn /s)
Fig.3. 15 Conditions for USGA Aerodynamic Test (The latest conditions)
0.30
0c9
0c8
"'u □ c7
[)
°t:
.i:;
0cb
~ 0cS - B
QJ
0
u 0c4 C
!"' 0c3
0c2
0c1
Oe□
0 30 60 90 120 150 180
C arry (ydl
Fig.4. Variation ofCD on Flying Golf Balls

The Aerodynamic Influence ofDimple Design on Flying Golf Ball 147
024
023
022
...1 021
u
'.: 020
~
.i:;
~ 0J.9
., 8
C
S 0J.8
~
--' 0J.7
0J.6
0-15
0J.4
0 30 60 90 120 150 180
C arry(ydl
Fig.5. Variation ofCL on Flying Golf Balls
4.2 Outdoor test by ball launcher

Test balls are launched outside to measure actual carry distance and trajectory under same
initial condition as indoor test (Initial ball speed: 58rn/s, Back spin rate: 2400rpm, Launch
angle: 12deg). Measurement is carried out by TRACKMAN. TRACKMAN is the flying ball
following system with milli-wave radar. Six balls of each sample are launched. In addition,
while testing wind is against the balls and temperature is 5degees centigrade. Figure 6 shows
the side view of actual trajectory. Radar unit is located 4yds in front of ball launcher. Actual
carry distance is shown in Figure 6 is the distance between TRACKMAN and the landing
point of ball. As dimple depth becomes shallower, carry distance becomes greater and apex of
trajectory becomes higher. The difference of carry distance is about 3 or 4 yards amongst test
balls.
2s~------------------------------
20 --
--,]
Ca:z::cy D Eltance (ave:cage) A :1.90yds B :l.86yds C :183yds
o~--~--~--~--~--~--~--~---~--'11"'""'-'"--,
0 20 40 60 80 100 120 140 160 180 200
Car,y (yd)
Fig.6. Actual side view oftrajectory at Outdoor Test

148 Takahiro Sajima
5 Conclusion
The analysis under initial launch condition of typical average golfer shows that as dimple
depth becomes deeper, CD becomes greater and CL becomes smaller. If dimple depth is
changed in the dimple design, carry distance and trajectory will be changed. From CL analysis,
as dimple depth becomes deeper, trajectory will become lower, and as dimple depth becomes
shallower, trajectory will become higher. lt is important to understand the balance of CD and
CL for distance and feedback the results to dimple design. The tendency of numerical analysis
results is similar to experimental indoor test results. To control carry distance and trajectory
with dimple design, it is important to understand not only tendency of initial launch condition
but also values of CD, CL change all over the trajectory. In this study, as dimple depth
becomes deeper, CD becomes greater from 0yd to about 80yd after fired but the tendency
becomes opposite after that by the identification of USGA software. Results of outdoor
test clarifies the difference of actual carry distance and trajectory while all flying among
different dimple depths. As dimple depth becomes shallower, carry distance becomes greater.
This result is concemed with the difference of CD on flying ball. As dimple depth is shallower,
trajectory before the apex becomes higher. This result is concemed with the difference of CL
on flying ball. The difference of dimple depth is only about 0.01mm among each samplc, but
the little difference makes great difference of carry distance.
References
Aoki, K., Oike, A., Nonaka, M. (2002) The Effect of dimple number on the flying
characteristics of a golf ball. The engineering of Sport 4. (Ed. By S. Uj ihashi and
S.J.Haake), pp330-336.
Ting, L.L. (2005) GOLF BALL AERODYNAMIC BEHA VIOR AS AFFECTED BY THE
DIMPLE DEPTH AND DIMPLE SHAPE CHANGES. The Impact of Technology on
Sport. (Ed. By A.Subic and S.Ujihashi), pp234-239.
Zagarola, M.V., Lieberman, 8., Smits, A.J. ( 1994) An indoor testing Range to measure the
aerodynamic performance of golfballs. Science and Golf II. (Ed. By A.J.Cochran and
M.R.Farrally), 53, pp348-354.
Experimental Verification of Trajectory Analysis of Golf
Ball under Atmospheric Boundary Layer
Takeshi Naruo 1 and Taketo Mizota2
1 Mizuno Corporation, Osaka, Japan, tnaruo@mizuno.co.jp

2 Fukuoka Institute ofTechnology, Fukuoka, Japan
Abstract. Aerodynamic forces and torque acting on the ball were measured under various
flight conditions in a wind tunnel. Using the aerodynamic force coefficients, mathematical
calculation of flight trajectory was made by time integral calculus. Three-dimensional flight
trajectory, changes in velocity as weil as rotation velocity wcre obtained. Furthermore the
logarithmic law was applied to trajectory formation of a golf ball in order to include influence
of atmosphcric boundary layer. Moreover, an experiment was conducted in order to verify the
logarithmic law and the trajectory formulation. Wind vclocity distribution in the vertical direc-
tion was measured. As a result, thc measured rcsult almost matched the logarithmic law. Golf
balls were hit undcr various initial launch conditions by a professional golfcr using various
golf clubs. Initial launch conditions of the golf balls were measured and wind velocity distri-
bution in the golf ball direction was also measured. The golf ball trajcctory under atmosphcric
boundary layer was calculated by using measured initial launch conditions and wind vclocity
distribution. The calculated drop positions by trajectory analysis agreed with the actual meas-
ured results.
1 Introduction
A maximum flight velocity of 80 m/s and a maximum spin rate of 10,000 rev/min
may be reached in a golf drive, but since the velocity and spin rate are high, aerody-
namic measurements under actual conditions are difficult. Furthermore, the golf ball
is strongly affected by the wind during its flight; however, previous studies have not
taken this into account.
A device to rotate an actual golf ball stably at a maximum of 10,000 rev/min in
an air flow of 80m/s was developed. Using the device, drag, lift, and aerodynamic
momentum acting on the ball were measured under real flight conditions. Trajectory
analysis was made by time integral calculus by the measured aerodynamic coeffi-
cients. To consider the effect by the wind, a flight analysis method was taken in
which the atmospheric boundary layer effect was included in the 3-dimensional
flight trajectory equation. According to these measured wind direction and wind
velocity, the flight trajectory analysis with the logarithmic law and a comparison of
analysis results with actual measurements were conducted.
150 Takeshi Naruo and Taketo Mizota
2 Results of Aerodynamic Forces Measurement
The golf ball-rotating device consists of a square frame structure with a D.C. motor
mounted on the top center of the frame. The ball was positioned at the center of the
frame with the steel wire passing through the center of the ball and fastened to the
ball. With this device, it became possible to rotate a golf ball stably at a high spin
rate of 10,000 rev/min. The Skyway SD432 ball (made by Bridgestone) was used for
the experiments. The rotating device is mounted on sliding air bearings and balanced
by pulling the springs in both the positive and negative directions of the force to be
detected.
Figure 1 shows the relationship between spin rate parameter Sp, drag coefficient
C0 and lift coefficient CL. The experiment results of Bearman et al. are shown re-
spectively in solid lines. In this experiment, Sp is made to change over a wide range.
During this time, the Reynolds Number is also changed but together with the results
of C 0 and CL, can be roughly expressed by one curved line.
01, 06
c, c.
□s 05
04
~,._
01!
□J 03
e:8.S1"10' O!!Flin~
:9!93"10• a lW'iSlKWt
De 02 ... IIF9S3~
::«: Re.=1.13xl0 5
J: BF1.U~
0 Re.=l2S ><1os
0 li1Fl.2Sllll0f
□ J. -a~.,,.ani:H....,,.y0.976l Ol --IINIH:i.~C.itl!
00
1 □-2 1□ -i
on
10' 10 W'
Sp
Fig. 1. Aerodynamic coefficients
3 Method of 3-D Trajectory Analysis of Golf Ball
3.1 Method of Trajectory Analysis

The force working on the ball during a certain time t is shown by the equations (1) to
(3).
Fx(t)= - l/2(C 0 (t)cosacosß+CL(t)(sinacos0+cosasinßsin0))pAU(t)2 (1)
Fy(t)= - l/2(C 0 (t)sina+CL(t) cosacosßcos0)pAU(t)2-mg (2)
Fz(t)= -l/2(C0 (t)cosasinß-CL(t) cosacosßsin0)pAU(t)2 (3)

Experimental Verification of Trajectory Analysis of Golf Ball 151
Also, during flight, aerodynamic torque reacts on the golf ball to decay spin
number. The decrease of spin number by aerodynamic torque may be shown by
formula (4).
N(t+.M)=-pAdCm(t)U(t) 2~t/(4nl)N(t) (4)
Where a:Launch angle, ß:Side deviation angle, U:Yelocity, d:Diameter, A:Cross
section area, p:Density ofair, N:Spin rate, !:Moment ofinertia
Figure 2 shows the wind velocity vertical distribution diagram used to embrace
the effects of the atmospheric boundary layer. Here, the logarithmic law is applied.
This logarithmic law closely matches the experimental value in the wind tunnel
boundary layer and it is said that it also coincides with the observation results in
surface boundary layer having smooth surfaces like those of
grasslands.
Where VO is the velocity at a height of I0m above ground,
Vy is the average velocity at a height of Y, V* is the friction
velocity, and t 0 is the shearing stress at ground sur-
'"-,ace, V· = vG
1
' 0 /J. Al so, K 1s
. Karman constant w1t. h an approx1-.
mate value of 0.4, while Y' is the roughness constant, but
here 0.09 is used assuming that the periphery of the course
consists of thick grown grass. H is the average height of
obstructions. To embrace the effect of the wind in trajectory
analysis, the components of the wind are calculated from the
angle A of the velocity components (clockwise direction from
X axis considered as +). The aerodynamic force applied on
the ball is obtained from the relative speed of the ball and air
current, while the position of the ball is obtained from the Fig. 2. Wind velocity
absolute speed. This is repeated until the ball lands on the vertical distribution
ground.
3.2 Confirmation of Logarithmic Law

The distribution of wind direction and wind velocity was measured in the vertical
direction to confirm whether the logarithmic law could be applied. The measure-
ments were made at the Shingu Coast of Fukuoka Prefecture where the ejection
experiment was conducted. Five vane anemometers were installed vertically from the
height of 1.5m to 5.5m at Im intervals. The wind velocity was measured 5 times at
1-minute intervals for 10 minutes and the average value was obtained. The appro-
priateness of the logarithmic law by these measurements was confirmed as shown in
Fig. 3.
Here, U* is the wind velocity measured at a height of 5.5m. By dividing wind
velocity U by U*, the horizontal axis is shown as dimensionless expression. Wind
velocity U was 4 - 6m/s. From the result, the wind velocities from the height of 0m
to 30m that was maximum height of golf ball flight by using a logarithmic law based
on the wind velocity at 5.5m was calculated.
152 Takeshi Naruo and Takcto Mizota
30 r;-:;:::::==:;::::==;---n
20 1
'/4
o L-_.___.____.__,::___.___, L'/
lu
0 02 04 06 08 1 12
Non-dimensional wircl velocity
Fig. 3. Rcsults ofwind velocity vcrtical distribution
4 Ejection Experiment and Trajectory Analysis with Study of the

Wind Effect
4.1 Method of Wind Direction and Wind Velocity Measurement

A golf ball flies the maximum 270m in approximately 6 seconds. lt is therefore
necessary to understand how wind velocity and wind direction vary with time and
position. First, 5 poles were set at every 50m from the reference point to 250m, and
the wind velocity and the direction were checked at 5.5m high. The measurements
were conducted over approx. 3 hours to take 50 sets of data. The difference in wind
velocity depending on position was measured. From the result, it was confirmed that
wind was blowing at a constant wind velocity, even at the point 250m away. The
time change barely affected the wind direction. Moreover, there was no change in
the wind direction by places, and the wind was mostly from south.
From this test result, while the ball flies a maximum of 270m in approx. 6 sec-
onds, there was no difference in the wind velocity and wind direction by time or
location, and it was found that flight trajectory analysis could be made, assuming that
wind velocity and wind direction are constant during the tlight.
4.2 Method of Ejection Experiment

Four poles equipped with vanes anemometers at the height of 5.5m were installed at
every 50m from the hitting position towards the flight direction. The wind direction
and the velocity were measured when the ball passed by the poles, and the average
values measured at four positions were used as the average of wind velocity and
wind direction. The test was made for which the ball direction was changed 30 de-
grees to have cross wind.
In regard to the launching ball, the initial values of launch angle u ,side deviation
angle p, ball velocity U0 , spin rate N 0 , and the rotating axis inclination 6 were meas-
ured by the launch monitor "Pythagoras (made by Mizuno)." In order to check under
Experimental Verification ofTrajectory Analysis ofGolfBall 153
various initial conditions, driver, 5 iron, 7 iron, 9 iron and pitching wedge were used.
All shots were taken by a professional golfer.
4.3 Results of Experiment and Analysis with Study of the Wind Effect
Two examples of the results are shown. The initial conditions of the ball, wind direc-
tion and wind velocity are shown in Table 1. Figure 4 shows the flight trajectory
analysis result of example 1 with the use of a driver. The flight trajectory was calcu-
lated using initial conditions, and the broken line indicates the case wherc the wind
was not considered while the solid line indicates the casc where the measured wind
direction and wind velocity were considered. The wind direction is shown on the
graph by an arrow. The solid circle mark shows the ball landing position measured.
Because of the tail wind, the analysis result which considered the wind showed a Sm
longer flight distance than the result which did not consider the wind. By consider-
ing the wind, it was found that the landing point could be adjusted to the actually
measured landing point.
Figure 5 show the tcst and analysis results of example 2 with the use of a pitching
wedge. The wind is roughly at a right angle to the direction of flight, and compared
with the case where the wind is not considered, it is seen that the ball deviatcs
approx. Sm to the left from the ball direction in case the wind is considercd. More-
over, the flight distance is about 3m longer. Also in this result, the calculated results
in consideration of the wind mect the actually measured value.
Figure 6 shows a comparison of the actual measurements and the calculated val-
ues of flight distance. The measurements with considering the wind are shown by
the solid square, while the results without consideration of the wind are shown by
circle. When the actual measurement and the calculated value coincide, the data is
appeared on y=x. When minimum squarc approximation is conducted on the result in
which wind is not considered, y=0.945x is obtained, but when the wind is consid-
ered, the result becomes y=0.981 x, i.e. it approaches y=x.
Figure 7 shows the comparison between the measurement of side deviation and
calculated value. The measurement with consideration for the wind is indicated by a
solid square mark, while the measurement without consideration for the wind is
indicated by a circle. lt is found that the data of the square approaches y=x more
than that of thc circle. However the slight dispersion is seen. The following reason
can be thought of. The actual wind changes for the approx. 6 seconds during ball
flying but we calculated assuming that there was no change in wind direction and
wind velocity during the flight, which might cause the difference.
Table. 1. Results of ejection experiment
tn/s)
exam :e 1 18 .9 6 l. 12 3 663 2293 157 52

exam :e 2 362 355 -253 328 8020 107 72
154 Takeshi Naruo and Taketo Mizota
Tmveme b)
1
-==-.:+
D Btance b)
0 50 100 150 200 250
Fig. 4. Result of example I with the use of a driver
T:m:veme 6:n)
::1:: ~............... f ..,.... r ...,... l. ------l

t ' ' ..s=;::==;:::::====:::==:.........·
20 ' - - - - - - - - - - - - - - - - - - - ~
Diltance 6:n l
0 20 40 60 80 100
Fig. 5. Result of example 2 with the use of a pitching wedgc
V
300
50
! 200
25
,ll
j 0
ju 100 - 5 -50 -'2"f 25 50 7

~ Measured d&t.a 6:n)
i5
-50
100 200 300

M easuDSd data tn )
Fig. 6. Results of distance X Fig. 7. Results of side dcviation Z
5 Conclusions
By applying the logarithmic law, trajectory formulation of a golf ball under atmos-
pheric boundary layer was shown. And the logarithmic law was verified by measur-
ing of wind velocity distribution to the vertical direction. Comparison of flight dis-
tance was made with numerical analysis results and actual measured values and an
extremely favorable matching of results was seen. Thus the validity of the trajectory
formulation of a golfball under atmospheric boundary layer was verified
References
Bearman P.W. and Harvey J.K. (1976) Golfball aerodynamics. Aeronautical Quarterly, 27,
pp.112-122.
Naruo, T. and Mizota, T. (2005) Trajcctory analysis of golfball under atmospheric boundary
laycr. Thc Impact on Technology on Sport, Mclboume, pp. 253- 260.
Validation of Accelerometers and Gyroscopes to Provide
Real-time Kinematic Data for Golf Analysis
Fitzpatrick, K. and Anderson, R.

University of Limerick, UK, Keith.Fitzpatrick@ul.ie
Abstract. Golfers continually demand more information about their performance

and require immediate feedback on their swing. This research attempts to measure
the pitch (Cl>) and yaw ('I') angle of the clubhead at ball-contact and compare this to
the same angle at shot set-up. Acceleration of the clubhead at ball-contact is also
measured. These three variables have been linked to golfing performance in the
literature. For the research two gyroscopes and one dual axis accelerometer were
incorporated with a putter head. One gyroscope was positioned on the head of the
club to measure deviations occurring between the yaw angle at set-up and the yaw
angle at ball contact (AcJ>). A second gyroscope was positioned to measure deviations
between the pitch angle at shot set-up and the pitch angle at ball contact (A 'I'). Each
gyroscope also provides angular velocities of the club throughout the swing. To
measure the clubhead acceleration (a) an accelerometer was positioned on the
clubhead parallel to the movement. The system and methodology has been verified
and tested for both accuracy and reliability in the field, it is those results that are
presented here. Several testing techniques were used to ensure the data obtained
from the sensors was accurate and compared weil with kinematic data obtained from
a motion analysis system. The reported RMSD error existing between the angle data
obtained from the sensors (A 'I' & A'I') and the motion analysis system was 0.2° for
both the yaw and pitch angle over any two second period (sufficient for a golf
stroke). The accelerometers illustrated potential benefit for use as feedback tool but
proved to be inadequate as a method for deriving velocities or displacements. The
output from the gyroscopes and accelerometer may be displayed in a manor such that
it is a feasible means for giving valuable feedback to a golfer.
1 Introduction
Many golfers in their post win interviews will hail their putting prowess in the tour-
nament while the loser will often lament poor putting as the cause of their less than
heroic performance. With 50% of prescribed shots in the golf allotted to putting it is
critical that, if a golfer is to succeed, this aspect of the game is performed to a higher
degree than any other. The literature emphasises the necessity to maintain the club-
156 Fitzpatrick K and Anderson R.
face perpendicular to the intended direction of ball at point of contact to propel the
ball in the intended direction (Brooks 2002; Pelz 2000; Rosburg 1963; Werner and
Greig 2001 ). lt is also clear from the literature that the golf ball should be putted
with the ball being hit slightly upwards (Pelz 2000; Werner et al. 2001). This up-
ward action allows for the ball to begin rolling quicker and avoid skidding across the
green (Pelz 2000). The literature also states that the clubhead should be accelerated
through the ball at contact (Foston 2001; Newell et al. 2004; Pelz 2000; Rosburg
1963). To provide information on these variables it is vital to measure the pitch (<1>)
and yaw ('I') of the clubhead and the clubhead acceleration. By combining Micro-
Electro-Mechanical Systems (MEMS) technology with golf clubs (specifically gyro-
scopes and accelerometers) golfers can acquire accurate and relevant information
regarding their golf swing.
Gyroscopes have many industrial applications and are used in the automotive
industry and image stabilisation. The basic principle of a rate gyroscope is it oper-
ates on the vibratory principle of a single proof mass suspended by flexures anchored
to the substrate. The flexures serve as the bendable suspension between the proof
mass and the substrate, allowing the mass to oscillate freely in two orthogonal direc-
tions (Acar and Shkel 2001). If the gyroscope is subjected to any angular rotation,
the Coriolis force is induced, the resulting oscillation amplitude is proportional to the
Coriolis force, and thus to the angular velocity to be measured as an electrical output.
Accelerometers are small, inexpensive, have low power requirements and are rela-
tively robust sensors that measure changes in acceleration. They have applications in
industry, in as diverse action as the release of airbags in cars to laptop hard drive
impact and theft protection. Many different types of accelerometer exist, strain
gauge, piezoresistive, piezoelectric and inductive, or in more recent times the move
has been towards capacitive. A centre sensor layer is built for the actual pendulous
mass and is bonded between two other silicon layers that form capacitive sensing
layers. The differential capacitor structure is decoded using phase sensitive demodu-
lators or other approaches on a separate chip. Using this technique, accelerometers
can be built that have excellent performance characteristics (AnalogDevices 2006).
The rate gyroscope is designed to measure the angular velocity of the clubhead
through out the putting stroke. The actual angle, calculated by integrating the angu-
lar velocity, can be used to determine the angle of the clubhead at any particular
phase of the putt and also provide results relating to the yaw and pitch angle at point
of contact. The data obtained from the accelerometer may be used to demonstrate
the acceleration pattern throughout the swing and indicate if the putter head is accel-
erating at ball contact.
One of the most common improvement methods used by golfers is video
based feedback. This method of feedback, based on your swing kinematics, offers
the golfer an opportunity to view his/her swing. lt can be relatively expensive, diffi-
cult to set up and often may involve the use of complicated computer editing. The
software may give an option to view the swing slowed down or compared to another
swing. One of the major draw backs of video analysis is it requires a trained or ex-
perienced person to pin point aspects of the swing that need to be adjusted. lt is
usually the coach that offers this advice. This can be an added expense and requires
the coach to be present, which can often mean sessions must be arranged around the
Validation of Accelerometers and Gyroscopes to Provide Real-time Kinematic Data 157
coach's availability. lt is anticipated to use the data from the MEMS to give mean-
ingful feedback to the golfer. The system in intended to be incorporated in a training
tool. The data from the MEMS could be displayed to the golfer in an attempt to
improve putting performance or identify flaws in the stroke. The data could poten-
tially be displayed in a means that is easily interpreted by the golfer or could be used
as an assistance to the coach. lt could be used to identify flaws in the swing not
visible to the naked eye or so minute that detecting their occurrence may not be pos-
sible.
2 Methods and Results
2.1 Attachment of sensors
The gyroscope chosen was the Analog Devices ADRXS150EB (Analog Devices,
Massachusetts). It was chosen because it was small, light weight, low cost, low
power supply required and could measure up to 150 degrees per second. The volume
of the sensor itself measures less than 0.15cc, weighs less than half a gram (Analog
Devices2004). The gyroscopes were positioned on a perspex board, which could be
easily attached to the putter head (see figure 1). and yaw Gyroscope A was setup to
measure the yaw angle ('P) and gyroscope B was setup to measure pitch angle ((f)).
The accelerometer chosen was the dual-axis Analog Devices ADXL202 (Analog
Devices, Massachusetts). lt was chosen because it was small, light weight, low cost,
low power supply required and could measure up to 2g in two perpendicular direc-
tions. The sensor measures 5mm x 5mm x 2mm. One axis of the accelerometer is
arranged to measure acceleration along the aim line of the putt, at point of contact
(the x axis). The second axis ofthe accelerometer is aligned perpendicular to the aim
line in a lateral direction the z axis; see Figure 2.
X axis 111d aim line of utt
Fig 1: The MEMS attached to the Perspex board that can slotted into a receiver on
the putter head (Right) and illustration of the axes that the dual axis accelerometer
measures (Left)
158 Fitzpatrick K and Anderson R.
A perspex board was attached to the sensors and could be fixed to the head of the
putter. The junction box on the perspex board slotted into a receiver that transmitted
the reading from the sensors along a series of wires up the shaft of putter, as shown
in figure 2. The sensor array was then connected to a computer (Powerlab, New
Zealand) to display and record the output from the sensors.
2.2 Compatibility of the Gyroscopes

Prior to the system being used in the field the limitations of its accuracy and reliabil-
ity must be assessed. The accuracy and reliability of the actual gyroscopes is re-
ported in the manufacturers data sheet (ADRXS l 50EB, Analog Devices, Massachu-
setts). To assess the accuracy of the gyroscopes in conjunction with the proposed
system a rig was designed to securely hold the golf club. The rig allowed the swing
rate of the club to be controlled and also restricted movement of the club-head to one
plane. Therefore data for both pitch angle or yaw angle could be obtained. Data was
acquired from the sensor array and also from a Peak Motus motion analysis system
(Peak Technologies, Colorado, USA) capturing kinematic data at 50Hz.
Retro reflective markers were placed at either side of the clubface, represent-
ing the putter face and, post calculation, the yaw angle (the movement specifically
measured by gyroscope A). Retro reflective markers were also placed at either end
of the shaft representing the pitch angle post-calculation (the movement specifically
measured by gyroscope 8),. When obtaining data related to the yaw angle the cam-
era was placed above the putter pointing towards the floor (i.e focused on the x-z
plane). This enabled the motion analysis system to calculate the angle between the
clubface and the x-y plane projected onto x-z plane (i.e the yaw angle). The second
sctup, with the camera focused onto the x-y plane enable the acquisition of the the
angle between the shaft and the x-z plane projected onto the x-y plane (i.e the pitch
angle).
Initial pilot work was undertaken and compared the data obtained from the
gyroscopes to the data obtained from the motion analysis. The movement of the club
was recorded using a 50Hz S-VHS video camera (Panasonic AGDP800, Matsushita
Electrical Industrial, Japan). The video images were digitised using motion analysis
software (Peak Motus 2000). The resultant coordinate data was filtered using a
Jackson Knee optimised Butterworth filter. During this data was acquired from the
sensors via a PowerLab data acquisition system (AD Instruments, New Zealand).
Pilot work and subsequent analysis of the angular velocity data identified that the
gyroscopes reached a plateau at certain stages throughout the movement. This can
be attributed to the movement velocity being greater than the range of the gyroscope
(150 degs.s" 1). Integration ofangular velocity curves, which exceeded this maximum
of 150 degs.s· 1 led to inaccuracies in the angle data.
Further trials were conducted at an angular velocity below 150 degs.s" 1 (more
representative of a 1Oft golf putt) to eliminate these shortcomings of the sensor. The
RMSD, between the integrated gyroscope data and the motion analysis angle values
was reported at 2°. At closer inspection of the angle data it was obvious that the
largest variance between the sensor data and motion analysis data occurs at the tum-
ing points of the curve. This notable divergence in two lines was inflating the
Validation of Accelerometcrs and Gyroscopes to Providc Real-time Kinematic Data 159
RMSD values artificially. This divergencc was decreased by fine-tuning the sensi-
tivity values of the gyroscopes within the ± l 0% range stated on the sensor data sheet
(stated sensitivity is 12.5 mVf/s ± 10%). For the two gyroscopes used in this re-
search the optimal sensitivity was calculated as being 11.9 mV fls. Final tests of
compatibility of the gyroscopes were completed over a number of durations to exam-
ine the effect, if any, of drift on the accuracy of the data. The results (illustrated in
table l) identify that the RMSD existing between the motion analysis data and the
sensor data was minimal. At the 2 s range (the typical length of a golf putt) the
RMSD was 0.21 °. In summary thc gyroscopes can offer accurate angular velocity
data when embedded in the club, this data can subsequently be used in a feedback
system.
Duration of
40sec 30sec 20sec 10sec 5sec 2sec
RMSD
Magnitude
0.5 l 0 0.53° 0.55° 0.52° 0.47° 0.21 °
ofRMSD
Table 1 RMSD values at a scnsitivity of 11.9 mV /s.
2.3 Compatibility of the Accelerometers
As thc system does not use absolute acceleration values, just the relative magnitude
and the time based curve that are created throughout a movement pattem, a true
calibration of the system is not required. The gravity based calibration procedure for
the accelerometers is outlined in the manufacturers data sheet and will not be dis-
cussed here (Analog, 2004). The accelerometers could adequatcly represent accel-
eration and therefore the change in acceleration pattems pre and post ball contact.
3 Discussion
In conclusion the sensor array presented here has been shown to be both accurate and
reliable and can offer the information required for a feedback mechanism for golf
putting. The yaw angle is crucial in insuring that the ball is putted in the intended
direction. Any deviation in the yaw angle results in ball being struck offline and the
likelihood of a missed putt is increased. The pitch angle insures that the control over
the putt is maximised. By striking the ball in a slightly upward direction, topspin is
created and the ball tends to skid less. Integration of the gyroscope's angular vcloc-
ity output, gives sufficient accuracy to determinc the yaw and pitch angle. This
information could be used as a valuable tool for coaches and golfers.
Accelcrometers can be of benefit to the putting stroke. They are capable of
indicating the presence of acceleration in the putt i.e. changes in acceleration can be
demonstrated. Therefore it is possible to demonstrate if the golfer accelerated at
point of contact with the ball or not. At this point quantifying the acceleration is not
possible, but accelerometers can be used to demonstrate the presence of acceleration.
Also it is feasible to use the acceleration data to represent a movement pattem.
160 Fitzpatrick K and Andcrson R.
Accelerometers have been shown to be a reliable tool in trying to replicate a move-

ment pattem (Foerster et al. 1999). This can be utilised in developing an ideal
movement pattem and then attempts could be made to replicate this. This in turn
could lead to a more consistent putting stroke. Initial results utilising the system
proposed herewith have proved positive on improving the putting performance of
novice golfers. However statistical data was not available at time of going to press.
Gyroscopes have been proven as a means of giving accurate and meaningful kine-
matic data on a golf putt. Accelerometers can be used to give meaningful indication
of kinematic movement. The ability to train with a club that can assess movement
pattems and movement pattems in real-time, is an exciting prospect.
Acknowledgements
Funding for this research is provided by the Irish Research Council for Science,
Engineering and Technology: funded by the National Development Plan oflreland.
References
Acar, C., and Shkel, A. (2001 ). A dcsign approach for robustncss of rate gyroscopes. Model-
ing and simulation of microsystems, 80-83.
AnalogDevices (2004). +/.300°/, single chip yaw rate gyro with signal conditioning, Vol. 2005.
Analog Devices.
AnalogDevices (2006). Using iMEMS Accelerometers in Instrumentation Applications. Ana-
log Devices.
Brooks, R. J. (2002). ls it a pendulum, is it a plane? - Mathematical models of putting, Science
and Golf IV Proceedings from the World Scientific Congress ofGolf, pp. 127-141.
Foerster, F., Smeja, M., and Fahrenberg, J. (1999). Detection ofposturc and motion by accel-
erometry: a validation study in ambulatory monitoring. Computers in Human Behaviour
15, 571-583.
Foston, P. (2001 ). The completc encyclopidia of golf tcchniques. Running press book publish-
ers, Pennsylvania.
Newell, S., Foston, P., and Atha, A. (2004). The complete golfer, 3 ed. Annes publishing
limited, London.
Pelz, D. (2000). Dave Pelz's Putting Bible. Doubleday, New York.
Rosburg, B. (1963). The putter book, Vol. 3. Nicholas Kaye Ltd, London.
Werner, F. D., and Greig, R. C. (2001 ). Setter Golf from New Research. Origin Inc, Jackson.
Investigation of Wrist Release During the Golf Swing
by Using a Golf Swing Robot
Yohei Hoshino 1, Yukinori Kobayashi2 and Soichiro Suzuki 3
Hokkaido University, Graduale School of Engineering, hoshinoy@eng.hokudai.ac.jp

Hokkaido University, Graduale School of Engineering.kobay@eng.hokudai.ac.jp
Kitami Institute of Technology. Faculty of Engineering. zuki@mail.kitami-it.ac.jp
Abstract. In this study, skill of the wrist release during golf swing is investigated by using a
golf swing robot. A torque plan of the shoulder joint of the robot is assumed as a two-steps
modulation torque that is a simplification of the acceleration torque induced by the weight
shift. rotation of torso and the shoulder of a golfer. Relationship between the torque plan and
the skill of the wrist release, such as natural release and late hitting. is investigated by using
the golf swing robot.
1 Introduction
Characteristics of swing style of professional and expert golfers appear in the wrist
release as a 'natural' or 'late' release. There have been studies on the effect of de-
layed wrist turn (Jorgensen 1970) and the wrist turn in the down swing by the three-
dimensional nature of the forces and torques applied to the club (Vaughan 1981;
Neal and Wilson 1985). The effect of 'natural release' and 'late hit' was investigated
numerically by using a two dimensional rigid double pendulum model (Pickering
and Vickers 1999). The skill of the wrist release in the down swing was analyzed by
using a three-dimensional dynamic model based on the double pendulum including
the effect of supination of a forearm and three-dimensional shaft vibration (Suzuki,
Haake and Heller 2005).
In this study. skill of the wrist release during golf swing is investigated by using a
golf swing robot (Hoshino, Kobayashi and Yamada 2005). The golf swing robot
consists of wrist and shoulder joints, two rigid links and a golf club. Although the
golf robot does not have a cubital joint and cannot rotate the golf shaft around the
axis of the shaft, the robot has a hook mechanism to lock the wrist joint. The hook
mechanism is important to evaluate the relationship between the wrist release and
torque plans for the shpulder joint. The swing robot pulls up the golf club (back-
swing) and then swings it in a swing plane. The backswing action is performed by a
digital servo controller. Feed-forward torque is applied only to the shoulder joint
during the downswing. A linear quadratic regulator (LQR) in consideration of the
vibration of the club shaft is employed to stop the robot during the follow-through
action after the impact for decreasing the degree of shock due to abrupt slowdown.
Since it is difficult to measure all of the state variables, a state observer is employed
for the state feedback control system.
162 Yohei Hoshino et al.
The torque plan of the shoulder joint is assumed as a two-steps modulation torque
that is a simplification of the acceleration torque that is induced by the weight shift,
rotation of torso and the shoulder of a golfer (Suzuki, Haake and Heller 2005). Rela-
tionship between the torque plan and the skill of the wrist release, such as natural
release and late hitting, is investigated by using the golf swing robot.
2 Equation of Motion of the Golf Swing Robot
Fig. 1. Swing plane of the golf swing
The swing robot and the system of orthogonal axes (x, y, z) are shown in Fig. 1,
and the robot swings a golf club in the swing plane o-xy. The angle between the
direction of gravity and the x-axis is a.. The robot has two joints and two rigid links
that are modeled on the arm of the golfer and the grip of the club. The simplified
governing equation of the system is needed to construct the control system of the
robot. Tue golf club is modeled as a flexible beam in consideration of bending and
torsional stiffness and in consideration of eccentricity of the center of gravity of the
club head on the shaft axis. Properties of the club are as follows: a uniform cross-
sectional area A 3, diameter d 3, length L 3 , Young's modulus E 3 , area moment of iner-
tia h, mass per unit volume p 3 and eccentricity of the center of gravity of the club
head L4 • Figure 2 shows the coordinate system in the xy-plane. Angles 01 and e are
the absolute rotating angles of the first joint and the second joint, respectively. The
club shaft is assumed homogeneous and initially straight along the ~raxis shown in
Fig. 2. The masses of the first link, the second link and the clubhead mass, respec-
tively, are m 1, m2, and m4, and the mass moments of inertia of the first link and the
second link are J 1 and J2, respectively. The mass moments of inertia of the clubhead
mass around the axes ~4, 174 and ( 4 are J 4,, J4q and J 41;, respectively. Tables 1 to 4
show parameters of the swing robot and golf club. In these tables, parameters L, ß
and c denote the length of link, the position of the center of gravity, the coefficient of
friction and Young' s modulus of the golf club, respectively. Deflection of the flexi-
ble link along the 17 2 axis is denoted by vq and torsion angle of the flexible link is rp.
A mathematical model of the golf robot is derived by Hamilton's principle in con-
sideration of bending and torsional stiffness and in consideration of eccentricity of
the center of gravity of the club head on the shaft axis. Natural frequencies and the
Investigation of W rist Release During the Golf Swing 163
Fig. 2. Coordinate system of the swing robot
Table 1. Parameters of the first link Table 2. Parameters of the second link
m 1 [kg] 2.92 m2 [kg] 0.650
J 1 [kg·m 2] 0.201 h [kg·m 2] 2.0xl0- 4
L1 Im] 0.310 L 2 [m] 0.050
ß1 0.774 c 2 [N · m · s/rad] 0. 102
c 1 [N·m·s/rad] 0.151
Table 3. Parameters of the club shaft Table 4. Parameters of the clubhead

p 3 [kg/m ] 1.50x 10 m4 [kg] 0.210
A 3 [m 2 ] 5.66xl0-5 J~ [kg·m 2] 4.5xl0- 4
/3 [m 4 ] 3.18xl0- 10 h,[~·~J l.~I~
L3 [m] 0.83 J4dkg·m 2J 2.0xl0- 4
100 L4[m] 4.00xl0- 2
vibration modes are obtained by numerical and experimental modal analyses. Using
a time function q(t) =sin OJI , we assume the bending displacement v, and the tor-
sional angle <p to be
= =
v,(i;,t) = Lq,(t)F,,;(<;), <p(_q,t) =Lq;(t)Fl>(i;). (1)
i=l i=I
where I',1; and I',p;, respectively, are the i-th mode shapes of bending displacement and
torsional angle, and w is the circular frequency. An approximate system of finite
dimensional equations is obtained using a finite number of modes in Eq. ( 1).
Considering first two modes, the non-linear equations of motion are obtained as
J(x)i+Dx + Kx +h(x,x)+ g(x) = pu + 11; (2)
where
x=[01 02 ql q2Y,u=k1 r2r,p=[l2x2 O2xif,
and 11; is the generalized force vector applied to the center of gravity of the club
head. J, D and K are inertia, damping and rigidity matrices, respectively, and h, g
and u are nonlinear force, gravity and input vectors, respectively. The nonlinear
force vector h includes these constraint forces.
3 Controller Design
The swing robot pulls up the golf club (backswing) and then swings it in a swing
plane. The backswing action is performed by a digital servo controller. Feed-forward
torque is applied only to the shoulder joint during the downswing. A linear quadratic
regulator (LQR) in consideration of the vibration of the club shaft is employed to
stop the robot during the follow-through action after the impact for decreasing the
degree of shock due to abrupt slowdown. Since it is difficult to measure all of the
state variables, a state observer is employed for the state feedback control system.
lgnoring the gravity in Eq. (2) and linearizing Eq. (2) around x = 0, we can obtain
the linear equations of motion as
Jx+Dx+Kx = pu+d (3)
where d is the equivalent disturbance vector, and K is the rigidity matrix . Equation
(3) can be rewritten as the state variable equation,
X= AX +Bu+Ed (4)
where
The discrete-time system of Eq. (4) is given as

xk+I =AdXk +Bduk +Eßk (5)
Control model of the backswing action is constructed on the system of Eq. (5). Here,
the output equation is
Y=[OI 02l1=CPX,Cp=[I2x2 02x2 :o2x2 02x2l (6)
Then we employ a type I digital servo system of the rotating angles 01 and (/2 for the
system of Eqs. (5) and (6). The state variables except for 01 and (/2 are estimated by a
state observer based on a type I digital servo system. The structure of this observer is
similar to that of a steady-state disturbance observer (Hoshino et al. 2005). The
robot also has a hook mechanism to lock the wrist joint for emulating the wrist cock-
ing action of a golfer (Suzuki and Inooka 1998). The hook can achieve wrist cock
action in the downswing.
4 Torque Plan and Wrist Release
Difference of swing motions between experts and beginners remarkably appears in

the wrist release. lt is important to utilize strain energy of a shaft and potential en-
ergy of an arm for fast head speed with natural and late wrist release in an actual
swing motion. Delay of the zero-crossing point where the shaft is straight and the
wrists then release is a way to achieve the fast head speed. We can evaluate the pro-
ficiency degree of the swing motion by using an efficiency index rthat is calculated
by ratio of kinetic energy of the head to work of the shoulder (Suzuki, Haake and
Heller 2005).
Investigation of Wrist Release During the Golf Swing 165
r r
z
z
<!) <!)
::,
::,
...
Tor,o
...0O" ""
;;
O'
~. Ol""
0
E-
. ' E-
0 Input time [sec! 0 Input time Jsec]

(a) Composite acceleration torque (b) Simplified two-steps modulation torque
of whole body motion
Fig. 3. Acceleration of the shoulder joint
TG Tc
~
~
<!)
::,
...O"'0 :;;""
Ol
E-
0 Input time Isec 1 0 Input time [sec]

(a) Triangular torque (b) Trapezoidal torque
Fig. 4. Triangular torque and trapezoidal torque
In order to delay the zero-crossing point, torque input of the shoulder joint has to
increase during the acceleration of the arm. The acceleration torque of the experts is
assumed to consist of weight shift, rotation of a torso and shoulder as shown in Fig.
3(a). Then, the acceleration torque was simplified to two-steps modulation torque as
shown in Fig. 3(b) to investigate the relationship between the zero-crossing point and
torque patterns. Validity of this torque function for delaying zero-crossing point was
examined by comparing with simple trapezoidal and triangular function as shown in
Fig. 4(a) and (b), respectively.
Figure 5 shows the simulation results in the case of the golf swing robot. Figure
5(a) shows the relationship between a shoulder joint angle at the wrist release 0,
fdeg) and QMAX [N-m] for three types ofthe function under the same condition ofthe
shoulder work. Figure 5(b) shows relationship between r and head speed at impact
Vh [m/sec]. We can know the position of the arm at the zero-crossing point from 0"
because the wrist is always released at the zero-crossing point by 'natural release'.
Although values of QMAX are not so !arge because of the restriction of the maximum
torque of the robot, results of two-steps modulation torque give tatest wrist release
for all values of QMAX• Therefore, the acceleration torque with whole body motion
can delay the zero-crossing point and achieve fast head speed efficiently.
Figure 6 shows the experimental results in the case of the golf swing robot. The
qualitative tendencies of the experimental results are agree with those of the simula-
tion. Therefore, the golf swing robot can emulate the human skill of 'natural release'
and 'late hitting' by actuating with two-steps modulation torque.
-30
-4 0
l -·-•···•· ......•
1
... ■
-----■-............
-■-
_,._ Triangular
2 stcps modulation
0--
-■-■-■-■-■
-50 ............. ~ 3.1
öii ....... •,.-o-Tmpe?Ddal
.5
~ -60 ;,-, ...................
g 2.9 .........
"'-70 -O-Tmpczoidal "
"ü
0-- ~!---..
-80 -90[degJ .Q·O·O·O·O·O·O
-=
Li3 2.7
-90 --------.l)-zyO.o-0•0:Q: ________ ~
0--0
-1000·0·0·0·0·0·0 J 2.5 ~ - ~ - - ~ - ~ - - ~ - - - - '
10 12 14 16 18 20 13.0 13.2 13.4 13.6 U.8 14.0
Q."' [N·m[ V, Jm/secJ
(a) Comparison of a shoulder joint (b) Comparison of efficiency index
angle at the wrist release
Fig. 5. Results of the wrist release and efficiency index in the case of golf swing robot
j
... .
2.0 ~ - ~ - - ~ - ~ - - ~ - ~
1.5 ,.:. ___ ., __ ::::::i:.:.·.::.:=:::~:.:::.:.:.!:.:.·

;,-,
g 1.0
_8 0 ···-■·-- 2 stcps modulation ;g" ---·•--- 2 steps ITKx.Julati(Xl

_90 -90 [degl -------=--~~--T_"'.'.,"~':'.____ _ lii O·5 - · - · · Trianh'lliar
-100 ~ - ~ - - ~ - ~ - - ~ - ~ 0.0 ~ - ~ - - ~ - ~ - - ~ - ~
10 12 14 16 18 20 13 14 15 16 17 18
Q."' [N·mJ V, [m/secJ
(a) Comparison of a shoulder joint (b) Comparison of efficiency index

angle at the wri st release
Fig. 6. Experimental results of the wrist release and efficiency index
References
Hoshino, Y., Kobayashi, Y. and Yamada, G. (2005) Vibration control using a state observer
that considers disturbances of a golf swing robot. JSME International Journal, 48, pp. 60-
69.
Jorgensen, T. ( 1970) On the dynamics of the swing of a golf club. American Journal of Phys-
ics, 38, pp.644-651.
Neal, R.J. and Wilson, B.D. (1985) 3D kinematics and kinetics of the golf swing. lnterna-
tiona/Journal ofSports Biomechanics, 1, pp. 221-232.
Pickering, W.M. and Vickers, G.T. ( 1999) On the double pendulum model of the golf swing.
Sports Engineering, 2, pp. 161-172.
Suzuki, S. and Inooka, H. ( 1998) A new golf-swing robot model utilizing shaft elasticity.
Journal of Sound and Vibration, 217, pp.17-31.
Suzuki, S. ( 1999) Torque planning for a new golf swing-robot emulating golfer's skill. Proc.
oftheJSME Symposium, No.99-41 (in Japanese), pp.44-47.
Suzuki, S., Haake, S. and Heller, B. (2005) Skill analysis of the wrist release in golf swing to
utilize shaft elasticity. The Impact of Technology on Sport (edited by Subic and Ujihashi),
Australian Sports Technology Alliance Pty Ltd, pp.188-193.
Vaughan, C.L. (1981) A three-dimensional analysis of the forces and torques applied by a
golfer during the downswing, Biomechanics Vll-B (edited by Morecki et al.), Polish Scien-
tific Publishers, Warsaw, pp. 325-331.
Segmental Sequencing of Kinetic Energy in the Golf Swing
Brady C. Anderson 1, Ian C. Wright 2 and Darren J. Stefanyshyn 1

1 University ofCalgary, Canada, brady@kin.ucalgary.ca
2 TaylorMade-adidas GolfCompany, USA
Abstract. This paper investigates the sequential transfer of kinetic energy from proximal to distal
scgments in the golf swing. Forty five male, scrateh golfcrs performed drivcr swings before a
multi-eamera motion analysis system. Position data recorded from the trials were used to scale and
update the movement of a kinematic golf model. The rigid bodies of the golf mo<lel were assigne<l
basic, homogenous, geometries with known inertial properties. The rigid bodies of the human-golf
club system were subdivided into 4 linked segments. Translational, remote rotational and loeal
rotational kinetie energy was calculated for each segment throughout the swing. lt was found that
the peak magnitudes of total kinetic energy increased sequentially from proximal tu distal
segments, while the timing ofthese peaks did not follow a sequential pattem. This paper describes
the use of a novel method of kinetic energy calculation in the golf swing, and its application in club
head speed generation.
1 Introduction
In order to generate a fast swing, a golfer interacts with the ground via his feet. Through
. a series of hip, torso and arm rotations, he generates both a translational and rotational
impulse at his hands on the club grip. Many studies have looked at how club head speed
at impact is generated by the kinetics of different segments of the golfer (Jorgensen
1994; Sprigings, Marshall, Elliot and Jennings 1994; Nesbit 2005). To date, there is still
no explanation available as to the transfer of speed between segments.
The summation of speed principal (Bunn, 1972; Putnam, 1993) has been an
influential concept in the literature, as a means of explaining how both angular and
translational velocities can be transferred in a linked system. This principal states that the
system will achieve an optimum end point velocity if the link velocities peak
sequentially. A link will begin motion at the instant its proximal neighbor achieves a
peak velocity, and so forth along the line. The distal end point of each link should then
incrcasc scqucntially along the systcm.
In opposition to sequential peaking of velocity, Van Gheluwe and Hebbelinck
( 1985) have proposed the principal of optimal coordination of partial momenta (Putnam,
1993). This states that optimal distal end velocity of a linked system is reached when the
angular velocities of the links peak simultaneously. Putnam ( 1993), has argued that this
168 Brady C. Anderson, Jan C. Wright and Darren J. Stefanyshyn
principal may not work in kinetic applications since the inertial propertics of the links
cause interactions which work against simultaneous peaking.
At present, it is not known how different segments contribute to impact speed in
golf. The purpose of this investigation was to explore the transfer of speed from the feet
to the club head. Kinetic Energy (KE) takcs the incrtial propcrtics of the links in account,
and is independent of direction, so there was no need to place constraints on the
kinematics of segmcnts. These motion constraints have been shown to lead to errors in
calculations of end point velocity; ic. Constraining long-axis rotation (Marshall and
Elliot, 1999).
2 Methods
Forty five male scratch golfcrs wcrc uscd as subjccts in this study. Kinematic data were
collected on the MATT motion analysis system (TaylorMade-adidas Golf, Carlsbad CA).
Subject motion was tracked using retro-reflective markers, placed on anatomical
landmarks and fixed to the subjects clothing. Position data from the markers were
collected at 110 Hz and used to scalc and update a kinematic golf modcl designed by
Motion Reality Inc. (Marietta, GA).
The volume of the kinematic model was scaled to the size of each human golfer.
Truncated cones and ellipsoids were used as the general geometric shapes to be fitted to
thc golf model rigid bodies. These shapes were modcled to be homogenous, and of
known volume and inertial properties. An example of the geometric shapes applied to
model the human golfer can bc sccn in figurc 1. Thc shapcs were given a density of 1.03
x 10 3 kg/m 3. as reported to be that of the human body, by Clauser, McConville and
Young ( 1969).
Body mass was seif reported by a subset of 21 male subjects. The seif reported mass
was on average 2.1 % lower than thc mass calculated by the model with a standard
deviation of 8.2% body mass. In order to ensure that our calculated mass was not
statistically different from our reported whole body mass, a studcnt t-test was employed
to test if the difference found was significantly different than zero. A p-value of 0.88 I 9
showed that the measured and modeled body masses were not significantly different.
The kinematic golf model was split into four segmcnts, to explore thc possibility of
proximal to distal pcak energy sequencing. Each of the segments containcd a number of
rigid bodies. Thc Hips segment contained the fcct, thc lcgs, and thc pclvis. Thc Torso
segment included seven independent scctions of the thorax, two sections of the neck, and
thc head. The Arms segment consisted of the shoulders, arms, hands, and fingers. The
club scgmcnt consistcd of a grip, 8 shaft sections, and the club head. The mass and
moments ofinertia ofthese rigid club bodies were measured directly previously.
Fig. 1. Example of the rigid body geometries applied to the kinematic golf model.
The total KE for each segment was calculated as the sum of the rotational and
translational kinetic energies of those segments ( eq.1 ). The translational KE was found
about the center of mass of each segment as a whole (eq.2), where the mass of the
segment was found as the sum of masses of the bodies comprising the segment (eq.3).
The velocity of the center of mass of each segment was calculated as the time derivative
of the position vector of the segment centroid (eq.4). This position vector is a mass
normalized sum ofthe positions of all the segment's rigid bodies (eq.5).
KEtotal = KEtranslational + KErotational {l) KEtranslational = -1 mcm ~-m

- '
- (2)
2
= Im;
n
V = d~-m (4)
msegment
i=I
(3)
cm dt
(5)
mcm
Rotational KE was calculated as the sum of local rotation of the rigid bodies
about their own center of mass, and remote rotation of the rigid bodies about each
segment center (eq.6).
KErotational = KEbody rotation + KEscgment rotation (6) KEbody rotation =-21 ol -J{ij
1 1 1
(7)
w/(3,2)]
tiJ; = [ w/(1,3) (8) m"- = dR; R. 1 (9)
' dt '
w/(2,1)
Local rotational KE of the individual bodies was calculating using the angular
velocity vectors and moment of inertia tensors for each body (eq.7). The inertial
properties of each body were derived from their geometry and the angular velocities of
170 Brady C. Anderson, !an C. Wright and Darren J. Stefanyshyn
each body were obtained from the skew-symmetric angular velocity matrix (eq.8). This
matrix was calculated by finding the first derivative of the 3x3 rotation transform at each
time step and multiplying by the transpose ofthat rotation transform (eq.9).
KEsegment rotation = I
i=l
..!._ m/7;t.tan 2
2
( J0) V6 , = dP; - dPcm
dt dt
(\ l)
V;t.tan =V;t.-V;t.rad (12) V;t.rad =f;u(f;u • V;t.) (13)

Remote segment rotation energy was found by calculating the tangential
velocity of each body i about the segment center of mass. The difference velocity vector
V"' is the difference between the velocities of each body minus the velocity of the
segment center of mass (eq.11 ). A radius vector r, from the segment center of mass
position to the body center of mass position was found at each time interval. The
component of the difference velocity along the radius vector ( VMact) was found as the
projection of the velocity on the radius (eq.13). The tangential component was then
found by vector subtraction (eq.12). This difference velocity tangential component vector
was then used to calculate remote segment rotational KE (eq.10).
Total KE traces were filtered using a 4th order low pass Wavelet filter at a cutoff
of20 Hz. Statistics were done using a multi-factorial one-way ANOVA test at a
significance level a=.05, with a Tukey post hoc analysis.
3 Results
Figure 2 shows total KE calculations for each segment during a single swing trial. Ball
contact is shown at time=0 sec. For this trial, the magnitude of the KE peaks increase
starting proximally at the Hips and moving distally to the Club. The Arms, Torso and
Hips segments all peaked at approximately the same time, prior to ball contact. The Club
segment KE peaked at impact. The total KE for the club segment peaked at roughly 2501 .
,...
s:
...
...
,
'°i
,.
Fig. 2. Total KE for segments throughout the golf swing for a single player during one trial. Time
= 0 sec. represents ball contact. KE is shown in joules.
3.1 Magnitude of Kinetic Energy Peaks

For each player, the segment KE peak magnitudes were expressed relative to the Club
KE peak. Figure 3 shows the means of these peaks and their standard deviations. Each
segmcnt's KE peak was significantly different from the othcrs. Thc magnitudc ofthc KE
peaks increased sequentially from the proximal to the distal segments.
0.8
0.6
0.4
02
*
0 ,..__.__L-_.......___ _.__ ___._.......___ _ _ __
Hps Torso CII.D
Relative magnitudes oftotal KE peaks for golfer segments throughout a golf swing as
Fig. 3.
compared to the golf club peak KE. Error bars reprcsent thc standard deviation. Thc astcrisk
symbolizes statistical difference from all other segments.
3.2 Sequential Timing of Kinetic Energy Peaks

The timing of the KE peaks were analyzed in a fashion similar to the peak magnitudes.
Timing was expressed relative to total Movement Time (MT) where O represents the top
of backswing and I represents ball impact. Figure 4 shows the group mean relative
timing for each segment KE peak along with the corresponding standard deviations. Only
the Club segment timing was statistically different from the other three. The three human
based segments peaked at approximately the same time while the Club segment, being
extemal to the player, peaked later in the swing.
12 , - - - - - - - - - - - - - - - - - - - - - - ,
08 *
0.4
0 _ ......._ _.__ _......._ ........_ _....__....__ _L-----''--~
Hlps To rso Arms Club

Fig. 4. Relative timing oftotal KE peaks for golfer segments as a function ofMovement Time. MT
at O represents the top of backswing, while MT at 1 represents ball contact. Error bars represent the
standard deviation. The asterisk symbolizes statistical difference from the other segments.
172 Brady C. Anderson, Ian C. Wright and Darren J. Stefanyshyn
4 Discussion
Peak magnitudes of total KE increased sequentially from proximal segments to distal
segments in the golf swing. This finding supports the principal of summation of speed.
The timing of these peaks however, did not occur sequentially. At approximately 80%
MT during the downswing, that the segments belonging to the golfers' bodies peaked
simultaneously. Shortly after, the Club segment reached a maximum KE. One would
expect that if energy were being transfered sequentially along body scgements, distal
links should experience an increase in kinetic energy at the expense of their proximal
neighbor. This decrease in proximal kinetic energy only seemed to happen when the
Club segment kinetic energy reached a maximal peak.
5 Conclusion
Neither the summation of speed principal nor the principal of optimal coordination of
partial momenta can fully explain the pattem of kinetic energy transfer observed in this
investigation. lt was found that the magnitude of kinetic energy increascs proximally to
distally, but there is an instant during the downswing when the kinetic energy of the
segments in the golfers body, tend to peak simultaneously. This leads to a later peak in
KE ofthe most distal link, the golfclub, at ball impact.
Acknowledgements
Wc would likc to thank thc TaylorMadc-adidas Golf Company and Biomcchanigg
Research Inc. for their support of this paper.
References
Bunn, J.W. (1972). Scientific Principals o(Coaching. Prentiee-Hall Inc., Eaglcwood Cliffs, NJ.
Clauser, C. E., McConville, J. T., & Young, J. W. (1969). Wcight, volume, and center ofmass of
segments of the human body. AMRL Technical Report (Rep. No. AD 710 622). Wright-
Patterson Air Force Base, Ohio: Aerospace National Research Laboratory.
Jorgensen, T. P. (1994). The Physics ofGolf. 2"'1 ed. AIP Press, New York, NY
Marshall, R. N. & Elliot, B. C. (2000). Long-axis Rotation: The Missing Link in Proximal-to-distal
Segmental Sequencing. Journal ofSports Sciences 18, 247-254.
Nesbit, S. M. (2005). A Three Dimensional Kinematic and Kinetic Study of the Golf Swing.
Journal of Sports Sciencc and Mcdicine 4, 499-519.
Putnam, C. A. ( 1993 ). Sequential motions of body segmcnts in striking and throwing skills:
Descriptions and explanations. Journal of Biomechanics 26, Suppl. 1, 125-135
Sprigings, E., Marshall, R., Elliot, 8. & Jennings, L. (1994). A three- dimensional kinematic
method for determing the effectivencss of arm segment rotations in producing racquet-head
spced. Journal of Biomcchanics 27, 245-254.
Van Gheluwe, 8. and Hehhclinck, M. ( 1985). Thc kinematics ofthe service movcment in tennis: A
three-dimensional cinematographical approach. Biomechanics IX-B., (Edited by Winter, D.A.,
Norman, R.W., Wells, R.P. Hayes, K.C. and Patla, A.E.). Human Kinetics Publishers lnc.,
Champaign, IL. 521-526
5 Gymnastics
Synopsis of Current Developments: Gymnastics
David G Kerwin
Cardiff School of Sport, UWIC, UK, dkerwin@uwic.ac.uk
Background to Recent Research in Gymnastics
Modeling has dominated research in gymnastics in the recent past. Peter Brugge-
mann' s Cologne (Germany) based group has concentrated on energetics in high bar
circling and release and regrasp skills whilst Fred Yeadon's Loughborough (UK)
group has focussed on momentum rings circling and in the preparation for high bar
dismounts (eg Hiley and Yeadon 2005). Another theme has been 'control' with re-
cent contributions from Jill McNitt Gray's group in Califomia (USA) focussing on
strategies for landing from rotating jumps whilst Fred Yeadon and colleagues have
addressed questions of control in twisting somersaults and hand balancing. This
group have also studied vaulting and tumbling using forward dynamics models to
examine both technique and control.
A common theme running through many of the recent studies has been the inter-
action between the gymnast and the equipment and inevitably therefore heightens the
impact that engineering can have on research in this sport. Engineering has also been
central to many of the gymnastics studies through the application of control theory to
the study ofhuman movements and the development offorward dynamics modelling
with Dads™ and Autolev III™ being the preferred options.
Papers Featured at the Conference
Two contributions in the current conference draw on the themes of control theory
and forward dynamics modeling. Wendy Kimme! and Mont Hubbard, reporting on
balancing on a horizontally compliant surface (trampoline), have developed single
and double segment models and showed that knowledge of foot position is not im-
portant in the feedback control system whereas segment orientation is necessary to
achieve a marginally stable closed loop system.
Alison Sheets, also with Mont Hubbard, used Autolev III to develop a forward
dynamics model of uneven bar giant circling. This is a combination of experimental
and modelling work, first to determine parameter values for the 'shoulder' then to
use these to enhance a four segment, rigid linked, planar model of circling. Interest-
ingly they report, no changes to the hand force related slippage limiting factor
176 David G Kerwin
(Sheets and Hubbard 2005), but a reduction in flight revolutions as a result of energy
dissipation during the swing.
Energy exchanges in bar swinging also features in the paper by Gareth lrwin and
David Kerwin. Using inverse dynamics, in a similar manner to the study by Aram-
patzis and Bruggemann (2003), male international gymnasts performing general
longswings (giant circles) on a high bar were studied. Customised inertia data for the
gymnasts have been combined with video analysis to develop a four segment planar
model. Based on observations from the earlier paper, the Jack of a compliant shoul-
der is likely to overestimate total energy, but serves to highlight the critical impor-
tance of the functional phase in the longswing.
The final paper in this section by David Kerwin and Gareth lrwin revisits the
problem of determining high bar forces in competition and examines the influence of
inertia and kinematic data on inverse dynamic determination of bar forces and con-
trasts this with estimates based on bar deflections.
Future Research Trends
Modelling will continue as a central theme of gymnastics based research. The control
in this environment makes the approach very powerful in examining aspects of tech-
nique, investigating control strategies and understanding mechanisms underpinning
injury without putting gymnasts at risk. Forward dynamic models which are torque
driven tend to be more stable than the angle driven variety, but as with all forward
dynamics models, the key challenge remains specification ofthe internal parameters.
Improvements in technology and the use of remote sensing will encroach more
into sport from engineering and the impact of increased volumes of fine grained data
will in the future underpin performance profiling, retrospective and prospective stud-
ies of injuries and form the core of future coaching knowledge.
References
Arampatzis, A., and Brüggemann, G.P. (2003) Mechanical energetic processes during the
giant swing bcfore thc Tkatchcv exercise. Journal of Biomechanics, 34, 505512.
Hiley, M.J. and Yeadon, M.R. (2005) The Margin for Error Whcn Releasing the Asymmetrie
Bars for Dismounts. Journal of Applied Biomechanics, 21, 223-235.
Sheets, A.L. and Hubbard, M. (2005) Effects of low bar avoidancc and gymnast sizc on high
bar dismount performance. Proceedings of XXth Congress of International Society of
Biomechanics.
Effect of Shoulder Compliance on Peak High Bar Forces
During the Giant Swing
Alison L. Sheets I and Mont Hubbard2
1University ofCalifomia, Davis, Mechanical Engineering, alsheets@ucdavis.edu

2University of Califomia, Davis, Mechanical Engineering, mhubbard@ucdavis.edu
Abstract. When a female gymnast model does not include shoulder compliance, a simulated
optimal performance of a giant swing on the uneven parallel bars is limitcd by the maximum
force that can be exerted on the hands without slipping from the bar (Sheets and Hubbard
2005). To determine the effect of shoulder compliance on hand force, two four-segment gym-
nast models including an arm, a torso/head and two leg segments are compared: one in which
the shoulder is a rigid pin joint, and one which includes shoulder compliance. Experimental
values of shoulder stiffness and damping are determined to best describe 1-D vcrtical damped
oscillations of the gymnast/bar system. Optimum shoulder and hip motions during the swing
are calculated to complete the most dismount flight revolutions prior to the mass center (CM)
passing a specificd landing height. Optimization constraints includc maximum bar/hand force,
physiologic joint limitations, low bar avoidance, and minimum landing distance from the bar.
Shoulder and hip motions result from time varyingjoint torques that are limited by joint angle,
angular velocity, isometric strength, and activation factors. Bar relcase time and joint torque
activations at ten nodes equally spaced throughout the swing are optimized using the downhill
simplex method. Joint torque activations at all other times arc approximated by cubic splines
fit to the tcn nodes. Performance limitations due to the slipping constraint are not reduced in
the compliant modcl, even though it is active for a shorter period. The compliant shoulder
model also produced fewer tlight revolutions than the rigid one, 1.417 vs. 1.4 78, because the
shoulder dissipated energy during the swing and stored it at bar release.
1 Introduction
The giant swing (giant) begins and ends in a handstand and allows the gymnast to
connect skills, or increase angular velocity to be able to execute more complex dis-
mount or release-regrasp moves (Hiley and Yeadon 2003).
Because of the importance in a routine, a giant swing simulation using a rigid
four segment gyrnnast model is created (Sheets and Hubbard 2005). The model's hip
and shoulder movements during the swing are optimized to maximize the number of
flight revolutions after bar release. Thc optimal performance is subject to constraints
on gymnast joint torque, angular velocity, and ranges of motion. lnterestingly, none
of these constraints is active when the maximum force that can be exerted on the bar
without slipping is limited to 4*mg (Witten 1996). This limitation is important be-
cause force is transmitted through the hands and onto the bar to accelerate the CM,
which occurs when the gymnast changes body configuration to avoid the low bar,
transfers angular momentum between segments or increases the system's mechanical
178 Alison Lynn Sheets and Mont Hubbard
energy using muscular work. Therefore, the bar force constraint inherently limits
maximum joint angular velocities, possible ranges of joint angles, and the gymnast's
ability to do work. lt also reduces the influence of increased hip and shoulder
strength on performance because the gymnast must remain in contact with the bar.
lt is possible that optimizations performed using a rigid model may overestimate
the importance of the slipping constraint. When Brewin et al. (2000) included shoul-
der compliance in the model of a gymnast performing a giant swing on the men's
ring apparatus, the calculated peak shoulder force was reduced; which also reduces
hand force on the ring. This paper aims to determine experimentally gymnast shoul-
der stiffness and damping. These values will be used to ascertain ifthe rigid shoulder
assumption used in previous models overestimated the bar force and thus the bar
force constraint's importance. Maximum bar deflection, and maximum number of
dismount revolutions are compared to quantify effects of shoulder compliance.
2 Methods
2.1 Gymnast and Uneven Parallel Bar Model

Two four-segment models of a collegiate gymnast (m = 69.3 kg, and h = 2.05 m with
arms above head) include head/torso, arm, and two leg segments. Each segment's
inertia, mass and CM location is determined using a "stadium solid" model. Both
gymnast models assume the hip is a rigid ball and socket, while one shoulder is a
rigid pin joint and the other includes a damped linear spring to represent connective
tissue. Arm and leg motion is driven by instantaneous torques determined by scaling
the maximum isometric joint torque by joint angle dependence, angular velocity
dependence, and joint torque activation A(t) (Sheets and Hubbard 2005). The scalar
function of time, A(t), varies over the range [-1, 1] and is the fraction of the maxi-
mum possible torque (created by fully activating all muscles crossing each joint at
the current state) applied to cause joint extension and flexion, respectively. The A(t)
time histories are approximated by cubic splines fit to ten nodes equally spaced
throughout bar contact. The simplex optimization method (Neider and Mead 1965),
is used to calculate ptimayoint torque acti ation at the nod and bar release time.
Fig. 1 Schematic of the compliant shoulder gymnast model

Effect of Shoulder Compliance on Peak High Bar Forces During the Giant Swing 179
The bar is modeled as a massless linearly elastic spring with an experimentally de-
termined stiffness of 15,000 N/m. The coefficient of friction between hands and bar
is measured tobe µ = 0.85. Bar damping is not included because the shoulder ac-
counts for most of the system damping (Hiley and Yeadon 2005).
The state vector defining the compliant gymnast and bar system is composed of
seven generalized coordinates (GC) [ß, 0,, 0hz, 0hy, x 6, y 6, y,] and their time deriva-
tives (Fig. 1). The rigid model needs only six GC's because there is no shoulder
stretch, Ys= 0. AutoLEV (Kane and Levinson 2000) is used to derive the equations of
motion, which are solved with a Runge-Kutta numerical integration using C.
2.2 Shoulder Stiffness and Damping Coefficients

Shoulder stiffness, ks, and damping, bs, are experimentally determined by initially
displacing a gymnast hanging from the bar and recording the bar and gymnast verti-
cal oscillations. Markers are placed on the bar center between the hands, on each
forearm and on the sternum. Ten trials are performed with three different initial con-
ditions. Four trials begin with the gymnast standing on an elevated surface with
extended shoulders, then the feet are unloaded and bar loaded. During three trials the
gymnast jumps vertically to catch the bar with fully extended shoulders. Finally,
three trials begin by hanging from the bar while a !arge downward force is applied to
the feet then released. Three motion analysis cameras film the resulting vertical os-
cillations of the bar, gymnast forearms and torso at 250 frames/sec.
The change in distance from the sternum to the forearm as a function of time is ap-
proximated by a damped oscillation and two decaying exponential functions.
F1 = c 1 sin(mdt ~)e -s"w„t + c 2e-t/r2 + c 3e-tfr3 (2)
where OJd = ✓ w~ (l - ?2 ) and the constants c 1, <j>, c2, c3, lll,, ?, r2, and f3 are deter-
mined using a least squares fit ofthe function F1 to the experimental data.
The calculated system natural frequency lll,, damping ratio ?, and time constants r2
and r3 are then used to determine k., and b Using measured values of bar stiffness kh,
5•
bar mass mh, gymnast arm mass m and remaining gymnast mass mg in the system
0
equations of motion
(mh +ma)xh =-khxb +k.,(xs -xh)+bs(X., -xh)+(mh +ma)g (3)
mgxg = -ks(xs -xb)-bs(xs -xh)+mgg (4)
the system characteristic equation is expressed in terms of the unknowns ks and b5 •
S 4 + 0. l 7lb5 S 3 + (2290.555 + 0. l 7lk 5 )S 2 + 42.574b,S + 42.574ks = 0 (5)
Initial guesses for ks and bs allow calculation of eigenvalues (roots of the characteris-
tic equation). Two real eigenvalues represent time constants, and the pair of complex
eigenvalues represent system natural frequency and damping. The values of ks and bs
are optimized to minimize the sum of squared differences between the calculated
eigenvalues and constants determined by the F1 optimization.
F2 = (OJn - OJnopt ) 2 + (10(? - Sopl )) 2 + (r2 - r2opl )2 + (r3 - r3opt ) 2 (6)
2.3 Giant Swing Optimization

The optimal solution begins (t 0 = 0) with the gymnast in a handstand rotating about
the deflected bar with an angular velocity of 1.8 rad/s (Hiley and Yeadon 2003). lt
ends at bar release (t1) and results in maximum straight body dismount revolutions,
with arms above the head, prior to CM passing height 0.4h above ground.
J 0 = (cd.J f )!J.t1 + y(t f ))/(2n-) (7)
where ol._ft) is release angular velocity, !J.t1 flight time and r(ft) angle shown in Fig. 1.
Optimization constraints include limitations on: joint ranges of motion, angular ve-
locity, rate change of muscle activation, bar force and minimum landing distance.
Using the optimization criterion in Eq. 7 and including constraints as penalty
functions to be subtracted from the objective function J 0 , the optimal giant swing
joint torque time history and release angle are calculated for gymnast models with a
rigid and compliant shoulder.
3 Results
3.1 Shoulder Stiffness and Damping Coefficients

The experimental data is weil approximated by a damped oscillation and two decaying expo-
nential functions (Fig. 2). There is little variation between the calculated natural frequencies,
-
damping coefficients, and time constants using the F 1 optimization even though three different
.
initial conditions are used (Table 1).
.,,_ ..,.,_ _,
i 0.6
r= w" ~
i
0
0
-0.6
t -0.6
·I
-...---
,_,
·1 • l.6
0 0.6 1.6 2 2.6 0 0.6 1.6 2 2.6
- E n..
o•
0-2
>
ia I
i
0 t
-02
-o•0 0.6 16 2 2.6 0 0.6 1.6 2 2.6

)
Fig. 2 A typical result for one ofthe ten trials using the F 1 curve-fit optimization
Effect of Shoulder Compliance on Peak High Bar Forces During the Giant Swing 181
Table 1: Natural frequency, damping ratio, and time constants from F1 optimizations
m., t; 'tz 't3 rms error (cm)
Average 14.713 0.114 0.239 0.343 0.137

cr 0.594 0.045 0.078 0.106 0.048
The coefficients calculated using F1 are consistent, meaning the standard deviations (cr) are
small compared to mean values, but the estimation of ks and bs gives less consistent results.
Although there is a !arge cr, the average ks and bs correspond fairly weil with those experimen-
tally determined by Hiley and Yeadon (2005) (Table 2).
Table 2: Shoulder stiffness and damping calculated using F2 optimizations
k,; (crk) Hiley ks b, (crb) Hiley bs

Average 22466 (12710) 25261 2432 (1042) 1003
Numerous other methods were tried, varying bar stiffness and including bar damping, but
none produced more consistent results. The average values were used in the giant simulation
to quantify the effect of shoulder stiffness on bar force.
3.2 Giant Swing Performance Effect Due to Shoulder Compliance

When the gymnast model includes shoulder compliance, the number of possible
flight revolutions following an optimized simulated giant swing is slightly reduced
from 1.478 (Sheets and Hubbard 2005) to 1.417. Possible reasons for reduced dis-
mount performance are that the damper dissipates energy when the shoulders stretch
(Fig. 4), and that the shoulders are stretched at bar release (Fig. 3) indicating that
kinetic energy has been stored and not retumed. The total energy during the rigid
model's giant swing (Fig. 4) is similar to that measured by Arampatzis and Brueg-
gemann (1999, Fig. 6a) during elite female performances.
0.1S
g 0.1
.c,
!.g 0.0S
"& 0
~
-0.0S
.3 -2 .] 0 -3 ·2 ·1 0
y (rad) ., (rad
Fig 3. Shoulder stretch as a function of Fig 4. Net energy change for rigid and
gymnast angle compliant shoulder models
The performance decrease caused by the bar force constraint was not reduced by the
compliant shoulder. Although the constraint is active earlier during the rigid body model's
swing, it still limits both performances when the gymnast's CM is below the bar (Fig. 5). The
rigid model's constraint becomes active when the gymnast's legs move to avoid the low bar
and the CM is accelerated, whereas the compliant model's shoulders stretch in addition to the
bar flexing to produce this motion (Fig. 3).
-·--
.
01
••
....,,
.,,
".L .,, 0 ., U
Fig 5 Both bar center paths, solid lines, approach bar deflection limits, dotted circle
The bar force result is consistent with the findings of Brewin et al. (2000) who de-
termined that, although elasticity of the shoulder and apparatus slightly decreased
peak shoulder forces, gymnast technique was far more important. Because the opti-
mized gymnast technique for the rigid and compliant models could not violate the
force constraint, it is not surprising that both maximum bar forces were identical.
4 Conclusions
lt is possible to determine shoulder stiffness and damping by recording vertical gym-

nast oscillations on the uneven parallel bars. Although the system natural frequency,
damping ratios and decay time are consistent over all trials, the resulting values of ks
and bs varied more. The addition of shoulder compliance, using average values,
decreased giant swing performance. Energy was dissipated during the swing by the
damper, and the gymnast releases the bar with energy stored in the compliant shoul-
der structures. The bar force constraint continued to be active during the swing.
References
Arampatzis, A, and Brueggemann, G.P. (1999) Mechanical energetic processes during the
giant swing exercise before dismounts and flight elements on the high bar and uneven par-
allel bars. J of Biomech 32, 811-820.
Brewin, M.A., Yeadon, M.R., Kerwin, D.G. (2000) Minimizing peak forces at the shoulders
during backward longswings on rings. Human Movement Science 19, 717-736.
Hiley, M. J. and Yeadon, M. R. (2003) Optimum technique for generating angular momentum
in accelerated backward giant circles prior to dismount. J Appl Biomech, 19(2), 119-130.
Hiley, M.J., Yeadon, M.R. (2005) The margin of error when releasing the asymmetric bars for
dismount. J Appl Biomech. 21, 223-235 .
Kane, T. R. and Levinson, D. A. (2000) Dynamics Online: Theory and Implementation with
AutoLEV. Online Dynamics, Inc., Sunnyvale, CA.
Nelder, J. A. and Mead, R. (1965) Computer Journal. 7, 308-313.
Sheets, A.L. and Hubbard, M. (2005) Effects of low bar avoidance and gymnast size on high
bar dismount performance. Proceedings ofXXth Congress oflnt Soc ofBiomechanics.
Witten, W. A., Brown, E. W., Witten, C. X., Wells, R. (1996) Kinematic and kinetic analysis
ofthe overgrip giant swing on the uneven parallel bars. J Appl Biomech, 12(4), 431-448.
Effect of Horizontal Surface Compliance on Balance
Strategies
Wendy Kimmel and Mont Hubbard
University of Califomia, Davis, Mechanical and Aeronautical Engineering,

wlkimmel@ucdavis.edu, mhubbard@ucdavis.edu
Abstract. Balancing on a horizontally compliant surface is examined by comparing the task

with successively more complex dynamic models. 1,2 and 3 DOF systems are used to examine
the balance strategies (state feedback control laws) needed to maintain stability. The 1 and 2
segment systems use a spring attached at the lowest point of the (lower) segment to model
horizontal compliance. Velocities are neglected and only position and angular displacement
state variables of the system are fed back to achieve balanced, marginally stable closed loop
systems. Regions of stability in the feedback gain space are compared using the Routh-
Hurwitz criterion, root locus and other graphical techniques. Boundaries of the stable regions
in the feedback gain space are provided by the inequalities satisfied by the gains for positivity
ofthe leading column ofthe Routh array. Knowledge and feedback ofhorizontal foot position
is not important since it is possible to achieve stability by neglecting it entirely.
1 Introduction
Balance is important enough to be leamed early in our lives. lt plays a crucial role in
posture and bipedal locomotion and is also relevant in many sports tasks. lt consists
of a dynamic feedback control process, in which the state of the body (orientation,
position and their rates of change) is sensed and compensatory muscular torques are
applied to restore the body to its inverted pendulum equilibrium position. We are
'balanced' ifwe are able to retum to vertical after a perturbation is applied.
Considerable research has occurred on the mechanisms and strategies of the bal-
ance task (Pai and Patton 1997; Pai 2003; Patton, Pai and Lee 1998) in standing and
locomotion. Virtually all these studies have assumed that the supporting surface is
rigid and the horizontal forces induced at the support point are provided by friction
between the rigid surface and the foot. Almost no work has yet examined the bal-
ance task on a surface with either vertical or horizontal compliance, yet this task is
important in some sports activities (e.g. trampoline ).
This paper examines balance on horizontally compliant surfaces by comparing
several models of increasing complexity. Active ankle torque determined from a
linear feedback control law is applied to the joint between the body and the compli-
ant surface in order to maintain stability of the system.
184 Wendy Kimme! and Mont Hubbard
2 Methods- Single and Multi-Segment Body Models

For each system, the equations of motion are derived using the AutoLev symbolic
dynamics computer program (Kane and Levinson 2000). System dynamics are also
linearized in AutoLev using Taylor series approximations. The linearized systems
are used to derive feedback strategies. The more complex the system, the more
states which must be included in the control law for stability. Although füll state
feedback including velocities as well as position states would allow·exact placement
of closed loop eigenvalues, here only position feedback is considered. Thus the
maximum stabilization that is possible with this limited feedback scheme results in
marginal stability (closed loop eigenvalues on the imaginary axis). The Routh-
Hurwitz Criterion is used to determine limitations on the feedback gains for stability.
Of interest is how stable regions in the feedback gain space are related to those for
more complex models with more segments and to those for simpler systems without
compliance.
2.1 One Segment Models

The simplest balance model including horizontal ground compliance consists of a
single uniform segment of length 2L, mass m, and central moment of inertia /, sup-
ported by the vertically rigid ground with constant surface stiffness k in the horizon-
tal direction (Fig. 1b). Because the segment can tilt 0 and its base can translate xr,
this system has two degrees offreedom (DOF). A similar single DOF system without
horizontal ground compliance, a one segment body with its base fixed on a com-
pletely non-compliant surface (Fig. la), is useful for comparison to understand the
effects of compliance. This body also rotates through 0 but motion of its base is
prevented. In both cases the control variable responsible for achieving balance is a
torque T applied between the ground surface and the body, as a model for ankle
torque in the human.
The system without compliance has linearized equation of motion
I jj - mgL 0 = - T = - K 0 0 . A root locus versus the gain K 8 shows that there are
two open loop real roots of equal magnitude (mgLII/ 12 , one stable and one unstable.
The active torque T= K 80, that feeds back the angular rotation of the segment, is able
to achieve marginal stability when K 8 > mgL, that is when the feedback term out-
weighs the destabilizing effect of gravity. For values of parameters m=60 Kg, g=9.8
m/s 2 and L=0.8 m, this inequality becomes K 8 > 470 N/m.
This 1 DOF, one segment model on a rigid surface without horizontal compliance
can be compared to its counterpart on a compliant surface (Fig. 1b). The stabilizing
feedback is again in the form of a control torque T= Kxxf + K 80between the segment
and the surface. In this case it is a function of two position states x1 and 0, where Kx
is the gain constant for the changes in the position x1 and K 8 is the gain constant for
changes in the body rotation angle 0. Again this torque is similar to an ankle torque.
Effect of Surface Compliance on Balance Strategies 185
Fig. la. 1 DOF, 1 segment, b. 2 DOF, 1 segment, c. 3 DOF, 2 segments,

non-compliant surface compliant surface compliant surface
The only other forces acting on the segment include the force of gravity at the center
of mass and the spring force acting at the 'foot' point f defined as
(1)
where k=24000 N/m is the spring constant.
The two linearized equations of motion for the 2 DOF system can be written as
mxcm =Fr
(2)
I cm ii = Fr L + mgL 0 + T
The linearized kinematic relation relating motion of the center of mass and the
foot, Xcm=x1 - L 0, allows the first of Eqs. 2 to be rewritten in terms of the variable x1
. Substituting it and Eq. 1 into Eqs. 2 yields
mxr - mLÖ = -kxr
(3)
lcm0 = -kxfL + mgL0 + Kxxf + K 0 0
After taking the Laplace transform of the differential Eqs. 3, the characteristic
equation for the 2 DOF system can be shown to be
mli +(m"Ke+kl+mkE+mLK,-m2 gL)s+(kKo-mgkI)=O (4)
First the gain Kx for the x foot position was ignored in order to determine whether
feedback of only angular position of the body is enough to stabilize the system. A
root locus can be used to visualize the effect that feedback of angular rotation of the
segment has on stability. When Kx=O, the root locus versus K 0 indicates that the
unstable system can be made marginally stable when the angular position ofthe body
is fed back with a sufficient gain, even when the motion of the foot is neglected. As
in the 1 DOF system, there are two real roots, one stable and one unstable. In addi-
tion, there exist two purely imaginary open-loop (OL) poles and a pair of lower fre-
quency purely imaginary OL zeros. Again as in the 1 DOF system, the root locus
indicates that the 2 DOF system cannot achieve füll stability in the sense of arbitrary
186 Wendy Kimme! and Mont Hubbard
pole placement, but rather only marginal stability, since the roots lie on the imagi-
nary axis even with infinite gain.
Knowing that the system can be stabilized with feedback of only one of the
states, the Routh-Hurwitz criterion (Nise 2000) was applied to more fully examine
stabilizability by applying it to Eq. 4, the characteristic equation of the 2 DOF sys-
tem. According to the Routh criterion (Nise 2000), each coefficient in the first col-
umn of the Routh array must be positive to guarantee system stability. In general,
because each EOM is second order, for an n DOF system the Routh criterion yields
2n-1 inequalities. Each of these inequalities involving the feedback gains, when
viewed as an equality, results in an equation for a boundary ofthe stable region, and
the set of gains resulting in a marginally stable system will lie on one side of the
boundary. The stable region is then the intersection of the three regions determined
by the boundary equations, the equalities related to the three Routh inequalities:
K 0 > -(-m 2 gL+kl +mL2 +mLKx)/m
K 0 > -Jm 2 gL-mL2k-mLKx + kl±2(-k 2 lmL2 -klmLKJ Im (S)

K 0 >mgL
This stable region is shown in Fig. 2. Note that one ofthe boundaries (Eq. 5c) is
identical to the stability requirement for the 1 DOF system. However, the region of
stable gains is further limited by the other two boundaries. The logical intersection
of these requirements indicates that the uppermost curved line, which results from
Eq. 5b, is the lower boundary of acceptable gains. There is no upper limit on values
of either gain. There is no value of Kx, or any amount of foot positional feedback
that can successfully stabilize the system alone. The gain Kx can even take on !arger
negative values for a sufficiently !arge K0 . But for any positive Kx, as long as K0 is
greater than the constant mentioned earlier, the system can be stabilized.
4
2 X 10
0 """~ *
*
Region of Stability
~ *
-2 'I;.
*~
-4 * ,>
:2 * "' ""
-6 *
*
"'
""
"'
-8
* "' ""
-10
*
*
"'
"' ""
-12
-5 0
*
5
"'
"' 10
K0 X 104
Fig. 2. Acceptable region in the Ke, Kx feedback gain space for a stable 2 DOF system
Effect of Surface Compliance on Balance Strategies 187
2.4 Two Segment Body Model

The next level of complexity involves the addition of a second segment (Fig. lc) by
splitting the body at the hip into legs (m 1= 26 Kg, 2L 1= 0.9 m) and a torso (m 2= 36
Kg, 2L 2= 0.8 m) Again the system is unstable without control, requiring an active
ankle torque between the lower segment and the surface that feeds back, in general,
all 3 position states; x1, 01 and 02.
(6)
Analogous to the procedure used above, the Routh-Hurwitz criterion for the char-
acteristic equation is used to derive five inequalities from positivity of the first col-
umn ofthe Routh array, the satisfaction ofwhich is necessary for stability. The com-
plexity of the inequalities motivates that this be done by computer using a symbolic
manipulator; we used Autolev. The boundaries are now 2-D surfaces in the 3-D gain
space. For the numerical values above, three ofthese inequalities (7a, b, c) are
K 81 > 651.40- 0.57965k - 0.70360Kx + 0.49796K 02
K 81 > -0.453k + 498.89 + 0.498K 02 - Ü.704Kx ±
(-1310.7k 2 -5.560e-38k-3512.8kK 02 -1906.3kKx - (7)
l.927e8 - l.520e6K 02 - 3.977e5Kx
KBI > (51007k + 0.6483Kx + l.159e - 3kK02) / k
Although one of the other two inequalities (7d) is too lengthy tobe shown here,
the last one (7e) is identical to the first (7a) above. Again a reasonable choice for
parameters allows numerical visualization of the five boundary surfaces and the
stable region in the gain space that satisfies the above boundary equations (7a, b, c)
(and the two not listed). Since they are difficult to portray on paper we choose not to
present the surfaces and stable regions in 3-space here, electing rather to show sec-
tions of these surfaces.
One such interesting section is that corresponding to Kx=O. In Figs. 3a and 3b be-
low, the stable subsection of marginal stability for this section is illustrated by the
diamond shapes bounded by several of the inequalities (other inequalities are irrele-
vant for these parameters and Kx=O) . Thus as in the 2 DOF case, it is possible to
neglect foot position entirely and, with the appropriate choice of gains K 01 and K 0:z,
feeding back only segment orientation is sufficient for stability. The inequality (7d)
not shown above has been plotted with the ' - -' shape in Fig. 3b, an enlarged portion
of Fig. 3a that allows the complexity of the inequality tobe seen. For this particular
inequality (7 d), as K 02 varies, K 01 has either 1, 3 or 5 real roots, resulting in 1, 3 or 5
branches ofthe surface seen in Fig. 3b.
188 Wendy Kimmel and Mont Hubbard
5
0. 5 X 10 X 104
1
/
/
➔:
0 0.5
0 a,el,
-0.5
◊ /
◊ -0.5
◊
-1 ◊◊
!:il
:,.:
◊◊
◊◊
Sj -1
,,:
◊ ◊
◊◊◊
-1.5 ◊◊◊
◊◊◊ -1.5 ◊ ◊
◊◊◊◊
◊◊◊◊
/,
-2 ◊◊◊◊
◊◊◊◊◊
-2 /, ◊ ◊
◊◊◊◊◊ /,
◊◊◊◊◊◊
-2.5 ◊◊◊◊◊◊ -2.5 ◊ ◊
◊◊◊◊◊◊
◊◊◊◊◊◊◊
-3 ◊:
◊◊◊◊◊◊◊ ◊
◊◊◊◊◊◊◊
-3
-2 -1 0 -3 -2 -1 0
K&l
K91 * 105 X 104
Fig. 3a. Section of 3 DOF system stability. 3b. Exploded region in Fig. 3a.
3 Conclusions
lt is possible to examine regions of gain space for 1 and 2 segment systems on a

non-compliant surface in order to determine the necessary feedback control to main-
tain marginal stability. For the two simplest dynamic systems representing balance,
and for the realistic parameters chosen, no account needs to be taken of foot position
in the feedback law. Only feedback of segment orientation is necessary to achieve a
marginally stable closed loop system. Further extensions of these systems to include
more segments and surface compliance in the vertical as well as horizontal planar
dimensions will allow closer comparisons to realistic human- sport surface interac-
tions.
References
Kane, T.R. and Levinson, D.A. (2000) Dynamics Online: Theory and Implementation with
AutoLev. OnLine Dynamics, Inc., Sunnyvale, Califomia.
Nise, N.S. (2000) Contra/ Systems Engineering. John Wiley, New York.
Pai, Y. and Patton, J. (1997) Center ofmass velocity- position predictions for balance control.
Journal ofBiomechanics 30 (4), 347-354.
Pai, Y. (2003) Movement termination and stability in standing. Exercise and Sport Sciences
Reviews 31 (1), 19-25.
Patton, J.L., Pai, Y., and Lee, W.A. (1999) Evaluation of a model that determines the stability
limits of dynamic balance. Gait and Posture 9, 38-49.
Predicting High Bar Forces in the Longswing
David Kerwin and Gareth lrwin

University of Wales Institute, Cardiff, dkerwin@uwic.ac.uk
Abstract. The longswing on high bar in men's artistic gymnastics is the core skill within all
competitive routines. The forces applied to the bar by the gymnast are important when study-
ing a performer's technique or when examining injury mechanisms. Previous studies have
used video measurements ofthe bar's motion to predict bar forces to within 7% ofthe range of
directly recorded forces. Also, by assuming zero extemal forccs at thc gymnast's feet, inverse
dynamics havc been used to predict bar forces, but previously this mcthod has resulted in
errors greater than 20%. A study, employing 2D DL T techniques and customized inertia data
for four elite male gymnasts performing thrce longswings on a strain gauged high bar, was
conducted to enable the two methods for estimating bar forces to be dircctly compared. Digital
video images were recordcd at 50Hz from which the bar ccntre, hcad centre and the nearest
wrist, elbow, shoulder, hip, kncc, anklc and toe were digitized, starting before the gymnast
reached the handstand, continuing throughout one revolution and ending once the gymnast
passed beyond the handstand (-400°). All video and forcc data ( 1000Hz) were interpolated
within a single 360° longswing at 1° intervals and root mean squared differences (rmsd) be-
tween the measured bar forces and those predicted by bar deformation and inverse dynamics
were compared. Tracking thc motion of the bar in 2D was poor in comparison to the 8% rmsd
whcn using inverse dynamics. In the latter technique, deliberately swapping inertia data scts
between the subjects increased errors. Inverse dynamics data werc sensitive to kinematic and
inertia data errors but the use of the 2D DL T and the inclusion of personalized body segmcnt
parameters contributed to an overall reduction in error comparcd to prcviously reportcd data.
When direct bar force measurement cannot be obtained, the bar deformation technique is
recommended providing that 30 video is used with a pre calibrated bar. Altematively, with
appropriatc inertia data and DL T processing, the inverse dynamics techniquc can be em-
ployed, albeit with a slight loss ofoverall accuracy in predicting thc precisc profile of the high
bar forces.
1 Introduction
The longswing on high bar in men' s artistic gymnastics is the core skill within all
competitive routines. The forces applied to the bar by the gymnast are important
when studying a performer's technique or when examining injury mechanisms. The
longswing has been a subject of much research but studies dealing specifically with
the bar forces are limited. Kopp and Reid ( 1980) used a strain gauged bar to measure
the forces directly. Okamoto, Sakurai, Ikegami and Yabe (1987) and Irwin and Ker-
190 David Kewin and Gareth lrwin
win (2006) described the kinematic and kinetic profiles of the hip and shoulder joints
through an inverse dynamics approach of the longswing and more recently Arampat-
zis and Brüggemann ( 1998) and Yeadon and Hiley (2000) have made detailed stud-
ies of 'general' and 'accelerated' longswings using forward dynamics. A study using
video measurements of the bar's motion to predict bar forces to within 7% of the
range of directly recorded forces was reported by Kerwin and Hiley (2003). Also, by
assuming zero extemal forces at the gymnast' s feet, inverse dynamics have been
used to predict bar forces, but previously this method has resulted in errors greater
than 20% (Gervais 1993). Challis and Kerwin (1996) analysed the sources oferror in
inverse dynamics analyses and highlighted the influence of the kinematics and in
particular the treatment of the raw digitizer data when determining segmental accel-
erations. During gymnastic competitions, direct measurement of the bar forces is
difficult although not impossible. However, there are many occasions when video
data are available but without force measurements. The bar in a competition can be
calibrated in the 'field' as described by Kerwin et al. (2003) and used by Hiley and
Yeadon (2005), but there are many occasions where these data are not available.
The purpose of this study was to revisit the use of inverse dynamics as a method
for predicting high bar forces in comparison to the values obtained by measuring the
displacement of the bar. Both data sets were evaluated against directly measured
forces using a strain gauged high bar.
2 Method
2.1 Data Collection

All testing was performed in a gymnastic arena on a standard competition high bar
(Continental Sports, Huddersfield, UK). Four men from the National Gymnastics
Squad participated in this study (age = 22.5 ± 4.1 yrs, mass = 66.4 ± 7.2 kg, and
height = 1.69 ± 0.05 m). Customized body segment inertia parameters were devel-
oped for each gymnast using the methods of Yeadon ( 1990). A digital camcorder
(Sony DSR-PDI I00AP, 3-CCD, Japan) was located approximately 40 m from the
centre of the high bar and angled at 80° to the plane of motion to avoid viewing the
gymnast's hands and bar centre through the bar supports. (The normal orthogonal
alignment requirement was negated by the use of 20 DLT procedures). Images of a
single calibration pole of height 5.176 m were recorded as it was sequentially placed
at three locations to create a vertical plane approximately 5 m x 5 m. Four equally
spaced spherical markers (diameter 0.10 m) were centered on the pole creating 12
known points within the field of view. Reaction forces on the bar were recorded
(1000 Hz) using strain gauges (CEA/09/280UW/120, UK) bonded in pairs to the
bar's surface. The bar had been pre-calibrated up to 3 kN and then back to ON in
each dimension and the two bar stiffness values (Kz and Ky) determined using linear
regression between the known loads and the bar's deflection.
Predicting High Bar Forccs in the Longswing 191
Each gymnast perfonned three series of five longswings. Images in the sagittal
plane were recorded at 50 fields per second with the electronic shutter set to 1/300 s.
Synchronization of the force and video data was achieved through the use of 20
LEDs (Wee Beasty Electronics, Loughborough, Leicestershire, UK) in the field of
view of the camera. A single trigger initiated force data capture and began a sequen-
tial illumination of the LEDs at 1 ms intervals. By identifying the single image in
which more than one and less than 20 LEDs were illuminated it was possible to
synchronize the data to an accuracy of -3 ms.
2.2 Data Processing

All digitizing was completed using the TARGET high resolution system (Kerwin
1995). Following six repeat digitizations of the calibration images, three of each
gymnast's longswings from the three sequences were selected for analysis. The
longswing begins with the gymnast in a handstand on top of the bar and ends when
he has rotated through 360° and is back in the handstand position. For a gymnast in a
sequence of swings, the gymnast passes through rather than holds the handstand
position, and so images ten fields before 0° until ten fields after 360° were digitised.
In each field, the bar's centre, gymnast's head centre and his wrist, elbow, shoulder,
hip, knee, ankte, and toe on his right side (nearest the camera) were digitized. Data
reconstruction was achieved using an eight parameter, 2D direct linear transfonna-
tion (DLT) algorithm (Kwon 1999). A low pass digital fitter with cutoff frequency
set to 6 Hz was used for all data. Later residual analysis (Winter 2005) was used to
customize cutoff frequencies for each trial and for selected data points.
Each gymnast was modelled as a pin-jointed four link system comprising anns
(including hands), trunk, thighs and shanks (including feet). Customized segmental
inertia profiles of mass and centre of mass location for each gymnast were produced.
These were based on volumes estimated from the anthropometric measures for the
individual gymnasts and intrinsic density data for each segment. Minor proportional
adjustments were made to the segment mass values so that the predicted and meas-
ured whole body mass values agreed.
The recorded strain gauge data were converted to force units using the pre-
detennined calibration coefficients to produce the measured force values (Fz and
Fy). All subsequent calculations were completed in Mathcad 13™ (Adeptscientific,
UK). For inter and intra subject comparative purposes all data were interpolated at 1°
intervals from 0° to 360°. Bar detlections were combined with the stiffness coeffi-
cients (Kz and Ky) to predict the vertical and horizontal bar forces. Externat forces at
the toes were assumed to be zero and inverse dynamics applied to estimate the forces
at the knees, hips, shoulders and finally the hands/bar interface. Quantification of the
level of agreement between the directly measured forces and the two sets of esti-
mated forces was developed around two scores; the root mean squared differences
(nnsd) between the measured vertical and horizontal force profiles, and the differ-
ences at the peaks of the measured forces. Finally perturbations to the inertia data
and the data filtering procedures were made to check the sensitivity ofthe analyses.
192 David Kewin and Gareth lrwin

The DLT reconstruction errors for the calibration points were ±0.004 m (y) and
±0.003 m (z). The ranges of forces determined directly from the strain gauge meas-
urements were -1991 N to +1993 N horizontally and -340N to 2760N vertically.
Table 1 shows the percentage root mean squared differences between the measured
forces and those predicted from the 2D bar displacement to range from 10 to 15% (y)
and from 12 to 17% (z). The percentage rmsd values for the inverse dynamics data
were from 4 to 5% (y) and from 5 to 8% (z).
Table 1. Differences between the measured force data and values predicted from bar deflec-
tions and inverse dynamics analysis as rmsd (%) and peak forces (%)
Bar Deflection Inverse Dynamics
Gymnast rmsd(Fy) rmsd(Fz) rmsd(Fy) rmsd(Fz)
A 15.0 (2.5) 13.8 (0.8) 3.9 (0.4) 8.0 (2.1)
B 10.2 (2.0) 12.1 (1.6) 5.2 (1.2) 7.7 (0.9)
C 12.4 (0.5) 13.2 (3.3) 3.6 (0.8) 5.3 (0.6)
D 14.8 (1.0) 17.0 (4.5) 4.7 (1.5) 6.7 (1.3)
peak(Fy) peak(Fz) peak(Fy) peak(Fz)
A 32.5 (3.4) 4.3 (1.6) 1.3 (5.5) -6.0 (6.5)
B 23.7(7.1) 17.7(3.8) -1.9(4.1) 13.4 (2.6)
C 17.7 (10.0) 15.5 (7.0) -1.6 (4.4) 5.8 (1.7)
D 13.4 (16.6) 7.0 (14.9) -0.3 (7.3) 4.0 (1.2)
The corresponding comparisons for the differences in peak values were from 18
to 33% (y) and from 4 to 18% (z) for bar deflection forces and from -2 to 1% (y) and
from -6 to 13% (z) for bar forces derived via inverse dynamics.
-3000 0
90 180 270 90 180 270 360
-Stra:inGauge -Stra:inGauge
·· InverseDynamics · · · lrcverse Dynamics
- · BarDeflection - • BarDeflection
Fig.1 Force profiles (N) for one longswing (gymnast C) with personalized inertia data.
Figure 1 shows that the overall match between inverse dynamics data and the
strain gauge data was consistently superior to that for the bar deflection data. In a
previous study by Kerwin et al (2003) very close agreement was found between bar
Predicting High Bar Forces in the Longswing 193
deflection data and the measured forces and so it would appear that difficulties in
digitizing the bar centre from the oblique camera view, even with the inclusion ofthe
DLT procedures, resulted in data of poor quality. The inverse dynamics data gener-
ated from this 2D analysis produced a better match to the measured forces than had
previously been reported and in all cases showed rmsd agreement of 8% or less. Peak
values did not agree as closely in time or magnitude with the measured data, with bar
deflection peaks varying from 13 to 33% (y) and from 4 to 18% (z). With inverse
dynamics these values were lower; from 0 to 2% (y) and from -6 to 13% (z). Thus,
although the overall level of agreement appears to be better than the previously re-
ported 20% of the range, the subtleties of the fluctuations in the forces, particularly
around 180° of rotation, were masked.
The sensitivity of the inverse dynamics analyses was investigated through vary-
ing the segmental inertia profiles and manipulating the cutoff frequencies. Three of
the gymnasts used were of similar size (70.39, 70.03 and 68.45 kg) whilst the fourth
was smaller (55.56 kg). The respective customized inertia data sets were swapped
within the inverse dynamics calculations and when the kinematics for the largest
subject were used with his own inertia data, the differences with respect to the meas-
ured forces were 3.6% (rmsdy) and 7.2% (rmsdz). Almost identical values were
observed for the two other gymnasts of similar masses, but for the smallest gymnast,
the corresponding differences increased to 6.5% and 12. 7% respectively. The corre-
sponding differences in the peak force values changed from 4% to 15% (y) and -13%
to 8% (z) when the small gymnast's inertia data were included.
Changing the cutoff frequencies in the Butterworth filter for the bar deflection
data from I Hz to 20 Hz, altered the level of agreement bctween the measured and
predicted forces with the residual analysis indicating an optimized cutoff frequency
of 4.5 Hz. Similar tuning of the movement data within the inverse dynamics analy-
ses resulted in cutoff frequencies ranging from 3.6 Hz (y and z) at the wrists to
6.9 Hz (y) and 7.6 Hz (z) at the toes, but had minimal influence on the level of
agreement between the predicted and measured bar forces.
4 Conclusions
Predicting high bar forces using video analysis of bar displacement has previously
been shown to produce good results in circumstances when the bar has been appro-
priately calibrated. Digitized data from a single camera view, even allowing for the
2D DLT analyses, was unable to reproduce data of the quality previously reported
when using 3D image data. Inverse dynamics analysis does not require knowledge of
the characteristics of the apparatus and so could be very useful in competitions, but
does rely on knowledge of subject inertia data to determine segment masses, mass
centre locations and hence, by differentiation, segmental accelerations. Poor levels of
agreement (-20%) have previously been reported for bar forces determined by in-
verse dynamics. Careful selection of customized inertia profiles and tuning of fiiter-
ing procedures has becn shown to improve the overall agreement between measured
and predicted forces to 8%. Peak forces also appeared to be predicted better by in-
194 David Kewin and Gareth Irwin
verse dynamics analyses. Comparisons of the time histories revealed that neither
method could be used to replicate the fine changes in the bar forces which were
particularly evident as the gymnast was passing under the bar. If the overall magni-
tude and general shape of the force profile is required, inverse dynamics is suitable
for predicting bar forces. When more detailed tracking of small bar force fluctuations
is required, particularly for example in detailed technique analyses; computer simula-
tion model evaluation or when considering injury potential, then either the forces
need tobe measured directly or very precise 3D bar displacement time histories need
to be obtained.
References
Arampatzis, A., and Brüggemann, G.P. (1998) A mathematical high bar- human body model
for analysing and interpreting mechanical- energetic proccsses on the high bar. Journal of
Biomechanics, 31, 1083-1092.
Challis, J.H. and Kerwin, D.G. (1996) Quantification ofthc uncertainties in resultantjoint
moments computed in a dynamic activity. Journal of Sports Sciences, 14, 219-231.
Federation International de Gymnastique (FIG) (2001) Code of Points, artistic gymnasticfor
men, Switzerland.
Gervais, P. (1993) Calculation ofreaction forces at the hands on the horizontal barfrom posi-
tional data. In S. Bouisset, S. Metra! and H. Monod, (Eds.) Proceedings of the X!Vth Con-
gress of the International Society of Biomechanics, 468-469. University of South Paris,
Paris, France.
Irwin G. and Kerwin, D.G. (2006) Musculoskeletal work in the longswing on high bar. 6 th
Interantional Sports Engineering Conference, Munich, Germany.
Kerwin, D.G. ( 1995) Apex/Target high-resolution video digitising system. In J. Watkins
(Eds.) Proceedings of the Sports Biomechanics section of the British Association of Sports
and Exercise Sciences, Leeds, UK. pp. 1-4.
Kopp, P.M. and Reid, J.G. (1980) A force and torque analysis ofgiant swings on the horizon-
tal bar. Canadian Journal of Applied Sport Science, 5, 98-102.
Kwon, Y.H. (1999) 2D Objcct plane deformation duc to refraction in two-dimensional under-
water motion analysis. Journal of Applied Biomechanics, 15, 396-403.
Okamoto, A., Sakurai, S., Ikegami, Y., and Yabe, K. (1987) The changes in mcchanical en-
ergy during the giant swing backward on the horizontal bar. In In L. Tsarouchas, J. Te-
rauds, B. A. Gowitzke, & L. E. Holt (Eds.) Biomechanics XIB, International Series on
Biomechanics, Amsterdam: Free University Press, pp. 338-345.
Winter, D.A. (2005) Biomechanics and motor control of human movement, Third Edition,
Wiley Science, Hoboken, New Jersey.
Yeadon, M.R. ( 1990) The simulation of aerial movement. Part II: A mathematical inertia
model ofthe human body. Journal ofBiomechanics, 23, 67-74.
Yeadon, M.R., and Hilcy, M.J. (2000) The mechanics ofthe backward giant circle on the high
bar. Human Movement Science, 19, 153-173.
Hiley, M.J. and Yeadon, M.R. (2005) The margin for error when releasing the asymmetric
bars for dismounts. Journal of Applied Biomcchanics, 21, 223-235.
Musculoskeletal Work in the Longswing on High Bar
Gareth Irwin and David G Kerwin
UWIC, CardiffSchool ofSport, Cardiff, Wales, girwin@uwic.ac.uk, dkcrwin@uwic.ac.uk
Abstract. The aims of this study were to determine the contributions of the gymnast' s muscu-
loskeletal system during the execution of a gcneral longswing on high bar and to evaluate the
overall interaction between the gymnast and the elastic bar. Images of four international gym-
nasts were recorded (50Hz) performing three series offour longswings on a strain gauged high
bar (I000Hz). Real world coordinates were reconstructed using 2D DL T and synchronized
with the force data. Inverse dynamic analyses were cmploycd to determine joint kinetics dur-
ing each longswing. Analyses were performed on the whole longswing and on the hip and
shoulder 'functional phases' dcfined as maximum extcnsion to flexion at the hips and maxi-
mum flexion to extension at the shoulders respectively. Thc muscle momcnts and powers at
the shouldcrs werc consistently found to be dominant, with maximum values at the shoulders
being 4.5 ± 1.70 Nm·kg· 1 & 14.4 ± 6.7 W·kg· 1 and 2.3 ± 0.5 Nm-kg·' & 6.0 ± 1.7 W.kg· 1 for
the hips. In all cases the pcak values within the muscle momcnt profiles occurred within the
functional phases highlighting the importance of these active phases to thc overall skill. The
corresponding muscular work profilcs highlighted that an average of 71 % ± 6% of the total
work occurred during the functional phases of the longswing. Quantification of bar strain
energy, based on bar deformation, enabled an cnergy deficit to be determined. This deficit
arose from frictional losses at thc hand bar interface, air resistance and bar hysteresis and
hence defined the minimum work that the gymnast nceded to contribute to complete the circle
successfully. These analyses highlighted the dominance of the contribution made by the gym-
nast between 200° and 240° of rotation, during a successful longswing.
1 Introduction
Longswings on the high bar fall into two categories the 'general' and the 'acceler-
ated'. The general longswing is leamed before the accelerated and is used to link
other skills. The accelerated longswing precedes complex release and re-grasp skills
and dismounts. Over the last two decades the majority of high bar related research in
sports biomechanics and engineering has focused on the accelerated longswing.
Since 1990 this has been dominated by two research groups, in Loughborough
(Yeadon and Hiley 2000; Hiley and Yeadon 2003) using a forward dynamics ap-
proach to investigate optimizing the longswing and in Cologne (Arampatzis and
Brüggemann 1998; 1999) using an energetic approach to explain the interaction
between the gymnast and the elastic bar during the longswing. These groups pro-
vided kinematic and kinetic descriptions of the hip and shoulder joints during the
196 Gareth lrwin and David G Kewin
accelerated longswing and identified the characteristics of optimum technique

(Yeadon et al. 2000). Okamoto, Sakurai, Ikegami, and Yabe ( 1987) described the
kinematic and kinetic profiles of the hip and shoulder joints through an inverse dy-
namics approach of the general longswing but there has been little recent research
into the general longswing, particularly considering the number of changes in the
rules, the apparatus and the skill levels of gymnasts (FIG 2001 ). The early work on
the general (Okamoto et al. 1987) and the later work on the accelerated longswing
(Y eadon et al. 2000; Arampatzis et al. 1998; 1999) established the importance of the
hips and shoulders, particularly as the gymnast passes under the bar. This phase of
the skill is characterized by a rapid hyper extension to flexion of the hips and hyper
flexion to extension of the shoulders. Quantifying the specific musculoskeletal de-
mands on current performers during the general longswing would provide useful
information for the development of this skill and may subsequently inform the de-
velopment accelerated longswing. Therefore, the aims of this study were to deter-
mine the contributions of the musculoskeletal system during the performance of the
general longswing and explain the overall interaction between the gymnast and the
elastic bar.
2 Method
2.1 Data Collection

Four members of the Men's UK National Gymnastics Squad participated in this
study (age = 22.5 ± 4.1 yrs, mass = 66.4 ± 7.2 kg, and stature = 1.69 ± 0.05 m). An-
thropometrie data were collected for use with a geometric inertia model (Yeadon
1990) to obtain subject specific body segment inertia parameters. All testing was
performed in a gymnastic arena on a standard competition high bar (Continental
Sports, Huddersfield, UK). Each gymnast performed three series of four general
longswings. Images in the sagittal plane were recorded using a digital camcorder
(Sony DSR-PD 11 00AP, 3-CCD, Japan) placed approximately 40m from the centre
of the activity at a height of 5 m with its optical axis at 80° to the plane of motion.
This provided a clear image of the functional phases of the longswing and particu-
larly limited the obstruction of the support upright of the high bar as the gymnast
passed the lower vertical. The camera was operated at 50 fields per second with the
electronic shutter set to 1/300 s. Calibration of the performance area was achieved by
placing a single calibration pole of height of 5 .176 m, containing four 0. 10 m spheri-
cal markers, at three pre-marked locations to form a vertical plane of approximately
5 m x 5 m. Reaction forces on the bar were recorded (1000 Hz) using strain gauges
bonded in pairs to the bar's surface. Calibration was performed by loading and
unloading the bar with known loads and recording the average voltages for each
loading condition. Vertical and horizontal bar stiffness (Kz and Ky) were used in
combination with linear regression equations to predict vertical (F 2 ) and horizontal
(F y) bar forces (Kerwin and Irwin 2006). Synchronization of the force and video data
was achieved through the use of 20 LEDs (Wee Beasty Electronics, Loughborough,
Leicestershire, UK) in the field of view of the camera which were sequentially illu-
minated at lms intervals. The force data capture and the LEDs were triggered simul-
taneously, enabling the force and video data to be matched to within 3 ms.
2.2 Data Processing

The images of the calibration object and the gymnast were digitized using the high
resolution TARGET motion analysis system (Kerwin 1995). Camera calibration was
achieved using an 8 parameter direct linear transformation (DLT) algorithm (Kwon
1999). In each field the centre of the bar, the centre of the gymnast's head and his
right wrist, elbow, shoulder, hip, knee, ankle, and toe were digitized. Based on Win-
ter's (1990) residual analysis, a digital low pass filter (6 Hz) was used to remove
random error from the reconstructed co-ordinates. Joint kinetics were determined
through the application of Newton's 2nd law of motion. The human performer was
modelled as a pin-jointed four link system comprising arms, trunk, thighs and
shanks. In order to minimize the propagation of errors, the closest known forces were
used to calculate the intemal joint forces. As such a combined approach of 'bar
down' to calculate the shoulder and hip forces and a 'toe up' to calculate the knee
and hip forces was used. The average of the two estimated hip forces was used
throughout the subsequent analyses. Muscle power (MP) was calculated as the prod-
uct of the muscle moments (MM) and angular velocity (ro) providing a measure of
the rate of work done. The mechanical work was calculated from the time integral of
the MP profiles for each joint and enabled the type of muscle action at each joint to
be specified. Muscle moments, powers and work dorre at the shoulders and hips were
calculated for each long swing. The total biomechanical energetic processes of the
gymnast performing the long swing were calculated using the relationship shown in
(Eq. 1). Equation 1 incorporates three major components including bar energy (Eq.
2), gymnast energy (Eq. 3) and by subtracting the combined bar and gymnast energy,
a value ofnet energy was calculated. For the gymnast in the handstand position on
top ofthe bar, the angle between his mass centre (CM) and the bar was set at 0°. To
compare within and between trials all data sets were interpolated in 1° intervals from
0° to 360° using a cubic 'lspline' function, (Mathcad 2001i, MathSoft Engineering,
lnc. Surrey, UK).
E total = -1 Im 2 + mgh + -1 m v
2
(1)
2 2
= -1F_Y_ + -1F_z_
2 2
E bar (2)
2 Ky 2 KZ
= L fMPi . dt
3
Egymnast
(3)
i=l
198 Gareth lrwin and David G Kewin
During the initial 90° of the descent, minimal muscular activity at any joint was
found, which corresponds to the findings of Okamoto et al. (1987). From 90°, an
extension-to-flexion moment at the hip joint precedes the ascending phase, which is
also reflected in the muscle power (Fig. 1.) This pattern concurs with Arampatzis et
al. ( 1998) although the magnitudes of their values are higher due to the fact they
investigated the accelerated longswing. In all cases the shoulders played a dominate
role particularly in the ascending phase. The peak hip moment was 48% of the peak
shoulder moment whilst the dominance of the shoulders was further emphasized with
the hips generating 40% ofthe peak muscular power at the shoulders (Fig. 1.).
5 ' 16
MMh MPh
14 -MPs
12
90 T80 270 360

-2 i
Angular Position ( 0 )
Angular Pos11ion ( 0 )
Fig. 1. Average Muscle Moment (MM) and Power (MP) at the hip (h) and shoulder (s) during
the general longswing on high bar.
The !arge extension (positive) moment at the shoulders (4.5 ± 1.7 Nm·kg" 1) and cor-
responding !arge positive powers demonstrate a concentric contraction around the
joint. Similar joint kinetics were reported by Arampatzis et al. ( 1998), but compared
to the study by Okamoto et al. ( 1987) the current study reports values 42% higher
which may retlect differences in modern technique and equipment. The majority of
work done by the performer occurred in the ascending phase with peak values of
0.81 ± 0.10 J.kg·' and 1.56 ± 0.76 J.kg" 1 at the hips and shoulders respectively. The
total energy of the bar-gymnast system is illustrated in Fig. 2.a. The maximum en-
ergy is achieved at approximately 160° with a value of 19 ± 4.1 J.kg" 1 which is com-
parable in magnitude to that reported by Arampatzis et al. (1999). The difference in
the total energy at the start (0°) and end (360°) of the longswing provides an indica-
tion of the success of the skill. Based on the conservation of mechanical energy the
difference must be equal to or greater than zero in order for the performer to retum to
the handstand position.
23
21
19
: 17
~ 15
i 13
>,
'° 11
90 180 270 360

90 180 270 360 -1 Angular position (°)
Angular Position ( 0 )
Fig. 2. (a) Average Total Energy and Net Energy (Enet) ofthe gymnast-bar system. (b) Aver-
age energy contributed by the gymnast's (Egymnast) musculoskeletal system.
lncorporated into the total energy profile (Fig. 2.a) is the strain energy at the bar,
the gymnast's musculoskeletal energy (Fig. 2.b) and the net energy that the swinging
gyrnnast possesses due to previous motion about the bar (Ene1). The findings of this
study show that 70% of the work contributed by the gyrnnast occurs relatively late in
the ascending phase.
4 Conclusions
This study has shown the gyrnnast's physical input into the longswing is a funda-
mental component of successful performance. The joint kinetics play a vital role in
understanding these variables and provides technical information relating to the
muscle actions and hence the physical demands placed on the gyrnnast. The gym-
nast's energy is required to compensate for friction at the bar hand interface, air
resistance and lasses of energy due to the bar not being perfectly elastic. In addition,
minor changes in timing of hip and shoulder actions, as explained by Hiley et al.
(2000), can remove energy from the system. The key active phase for the general
longswing for the shoulder and hip joints occurs consistently between 200° and 240°
of rotation about the bar.
References
Arampatzis, A., and Brüggemann, G.P. (1999) Mechanical energetic processes during the
giant swing exercise before dismounts and flight elements on the high bar and the uneven
parallel bars. Journal ofBiomechanics. 32, 811-820.
Arampatzis, A., and Brüggemann, G.P. (1998) A mathematical high bar- human body model
for analysing and interpreting mechanical- energetic processes on the high bar. Journal of
Biomechanics. 31, 1083-1092.
Federation International de Gymnastique (FIG) (2001) Code ofPoints, artistic gymnasticsfor
men. Switzerland.
Hiley, M.J., and Yeadon, M.R. (2003) The margin for error when releasing the high bar for
dismounts. Journal ofBiomechanics. 36, 313-319.
200 Gareth Irwin and David G Kewin
Kerwin, D.G. (1995) Apex/Target high-resolution video digitising system. In J. Watkins

(Eds.) Proceedings ofthe Sports Biomechanics section ofthe British Association ofSports
and Exercise Sciences. Leeds, UK. pp. 1-4.
Kerwin, D.G., and lrwin, G. (2006) Predicting high bar forces in the longswing. 6th Interna-
tional Sports Engineering Association Conference, Munich, Germany.
Kwon, Y.H. (1999) 2D Object plane dcformation due to refraction in !wo-dimensional under-
watcr motion analysis. Journal of Applied Biomechanics. I 5, 396-403.
Okamoto, A., Sakurai, S., Ikegami, Y., and Yabc, K. (1987) The changes in mechanical en-
ergy during the giant swing backward on the horizontal bar. In In L. Tsarouchas, J. Te-
rauds, B. A. Gowitzke, & L. E. Holt (Eds.) Biomechanics XIB, International Series on
Biomechanics. Amsterdam: Free University Press, pp. 338-345.
Readhead, L. ( 1997) Men 's Gymnastics Coaching Manual. Huddersfield, UK.
Yeadon, M.R. (1990) The simulation of aerial movement. Part II: A mathematical inertia
model ofthe human body. Journal ofBiomechanics. 23, 67-74.
Yeadon, M.R., and Hiley, M.J. (2000) The mechanics ofthe backward giant circle on the high
bar. Human Movement Scicnce. 19, 153-173.
6 Lawn Sports
Synopsis of Current Developments: Lawn Sports
Matt Carre
Sports Engineering Research Group, University of Sheffield, m.j.carre@sheffield.ac.uk
The topic of lawn sports could arguably include any sport that is played on natural or
synthetic turf, including those as diverse as lawn bowls, tennis and American foot-
ball. The sports that are included in this section are related to only three: cricket,
field hockey and rugby football, as the subject of soccer has generated enough papers
to warrant its own section. However, the research areas covered in this section dem-
onstrate the wide variety of expertise applied in the field of sports engineering, in-
cluding soil mechanics, impact modelling and optimisation, amongst others.
The one factor that all lawn sports have in common, of course, is the lawn (or .
turt) itself. Four papers in the following section concentrate on improving the under-
standing of the performance of turf surfaces. The paper by McLeod et al. examines a
new method for quantifying the amount of wear in synthetic turf surfaces. In recent
years, synthetic surfaces have seen increased use, due to advances in technology as
weil as changing lifestyles. Great effort is concentrated on designing the complex
systems that make up these surfaces, as weil as applying the experience required to
install a quality product. However, less is known about how different designs of
surface degrade over time, both through repeated use and exposure to the elements.
A better understanding of these issues will undoubtedly be improved by a measure-
ment technique such as the one proposed here.
Young and Fleming also discuss measurements of synthetic turf, in this case, the
type designed specifically for world-class field hockey, water-based pitches. Their
paper contains an in-depth critique of the test devices used to predict playing per-
formance of such a surface, including player and ball interactions and conclude with
some sound recommendations for future study.
The behaviour of natural turf, is equally as complex as a layered, synthetic sys-
tem, but requires a different kind of expertise. In the case of a cricket pitch, the play-
ing surface is designed mainly for interaction with the ball, providing a hard, con-
solidated surface which would appear alien to players from most other sports that use
turf. However, the performance of the pitch has huge implications for the way a
game of cricket is played. The paper by Shipton et al. examines how the mechanical
behaviour of soil changes through repeated rolling; one of the key elements in pitch
preparation. This fundamental research is vital to lead to a better understanding of
cricket pitch performance.
James et al. also examine cricket pitch performance, but their study is related di-
rectly to the interaction between ball and surface. Their paper describes a model of
oblique impact that can be used in conjunction with two relatively simple tests, to
204 Matt Carre
predict how a cricket ball rebounds off the surface; or in cricketing parlance, 'pace'
and 'bounce'. lt is hoped that this knowledge and technology can be used to aid
groundsmen in their preparation of quality pitches.
The remaining three papers are less concemed with what happens at the surface,
but rather what happens to the ball during play. Still on the theme of cricket, Justham
et al. discuss the quantification of a bowling delivery, one of the key factors in the
game. Using data collected during the thrilling Ashes series in 2005, fought between
England and Australia, they examine key aspects of professional deliveries and use
this information to aid the design and manufacture of a bowling machine.
Rugby football is the subject of the paper by Holmes et al., which again uses
measurements taken from professional sportsmen, but in this case a range of kicks
and passes are studied, which are all important in an actual game situation. This
. study results in the generation of an extensive data set of flight characteristics, im-
mediately after ball launch (velocity, spin and angle) which will serve as being very
useful for future studies of ball aerodynamics, ball-boot interaction and ball han-
dling.
Seo et al. examine the flight of a rugby ball after being kicked, for three different
kicking scenarios. They use multi-optimisation techniques to predict the best condi-
tions to be adopted in each kick, to obtain the desired results. This kind of research
has great implications for providing strategies that can be used by coaches and play-
ers alike.
In summary, this section demonstrates the wide variety of expertise, knowledge
and understanding in different areas of sports engineering. Once applied, this will
have a very positive impact on a range of exciting lawn sports.
Quantification of the Cricket Bowling Delivery; a Study
of Elite Players to Gauge Variability and Controllability
Laura Justham, Andrew West, Andy Harland, Alex Cork

Loughborough University, UK, L.Justham@Lboro.ac.uk
Abstract. The bowling delivery has been recognized as an important factor in cricket. The
batsman faces each delivery and attempts to read the bowler's actions to predict the type of
delivery and to avoid making an error in judgment which could cost the game. Numerous
studies have been carried out to investigate factors such as the biomechanic and kinematic
aspects of the bowling delivery and what information the batsman is able to pick up from the
delivery sequence. However the factors which constitute the bowling delivery, the mecha-
nisms adopted to bowl the ball and how a subtle variation in the ball release affects the deliv-
ery, have not been studied in such detail. This research is focused on understanding how the
bowler is able to control and vary their delivery pattems. Using performance analysis data
collected from the sccond Ashes test held between England and Australia in August 2005 two
bowlers have been studied over a six over bowling spei!. Information regarding the variability
within cach over has been analyzed to help quantify thc mechanics ofthe bowling delivery.
1 Introduction
The bowler is a key player in any cricket match as they can alter the outcome simply
by the way they deliver the ball. The mechanisms involved with creating a bowling
delivery have been investigated in terms of kinetic, kinematic, biomechanic, physio-
logical and anthropomorphic factors but not in terms of what the bowler actually
does to create the delivery or how consistently they are able to bowl over a pro-
longed period (Elliott I 986; Elliott 1993; Bartlett 1996; Glazier 2000; Noakes 2000).
Coaching manuals mention a correct grip, economical run-up, balanced delivery
stride at the crease and a fluent follow through but they omit to mention how to con-
struct a delivery from them. (Khan 1989; Ferguson 1992). Generally the bowling
delivery is classified by the speed of the ball at release, as shown in Table 1, with a
recognized range of speeds for each bowling type. Any further generalized classifi-
cation is avoided due to the unique features of each bowler.
A fast paced bowling delivery reaches the batsman in less then half a second,
which does not give him long enough to view the ball, work out his shot selection
and move in preparation. He must supplement the information available from the
ball's flight with information provided by the bowler during the preparatory stages of
the delivery. The batsman watches for variations in the length of the run-up, the
position and angle of the arm and hand as the ball is released and the grip on the ball
206 Laura Justham et.al.
as each of these have an effect on the delivery characteristics (Abemethy 1984; Pen-
rose 1995; Renshaw 2000). Bowlers will practice to make the difference between
their stock delivery and any variations as small and undetectable as possible. lt is
therefore not simply the speed of the ball or the mechanism of ball release which
constitutes the complete bowling delivery. There are a number of other factors to
consider which combine to formulate the complete delivery sequence. The purpose
of this research is to begin the process of understanding and quantifying the unique
and common ball release characteristics of elite cricket bowlers.
Bowler Classifica- Transit time (ms) Ball velocity Ball velocity (m/s)
tion (mph)
Fast 528-396 75-100 33.5-44.7
Medium 660-528 60-75 26.8-33.5
Spin 988-660 40-60 17.9-26.8
Table 1. A classification ofbowling with respect to the speed of delivery. The transit
time is the time taken from ball release to reaching the batsman (17. 7m)
Player Performance analysis is becoming increasingly important in all sports. The

Hawk-Eye ball tracking system is used as a television commentary tool and also as a
performance analysis tool. lt uses three orthogonal cameras to track the ball from the
moment of release to just before it impacts the bat. The speed of the ball and its
trajectory is recorded so the position of where the ball bounces and any swing or
deviation in the flight path may be calculated. This can show how the bowler's per-
formance changes over a prolonged period and under the pressures associated with a
match. Feedback Cricket is a video based analysis tool. From each delivery the area
of the pitch where the ball bounces, the shot selected by the batsman, where the ball
was hit and any runs scored from the ball are recorded.
The Ashes is a biennial series of 5-day test matches taking place between Austra-
lia and England. Access has been granted to data collected from the second test
match of the series, which was held at the Edgbaston ground in Birmingham, Eng-
land from Thursday 4th to Monday 8th August 2005. Data collected from the Hawk-
Eye and Feedback cricket systems have been analyzed for two right handed fast
bowlers over a six over spell in the first innings of the match to understand the vari-
ability and controllability that each player possesses.
3 Results
Hawk-Eye is a valuable tool, used here to compare the average delivery characteris-
tics of each bowler for every over in the bowling spei!. Figure 1 shows the average
speed of the ball release with respect to the length at which it pitched during the
delivery. The speed ofthe ball at release seems tobe weil controlled by both bowl-
Quantification ofthe Cricket Bowling Delivery 207
ers. Bowler 1 bowls with a speed range of ±3 mph for every over except over 6 when
he bowls one much slower ball. This slower ball is a deliberate variation on the stock
delivery and is used in an attempt to wrong-foot the batsman. Bowler 2 is more con-
sistent and bowls with a speed range of ±2 mph except during over 5 when he bowls
two slightly quicker balls as a variation to his stock delivery. There is not any notice-
able drop off in consistency of the speed of delivery over the course of the bowling
spell. Figure 1 shows that bowler 1 tends to bowl slightly more quickly than bowler
2, but most deliveries are clustered around the same speed of 84 mph to 86 mph
(37.6 m/s to 38.4 m/s) for both bowlers.
,. ,. 0G 1200 „ . .
Fig. 1. Hawk-Eye analysis comparing the average speed ofthe ball release with the
length at which it is pitched each over bowled.
Variations to the stock delivery come in many forms and the bowler practices to
make every delivery sequence as similar as possible in a bid to prevent the batsman
being able to perceive any difference in the delivery until the ball is released. The
major variations are in the pitching length and the speed at which the ball is released.
Often a quicker ball will be pitched shorter such that it will bounce up towards the
batsman more aggressively. Bowler 1 bowls at a consistent speed whilst altering the
pitching length of the ball much more frequently, for example in over 3 the speed is
consistent to within 2.6 mph (1.16 m/s) but the pitching position varies between
4.13 yards to 12.63 yards (3.78 m and 11.55 m).
Throughout the bowling spell both bowlers were predominantly bowling to right
handed batsmen. Figure 2 shows that they consistently pitched the ball on the off
side of the batsman. The variability in pitching length and line are larger due to ex-
temal effects acting on the ball during flight, for example airflow around the ball.
Bowler 1 varies the pitching length regularly and variations up to 8.5 yards (7.8 m)
in the pitching length and 0.6 yards (0.55 m) in the width are seen. For Bowler 2 the
pitching length varies by up to 6.6 yards (6.0 m) with the same variation in pitching
line as Bowler 1. Figure 3 contains images from the Hawk-Eye virtual environment
and shows how Bowler 1 and 2 vary in their deliveries. The images show over 3 of
the bowling spell, whereby bowler 1 varies the length of the ball considerably and
bowler 2 maintains as much consistency as possible.
1
1
i
!
t!'
i
.., 00
'1kNn911N .... 1Mcllftln'MO.llllf)Wdl.J
Fig. 2. Hawk-Eye analysis comparing the average length and line ofthe ball pitching
position both bowlers.
Fig. 3. Images from the Hawk-Eye virtual environment which shows the third over
bowled in the spell. Bowler 1 is on the left and bowler 2 is on the right.
Feedback cricket can be used to understand how the bowler creates their delivery
and the variations resulting from any alterations. The basic actions of both bowlers
are the same. They are right handed with a long run up, a front-on delivery action
and a pronounced follow through to dissipate excess momentum. The bowling arm is
kept very straight throughout the delivery process and the ball is released close to the
apex of the curve of the arm. Bowler 1 tends to vary the release position of the ball
from the apex of the curve to just following it which results in the large variations in
Quantification ofthe Cricket Bowling Delivery 209
pitching positions. A ball which is released later will pitch shorter than a ball re-
leased earlier. Bowler 2 tends to release the ball as close to the apex of the arm curve
as possible, resulting in the slightly more consistent pitching length.
The reaction of the batsman for each bowling delivery and how many runs were
scored is also recorded using Feedback Cricket. A shorter bowled delivery will gen-
erally force the batsman to play off the back foot and a more fully bowled delivery
will allow the batsman to come forward and play off the front foot. A good length
ball is one which causes the most indecision about whether to play forwards or
backwards. Figure 2 shows that both bowlers pitched the ball between 5 yards
(4.6 m) and 9 yards (8.2 m) from the batsman's stumps and from the data collected
from feedback cricket this shows that the ball was pitched at a good length to slightly
short of a length.
4 Discussion
Playc : II Machinc:
•Vory thc rein • position .... bno
,.mhu ·Control spccd of ball rein c
•Vary thc lcngth ofthc
dclivcry •Con I pitching linc and
• V ary thc pccd of thc
lcnath
dclivcry
• Prognmmablc voriations 10
· lntrodu c v1ri1tion to thc imuhlc hwnan bowling
lock dclivcry
•Bowler \Oruali rion 10 hidc
thc ma hinc
Fig. 4. The novel bowling machine has becn designed and developed using informa-
tion gathered about real bowling spells to ensure a realistic training environment.
Key aspects of both bowlers' deliveries have been considered using Hawk-Eye and
Feedback Cricket. The speed of ball release remains consistent throughout the bowl-
ing spell with variations being observed in an attempt to take a wicket. The position
where the ball pitches has greater variation duc to factors outside of the bowler's
control such as the motion of the ball through the air between being released and
impacting the pitch. The bowler tries to wrong foot the batsman and force them to
make errors with every delivery. However the Hawk-Eye analysis has shown that the
bowler does not appear to aim for the batsman's stumps, rather for the batsman
themselves, which entices them to attempt to play each delivery.
Even on a small scale it is very difficult to. quantitatively analyze bowling deliv-
erics. Each bowler subtlety varies each delivery and therefore unknown factors are
continually at play throughout any match. The bowler considers the score line, the
batsman who is facing him, the positioning of the fielders, the time of the day, how
many overs he has already bowled or is likely to still need to bowr and many other
critical aspects unique to the individual match. This initial study has shown that
modern bowlers, who have a similar playing history and coaching input, will main-
tain common playing techniques but will develop personal methods to vary the de-
livery, deceive the batsman into misjudging their movements and take wickets. Mod-
em players must be successful tacticians to reach international standard. Further
work must be carried out looking more in depth at specific aspects of the bowler's
delivery under more controlled conditions before further conclusions may be drawn.
Knowledge and understanding about the delivery of elite level cricket bowlers
has aided the manufacture of a bowling machine system at Loughborough University
which is able to recreate realistic delivery pattems as shown in Fig. 4. Existing bowl-
ing machines may simulate bowling deliveries but they do not have the integrity of
the additional background knowledge from the studies undertaken to understand the
bowling delivery. This analysis has enabled a better understanding of the ranges of
capabilities required in the novel bowling system as weil as the variations which
would generally be observed throughout an individual over or bowling spei!.
Acknowledgements
The authors would like to acknowledge the financial support of the Engineering and
Physical Sciences Research Council of Great Britain (EPSRC) and the IMCRC at
Loughborough University. They would also like to thank the players and coaches at
the ECB-NCC at Loughborough University.
References
Abernethy, B., Russell, D.G. (1984). "Advance cuc utilisation by skilled cricket batsmen."
Australian Journal ofSciencc and Medicine in Sport 16(2): 2-10.
Bartlett, R. M., Stockill, N.P., Elliott, B.C., Burnett, A.F. ( 1996). "The biomcchanics of fast
bowling in men's crickct: a revicw." Journal ofSports Sciences 14(5): 403-424.
Elliott, B., Baker, J., Foster, D. ( 1993). "The kinematics and kinetics of the off-drive and on-
drive in crickct." Australian Journal of Science and Medicinc in Sport 25(2): 48-54.
Elliott, B. C., Foster, D.H., Gray, S. (1986). "Biomechanical and physical factors influencing
fast bowling." Australian Journal of Science and Medicine in Sport 18( 1): 16-21.
Ferguson, D. (1992). Cricket: Technique, Tactics. Training, The Crowood Press Ltd.
Glazier, P. S., Paradisis, G.P., Cooper, S-M. (2000). "Anthropometrie and kinematic influ-
ences on release speed in men's fast-medium bowling." Journal of Sports Sciences 18(12):
1013-1021.
Khan, K. (1989). lmran Khan's c-ricket skills, The Hamlyn Publishing Group.
Noakes, T. D., Durandt J.J. (2000). "Physiological rcquirements ofcrickct." Journal ofSports
Scicnces 18( 12): 919-929.
Penrosc, J. M. T., Roach, N.K. ( 1995). "Decision making and advanccd cue utilisation by
cricket batsmen." Journal of Human Movement Studies 29(5): 199-218.
Rcnshaw, 1., Fairweather, M.M. (2000). "Cricket bowling deliveries and the discrimination
ability ofprofessional and amateur batters." Journal ofSports Scienccs 18(12): 951-957.
Ball Launch Characteristics for Elite Rugby Union
Players
Christopher Holmes, Roy Jones, Andy Harland and Jon Petzing

Loughborough University, c.e.holmes@lboro.ac.uk
Abstract. The role played by a team's kicker in determining the outcome of a rugby union
match is becoming increasingly important. However, unlike in other sports, there is no exist-
ing data regarding the kicking and passing abilities of elite rugby players. The objective of this
study was to determine the launch characteristics of a place kick, drop kick, spiral kick (kick
to touch) and spin pass. Testing was carried out at senior English League rugby union clubs,
and data from 14 elite kickers were evaluated including current international playcrs. The
subjects were asked to perform the different kicks on a specially markcd rugby ball at a dis-
tance of 60 m from the posts. Each skill was performcd until they had achieved five 'good'
strikes or passes. A high speed camcra (NAC 500), operating at 500 frames per second was
used to record the ball velocity, spin and launch angle. The data presented shows that players
are able to achieve velocities of 38.1 m/s whilst imparting 405 rpm to a rugby ball (drop kick).
The maximum spin rates seen in the other types of kick are considerably lower. The study of
the spin pass has shown that whilst players impart considerably lower lcvels of velocity to the
ball (18.3 m/s), they are capable of achieving spin rates similar to those seen for a place and
spiral kick.
1 Introduction
The specification for a 'match ball' (rugby) is defined by the governing body, the
International Rugby Board (IRB). The ovoid ball specification falls into three main
categories, the dimensions of the ball, weight of the ball and pressure of the ball at
the start of play. The parameters defined are all static values, and dynamic ball per-
formance is not measured. This is similar to the majority of other ball sports, how-
ever the balls behaviour during play is of importance and other governing bodies are
introducing dynamic performance criteria. In order to develop dynamic ball assess-
ment procedures it is necessary to appreciate the capabilities of elite players. This
data can then be used to define the parameters for the dynamic ball tests. The aim of
this study was to obtain the ball launch characteristics of a spin pass and three differ-
ent types of kick, using professional players from senior English League rugby union
clubs. The three kick types selected arc the place kick, drop kick and spiral kick.
There is currently no literature detailing the initial launch characteristics of a
rugby ball during various types of kicks or passcs. Macmillian ( 1975) used three
212 Christopher Holmes
skilled footballers to perfonn three different types of kicks seen in Australian Rules
football. The drop punt and drop kick are similar to the spiral and drop kick investi-
gated during this study. A high speed camera operating at 400 frames per second
with a shutter speed of l/1200s was used during the testing. Ball velocities of 24.9
and 27.2m/s were stated for the drop kick and punt respectively. Hartschuh (2002)
studied the ball launch characteristics during an American football punt. 13 punts
were filmed using a 30 frames per second video camera. A ball velocity of
24.5±0.6m/s and an angle of 49.4± 1.5° were stated.
A number of studies have examined ball velocities achieved during the kicking of
a soccer ball. Neilson (2003) carried out a comprehensive study of 25 professional
players, at five senior English football clubs. A high speed camera operating at 500
frames per second with a shutter speed set to 1/1 000s captured the initial trajectory
of the ball after impact. A maximum-recorded velocity of 33.1 m/s was stated for a
füll power kick, with a maximum ball spin of 833rpm achieved during an instep and
outstep swerve kick. The velocities stated by Neilson (2003) were similar to the
maximum measured during a study by Asami and Notle ( 1983), 34.0 m/s. Of signifi-
cance is that the majority of studies do not include spin data. The study aims to ob-
tain comprehensive launch data which can be used for fürther scientific study.
2 Method
Player testing was carried out at six professional English rugby clubs, with data from
14 players analysed and presented. All players were established kickers and had
international representative honours, including four füll intemationals. The clubs
consisted of five English Premiership clubs and a national division one club. The
testing procedure was carried out on natural turf training pitch environments. Seven
unbranded 'match balls' were used, to help negate any player brand bias that may
exist. The unbranded balls were marked with a series of constant and dashed lines to
enable the accurate detennination of spin during the digitising process, "Fig I ".
The subjects were requested to perfonn three different types of kick at füll
power, enabling maximum velocity to be achieved, at a distance 60m from a set of
posts. The first kick was a place kick, the second a drop kick and the third a spiral
kick. The subjects were asked to perfonn each skill until they had achieved five
'good' strikes or passes. The test players all wore their own boots and supplied kick-
ii:ig tees. Only one player analysed during the study kicked with their left foot. The
player details are shown in "Table I ".
Table 1. Mean (±SD) tcst subject data (n=l4)
Age (years) 25± 3.74

Body mass (kg) 88.06 ± 5.72
Body height (m) 1.78 ± 0.03
Boot size 9.71 ± 1.14
The test procedure was approved by the University Ethics Committee and players
were informed of the test procedure and their rights prior to testing.
The initial movement of the ball after impact or throw was captured using a high-
speed camera (NAC 500) operating at 500 frames per second with a shutter speed of
III 000s. A composite high-speed video image can be seen in "Fig l ". The high-
speed camera footage was digitised using Image Pro Plus software, and the ball
velocity and launch angle numerically calculated. The software allows the displace-
ment of the centre of the ball to be calculated, over a given time interval.
Fig 1. Composite image of place kick
The uncertainty of the measurement, Q TOTAL, can be defined using Eq 1.

The uncertainty of the device, Q DEVICE, is assumed to be very small by comparison to
other uncertainty that exists within the measurement procedure. The uncertainty of
the setup, Q SETUP, is attributed to the ball flight not being perpendicular to the cam-
era placement. Q SETUP, is difficult to measure and was controlled by accurate align-
ment of the cameras position, and requesting players to strike the ball along a given
trajectory. Uncertainty of the analysis, Q ANALYSIS, can be described as a measure of
repeatability. The repeatability of the analysis was calculated and defined for veloc-
ity as ±1.33m/s (95% confidence) and for launch angle ±0.29° (95% confidence).
Q TOT AL = Q SETUP + Q DEVICE + Q ANALYSIS (!)
Eqs 1. Uncertainty of the mcasurement procedure
The measurement of spin has been analysed by evaluating the number of

frames required for a quarter rotation of the ball. The spin defined for the place and
drop kick is tumble axis backspin, whilst rifle spin is the main component of a spiral
kick and spin pass. The distance to the first bounce of each of the kicks is measured
using the Bushnell Yardage Pro 1000 laser range finder to an accuracy of ± 1m.
214 Christopher Holmes
The test results are summarised in table 2 and shown in graphical form in "Figs 2-4".
Table 2. Mean (±SD) test results for all kicking and passing data
Kick/ Pass Velocity (m/s) Ball Spin (rpm) Launch Angle (0 ) Distance (m)
Place 26.44 ± 2.97 238.10 ± 44.92 30.22 ± 4.41 53.74 ± 5.72

Drop 25.60 ± 3.77 234.25 ± 66.57 35.76 ± 4.28 51.30 ± 5.70
Spiral 28.06 ± 3.70 216.41 ± 46.11 43.91 ± 4.55 55.42 ± 7.22
Spin Pass 13.79 ± 1.48 219.09 ± 32.98 12.19 ± 5.26
The average ball velocities achieved during the different types of kicks were similar.
A maximum recorded velocity of 38.lm/s was achieved during a drop kick. Macmil-
lan (1975) stated a mean velocity of 27.2m/s (89.2 ft/s) during an Australian rules
drop kick. This lower value could be due to the differences between the types of
balls investigated, and an increase in player development. The maximum recorded
velocity for the spiral and place kick were similar, 33.6 and 33.5 m/s respectively.
The maximum velocities recorded were 5m/s higher than the füll power soccer kick
recorded by Neilson (2003). The differences between these values could be due to
the differences between balls being kicked and the greater mass of the rugby players.
The average velocity value for the spin pass was considerably lower than the kicking
actions, but the low standard deviation illustrates the repeatability of the simple ac-
tion. A maximum velocity of 18.3 m/s was recorded for the pass.
40
400
E
_§: 300
.,
ni
II::
.!: 200
Q.
U)
10
Pl■c• kick Drop koclc Spu1I kick Spin pass Pl■ ca kick Drop kick Spiral kick Spin pass
Fig 2. a) Boxplot ofball velocity data b) Boxplot ofball spin rate data
The average ball spin produced during a spiral kick was lower than that of a drop
or place kick. This maybe attributed to the fact that the spiral kick had a higher ve-
locity. During the natural kicking motion of a drop or place kick, tumble axis spin is
imparted onto the ball as it is struck below centre. Large amounts of rifle spin could
be imparted onto a spiral kick, but this would cause the player to kick across the ball,
reducing the velocity and overall distance. The maximum spin value of 405rpm was
recorded during a drop kick. The average rifle spin imparted onto a ball during a
pass is larger than that of a spiral kick. The maximum ball spin for a spiral kick and
spin pass were 341 and 288rpm respectively.
60
: so
~
Cl
~ 40
.,
"gl30
<(
C
.Q 20
(Ö
.,
>
iij 10
Ptace kick Dfop kick Spiral kick Spin P"'•
Fig 3. Boxplot ofball elevation angle data
70
f., so
.
u
C
~ 50
C
40
Ptace kick Drop kick Spiral kick
Fig 4. Boxplot ofball distance data
The lowest angle recorded was for the place kick. During a drop and spiral kick
the ball is dropped before contact with the foot, which may account for this increase.
216 Christophcr Holmes
The increase in launch angle between a spiral and drop kick was noted by Macmillan
(1975) whose results state that the launch angle of a spiral kick was 7° greater than
that of a drop kick.
The maximum distance achieved during any kicking action was 71 m (spiral
kick). During the investigation players found the spiral kick the most difficult kick-
ing motion to replicate, which accounts for the increased standard deviation. The
maximum distances for the place and drop kick were 63 and 60m respectively.
4 Conclusion
The present study provides the first comprehensive data for launch characteristics of
elite rugby players. The mean initial velocity for the three types of kicks ranges be-
tween 25.60 and 28.06 m/s, depending upon kick type. The mean velocity for the
spin pass was significantly lower at 13.79 m/s. Maximum velocities of 38.05 and
18.30 m/s were recorded for a kicking action (drop kick) and spin pass respectively.
The maximum spin rate obtained by a professional player in this study was 405 rpm
(drop goal), which was considerably higher than data recorded for the pass (288
rpm). The launch angle increased during a drop goal to 43.91± 4.55°, in comparison
to the other kicking actions, which compares to results stated by Macmillian ( 1975).
The most difficult skill to replicate during the testing procedure was the spiral kick.
This accounts for the higher standard deviation when examining the distance meas-
urement. The maximum distance measured was 71 m (spiral kick) to the first bounce.
The data from this study can be used to define the parameters of any dynamic
tests that are to be developed in the future. The study used elite players, which en-
sures that the values obtained are close to the maximum achievable.
Acknowledgements
The authors would like to thank the rugby clubs, and players who agreed to partici-
pate in this study, and would also like to thank adidas for supplying the balls. Thanks
also to the EPSRC Advance Instrument Pool for the use of the high speed camera
References
Asami, T and Nolte, V (1983). "Analysis of Powerful Ball Kicking". Biomechanics

VIII-B, 675-700
Hartschuh, R. (2002). "Physics of Punting a Football."
http://www.wooster.edu/physics/Jr!S/Files/Ryan.pdf (Accessed - 07 /07 /05)
Macmillian, M. ( 1975). "Determinants of the Flight of the Kicked Football." Re-
search Quartley 46: 48.
Neilson, P.N. (2003). "Dynamic Soccer Ball Performance Measurement." 5th World
Congress ofScience and Football (Lisbon, Portugal, 11-15 April)
A Novel Quantitative Method for the Determination of
Wear in an Installed Synthetic Turf System
Andy Mcleod, Iain James, Kirn Blackbum and Gavin Wood

Cranfield University, UK, a.j.mcleod.s0 l@cranfield.ac.uk
Abstract. This study focuses on the initial development of an image analysis meth-
odology for quantifying the wear and degradation of synthetic sports turf, post instal-
lation, where the carpet/infill system is subjected to systemic abrasion and wear from
play and maintenance. The pilot study images the surface of polypropylene fibres,
which have been agitated with differing sand infill types, with a scanning electron
microscope. The resultant images were analysed to determine the degradation of the
extrusion features evident in virgin fibre, and it was found that there was significant,
quantifiable wear of the turf fibres after seven days with all test sands. The image
data for fibres between 7 and 28 days was dependent upon sand type. Further devel-
opment of the technique is required for determining the next stage of wear - charac-
terized by pitting of the fibre surface by the sand.
1 Introduction
The use of a quantitative model of fibre wear within a synthetic sports turf system
will enable identification of the main causes of wear and, by a change of materials
and management techniques, will allow significant advances in the financial sustain-
ability of synthetic turf surfaces.
At present, within the sports industry, the measurement of the resistance to wear
of synthetic turf sports surfaces is carried out pre-installation. Normal testing of the
integrity of the carpet system is carried out by measuring the tensile properties of
individual fibres, the effect of ultra violet light and aging. Further fibre is subjected
to simulated wear by meta! blades, abrasive wheels and the reproduced action of
football studs (FIFA, 2005; 8S7044, 1990). These methodologies were designed for
testing the compliance of a product with the relevant specifications for installation
and performance; such approaches to testing do not characterize the wear mechanism
of the whole system of polymer carpet fibres and infill materials in the field envi-
ronment
The usual method for wear quantification is by mass loss measurements, which is
suitable when the wom material components are detached from the main sample.
Due to the low mass of the wom turf fibre components there is a loss of resolution
when weighing, even on a high precision balance, which makes this method less
effective for use with synthetic sports turf. In addition, this method provides no
information on the distribution of wear over the component and will not show the
level of wear in areas of fibre structural failure (Gahlin & Jacobsen, 1998).
218 McLeod et al.
This paper will show the results of a pilot study for a field assessment method, in
development, which aims to objectively quantify the 'degree of wear' of fibre sam-
ples and identify the processes causing the wear of the fibre, using surface image
analysis techniques. The aim of this experiment was to quantify the effect of sand
infill abrasion on the fibres of an artificial turf surface.
2 Materials and Methods
2.1 Test Materials

The fibres used were from a stock roll of 4 mm wide, extruded monofilament poly-
propylene used in the manufacture of synthetic sportsturf. Fibres were cut by scalpel
to a length of 23 mm.
The infill materials comprised two rounded sands commonly used in sand filled
(2 nd generation) synthetic turf surfaces in the UK, with the trade names 'No 21' and
'2EW'; in addition a sharp sand was used as a contrast.
2.2 Test Method

50 g of each sand was placed with three of the 23 mm long fibres specimens into a
250 ml polypropylene screw capped bottle. The bottles were placed on an over-and-
under shaker and rotated through their long axis at 28 rpm. Treatment periods were
7, 14, 21 and 28 days, with an untreated control. Each treatment combination of
sand and period was replicated five times. On removal from the shaker the fibre
samples were washed gently with distilled water to remove sand particles and al-
lowed to air dry.
2.2 Fibre imaging technique

Initial investigations of images recorded with high power optical microscopy, using
transmitted and reflected light, determined that such techniques were not appropriate
for quantification ofthe degree ofwear.
Subsequently, the fibres were imaged using an FEI XL30 SFEG scanning elec-
tron microscope (SEM) to investigate the pattem of abrasion and how this contrib-
uted to surface deformation and ultimately plastic and brittle failure of the turf fibres.
Before the fibres were viewed using the SEM they were sputter coated with a
gold/palladium (80/20) mix, to increase the conductivity of the sample. The SEM
method ensured consistent resolution, magnification (x200) and luminous flux den-
sity among the replicates of each treatment. Images were captured from each repli-
cate. Output images were 256 grey scale, 712 x 484 pixel in tagged image file for-
mat.
A Novel Quantitative Method for the Determination ofWear 219
Fig. 1. SEM images of the control and after 7 and 28 days of continuous agitation with the
2EWsand
10000
9000
8000
----~-
7000
6000
g
III
:, 5000
!
1.1.
4000
3000 - 7days
- - - - 14 days
2000 - - · - 21 days
- - - 28days
1000
- - • • • • Control
0
N
0
ao
;j; .
1D
<';'
.
";
,.._
N
N
~
N ao
Angle
Fig. 2. Frequency distribution of image processing output: frequency of orientation in each 2°

angle class within a 7 x 7 window for the mean ofthe five 2EW sand images. Maximum
frequency decreases with increasing wear period; maximum frequency for the control is 11922
2.3 Image analysis

Initial visual inspection of the images determined that the linear pattems on the vir-
gin fibre, resulting from the extrusion process were degraded into random pattem
orientations with increased wear period (Fig. l ). In extremis this resulted in punc-
tured fibres with point defects that were related to fibrillation in SEM images of field
samples of synthetic turf of 5 years age.
220 McLeod et al.
To detennine objectively whether this pattem of degradation was related to sand

type and wear period, a filtering process was used to quantify 'linearity' in each
image using the Leica Erdas lmagine v8.7 image analysis software.
The fitter comprised two 7 x 7 matrices that measured horizontal and vertical
continuity in each image. The resultant of each component was detennined as an
angle, and a frequency distribution of pixels within 2° classes, between 0 and 90°
detennined. Images with a high linearity were characterized by a distinct peak at an
angle between 2 and 4° (the specific angle is an artefact of fibre alignment on the
SEM, it is the peak that is significant, see Fig. 2). Images with less linearity were
characterized by a more unifonn distribution of pixels between 0 and 90° (see Fig.
2). To quantify this characterization, the ratio ofthe maximum frequency to the total
number of pixels within the distribution was detennined, referred to as the peak ratio
(PR). lt was hypothesized that PR would decrease with increasing wear period.
3 Results
Table 1. Mean pcak ratio of linearity in SEM images of synthctic turf fibres subjected to wear
by three sands for four wear periods
Mean peak ratio

Wear period. days
Treatment 7 14 21 28
Control 8.76'
2EW 3.28' 3.ÜÜb 2.89'·d 2.82d
No21 2.87d 2.98b.c 2.93b,c 2.93b,c
Sharp 3. 19' 2.98b,c 3.25' 2.88'·d
Means with identical superscripts cannot be scparated by the LSD ofO. I at p=0.05
34
♦ 2EW
3.3 ♦
O No21
32 t:,.
i:i t. Sharp
I
!!:. 3.1
2 3
.
~
.,,_
Q)
D..
2.9
□
2.8
2.7
26
2.5
0 7 14 21 28 35
Wear penod, days
Fig. 3. Mean peak ratio for each sand - wear period. Error bar represents thc LSD at p=0.05
For all treatments, PR reduced significantly after 7 days ofwear, compared to the
control (p<0.001). The 2EW sand behaved as hypothesised with a significant de-
crease in PR until 21 days (Table 1). No 21 did not follow a consistent pattem, with
A Novel Quantitative Method for thc Determination ofWear 221
an anomalously low PR value at day 7 and no significant difference between days 14

- 28 (Fig. 3). The sharp sand followed a similar pattem to the 2EW except for an
outlier at 21 days.
4 Discussion
The results of this trial illustrate the potential of the SEM image analysis methodol-
ogy for the quantification of wear of synthetic turf in field samples of different ages.
A clear effect of abrasion by sand infill material has been shown and quantified.
For application in field samples, however, more development is required. In this
experiment the effect of abrasion on this PR index of 'linearity' was immediate. The
PR value measured the abrasion of the linear extrusion features on the fibre surface,
it did not reflect the severe 'pitting' of the surface at 28 days, which is not linear and
more irregular - further image processing techniques are being investigated to iden-
tify these features as they are believed to be critical in creating point defects in the
fibre.
Further work will also include a quantification of wear and an investigation of
wear mechanisms in the field for fibre collected from field sites of different usage
and maintenance. The final objective is to characterize and quantify the effect of
maintenance equipment on turf fibre and surface' performance.
5 Conclusion
A directional fitter image processing technique was successfully applied to scanning

electron microscope image of synthetic turf fibres following increasing periods of
wear with different sands used in the infill of sand filled synthetic turf surfaces.
Significant wear of the turf fibres was identified after seven days with all three
sands. Over a 7 to 28 day period, further wear was identified with the 2EW and
sharp sands; this effect was not observed with the No 21 sand.
Further development of the technique is required for determining the next stage
of wear - characterized by pitting of the fibre surface by the sand.
Acknowledgements
The authors gratefully acknowledge that this research was funded by the Engineering
and Physical Sciences Research Council of the UK and the Institute of Groundsman-
ship.
References
Adachi, K., Hutchings, I.M. (2005) Sensitivity ofwear ratcs in thc micro-scale abrasion test to
test conditions and material hardness. Wear. 258, 318-321
8S7044 (1990). BS 7044-2.3: 1990: Artificial sports surfaces methods oftest: methods for
detcrmination of durability. British Standards Institution, London, UK
FIFA (2005) FIFA Quality concept: Handbook of test methods and rcquiremcnts for artificial
football turf. FIFA Marketing & TV AG, Switzcrland.
222 McLeod et al.
Gahlin, R. & Jacobsen, S. ( 1998) A novcl method to map and quantify wear on a micro-scale.
Wear. 222, 93-102
Hertzberg, R.W. ( 1983) Deformation and Fracture Mechanics of Engineering Materials. John
Wiley & Sons, New York, 826 pp
Multi-optimization of Three Kicks in Rugby
Kazuya Seo 1, Osamu Kobayashi 2 and Masahide Murakami3
1 Yamagata University, seo@e.yamagata-u.ac.jp

2 Tokai University
3 University ofTsukuba
Abstract. What are the features ofthe optimal kicks in rugby? What is the best way to achieve
satisfactory kicks? The objective of this paper is to provide the answers to these questions by
optimizing the initial velocity and angular velocity vectors for three kinds of kick - the punted
kick, the kick into touch and the kick for goal.
1 Introduction
There are three kinds of kick in rugby, the punted kick (the ball not spinning), the
kick into touch (the ball spinning on its longitudinal axis) and the kick for goal
(spinning on its transverse axis). We have measured the aerodynamic forces for each
of these cases (Seo, Kobayashi, Gotsu and Murakami 2003) (Seo, Sakamoto, Koba-
yashi and Murakami 2005), and the results can be summarized as follows. In the case
of the non-spinning ball, the drag increases with increasing angle of attack, a, which
is the angle between the longitudinal axis and the flight path. The lift increases up to
about a = 60°, when the effects of aerodynamic stalling occur. The pitching moment
has positive values (nose-up) except when a = 0 or 90°. The side force depends on
the position of the lace since the flow around the ball, which is affected by the lace
and the 4 seams of the ball, is asymmetric. Therefore, it is considered that a punted
kick rotating at lower spin rates tluctuates in the lateral direction during the tlight. The
rolling moment and the yawing moment are assumed to be O (Seo, Kobayashi and
Murakami 2004). In the case of a ball spinning on its longitudinal axis, the drag, the
lift and the pitching moment are almost same as the non-spinning case, though there
is a slight quantitative difference for C 0 . The side force increases with increasing ffi
because of the Magnus force except when a = 0°. In the case of a ball spinning on its
transverse axis, the lift increases with increasing ffi because ofthe Magnus force. The
Magnus force is at its maximum when the spinning (transverse) axis is perpendicular
to the flight direction. The side force depends on the angle between the spinning axis
and the direction offlight. The side force is a maximum when this angle is 45°, while
it is Oat 90°.
In order to optimize the three kicks, the flight trajectories need to be simulated.
This can be solved by integrating the füll nonlinear six-degrees-of-freedom equations
of motion numerically (Seo, Kobayashi and Murakami 2006) on the basis of an
aerodynamic database constructed from wind tunnel test data. What are the best
224 Kazuya Seo
kicks to use? The objective of this paper is to calculate the features of the optimal
kicks in each case.
2 Modeling
The inertial coordinate system is shown in Fig.1. The origin is defined by the inter-
section between the player's own goal line and the left touch line, with the XE-axis
along the touch line towards the opponents half, the YE-axis along the goal line to-
wards the goal posts and the ZE-axis vertically downwards.
Tabl e 1 C ontro parameters

Control oarameters Abbr.
Velocity vector v.
Flight path angle Yo
Azimuth angle Yn
Angular velocity vector %
Elevation angle of ro 0 lo
Azimuth angle of ro0 Ko
Yaw angle 'Po
Pitch angle 0o
Roll angle <l>o
Fig. 1. Interna! coordinate system

XE
/10 (or%)
~o(or , 0)
<l>
½
Fig. 2. Velocity vector Fig. 3. Euler angles
(Angular velocity vector)
The control parameters for optimization are the initial conditions, which are the
magnitude ofthe velocity, the direction ofthe velocity, the magnitude ofthe angular
velocity, the direction of the angular velocity and the Euler angles as shown in Table
1. Figure 2 shows the velocity vector as weil as the angular velocity vector and fig-
ure 3 shows the Euler angles. The coordinate system in the frame of the ball is de-
noted by xh,Yh, and zh with the origin at the center of gravity of the ball. The xh axis
is along the longitudinal axis, while the Yb and zh axes are along transverse axes with
Multi-optimization ofThree kicks in Rugby 225
the zb-axis in the direction of the valve. The sequence of rotations conventionally
used to describe the instantaneous attitude with respect to an inertial coordinate sys-
tem is shown in Fig.3 (Stevens, 1992). Starting with the inertial coordinate system
the following sequence is followed; 1. Rotate about the ZE axis, nose right (positive
yaw 'P), 2. Rotate about the y 1 axis, nose up (positive pitch 0), 3. Rotate about the xb
axis, right wing down (positive roll cf>).
3 Multi-optimization of the Punted Kick
The punted kick rotating at low spin rates fluctuates during flight. The objective here
is to achieve the maximum number of extreme positions in the XE and Y E directions
as well as the longest hang time. This makes it more difficult for the opposition to
catch the fluctuating ball. A long hang time reduces the chances of the opposition
being able to mount a counter-attack by tackling themjust after the catch.
70
60
50
g
,,t 40 40
'<l
>-"' 30 30
:><"'
20
10
o L___.L-=====~==::::i:::....---'_ _J
34.7 34.8 34.9 35 35.1 35.2 35.3
Time(s) YE (m)
Fig. 4. Time variations of the position Fig. 5. Catcher's view
The punted kick from the center of the field is optimized. The initial position is
assumed tobe (XE, Y E, ZE) = (50, 35, -0.5). The ranges of the nine control parame-
ters in Table 1 are defined as follows: 15 <I v0 f< 25m/s, 15 <y 0< 85°, -90 <XJ.J< 90°, 0
<fw0 I< lrps (revolutions per second), -90 <1-0< 90°, -90 <1<o< 90°, -90 <%< 90°, -90
<0 0< 90°, & 0 <<t> 0< 360°. Three objective functions are considered, two of which
are the number of extrema in the XE-axis and the Y E-axis. The third objective func-
tion is the hang time. For optimization all the objective functions must be maxi-
mized. Since there are three objective functions, the problem is a multi-optimization
problem. To carry out optimization, the elitist multi-objective genetic algorithm was
applied (Deb, 2003).
The optimum flight trajectory is shown in Figs 4 & 5. The optimal initial condi-
tions were found tobe lv0 i=25m/s, yo=80°, XJ.J=-2.7°, Im0 i=0.94rps, 1-0=99°, Ko=-128°,
'!10 =85°, 0 0 =-84°, & <f>o=7.7°. The hang time is 5.4 seconds. The number of extrema
226 Kazuya Seo
in the XE-axis is 3, and that in the Y E-axis, 13. If the velocity vector is not included
in the XE-ZE plane, there are extremas in the XE-axis because of the side force. Fig-
ure 5 is the catcher's view. The amplitude of the fluctuations is from several to ten
centimeters. Since it must be difficult for the opposition to catch this kind of punted
kick, it becomes a powerful weapon in the game.
4 Multi-optimization of the Kick into Touch
Two kicks into touch from the behind the 22m line are optimized. Two images of the
flight trajectory are shown as I & II in Fig.6. The spin direction is assumed to be
clockwise as viewed from behind. This spin direction is normally given by the left-
footed kicker. The initial positions are assumed tobe (XE, Y E, ZE) = (20, 50, -0.5) for
I and (XE, Y E, ZE) = (20, 20, -0.5) for II. The flight time is denoted by tr and two
objective functions are considered, one of which is the distance achieved in the for-
ward direction at the end of the flight time, tr , and the other is the absolute value of
the difference in Y E between the ball and the touch line at tr, i.e., the distance be-
tween the touch line and the ball when it lands on the ground. Although there are
nine control parameters in Table 1, only six of these are taken into account. Since the
longitudinal axis coincides with the spinning axis, 1-0 & Ko are automatically deter-
mined from the Euler angles. Moreover, the initial roll angle <1> 0 is unimportant and is
assumed to be 0° in this case.
The results are shown in Figs.7 & 8. The data for I are shown by the open circles,
and those for II are shown by the open triangles. The two objective functions are
shown in Fig.7. The flight distance in the forward direction, XE-20, increases as the
absolute value of the difference in Y E between the ball and the touch line, Ll Y E, in-
creases. Although the ideal situation is given by the largest flight distance in the
direction of XE and the smallest value of Ll Y E, it is impossible for these conditions to
be satisfied simultaneously. lt can be seen that the flight distance of II is longer than
that of !; that is, the optimal kick made using the leg nearest the touch line produces
a greater distance than that made using the leg furthest from the touch line. However,
the range of optimal solutions for II is narrower. The relationship between the opti-
mal y0 and the optimal 0 0 is shown in Fig.8, with the solid line showing 0 0 = y0 • lt
was found that 0 0 is comparable with y0 in the case of I, while 0 0 is slightly greater
than y0 in the case of II. The condition where 0 0 = y0 means that there is no preces-
sion at t = Os. In the case of 0 0 > y0 for II, there is precession owing to the difference
in angle betwecn the velocity and angular velocity vectors, and the precession is
clockwise around the velocity vector from the kicker's point of view. This helps
prevent any lateral movement in the negative Y E direction in the very primary phase.
Multi-optimization ofThree kicks in Rugby 227
e
0
~
UJ
X
0 0.5 V (m) 1.5
Fig.7. Objecti e function
> ~
45 ~
. ·······J J ~0 070 1
0
....... 40
' 35 - 1
0 1
z 0 35 40 45 50 55
Yo ( )
Fig.6. Kick into touch Fig.8. The relationship between the initial
by a left-footed kicker angle of attack Yo and the pitch angle 0 0
5 Multi-optimization of the Kick for Goal
The kick for goal from 45° on the 22 meter line is optimized. The initial position is
assumed tobe (XE, Y E, ZE) = (78, 13, -0.145) and the time at which the ball passes
over the crossbar is denoted by tr. Two objective functions are considered. One is the
absolute value of the lateral deviation of the ball from the center of the goalposts at
tr, and the other is the difference in height between the ball and the crossbar at tr.
Since the transverse axis coincides with the spinning axis, to & Ko are automatically
determined from the Euler angles. Therefote, seven of the nine control parameters in
Table 1 are taken into account for this optimization.
Two objective functions are shown in Fig.9. Although the smallest lateral devia-
tion and the largest height difference give the ideal situation, it is impossible for
these conditions tobe satisfied simultaneously. The left limit solution in Fig.9 is that
in which the deviation is almost O (the ideal situation), but the height difference is
relatively small. The flight trajectory in this case is shown in Fig.10. The optimal
initial conditions are 1110 1 =25m/s, y0 =41 °, x0=45°, 1%1 = lOr.p.s, 'l'o =51 °, 0 0 =90°, &
<I>o =6°. lt can be seen that the projection of the flight path onto the surface of the
pitch is a straight line without any hook. The optimal y0 and optimal Xo should be in
the ranges of 41.2 < y0 < 42.5° and 41.9 < Xo < 45.0°. The deviation is almost 0, when
Xo = 45.0°. This means that the direction of the velocity vector points toward the
center of the goalposts, and no side force acts on the ball. If the ball tends to hook
because of a side force, it has to travel a greater distance and loses altitude. lt can be
concluded that the angular velocity should be perpendicular to the velocity as weil as
the longitudinal axis in order to meet the optimal conditions.
228 Kazuya Seo
9. 40 ..------------~110
35 10
9.6
0 100
e
- 94
i
2
~ 9.2
I 20
:;;
15
i.!:! 9.0
10
5 75
_______...___.._....__....._._,,___, 70
010 -15
0. 1 1 2
Qc..,auon (m)
Fig. 9. Objective functions Fig. 10. Flight trajectory
5 Conclusions
We have carried out a multi-optimization study of three kinds of kick in rugby. The
features ofthe optimal solutions are:
1. In the case of the optimal punted kick, the number of extrema in the lateral
directions is more than l 0, and the amplitude of the fluctuations is from
several toten centimeters.
2. In the case of the kick into touch, the optimal kick made using the leg near-
est the touch line produces a greater flight distance than the optimal kick
made using the leg furthest from the touch line. However, the initial win-
dow is narrower.
3. In the case of the optimal kick for goal, the angular velocity should be per-
pendicular to the velocity and also to the longitudinal axis ofthe ball.
References
Deb, K. (2002) Multi-Objective Optimization using Evolutionary Algorithms. Wiley, New

York.
Seo, K., Kobayashi, 0. Gotsu, A. and Murakami, M. (2003) Aerodynamic force data on a
rugby football. Sports Dynamics Discovery and Application, pp.289-294.
Seo, K., Kobayashi, 0. and Murakami, M. (2004) Regular and irregular motion of a rugby
football during flight. The Engineering of Sport 5, Vol.l, pp.567-573 .
Seo, K., Sakamoto, S., Kobayashi, 0. and Murakami, M. (2005) The initial window of a
successful kick for goal ofrugby football. The Impact ofTechnology on Sport, pp. 280-
285.
Seo, K., Kobayashi, 0. and Murakami, M. (2006) Flight dynamics ofthe screw kick in Rugby.
Sports Engineering, Submitted for publication.
Stevens, B.L. and Lewis, F.L. (1992) Aircraft Contra! and Simulation, Wiley, New York.
The Mechanical Behaviour of Cricket Soils During
Preparation by Rolling
Peter Shipton, Iain James and Alex Vickers

Cranfield Centre for Sports Surfaces, Cranfield University, UK,
i.t.james@cranfield.ac.uk
Abstract. The nature of the ball - surface interaction in cricket has been identified as critical
to the quality and safety of the sport. The requirement for even ball bounce and good pace
from a clay loam soil cricket pitch has been successfully characterized and has been observed
to be related to soil properties such as dry bulk density, moisture content and organic carbon
content. To achieve the required mechanical properties, practitioners manage the compaction
of a cricket pitch through the use of smooth steel-wheeled rollers. The relationship between
moisture content and the compaction and shear strength was determined for a typical clay
loam soil and was found tobe significant. The effect of subsequent passes of 4.75 and 5.71
kN on soil dry bulk density was also determined in the soil dynamics laboratory. Maximum
dry bulk density was achieved after 20 and 10 passes of each roller, respectively. The roller
did not have a significant effect on dry bulk density below 50 mm in thc profilc.
1 Introduction
There are few sports where the ballistics of ball trajectory prior to, during and post
interaction with the surface are as critical for the quality and safety of play as in
cricket. The two key parameters, describing this interaction are known within the
sport as 'pace' and 'bounce'; defined by James, Carre, and Haake, (2004) as the
velocity and trajectory of the ball post impact with surface. For both the batting and
fielding side to have an equal chance in the game, and for batsman safety, variation
in pace and bounce of pitch should be 'predictable' i.e. within acceptable limits of
play. Whilst these limits are subjective in their nature, the effect of adverse ball -
surface interaction is apparent to players, officials and spectators of the game and
will result in low scores, shortened games and risk of injury. Further, complete uni-
forrnity in pace and bounce will favour the batting team, desirable in short versions
of the game where a result is guaranteed but less so in 4 and 5 day games where a
result is dependent upon completed innings.
Variation in pace and bounce was studied by James et al, (2004) who developed
methods of measuring and predicting variation in pace and bounce in UK cricket.
The relationships among pace and bounce and soil physical/chemical parameters
such as particle size distribution, dry bulk density (the oven-dry mass of soil in a
known volume) and organic matter content were reported by Baker et al (2003), who
deterrnined that there was a positive correlation between pace and the dry bulk den-
sity and sand content of a soil, but there was a negative correlation between pace and
moisture content, silt content and organic matter. Whilst established relationships of
230 Shipton et al.
this type exist, these relationships observe the pitch in its 'match-ready condition',
i.e. they do not inform the pitch construction or preparation process beyond the re-
quired final physical condition of the surface. Furthermore, these relationships are
based upon soil-physical rather than soil-mechanical properties.
To achieve the required uniformity of surface mechanical properties, the soil is
compacted using a steel smooth-wheeled roller. This aims to increase bulk density
and shear strength through compaction and to produce a level, smooth surface.
Whilst in a compacted, dry state a clay soil provides high mechanical stiffness and
shear strength (a key property for resistance to wear), the shear strength of a clay soil
is known to be highly sensitive to moisture content (Henkel, 1959). Therefore, dur-
ing play, soil moisture content must be kept to a minimum. To achieve drying ofthe
complete soil profile, grass growth and deep rooting are essential as water is re-
moved through transpiration. Therefore bulk density and shear strength must not
exceed critical values which prevent grass growth for extended periods of time. The
requirements of a surface are bounded by a lower limit of sufficient surface shear
strength and stiffness for pace, bounce and wear resistance, and an upper limit of
shear strength for grass growth.
To translate any mechanical model into practical guidelines for the practitioner,
the optimum soil moisture and duration for rolling, and how rolling practices affect
surface mechanics must be investigated. This forms the key aim of a four year study
by the authors to model the mechanical behaviour of cricket soils during preparation.
This paper reports on the first year of this project which aimed to determine key
mechanical properties of a typical soil used in cricket pitch construction in the UK
and the behaviour of these soils during construction and compaction by a steel
smooth-wheeled roller.
2.1 Test Soil Characterization and Selection

The soil (labelled C/30 here) was sourced from Essex, UK and had a particle size
distribution of 30% sand, 40% silt, 30% clay. lt is typical of soils used at elite and
weil resourced recreational levels of the game.
2.2 Determination of the Optimum Moisture Content for Compaction of

Each Soil
All moisture contents in this paper are reported on a gravimetric (mass) basis. The
optimum moisture content for compaction of the soils was determined using the
'Proctor Test' (Proctor, 1933). Soil specimens of a fixed bulk volume 929 ml, were
prepared at a range of gravimetric moisture contents between 5 and 30%. Each
specimen was constructed in 3 layers, each layer receiving 27 blows from a 4.5 kg
hammer, dropped from a height of 450 mm. The resultant dry bulk density (pb) of
each specimen was determined by measuring the oven dry mass of the soil in a
known volume. The proctor optimum moisture content for compaction was deter-
mined as the peak of the moisture content - dry bulk density curve.
2.2 Determination of the Shear Strength
Shear strength, the shear stress at failure of a test specimen was determined using
two methods: the translational shear box and the quick, undrained triaxial shear test.
For the translational shear box, specimens were prepared at 6% and 18% moisture
content. The 60 x 60 x 20 mm samples were then placed in the translational shear
box and subjected to an incremental strain at a rate of 1.25 mm/min. The shear stress
at failure was determined for a range of normal stresses between 15.7 kN/m 2 and
70.2 kN/m2 .
For the quick, undrained triaxial shear test, 38 mm diameter x 72 mm length cy-
lindrical samples were failed in shear at total confining stresses of 35, 69, 103 and
138 kN/m 2, at a strain rate of 1.5 mm/min. Samples were prepared at 15, 18 and
23% and tests were conducted in triplicate. In both methods, dry bulk density was
constant at 1500 kg m-3.
Linear models of soil failure at a shear stress ('C, kN m-2), as a function of normal
stress (cr, kN m- 2) were determined by the Coulomb theory of soil failure (Lamb &
Whitman, 1969).
r= c + otanip (1)
This model, described in Eq. 1, was determined by linear regression in the transla-
tional shear box test and by the construction of Mohr's circles for the triaxial test.
Mean values of cohesion (c, kN m- 2) and intemal angle of friction ( ip, degrees) were
analyzed by ANOV A.
2.3 The Effect of Successive Passes of a Roller on Dry Bulk Density

A test surface of C/30 soil (10 m length, 1.8 m width, 0.2 m depth) was constructed
in the Cranfield University Soil Dynamics Laboratory in 50 mm compacted layers.
Initial Pb was 1200 kg m- 3 at a moisture content of 20.5%. A smooth steel-wheeled
roller, typical of those used in the preparation on cricket pitches (diameter 0.3 m,
width 1.2 m), was towed at two speeds, 0.28 and 0.56 m s- 1 over the soil surface.
The experiment was conducted at roller weights of 4.75 and 7.51 kN.
Pb was measured for subsequent passes of the roller at 50, 100 and 150 mm
depths within the profile. The effect of roller weight on bulk density at each depth
was determined by ANOVA.
3 Results
3.1 Compaction and Moisture Content

In the compaction test, there was a characteristic increase to a maximum, and then
decrease in dry bulk density as moisture content was increased (Fig. 1). A signifi-
cantly greater maximum Pb was achieved with the heavier hammer ( 1850 kg m- 3 at a
moisture content of 15%) than with the lighter hammer (1650 kg m- 3 at 20%). The
difference between hammers was expected due to the increased work done on the
232 Shipton et al.
soil from the greater mass and this is typical of proctor test results. Beyond maxi-
mum Pb, compaction was limited by the pore water in the soil and thus the curves
were similar from 20% moisture content.
1900 ,
•••• •
1
1850 ♦ Cl30 (4 5 kg)
1800
<> Cl30 (2 5 kg)
.~
"'E
~
,:-
.iii
1750
1700 •
1$.
C:
1650
"'
~
1
.X
:i 1600 j
.c
•
1550 ◊
i!' ◊◊
0
1500 ◊
1450 ◊o
◊
1400 ♦
0 5 10 15 20 25 30 35 40
Moisture, % mim
Fig. 1. Proctor compaction test rcsults for the Cl30 soil with a 2.5 kg & 4.5 kg hammer
3.2 Shear Strength Parameters
Table 1. Mohr-Coulomb shear strength parameters for the C/30 soil
Soil Gravimetrie moisture Test method Cohesion (c), Angle im. friction ( efJ),
content, % kN m·' de >rees
C/30 8 TSB <0.1 51.8
C/30 15 TSll 30.8 41.1
C/30 15 QTxS 34.2 33.2

C/30 18 QTxS 27.2 19.4
C/30 23 QTxS 31.6 5.9
TSB - translational shear box; QTxS - quick undraincd triaxial shear
Variation among cohesion (c) and intemal friction angles ( </J) is illustrated in Table 1.
In the translational shear box experiment, c was significantly greater at 15% than 8%
moisture content (p<0.05). In the triaxial shear experiment, there was a significant
decrease in </Jas moisture content increased (p<0.05), due to lubrication by soil wa-
ter. The liquid limit of this soil was determined to be 20%. Whilst there was a sig-
nificant relationship between c and moisture content (p< 0.05), it was not consistent.
The only direct comparison between the two test methods was for the C/30 soil at
15%; there was no significant difference in c and </J between methods.
3.3 The Effect of Subsequent Passes of a Roller on Soil Bulk Density

Rolling increased Pb significantly in the first 0-50 mm of the profile. At 50-150 mm
there was no consistent pattem of Pb increase with successive passes of either roller.
The increase in Pb at 0-50 m with subsequent passes of a roller was characterized by
an increase to a maximum for the first passes and then a plateau for subsequent
passes. For the 4.75 kN roller this peak occurred at 1470 kg m- 3 after 20 passes; for
the 7 .51 kN roll er the peak was 1540 kg m- 3 after 10 passes.
1700 1700
♦ 0-50 X50-100 □ 100-150
I I. •• • • • •
1600 1600
•• • • "E
•• R •
11500
;..
al •••• •
~
i:-
·;;;
1500
X X
~ 1400
-"
•
, xxx
X
X
ai 1400
'O
~
□
iX X□~
□
X
~
'5
; 1300 □R .0
1:- 1300
II
0 ~ iS
0
□ □
□ □
1200
□□ 1200
1100 1100
0 20 40 60 80 0 20 40 60 80
(B) Number of passes (b) Number of passes
Fig. 2. Mean dry bulk density of C/30 at 0-150 mm depth, following successive passes of a
4.75 kN (a) and 7.51 kN (b) smooth steel wheeled roller at 20% gravimetric moisturc content.
Error bars reprcsent the LSD at p=0.05
4 Discussion
The results of both the compaction and shear strength analyses illustrate the sensitiv-
ity of both properties to moisture content. Depending upon the magnitude of the
load applied, there is an optimum moisture content for the compaction ofthe soil and
a maximum bulk density that can be produced. For effective rolling of cricket
pitches soils should be as close to the optimum moisture content for the load being
applied.
In compaction, an increase in moisture content eventually reduces the effect of
normal stresses due to the low compressibility of water. In shear, an increase in
moisture content was shown to initially increase shear strength due to an increase in
cohesion caused by the interaction between clay particles and thin films of water. As
soil moisture content increased above 15%, however, cohesion did not increase but
friction between soil particles reduced significantly, reducing soil strength.
When rolling cricket pitches the shear strength of the soil has to be exceeded by
the rolling stresses for compaction to be achieved. Whilst the principal objective is
normal loading of the soil, horizontal stresses are inevitable and are considerable in
driven rollers where torque can cause wheel slip in excess of shear strength. In these
conditions the soil is at significant risk of undesirable surface damage that could
have an adverse effect on ball pace and bounce.
The data in Fig. 1 showed that maximum compaction occurred with the greater
load at 15%, which coincides with the maximum shear strength of the soils in Table
1. The data in Fig. 2, however, showed that even the heaviest roller (some rollers
currently in use in the field may exceed this weight) did not cause compaction be-
yond 1540 kg m-3, less than the maximum with the lighter hammer in Fig. 1 at simi-
lar moisture contents. Field data have shown that Pb can reach the 1850 kg m- 3 ob-
served in Fig. 1 due to consolidation from removal of water over longer periods of
234 Shipton et al.
time. At such a value of Pb, air filled porosity can be as low as 5%, to reduce poros-
ity further will limit grass growth and the effect of grass on both moisture content
and shear strength by root reinforcement has not been considered here.
lt is also important to note that maximum compaction was achieved after only 20
passes of a 4. 75 kN roll er and 10 passes of a 7.51 kN roll er. If this result is observed
in similar experiments with grass surfaces, currently in progress, then it will have a
significant effect in reducing the time spent rolling cricket pitches in practice. Also,
the roller was only effective in compacting the first 50 mm of the profile. This has
direct implications for the construction of cricket pitches where it is apparent that
only consolidation will increase the bulk density of the soil below 50 mm depth. lt
should be noted that the critical compacted surface depth for ball interaction is not
weil understood or quantified in the literature.
5 Conclusion
The maximum dry bulk density (pb) in a proctor compaction test of a typical soil
used in first class cricket in the UK, was 1850 kg m· 3 at 15% gravimetric moisture
content using a 4.5 kg hammer and 1650 kg m· 3 at 20% using a 2.5 kg hammer. The
shear strength of the soil was greater at 15% than 18 and 23%. As soil moisture
content increased from 8 to 15 %, cohesion increased significantly, but was not sig-
nificantly different from 15 to 23%. Interna! friction decreased significantly from 8
to 23% due to lubrication of soil particles by water.
lt was shown that after 20 passes of a 4. 75 kN roller, typical of the type used in
cricket pitch preparation, Pb increased significantly from 1200 to 1470 kg m· 3 but that
there was no significant increase for subsequent passes. Likewise, for a 7.51 kN
roller, a maximum bulk density of 1540 kg m· 3 was achieved after only 10 passes.
These data form part of the development of a model to improve the construction
and preparation of cricket pitches.
Acknowledgements
The authors gratefully acknowledge that this research was funded by the UK
EPSRC and the England and Wales Cricket Board. Thanks to R. J. Godwin, M
Freeman, R Newland, S Stranks and R Swatland for their input and assistance.
References
Baker, S.W., Harnmond, L.K.F., Owen, A.G. and Adams, W.A. (2003). Soil physical proper-
ties offirst class cricket pitches in England and Wales. 2. Influence ofsoil type and pitch
preparation on playing quality. Journal ofthe Sports Turf Research Institute, 79, 13-21.
Henkel, D.J. ( 1959). The relationship between the strength, pore-water pressure and volume-
change characteristics of saturated clays. Geotechnique, 9, 119-135.
James, D. M., Carre, M. J., Haake, S. J. (2004). The playing perforrnance of county cricket
pitches. Sports Engineering. 7( 1): p. 1-14
Lamb, T.W., Whitman, R.V. (1969) Soil Mechanics. John Wiley & Sons, New York.
Proctor, R.R. (1933). The design and construction ofrolled earth dams. Engineering News-
Rec, 111 (9), 245-248
Studies on the Oblique Impact of a Cricket Ball on a
Cricket Pitch
David James, Matt Carre and Stephen Haake

Sports Engineering, CSES, Sheffield Hallam University, d.james@shu.ac.uk
Abstract. The cricket pitch is a carefully prepared strip of natural turf of fundamental
importance to the play of the game, yet the understanding of the factors that lead to good pitch
construction remains limited. In order to ascertain why some cricket pitches are perceived to
perform weil whilst others cause difficulties, the mechanics of the oblique impact of the ball
on the pitch have been explored. This study presents the development of a normal impact
model to the oblique impact scenario. A coefficient of dynamic friction and an analysis of the
pitch crater were used to predict the ball's rebound dynamics. Model parameters were
determined from simple surface testing procedures and model predictions were found to be
within an acceptable range of divergence from experimental impact measurements.
1 Introduction
Previous attempts at modelling sports ball-surface interactions have generally as-

sumed that either the ball or the surface could be considered rigid, due to the differ-
ence in stiffness of the two parts. For instance, studies of tennis ball-court interac-
tions have assumed the surface to be rigid whilst the ball deforms in a non-linear
elastic way (Haake et al., 2005). In these models, the surface simply acts to allow a
frictional force to develop on contact. In other impact situations, such as golf
ball/green, the ball has been assumed tobe rigid compared with the surface (Haake,
1989).
Modelling of a cricket ball-pitch interaction is complex due to the fact that both
the ball and the surface de form to some degree. Carre et al. (2004) developed a
spring-damper normal impact model of a cricket ball impacting on a rigid surface
using experimental data. The model was then used to predict oblique impacts by
including a coefficient of friction. The model predictions compared weil with ex-
perimental data for rebound speed and spin generation, but not with rebound angle.
More recently, James et al. (2006) showed that it is possible to predict the 'pace' of a
cricket pitch (oblique rebound speed) by combining two simple measurements of
restitution and friction in a rigid-body model. However, similar to Carre et al. (2004)
the model was unable to predict the angle of rebound. lt is believed that deformation
of the pitch can cause a steeper rebound than one might expect due to the ball im-
pacting on the front of the depression that it has made (Carre et al., 1999).
236 David James, Matt Carre and Stephen Haake
2 Normal Impact Model
James et al. (2004) described the normal impact of cricket ball on a cricket pitch by
suggesting a two degree of freedom system that allowed for both deformations of the
ball and the pitch. This paper will use their suggested normal impact model as a basis
for understanding the more complex oblique impact.
m - crici<et ball mass

kb - ball sliffness parameter
cb- ball damping parameter
kp- pitch sliffness parameter
Cp- pitch damping parameter
Y,
Yb- displacement of ball's centre of mass
yp- dlsplacement of pitch
Fig. 1. A two degree of frecdom spring damper model of a normal ball impact.
Figure 1 shows a schematic of the normal impact model. Since the force acting
on the ball and the pitch are equal, the impact force can be determined at any time
interval, t, during impact using Eqs. 1-3.
(F,,), = m(j\)i (1)
(Fv), = (kh)r-t.,(Yh -yp)r +(ch)r-t.r(Yb - .Yp), (2)
(3)
The force equations assume that both pairs of stiffness and damping coefficients
change during impact. However, their values at each time interval, t, are based on the
ball conditions at the previous time interval, t-1':it. In this way, the model can be
solved using the finite-difference method. However, the displacement of the pitch,
yp, is unknown and a further relationship is required to define it. Combining Eqs. 2
and 3 and using finite difference definitions,
(yp)I
(kh\-L'J(Yb), +(eh\-"" +(yp\-L'J[<ch),_"" +(cp)r-L'J i:v-i (4)
(kp)r-L'J +(kh)r-L'J +[<cb\-L'J +(cp)r-L'J] Af- 1
Equation 4 can describe the displacement of the pitch in relation to the motion of
the ball 's centre of mass if all stiffness and damping parameters are known. Parame-
ters were determined by dropping a cricket ball onto a rigid load cell, and then drop-
ping a rigid impact hammer onto a cricket pitch (James et al., 2004).
3 Oblique Impact Model
The two degree-of-freedom normal impact model was extended to describe the
oblique impact. Notably, the interaction between the front face of the deformed ball
and the crater that is formed during impact was considered.
r,x
Fig. 2. A schematic of the forces acting on a cricket ball during impact.
A schematic of the forces acting on the cricket ball during impact with a deform-
able pitch is shown in Fig. 2. The normal reaction force, Fnorm, is calculated using the
same method as for the normal impact model (Fy in Eq. 1) and the corresponding
frictional force is the product of the normal reaction force and a coefficient of fric-
tion, µ. This was measured using a weighted sledge supported by three cricket balls
that was pulled along the pitch surface using a constant torque motor (see James et
al., 2006).
As shown in Fig. 2, the ball impact produces a crater in the pitch that generates
an additional resultant force, Fdef, accompanied by a corresponding perpendicular
frictional force, µFdef• The direction and position ofthese forces can be calculated by
considering the variation in force acting over the ball/crater contact surface.
The oblique impact model assumes the area of contact between the ball and pitch
crater to be represented by the geometry shown in Fig. 3 (i.e. the truncated base of
the deformed ball plus the front half of a slice of the ball with a depth equal to that of
the pitch deformation). Experimental results using the SERG impact hammer showed
the cricket pitch to have a slow recovery from deformation (James et al. 2004). The
back half of the ball was therefore considered to be free from contact with the pitch.
(a) (b)
Fig. 3. The area of contact between the ball and pitch during an oblique impact (deformations
are exaggerated for clarity) with (a), the reaction force from one segment ofthe sphere's annu-
lus, (F def)0. And (b ), the resultant force, Fdef calculated by integration over the contact area.
An estimation ofthe magnitude ofthe force, Fdefwas taken as the damping force
produced by the area of material in front of the ball. Difficulties arise if a stiffness
parameter is also incorporated in this part of the model due to the large horizontal
deflections. During impact, the size of the area in front of the ball changes, as does
the ball's velocity. The oblique reaction force, Fdef, is therefore time dependent and
was approximated by;
(Fdef )1 = w11r(yp)1 ✓(xb 2 + Y;)I (5)
i.e. the product of a pitch damping coefficient, w (determined from impact hammer
tests), the area ofpitch in front ofthe ball and the ball's absolute velocity.
In order to calculate the position from where, the resultant force, Fdef acted, a
general force, (FdeJ)0, for a small segment of the sphere's annulus was considered
(see Fig. 3(a)). Since the direction of this reaction force is dependent on the location
of the segment it was necessary to determine the combined effect of the reaction
forces from all segments within the half annulus. This general reaction force located
at an angle, 0, and at time, t, can be estimated as follows:
(Fdef )0,1 = wr(y p)lc50✓(x/ + Y;)I - (6)
lt follows that the reaction force from all segments can be spilt into components
in the x, y and z directions (shown in Fig. 3(a)) by considering the angles of Band A.
(Fxdef )8,1 =-wr(yp) 1c50.j(x/ + Yh cos0cod (7)
(Fydef )e,t = wr(yp)lc50✓(x/ + Y;)I sinA (8)
(9)
The combined effect of the oblique reaction force in each of the x, y, and z direc-
tions can be determined by integrating Eqs. 7-9 with respect to 0between ,r/2 and -
m'2, so that:
(Fxdef )1 = -2wr(yp)i ✓(x/ + Y;)I COSA (10)
(Fydef )1 = 1ZWr(yp)1 ✓(x/ + Y;)I sinA (11)

(Fzdef )1 = 0 (12)
Fig. 3(b) shows the direction and location of the resultant reaction force, Fdefi due
to the combined reaction forces from all segments of the front of the pitch crater. The
angle at which the combined oblique reaction force, Fdef, acts to the horizontal was
determined by comparing the magnitude of its horizontal and vertical components
(Fxdef, Fydef)- Defining this angle as y(see Fig. 3(b)), it can be shown that;
r=Tan-1[;ranA] (13)
This angle changes throughout impact but can be calculated by considering the
deformation of the ball and the pitch, thus;
_1 r-yb +O.5yp
A Tan
= -,=========== . (14)
.Jr 2 -(r- Yb +O.5yp) 2
This analytical description shows the combined oblique reaction force Fdef, to al-
ways act at a point two thirds of the horizontal distance between the centre line of the
ball and its leading surface.
With the direction and magnitude of all forces known (Fnorm, µFnorm, Fdef, µFde()
the model was set to run with a time interval of 1 µs. The resultant vertical and hori-
zontal forces on the ball were calculated at each time interval along with displace-
ment, velocity and acceleration of the original geometric centre of the ball and the
pitch surface. The ball's angular velocity was also calculated at each time interval by
considering Eq 15.
(w) = CFx),[r-(yh)i]M +(w) (15)
t (/), t-!it
The moment of inertia of the ball was assumed to vary due to deformation and
was approximated as a solid sphere with a reduced radius so that:
(/) 1 = ½m[r-(Yb)½r (16)
The directions of the frictional forces (µFnorm, µFdef) were made to switch direc-
tion when the ball was deemed to be rolling on the surface. For the force µFnorm, the
direction switched when the equatorial velocity of the truncated section of the ball
exceeded that of the horizontal velocity of the ball's centre. The oblique impact was
considered to end when the ball and pitch lost all physical contact. lt was found that
the normal reaction force, Fnorm, retumed to zero before the end of impact. From this
point onwards, forces Fnorm and µFnorm were switched off and the ball's trajectory
was solely affected by forces Fdefand µFdef
4 Experimental Verification
Validation data for the oblique impact model was gathered by projecting cricket balls
obliquely onto a test plot of a professionally maintained cricket pitch. The impacts
were recorded using high speed video at 400 frames per second. A bowling machine
was set to simulate a range of bowler deliveries and the impact footage was analysed
using bespoke software. Model parameters for the cricket ball were considered to be
the same as those found by Carre et al. (2004). Model parameters for the pitch were
found by using the SERG impact hammer in the same method as described by James
et al. (2004). The coefficient of friction between the ball and pitch was determined
by using friction sledge apparatus as discussed in James et al. (2006).
Figure 4 shows the model predictions to be in a generally good agreement with
the experiment data. An equal ratio line is shown on both plots and it can be seen
that the model is able to predict rebound speed with a high degree of accuracy, whilst
predictions of rebound angle are more scattered. The model over-predicts the balls'
rebound angles to some degree, but it provides a significant improvement on previ-
ous models (Carre et al., 2004; James et al., 2006). lt is also worthy to note that the
rebound angles of cricket balls are inherently variable. The model 's predictions of
rebound spin rate were found to be in good agreement with the experimental results
(data not shown).
~
~~------------~
.
~
30~------------~
25
•
l..,
- 25
~
.!!
~ 20
§ 20
,,
0 §
..,e _& 15
~ ..,e
,;
e
$
u 10 - -'!'; • - -•_ - - - - - - - - - - - - - .. _j
0.
~
0. 1
+---~--~--~--~-----i
10 15 20 25 10 15 20 25 30
Actual rebound speed (ms·1 ) Actual rebound angle (degrees)
(a) (b)
Fig. 4. Comparisons between high speed video measurcments of the cricket ball rebound and
the corresponding model solutions for; (a) rebound speed and (b) rebound angle.
5 Conclusions
The model for a normal impact of a cricket ball on a pitch was developed to the
oblique impact scenario. The performance of the new model was evaluated by com-
paring experimental impact data to model predictions. The new model was shown to
perform generally weil, and in the case of predicting rebound angle, it performed
significantly better than previous models. The new model provides the ground work
for a method of predicting the playing performance of cricket pitches; a tool which
would be a valuable asset to Ground Staff.
References
Carre, M.J., Baker, S.W., Ncwell, A.J. & Haake S.J. (1999). The Dynamic Behaviourof
cricket balls during impact and variations due to grass and soil type. Sports Engineering,
3, 145-160.
Carre, M.J., James, D.M. & Haake, S.J., (2004). Impact of a non-homogenous sphere on a
rigid surface. Proc. lnstn. Mech Engrs. Part C: Journal of Mechanical Engineering Sci-
ence, 218, 273-281
Haake, S.J. ( 1989). Apparatus and test Methods for measuring the impact of golf balls on turf
and their application in the field. PhD thcsis, The University of Aston in Birmingham.
Haake, S.J., Carre, M.J., Kirk R. and Goodwill, S.R. (2005) Oblique impact ofthick wallcd
pressurized spheres as used in tennis. Proceedings for the Institution of Mechanical Engi-
neers, 219 (C), 1179-1189.
James, D.M., Carre, M.J. and Haake, S.J. (2004) The normal impact of a cricket ball on a
cricket pitch. In: The Engineering of Sport 5: Proceedings of the 5th International Confer-
ence an the Engineering of Sport (eds M. Hubbard, R.D. Mchta & J.M. Pallis) Vol 2, pp.
66-72, ISEA, UK.
James, D.M., Carre, M.J. and Haake, S.J. (2006) Predicting the playing characteristics of
cricket pitches. Sports Engineering (In press).
Test Devices for the Evaluation of Synthetic Turf Pitches
for Field Hockey
Colin Young, Paul Fleming and Neil Dixon

Loughborough University, C.Young@lboro.ac.uk
Abstract. Many existing tests for field hockey can be categorized into; how the ball, and how
a person interacts with the surface. Interactions during sporting activities can significantly
influence how a game is played from both a technical and tactical perspective. Understanding
interactions of this nature and identifying factors that can influence and control their perforrn-
ance is essential to comprehend the mechanical behavior of a sports surface. However, syn-
thetic turf pitches are complex structures, comprising several layers, all of which contribute to
their composite behavior. Therefore, the mechanical response of the surface to interactions is
difficult to measure. lt has bccn argued by many researchers that mcchanical tests are inappro-
priate to simulate in-game conditions and their suitability has been brought into question.
Furtherrnorc, there is a lack of good quality peer reviewed data on the mechanical behavior of
synthetic turf pitches. Test data arc collected by accredited laboratorics for the relevant sports
goveming body, with the data rcmaining unpublishcd, thus therc is no way to validate or
recommend improvements to these standards. Consequently, this paper presents results from a
comprehensive program of tcsting on six world class synthetic turf pitches used for field
hockey. Currcnt test equipment and methods employed by the govcming body for field hockey
(FIH) were validated and recommendations were forrnulated for their suitability. lt was found
that impact tests, including thc Berlin Artificial Athlete, provided a simple means to classify
one pitch against another and gave a significant difference between the six pitches. A review
of ball intcraction tests, including vertical ball rebound, and ball roll were found to be signifi-
cantly influenced by environmental factors such as moisture and wind, which highlighted the
importance of careful monitoring during testing to ensure pitchcs were evaluated in approved
conditions. In conclusion, currcnt mechanical tests provide a simple and cffective way to
classify onc pitch directly against anothcr. Howevcr, their use for deterrnining how the surface
bchaves in a 'real' gamc situation and the mechanical inforrnation obtained is considered
limitcd
1 Introduction
Synthetic turf pitches are complex structures with several layers, all of which con-
tribute to their composite behavior. Therefore, the mechanical response of the sur-
face to interactions from players, balls and sports equipment are difficult to assess.
Impacts involving sports objects, such as a ball or the player and the surface, can
242 Colin Young, Paul Fleming and Neil Dixon
affect the technique and tactics of a sports performer and the way in which the game
is played.
The Federation De Internationale Hockey (FIH) produced a !ist of requirements
to which a playing surface must adhere to in order tobe used for sanctioned competi-
tions. These standards are published in the 'handbook of performance requirements
and test procedures for synthetic hockey pitches - outdoor' (FIH, 1999). The objec-
tives of the standards are to ensure that field hockey competitions are played on
pitches which; provide a proper reflection of team merit, allow players to display
and develop their skills, offer comfort and limit risk to players, and extend playabil-
ity in adverse weather conditions. The handbook has three tiers of standards for
different levels of ability/competition: global, standard and starter. The 'global'
standard is the most stringent and is compulsory for international competitions and
only unfilled (or water based) systems can obtain this standard. However, there is
still a !arge range of acceptability even at this tier and there is a Jack of any good
quality peer reviewed research on pitch accreditation to validate the approach
(Young, 2006).
The most common device for measuring impact behavior on sports surfaces is the
Berlin Artificial Athlete. The Berlin is currently used by the FIH as a measure of
impact response. The peak impact force is measured, and surface cushioning (FJ is
presented as the percentage reduction compared with a rigid (normally concrete Fe)
surface.
Force Reduction = (Fe - Fi)/Fc (1)
There are two tests specified by the FIH to measure ball/surface interactions. The
first ball rebound (or rebound resilience) is a measure of the energy lost during im-
pact with the surface from a vertical drop. The second is a measure of the frictional
resistance of the ball as it rolls across the surface and is called ball roll distance (or
ball roll resistance). The roll resistance is defined as the force acting at the point of
contact between the ball and surface.
This paper presents results from a comprehensive program of testing on six
'global' standard field hockey pitches. Several of the current tests methods are
evaluated for their suitability to measure the behavior of synthetic turf pitches and
factors that influence the measurement are also assessed, including the effect of
surface water/irrigation and construction specification.
2 Methodology
This section outlines the test methods/equipment used to evaluate the behavior of six
'global' standard water based field hockey pitches. Pitch selection was based on
several criteria. Firstly, feedback given by players during interviews and question-
naires (Young, 2006) were analyzed and a shortlist of suitable pitches were identified
based on perceived playing characteristics. The shortlist was then reduced to pitches
that conformed to FIH 'global' standard accreditation. From the remaining !ist prior-
ity was given to the pitches with available construction specification to facilitate
understanding of the effects of different constructions. From the above criteria six
pitches were highlighted for field testing, due to data protection the pitches can not
be identified and henceforth shall be labeled pitches A to F. Details of the six pitches
are illustrated in Table 1.
Pitch
A B C D E F
Subbase 250mm 450 mm 200mm 250mm 200mm 250mm
Thickness
Asphalt 65mm 65mm 70mm 65mm 65mm 65mm
Thickness
Shockpad ln-situ & ln-situ & Integral In-situ Integral ln-situ
Type Integral Integral
Shockpad 15 mm' 15mm' 8mm 15 mm 6mm 15mm
Thickness
Pile Mate- Nylon Nylon Nylon Polypropylene Nylon Polypropylene
rial
Pile Height 12mm 12mm 11mm 15mm 11mm 13mm
Note: 1combination of 12 mm in-situ and 3 mm integral shockrads
Table 1. Construction details of the six synthetic turf field hockey pitches
Prior to testing each pitch was applied with a füll irrigation cycle to ensure it was
tested under similar conditions to what players experience during a game. This was
repeated every 40 minutes as would be standard during agame of field hockey. To
ensure that a good global coverage of the pitch was achieved during testing a grid
system was produced with 25 test locations evenly spread across the entire playing
area, this provided comprehensive coverage of each pitch.
The FIH outlines many tcsts for the accreditation of field hockey pitches, how-
ever, given this size and context of this paper three test methods are presented
within, these are described below.
2.1 Berlin Artificial Athlete

The Berlin Artificial Athlete consists of a falling mass of 20 kg that is electronically
released from a height of 55 mm onto a spring with a stiffness of 2000 kN/m- 1 that is
connected to a test foot of 70 mm diameter. The peak impact force is measured three
times, and surface cushioning is presented as the average pcrcentage reduction of the
second and third drops compared with a rigid (norrnally concrete) surface, as de-
scribed in the FIH handbook (1999). The requirement for 'global' standard pitches is
between 40 - 65 % force reduction.
2.2 Ball Roll Distance

Ball roll distance, was measured by rolling a ball down a standard inclined plane or
ramp. The ball (approved by the FIH) should roll a prescribed distance within a
max im um deviation of 3 ° from the straight line. The test was repeated in the opposite
direction and results were averaged, thus reducing the possible effects ofwind, slope,
wear, pile bias and smoothness. The test follows the procedure outlined in the FIH
handbook of perforrnance requirements ( 1999). The requirements outlined by the
FIH for 'global' standard pitches is between 9 m - 15 m ± I 0 % of the mean.
244 Colin Young, Paul Flcming and Neil Dixon
2.3 Ball Rebound Height

To detennine the ball rebound resilience a vertical drop test was used. The test fol-
lowed the procedure of the FIH standard (1999). lt consisted of releasing a ball from
a height of 1.5 m (surface to underside of ball) on to the test surface. The height of
rebound for 'global' standard pitches should be between 100 mm and 250 mm with a
maximum deviation of 20 % from the mean. The FIH specify that the test should be
'wet' and an approved hockey ball be used.
Pitch
Test Device A B C D E F
Berlin Artificial Mean 60.4 61.8 43.6 55.5 45.4 52.7
Athlete (%) so 3.0 3.8 1.7 3.5 2.1 4.2
cov 5.0 6.1 4.0 6.3 4.6 7.9
Ball Rebound Mean 32.8 36.8 20.7 41.1 26.2 32.2
Height (cm) so 2.4 3.2 5.2 1.0 0.7 2.2
cov 7.2 8.6 25.0 2.5 2.8 6.8
Ball Roll Dis- Mean 14.5 13.6 15.4 15.1 14.0 15.5
tance (m) so 1.0 2.1 1.8 1.6 1.1 3.1
cov 7.2 15.6 11.8 10.3 8.1 20.0
Note: SO= standard deviation, COV = coefficient ofvariance
Table 2. An overview of the results from six synthetic turf field hockey pitches
3 Results
The following section presents the results from the data collection on six 'global'
standard synthetic turf field hockey pitches. An overview of the results is presented
in Table 2.
3.1 Berlin Artificial Athlete

Measurements with the Berlin identified pitch C as the hardest pitch with a force
reduction of 43.6 %. Pitch B was measured as the saftest pitch with a force reduction
of 61.8 %. Table 2 illustrates the force reduction for all six pitches. From the 25 test
locations it was found that pitch C had the least variability with a COV (coefficient
of variance, standard deviation / mean) of 4.0 % compared with pitch F which had
the most at 7.9 %.
3.2 Ball Roll Distance

A small difference was measured between the six pitches with the ball roll test. Pitch
B had the shortest measured distance of 13.6 m and Pitch F had the langest with 15.5
m. Three of the pitches (C, E & F) feil narrowly outside the FIH 'global' specifica-
tion. A !arge directional difference was noticed during testing, on pitch B there was a
difference from 11.84 m (north to south) to 18.12 m (south to north). This difference
was attributed to the influence of the wind.
10 □ Dry
◊ Match Conditions
o Saturated
~Nw~~~Nw~~~Nw~~~Nw~~~Nw~~
w w w w w rr rr rr rr rr o o o o o
~ ~ ~ ~ ~ ro ro ro ro ro
Location
Fig 1. The influence ofmoisture ofball rebound height on pitch A
3.3 Ball Rebound Height

A large spread of measurements were taken from the ball rebound height tests. Pitch
C had the lowest mean rebound height of 20.7 cm compared with pitch D 41.1 cm
which was the highest. Five of the six pitches rebound height fell outside the FIH
guidelines for rebound height. lt was noticed whilst testing that the degree of water
on the surface significantly influence the rebound behavior of the ball. Hence, pitch
A was tested under three different levels of saturation (dry, match and saturated).
Figure 1 illustrates the magnitude of difference for each ofthese conditions.
4 Discussion
The amount ofwater on the pitch was shown to significantly influence the behav-
ior of the ball during impact. Thus the uniformity of the watering system to apply a
even application of water to the whole playing area is vital to ensure the behavior of
the pitch is consistent. The surface water appears to dissipate the impact energy of
the ball resulting in less energy being retumed to the ball and hence a lower rebound
height. The Berlin and ball roll tests did not measure a difference for each moisture
level.
The impact behavior of the surface measured with the Berlin was most dependent
on the shockpad and carpet layers. lt was found that the pitches evaluated with a
246 Colin Young, Paul Fleming and Neil Dixon
relatively thin integral system (C & E) had a much higher stiffness than the pitches
with an in-situ shockpad system. From these data there was no obvious link between
subbase and asphalt layers and pitch behavior, it can therefore be assumed that the
carpet/shockpad combination are more influential to pitch performance.
The six pitches fell within the FIH specifications for impact behavior. However,
for ball behavior three pitches failed roll distance and five pitches failed the ball
rebound test. This suggests that the pitches are outside the requirements of the FIH
and hence not suitable for 'elite' field hockey. However, all of the six pitches ini-
tially passed the accreditation process. lt is unclear if the pitches behavior have
changed over time (between the accreditation testing and this testing) or if the
equipment/methodology used was different. This raises the issue of regular re-
accreditation to ensure pitches remain within the required standards.
In the past it has been argued that these mechanical tests do not fully represent
what a player or ball experiences during a game situation and that to fully understand
the complex mechanism of pitch behavior test methods are required that more
closely simulate these conditions. However, given that the existing test methods are a
suitable way to index/classify pitches, for the purpose of surface accreditation, they
are considered appropriate.
5 Conclusions
Measurements from these test devices have established that )arge differences exist
between pitches. These differences can be attributed to their construction specifica-
tions and environmental influences.
Future measurements are required to determine the influence of 'ageing' and how
the pitches performance changes over time. This can also be linked to the mainte-
nance ofthe pitch which should be investigated.
A more fundamental study into the precise influence of water is required to better
understand its effects. Similarly, the rate of drainage/evaporation of water from the
surface can influence the pitches behavior during the course of a game, especially in
warm weather conditions and this problem needs to be evaluated.
References
Dixon, S. J., Batt, M. E. and Collop, A. C (1999) Artificial Playing Surfaces Research: A
Review ofthe Medical, Engineering and Biomcchanical Aspects. International Journal of
Sports Medicine 20, 209-218.
Handbook of Pe,jormance Requirements - Outdoor ( 1999) Federation Internationale de
Hockey, Brusscls, Bclgium.
Y oung, C (2006) Mechanical and Perceived behavior o(synthetic tur(pitches(orfield hockey.
Unpublishcd PhD Thesis, Loughborough University.
7 Skiing, Snowboarding and Ski Jumping
Synopsis of Current Developments: Skiing,
Snowboarding and Ski Jumping
Veit Senner
Technische Universität München, Department Sport Equipment and Materials,
senner@sport.tu-muenchen.de
A total of 11 contributions demonstrate the diversity of research in the field of ski-

ing, snowboarding and ski jumping. This shows how engineering methods can help
to better understand the complex interaction between equipment and athlete and to
better describe the physical phenomena behind these sports.
Snow Friction and Skiing
Four research groups address the topic of ski snow tribology: Paul MILLER et al.
(USA) discuss the development of a system for measuring the kinetic coefficient of
friction between the bottom surface of a ski and snow.
A large-scale laboratory tribometer, which measures the friction of a small ski
sample on ice or snow was built and validated by Mathieu FAUVE (Switzerland).
The scientists in the group of Qianhong WU (USA) present a simplified mathe-
matical model to describe lift mechanics due to both the transiently trapped air and
the compressed snow crystals.
An interesting camera-DLT-based method to determine friction and reaction
forces in competition skiing without interfering with the athlete is proposed by Mi-
chael SCHIESTL and his co-workers (Austria).
Ski Design and Performance
Mechanical properties of the ski are focused in three contributions coming all from
Switzerland.
The investigation of Anton LÜTHI and his colleagues is dealing with the effect
of different bindings and plates on the mechanical behaviour of the ski. Combining
their laboratory measurements with performance tests on the slope they were able to
perform correlation analysis between subjective tester ratings and the measured
physical characteristics.
A similar question is treated by Peter FEDEROLF et al. whose study underlines
considerable differences in the subjective assessment ofthe ski's performance. Nev-
ertheless some correlation between subjective tester ratings and the bending and
torsional stiffness of the ski were found.
250 Veit Senner
Finally the dynamic response of the ski as a function of temperature is examined

by Christian FISCHER and his colleagues. The researchers attribute increased damp-
ing of a ski to the existence of the polyamide top layer which seems to be most effec-
tive at temperatures around O °C.
Skiing and Ski Jumping Motion
The long turn of an expert skier is examined in the Japanese - Australian co-
production of Takeshi YONEY AMA and Nathan SCOTT. They determine vertical
binding forces and plantar foot pressure and combine this data with joint angle
measurements and video recordings.
Two research groups present their work on ski jumping: Kurt SCHINDEL WIG
and Werner NACHBAU ER (Austria) analyse the accuracy of an instrumented jump-
ing hill in Innsbruck. Satisfying accuracy was obtained for the joint angles; im-
provements are needed regarding velocity calculation of the centre of mass.
Another approach to flight dynamics of ski jumping is proposed by a Japanese
researcher group around Yuji OHGI. Their method uses a tri-axial accelerometer and
a single axial gyroscope attached to the jumper's body. Comparing these measure-
ments to the results of high speed video-grammetry the authors conclude that their
method was suitable to derive accurate enough motion data for the entire jump ac-
tion.
Equipment Related Safety Aspects in Skiing and Snowboarding
Equipment related safety aspects in Skiing and Snowboarding are addressed in three
contributions. Jasper SHEAL Y (USA) examines the distribution of stated primary
cause of death as a function of helmet utilization for the last five winter seasons in
U.S. His results suggest that while helmets may be effective at preventing minor
injuries, they have not been shown to reduce the overall incidence of fatality in ski-
ing and snowboarding. (Please note: This paper is printed in the "Safety Section" of
these Proceedings ).
Richard GREENW ALD and colleagues (USA) present and validate a novel brace
for preventing wrist fractures in snowboarding. With impact testing they compared
the stabilizing effects of their prototype to those of a commercially available wrist
guard. (This paper is also printed in the "Safety Section" ofthese Proceedings).
An overview of 15 years of work in the field of skiing equipment is given by
SENNER and co-workers (Germany). They present their approaches to prevent knee
injuries, deal with the problem of ski bindings' inadvertent release and show some
aspects regarding the interaction between binding function and ski design. Future
prospects on ski equipment research are summarizing their contribution.
Laboratory Device for Measuring the Friction Between
Ski-Base Materialsand Ice or Snow
Mathieu Fauve, Lukas Bäurle, Hansueli Rhyner

WSL, Institute for Snow and Avalanche Research (SLF),
Davos, Switzerland, fauve@slf.ch
Abstract Today's knowledge of ski friction is mainly based on experience from field testing
with real skis. A systematic investigation of ski friction in field conditions however, is diffi-
cult, since snow propertics and weather conditions can vary rapidly during outdoor tests. An
efficient field test conducted with an accurate and continuous snow characterization is both
time consuming and expensive. In order to avoid the high unccrtainties of field testing, a
large-scale laboratory tribometer, which measures the friction of a small ski sample on ice or
snow was built. The sample is attached to the measuring arm which is placed on a revolving
rink of ice or snow. Additional wcights can be mounted on the arm to vary the load. The oper-
ating temperaturc can be set bctween -20°C and 0°C. Yclocity at the position of the sample
can be varied between 0.5ms· 1 and 20ms· 1• lnfrared scnsors mcasure thc temperaturc evolution
of the track ovcr time during the friction tests. In addition, proccdures were dcveloped to
create different ice roughness and snow hardness. Prcliminary results show a good correlation
between the ficld and laboratory measurements. This dcvice allows for morc accuratc, faster
and cheaper testing of ski-base materials or surfacc treatmcnts.
1 Introduction
Sliding on snow has been studied for a long time. Yet it was only 1939 when Bow-
den and Hughes showed that the water film that enables low kinetic friction is caused
by frictional melting. The influence of snow characteristics on the sliding friction
was analysed by sevcral researchers (Eriksson 1955; Colbeck 1992; Nachbauer
1996; Moldestad 1999; Buhl, Fauve and Rhyner 2001; Fauve et al. 2005). Most of
these studies rely on results of field testing. However, field testing remains time
consuming, is relatively expensive, and can lead to false judgement if conducted
without proper snow and weather characterisation during the testing (Fauve et al.
2005). Some groups have therefore set up experiments to study under laboratory
conditions the friction between different materials and ice or snow (Bowden and
Hughes 1939; Lethovaara 1987; Buhl et al. 2001; Ducret et al. 2004). lt was recog-
nized early that the wanning of the ice/snow track, as weil as the vibrations of the
device can present a problem. Thus small tribometers using low sliding velocities,
which are not representative of actual skiing conditions, were often used. Our aim
was therefore to built a laboratory device which can measure friction at velocities
252 Mathieu Fauve, Lukas Bäurle, Hansueli Rhyner
encountered in real skiing while at the same time measuring the track surface tem-
perature. The device should also be accurate enough to measure low differences
between current ski-base preparations. In order to analyse the adequacy ofthe device
and to optimise its development, comparative tests between field and laboratory
measurements were made.
2 Experimental Set-Up
2.1 General Description of the Tribometer

The tribometer consists of a 1.80m diameter rotating table filled with ice or snow, of
one arm for holding the sample and measuring the friction force and of another arm
for fixing snow and ice preparation tools (Fig. 1).
m uringarm
tecl bar for boldin the sample
ice preparation
Fig. 1. Ice and snow tribometer
In order to avoid undesired vibrations, the table is directly attached to the con-
crete ground. Normal force can be varied by applying dead loads of 10, 20 and 30N.
Frictional force is measured at a frequency of maximum lOOHz via a strain gauge
load cell. The relative velocity at the location of the sample can be varied between
0.5ms- 1 and 20ms- 1• The rotational speed of the tribometer is measured continuously
using a magnet encoder.
Laboratory Measuring Device ofthe Friction Between Ski and lce or Snow 253
The device is placed in a cold room whose temperature can be set between -20°C
and 0°C. The humidity in the cold chamber oscillates between 60% and 80%. Infra-
red temperature sensors measure the temperature of the track in front of and behind
the sample (fig. 2).
2.2 Sample Preparation

The samples consist of aluminium plates onto which different materials are glued
(Fig. 2). The size of the sample can vary between 40 to 200mm (length) and 5 to
70mm (width). A soft foam rubber is placed between the sample and the force sensor
in order to assure a flat, parallel contact and to damp the small unevenness of the
track and vibrations. A special equipment enables to grind the small size samples
using a traditional ski grinding machine.
Fig. 2. Sample fixed to thc friction force sensor of the measuring arm with infrarcd surface
temperaturc scnsors in front of and behind thc samplc.
2.3 Ice and Snow Preparation

A steel bar is used to cut a perfectly flat ice surface (Fig. I ). For the experiments, an
elevated track is created, in order to avoid the slider to cut into the ice. Special cylin-
ders with given surface roughness can be mounted to roll on the ice track and pro-
duce a defined ice surface topography. For measurements on snow, special proce-
dures were developed in order to obtain homogeneous and sufficient hard snow.
3 Results
In order to verify the usefulness and accuracy to the tribometer, tests were carried out
in the laboratory and on the field with the same sliding materials. During theses
comparative tests, three different polymer base materials, denominated A, B and C,
were tested on three downhill skis as well as on three 70mm x 20mm laboratory
samples. Skisandsamples were grinded and waxed the same way.
3.1 Testing Conditions

The downhill skis were tested in summer 2005 in Kaunertal, Austria during two test
days. The snow was hard and coarse grained on both days. The snow surface tem-
perature was constant at -8°C +/- 0.5°C during the first day and increased from
2°C to 0°C and was slightly wet at the end of the second day. The air humidity was
around 65% on both days. A top level skier tested the skis three times during each
day. The average gliding time was around 19.5 seconds on the first day and 21.1
seconds on the second day. In order to analyze both days independently of the glid-
ing times, the parameter Diff_ Time was introduced. This parameter corresponds to
the time difference per second of sliding between one ski and the mean sliding time
for all skis.
The laboratory tests were carried out on the tribometer at a speed of 7ms· 1 at -1 °C
and -10°C on a flat ice surface. The pressure applied on the small samples was
30kPa, which corresponds to the static pressure applied by a 80kg skier on a down-
hill ski. The air humidity in the cold room was 70%. The friction tests with the small
samples were repeated three times for each temperature. After a run-in period of
30 s, the friction coefficient was measured for 10 seconds and averaged. The meas-
ured coefficients of friction were between 0.026 and 0.082. The parameter Diff_COF
was introduced in order to minimize the influence of different ice surface prepara-
tions on the coefficient of friction. The parameter Diff-COF corresponds to the dif-
ference in percent between the friction coefficient measured for one sample and the
mean friction coefficient for all samples at a given temperature.
3.2 Comparison Between Field and Laboratory Measurements

The results of the field tests show a big difference in the sliding properties of the
three ski-base materials (Fig.3). Skis with the base material B were the fastest for
both cold and wet conditions (a negative value of Diff_Time indicating a faster ski
than the mean ski). Base material C was the slowest especially for cold conditions.
Figure 4 displays the results of the laboratory tests which show the same trend as
observed during the field tests. The field and laboratory testing conditions were not
exactly identical, nevertheless, the correlation coefficient between these tests for
warm and cold conditions is 0.93 and 0.84, respectively, which suggests very good
correlation.
Laboratory Measuring Device ofthe Friction Between Ski and Ice or Snow 255
-8°C, d snow 0°C, we snow
0.02
0.00 0.00
., .,
E E
~I
~I
i5 Q
-0.02 -0.02
-0.04 ~ - - - + - - - - - + - - - e - - - ~ -0.04 ~ - - - - - - - - + - - - - - - + - - ~
A B C A B C
Fig. 3. Field test results for cold and warm conditions.
0.1 0.1 + - - - - + - - - - - + - - - - - - + - - - - !
u.. u..
8
:r::I 0.0
0
~I 0.0
i5 Q
-0.1 -0.1
A B C A B C
Fig. 4. Laboratory test results for cold and warm conditions.

4 Conclusions
The tribometer enables an accurate measurement of the friction coefficient of differ-

ent ski-base materials on ice and snow at high sliding velocity with a good control of
the track temperature. Such testing at constant conditions presents a major advantage
in comparison to field testing, where to many factors can influence the results.
The comparison between field and laboratory tests shows a very good agreement.
Nevertheless, more comparative tests should be conducted on other snow conditions
in order tobe able to reproduce other real conditions in the laboratory.
The tribometer can be very useful for the selection of the best ski-base surface
treatment for specific snow conditions. lt can also be used for the testing of new base
materials or surface treatments. Apart from ski bases, other equipments like ski
edges or skins for back-country skiing can also be tested with the tribometer. In
parallel to such comparative tests, the device is also being used for fundamental
research with the aim of developing a snow and ice friction model (tobe published).
Acknowledgment
Authors would like to thank the companies Stöckli Skis and Toko for the good col-
laboration during this project and the Commission for Technology and Innovation
(CTI) for its financial support.
References
Bowden, F.P., Hughes, T.P. (1939) The mechanics ofsliding on ice and snow. Proc. R. Soc.
London, Ser A 2117, 280-298.
Buhl, D., Fauve, M., Rhyner, H.U. (2001) The kinetic friction ofpolyethylenc on snow: the
influence ofthe snow temperature and the load. Cold Regions Sciencc and Technology 33,
133-140.
Colbeck, S.C. ( 1992) A review of the processes that control snow friction. CRREL Mono-
graph 92-2, U.S. Army Regions Research and Engineering Laboratory, Hanover.
Ducret, S. et al. (2004) Friction and abrasive wear ofUHMWPE sliding on ice. Wear. Res. 99,
110-118.
Eriksson, R. (1955) Friction of runncrs on snow and ice. SIPRE Report TL 44, Meddelande
34/35, 1-63.
Fauve, M. et al. (2005) Influence of snow and wcather characteristics on the gliding properties
of skis. Science and Skiing 3, 401-410.
Kuroiwa, D. ( 1977) The kinetic friction on snow and ice. Journal of Glaciology 19 (81 ), 141-
152.
Lethovaraa, A. ( 1987) lnfluence of vibration on the kinetic friction betwcen plastics and ice.
Wear 115, 131-138
Moldestad, D.A. ( 1999) Some aspects of ski base sliding friction and ski base structure. PhD.
Thesis Norwegian University ofScience and Technology.
Nachbaucr, W. et al. ( 1996) Effects of snow and air conditions on ski friction. Skiing Trauma
and Safety: Tenth Volumc, ASTM STP 1266, 178-185.
Biomechanical Instrumentation of the Berglsel Jumping
Hili in Innsbruck and Exemplary Analyses
Kurt Schindelwig, Werner Nachbauer

University Innsbruck, Austria, kurt.schindelwig@uibk.ac.at
Abstract. During the rebuilding of the Bergisel ski jumping hill, a new biomechani-
cal measurement system was installed. The system consists of seven force platforms
which were placed about 10.5 m along the jumping table and six high speed cameras
to film the take-off and the flight phases. Five elite ski jumpers, each with three
training jumps, were analyzed. In order to assess the accuracy of the measuring sys-
tems, velocity of the centre of mass during take-off was calculated from the kine-
matically and dynamically measured data and then compared. Additionally, each
jump was digitized three times by two different persons to determine the error of
semimanual digitizing. The results showed that the velocity of the centre of mass
differed by an average of 5 % with a maximum difference of 30 %. The maximum
difference between the repeated digitized data of the joint angles was three degrees.
From the results we concluded that the data are good enough for precise describing
jumping movement of elite ski jumpers. However, the !arge differences between the
two methods of calculating velocities of the centre of mass must be further exam-
ined.
1 Introduction
The most recent studies of ski jumping describe the correlation between special
parameters such as take-off velocity or rotation impulse at take-off and the jumping
distance (e.g. Arndt et al., 1995; Schwameder & Müller, 1995; Müller et al., 1996;
Jost et al., 1997, 1998; Virmavirta et al., 2005). Before this project, a ski jumping hill
equipped with a kinematic and dynamic measuring system did .not exist in Austria.
The aim of this work was to develop a measurement system allowing precise and
rapid determination of kinematic and dynamic parameters at the Bergisel jumping
hill. Comparisons of the velocity of the centre of mass calculated by the kinematic
and dynamic measuring data were used to assess the accuracy of the measuring and
analysis system.
258 Kurt Schindelwig
2 Methods
2.1 Measurement Devices
For kinematic recording of take-off, a field of view of 7 m before to 2 m behind the

jumping table was filmed with two high speed cameras with a sampling frequency of
230 Hz and a resolution of 1024 x 512 pixel (Vosskühler, Typ HCC-1000 CMOS)
(Fig 1.1 ). For the flight phase, four high speed cameras with a sampling frequency of
60 Hz and a resolution of 1280 x 512 pixel (ISG LightWise, Typ LW-l.3-G-1394)
were mounted (Fig. 1.2). Three photoelectric beams were installed at the position
10.5 m, 6.5 m and O m on the jumping table (Fig. 1.3). They were used for triggering
the kinematic and dynamic measuring systems and determining the horizontal veloc-
ity. The three force components were sampled ( 1000 Hz) by seven force platforms
(KISTLER, Typ Z 18401-100), which were fixed about 10.5 m along the jumping
table (Fig. 1.4). Two wind gauges were installed next to the jumping table.
Fig. 1. Measurement devices at the Berg !sei jumping hill

Biomechanical Instrumentation ofthe Berglsel Jumping Hill 259
2.2 Data Collection
Five elite Austrian ski jumpers were recorded during three training jumps under
windless conditions (speed ofwind less than 0.5 m/s). The force data were smoothed
with cubic splines. A special software program using Labview 7.1 (National Instru-
ments) was developed in order to digitize the image coordinates of the joint points.
The software permits automatic tracking with manual correction. From the image
coordinates the 2-d coordinates of the joint points were determined by the DLT-
method. The 2-d coordinates were smoothed with quintic splines.
Every jump was digitized three times by two different persons. The velocity of
the centre of mass was determined in two different ways. In the first method, a six
segment model of the ski jumper's body was used to compute the centre of gravity
and its velocity was determined from the first derivative of the smoothed coordi-
nates. In the second method the measured force normal to the jumping table was
integrated numerically. The dynamic calculation did not include the aero dynamical
lift.
3 Results
The maximum difference between the repeated digitized data of the joint angles was
three degrees (Fig. 2). The maximum velocity difference of the centre of mass be-
tween the multiple digitalisations was 0.15 m/s.
m rrom ddference
35
30
25
1;20 t - - - -
lii 15
10
05
00
Fig. 2. Maximum difference between the repeated digitized data ofthe joint angles
The difference of velocity of the centre of mass, calculated kinematically and dy-
namically, was on average 5 % with a maximum of 30 %. Some trials showed a
difference of only 0.12 m/s (Fig. 3).
260 Kurt Schindelwig
3
2,5
+-
1
~ 1,5
~ l kinematic
~ dynamlc '\
~ 0,5
--0,5
-1
7 6 5 4 3 2 l 0 -1
distance Im l
Fig. 3. Maximum difference between the repeated digitized data ofthe joint angles
4 Discussion
The results show that the system described precise the angles of the body
segments during ski jumping. However, the large differences between the
two methods of calculating velocities of the centre of gravity must be further
examined. One possibility is the large influence of the angle of trunk
inclination on lift. The influence of lift will be determined through wind
tunnel readings in future projects.
References
Arndt, A., Brüggemann, G.-P., Virmavirta, M., Komi, P.V. (1995) Techniques used by 0lym-
pic ski jumpers in the transition from take-off to early flight. Journal of Applied Biome-
chanics 11, 224-237.
Jost, B. Kugovnik, 0., Strojnik, V., Colja, I. (1997) Analysis ofkinematic variables and their
relation to the performance of ski jumpers at the World Championship in ski flights at
Planica in 1994. Kinesiology 29 (!), 35-44.
Jost, B. Vaverka, F., Kugovnik, 0., Coh, M. (1998) Differences in selected kinematic flight
parameters of the most and the least succesful ski jumpers of the 1996 World Cup
competition in Innsbruck. Biology of Sport 15 (4), 245-251.
Biomechanical Instrumentation ofthe Berglsel Jumping Hili 261
Müller, W., Platzer, D., Schmöltzer, B. ( 1996) Dynamics ofhuman flight on skis: improve-
ments in safety and fairness in skijumping. Journal ofBiomechanics 29, 1061-1068.
Schwameder, H., Müller, E. (1995) Biomechanische Beschreibung und Analyse der V-
Technik im Skispringen. Spectrum 1, 5-36.
Virmavirta, M., Isolehto, J., Komi, P., Brüggemann, G-P., Müller, E., Schwameder, H. (2005)
Characteristics of the carly flight phase in the Olympic ski jumping competition. Journal
of Biomechanics 38, 2157-2163
Dynamic Properties of Materials for Alpine Skis
Christian Fischer1, Mathieu Fauve2, Etienne Combaz1.2,

Pierre-Etienne Bourban 1,Veronique Michaud 1, Christopher J.G. Plummer 1,
Hansueli Rhyner2 and Jan-Anders E. Mänson 1
1 Laboratoire de Technologie des Composites et Polymeres (L TC), Ecole Polytechnique
Federale de Lausanne (EPFL), Switzerland, christian.fiseher@epfl.ch
2 WSL, Institute for Snow and Avalanche Research (SLF), Davos, Switzerland
Abstract. The aim ofthe present research has been to quantify the influence ofdifferent mate-
rials on the global dynamic responsc of a ski. The properties of the individual constitucnt
materials were characterized as a function of temperature and frequency using dynamic me-
chanical analysis. At thc same time, the dynamie behavior of skis with different designs was
investigated in a cold room at between -15 and 25 °C using specially developed apparatus. The
results indicatcd the overall behavior to bc influenccd significantly by the polymeric topsheet,
whieh showcd a strong damping peak at about O 0 C. Elasticity-based FEA aecounted weil for
the experimental results for the two first flcxural vibrational modes of the skis. For higher
modes, however, the viseoelastic nature of the polymeric components led to significant dis-
crepancies between the predicted and obscrvcd behaviors.
1 Introduction
A principal aim of ski manufacturers in recent years has been to reduce vibration in
various ways. Skis completely devoid of vibration nevertheless do not procure good
sensations for the athlete, so that it is crucial to discriminate between those frequen-
cies that should be damped to increase performance, and those that are important for
the skier's "feel". New test methods have therefore been developed to analyze the
dynamic properties of skis. Understanding how skiers apply forces and how vibra-
tion pattems affect the feel of a snowboard or a ski should allow a more rational
approach to design and improve overall performance.
The dynamic properties of skis were first investigated systematically in 1972
(Piziali and Mote 1973). Glenne et al. (Glenne, Jorgensen and Chalupnik 1994) sub-
sequently compared different measurement devices, showing that small amplitude
tests such as the /SO test may not be representative of field conditions. More re-
cently, various groups have integrated the boot/binding system into their analyses so
as to reproduce real skiing conditions more accurately (Glenne, DeRocco and Foss
264 Christian Fischer et al.
1999; Casey 2001 ). Comparison of the results with results obtained from free-
suspension tests demonstrated the important role of the boot and binding in cutting
off high frequencies. The behavior of skis on snow has also been studied in situ
using accelerometers (Nemec 2001 ), leading to the conclusion that carving skis re-
sult in less vibration during tums by preventing skidding. Nevertheless, resonance
may still occur and is often detrimental to performance in that it reduces ground
contact, so that the skier is no longer able to continue the carved turn. High-
amplitude deformations at low frequencies are of particular concem in this respect.
Comparatively little attention has so far been paid to the influence of the con-
stituent materials on the dynamic response of skis, although it is known that they
may play a significant role (Scherrer, Bidaux, Kirn, Mänson and Gottardt 1999). The
aim of the present work has been to investigate the contribution of selected compo-
nents to the global behavior, with emphasis on its temperature dependence. To this
end, the mechanical properties of the different materials that constitute the ski have
been characterized independently and then linked explicitly to experimental results
for the vibrational properties ofthe entire ski by finite element analysis (FEA).
TopahMt
Uppar Face
Core
Fig. 1. Sandwich structure of a ski, showing the main elements of its construction. For clarity,
glue layers are not indicated, but are present at each of the interfaces shown.
A schematic of the cross-section of a competition ski is given in Fig. 1. Each part of

the sandwich structure has a specific function. The topsheet is intended as a protec-
tive layer. The wood core, which has a non-uniform thickness, giving a smooth
bending profile, plays an important role in stiffness and damping. The glued alumi-
num alloy/composite (usually a glass fiber-epoxy laminate) stack that constitutes the
upper and lower faces of the ski determines stiffness in bending and torsion. The
only material whose characteristics are essentially identical in different skis is the
aluminum alloy, specifically developed for ski applications and said to show an
optimum combination of stiffness and weight. In the course of the present investiga-
tion, skis with various types of sandwich structure were considered, differing in the
construction of the upper and lower faces, the face thicknesses and presence or ab-
Dynamic Properties of Materials for Alpine Skis 265
sence of the polyamide topsheet. However, the following discussion will concentrate
on the influence of the topsheet and the structure of the upper face. The ski with the
topsheet is referred to as ski A, and an identical ski without the topsheet (replaced by
a simple powder surface coating), is referred to as ski B. Ski C contained an addi-
tional layer of cured epoxy impregnated glass fiber mat glued to the aluminum Iayer
of the upper face. Ski D was identical to ski C without the additional composite layer.
2.2 Dynamic Mechanical Analysis (DMA)

Specimens from the topsheet were clamped at both ends and subjected to dynamic
deformation in torsion, using a strain-controlled Rheometrics ARES rheometer. The
remaining materials, which were significantly stiffer than the topsheet, were tested in
three-point bending using a Rheometrie Solids Analyzer (RSA). 52 x 12 mm 2 speci-
mens were used throughout, but their thickness varied according to the material.
2.3 Ski Testing

The Skitester 2004 is a device specifically developed for dynamic testing of whole
skis and snowboards. The principle is to measure the acceleration after a standard-
ized shock imparted to a ski clamped at the position of the binding. A small hammer
hits the front of the ski, close to its edge, exciting both torsional and flexural vibra-
tional modes. The corresponding displacements are measured by laser reflection. The
experiments were performed at temperatures ranging from -15 to 25 °C.
2.4 Finite Element Analysis (FEA)

Elasticity-based FEA (Ansys 8. J) was used to verify the measurement technique and
the experimental results. The ski was represented by a multi-layer mesh incorporat-
ing elastic material parameters extracted from DMA measurements at room tempera-
ture and a frequency of 1.5 Hz. lt was composed of perfectly bonded 3D elements
whose thickness corresponded to that ofthe thinnest layer ofthe structure.
3.1 Vibrational Modes and FEA

The first five resonance frequencies for ski A obtained experimentally at 25 °C and
calculated using FEA are shown in Table 1. The agrcement between the calculated
and obscrved behavior for the first two vibrational modes was good. For higher
modes, however, it was poorer. This was due to the viscoelastic nature of the poly-
mer-based materials in the sandwich structure (glues, composite laminates, wood, the
topsheet). Viscoelastic effects have not so far been incorporated into the calculations,
but become increasingly important as the frequency increases.
3.2 lnfluence of the Topsheet

Fig. 2 shows the 151 resonant frequencies of ski A and ski B at different temperatures,
showing consistently lower values for ski A than for ski B. This was attributed to the
increased damping of the ski from the polyamide topsheet, this generally tending to
reduce the resonant frequency (Balachandran and Magrab 2004). The resonant fre-
quency of ski B decreased monotonically with increasing temperature. However, ski
A showed a significant minimum at 0 0 C.
Table 1. Rcsonant frequencies for ski A at 25 °C dctermined experimcntally and predicted
using FEA.
Mode Experimental FEA

151 flexural 13.4 Hz 13.6 Hz
2nd flexural 48.6 Hz 46.1 Hz
151 torsional 93.9 Hz 72.6 Hz
3rd flexural 128.1 Hz 104.4 Hz
41h flexural 234.5 Hz 171.1 Hz
102~--------------~
c::::J 25'C
10 o o·c
- -20'C
98
"
~ 9.6
?;
~ 94
:,
I 92
9.0
8 8 4-'---.__.....-
SktA SkiB
Fig. 2. Comparison ofthe 1' 1 resonant frcqucncy of ski A (lcft) and ski B (right) at different
temperatures. These measurements were carried out using accelerometers placed on the ski
rather than laser reflection. The added weight ofthe accelerometer also reduced the resonance
frequencies somewhat (cf. Table 1), but had no effect on the relative influcnce of the topshcct.
The mechanical response of viscoelastic materials is sensitive not only to tem-

perature but also to strain rate. This is illustrated in Fig. 3(a), which shows tanövs.
temperature for the topsheet at different constant angular frequencies, {JJ. The tanö
values reflect the damping capacity of a material and are equivalent to the loss factor,
17, at low damping levels (Graesser 1992). The results in Fig. 3(a) clearly show the
damping peak, associated with the glass transition in polyamides, to shift to higher
temperatures as {JJ is increased. This peak dominates the temperature regime charac-
Dynamic Properties ofMaterials for Alpine Skis 267
teristic of service conditions for an alpine ski, which explains the minimum in the
resonant frequency for ski A at 0 °C (cf. Fig. 2).
T o gain an idea of the dynamic properties of viscoelastic materials at frequencies
beyond those that are directly accessible experimentally, time-temperature superposi-
tion is often used (Ferry 1980). Fig. 3(b) shows time-temperature superposition of
the elastic modulus G ·, the loss modulus G ", and tanö for a reference temperature of
0 °C. The damping peak appeared at w in the range 10 - 90 rad/s (fbetween about 1.5
and 15 Hz). This was consistent with the dynamic behavior of the skis in Fig. 2. Ski
A, with the topsheet, showed a maximum loss factor at 0 °C for the first resonant
frequency (84.2 rad/s or 13.4 Hz), whereas ski B, without the topsheet, showed a
continuous decrease in loss factor as the temperature decreased from 25 to -20 °C.
.
-
__0_,,..
025
- 1~ 1ED
.. 10•• G'
-·-·- --- ........

0 G"
-·- 100,..., .,,.
' ,
T - - - ..
:;:: 0.20
'l! ' 'ii' 02::;:
.!I !!,.
1
j 015
C)
C)
1EI
l5
ü
J
i
.!!
J
0 10
-15 ,10 .5 0 10
l• I
,s
1E7 "o ., 0,
001 01 10 100 1000
1emperature. T rCJ
frequency. oo (nodls)
Fig. 3. (a) Evolution ofthe main damping peak ofthe polyamide topsheet as a function of
applied frequency and (b) time-temperature superposition for a reference temperature ofO °C.
10•
r:::::l25'C
o·c
,., 008
007
c::::J2s·c
o·c
(b)
- -20·c - -20'C
0011
102
r::
::;: 005
I ~ll 004
j 003
J 002
001
98
000
S>J C
Fig. 4. (a) 1st resonant frequencies of ski C and ski D at different temperatures and (b) the
corresponding damping properties.
3.3 Influence of Variations in the Structure of the U pper Face

The influence of the additional epoxy-glass fiber mat in the upper face of ski C is
shown in Fig. 4, which also give results for ski D. In this case, although the increase
in stiffness associated with the presence of the additional glass mat clcarly increased
the structure's I st resonance frequency (Fig. 4(a)), it had relatively little intluence on
damping behavior (cf. Fig 4(b)). The stiffness could also be increased by using
thicker aluminum plates for the sandwich faces. However, this led to an undesirable
increase in the overall weight ofthe ski and to significantly decreased damping.
4 Conclusions
The influence of the constituent materials properties on the overall dynamic behavior
of skis has been investigated. The results indicated the overall behavior to be influ-
enced significantly by the polymeric topsheet, which showed a strong damping peak
at about O 0 C. Elasticity-based FEA accounted weil for the experimental results for
the first two flexural vibrational modes of the skis. For higher modes, however, the
viscoelastic nature of the polymeric components led to increasing divergence be-
tween the predicted and observed behaviors. In futurc work, it will therefore be nec-
essary to introduce more complex models that take into account these effects.
Acknowledgements
The authors gratefully acknowledge the Stöckli Ski Company for materials and skis,
the Interstate University of Applied Sciences of Technology Buchs for designing the
Skitester 2004, and the Sports and Rehabilitation Engineering (SRE) program of the
EPFL for financial support.
References
8alachandran, 8. and Magrab, E.8. (2004) Vibrations, 8rooks Cole, Pacific Grovc, CA.
Casey, H. (2001 ). Materials in ski design and development. In: Materialsand Science in
Sports, Frocs, F.H. and Haake, S.J. (Eds.), TMS, Warrendale, PA, pp. 11-17.
Ferry, J.D. (1980) Viscoelastic Properties of Polymers, Wilcy, New York.
Glenne, B., DeRocco, A. and Foss, G. (1999). Ski and Snowboard Vibration. Sound Vib.
33(1), 30-33.
Glcnnc, 8., Jorgensen, J.E. and Chalupnik, J.D. (1994) Ski Vibrations and Damping. Exp.
Tcchniqucs 18(6), 19-22.
Graesser, E.J. and Wong, C.R. (1992) The Relationship ofTraditional Damping Measures for
Materials with High Damping Capacity: A review. In: M'D: Mechanics and Mechanisms
o/Material Damping, Kinra, W. and Wolfendcn, A. (Eds.), ASTM, Philadelphia, pp. 316-
343.
Nemec, B., Kugovnik, 0., Supcj, M.(2001) Intlucnce ofthe Ski Sidc Cut on Vibrations in
Alpine Skiing. Sei. and Skiing II, 232-241.
Piziali, R.L. and Motc, C.D. (1973). Snow Ski as a Dynamic System. Mech. Eng. 95(2), 52-
52.
Scherrer, P., Bidaux, J.-E., Kirn, A., Mänson, J.-A.E. and Gottardt, R. (1999) Passive vibration
Damping in an Alpine Ski by Integration of Shapc Memory Alloys. J. Phys. IV 9, 393-
400.
Calculation of Friction and Reaction Forces During an
Alpine World Cup Downhill Race
Michael Schiestl 1, Peter Kaps', Martin Mössner2, Werner Nachbauer2
1 University oflnnsbruck, Department ofEngineering Mathematics, Geometry, and Computer

Science, Michael@Schiestl.name
2 University of Innsbruck, Department of Sport Science, Wemer.Nachbauer@uibk.ac.at
Abstract. Understanding friction and reaction forces involved in Alpine Skiing is of great
theoretical importance for sport science. We have developed a method to analyze a skier's
motion during a downhill race from video data taken by a single camera. This may help to
compare the technical equipment and the skills of different skiers.
1 Introduction
In the following we will present a method to reconstruct the trajectory of a skier from
motion pictures taken by a single camera. This will allow us to set up the equations
of motion on the measured path in order to calculate the friction and reaction forces.
Finally, we will present the results of the analysis for Stefan Eberharter's downhill
race during the Alpine World Cup in Kitzbühel in 2002.
2 Method
2.1 Surface generation

The calculation of an object's position visible on just a single picture requires the
object to be located on a known surface - in our case the skiing slope. Therefore, the
first kilometer of the famous skiing slope Streif in Kitzbühel was geodetically sur-
veyed by measuring the position of approx. 550 terrain points with a theodolite.
Then, a mesh describing the snow surface was generated by triangulation of the
terrain points (see Fig. 1).
To guarantee realistic results it is of importance that the skier's trajectory is re-
constructed on a preferably smooth surface or - more precisely - on a surface with
continuous first derivatives, also called G 1-surface. In our case the terrain points
were distributed irregularly - a fact that restricts the possible methods. We chose the
Clough-Tocher algorithm which generates a G 1-surface from any given triangulation
270 Michael Schiestl et al.
(Hugentobler 2004). This way we obtained a parameterization ofthe snow surface of

the form z = z(x, y) with (x,y) being the horizontal, and z the vertical Cartesian coor-
dinate (terrain altitude).
2.2 The DLT

The direct linear transformation (DLT) is a well known photogrammetric method for
reconstructing an object's position from image data. However, the calculation of3-D
coordinates of a moving object is only possible if either multiple pictures are avail-
able which are simultaneously taken by several cameras or the object is located on a
known surface. Furthermore, a minimum number - depending on the type of algo-
rithm - of visible pass points with measured position coordinates is required. Gener-
ally, the more pass points on an image the better the precision ofthe result.
We have modified the DLT algorithm as described e.g. by Kwon (1998) in two
respects: Firstly, we solved the DLT equations on the Clough-Tocher G 1-surface
describing the skiing slope. Secondly, we improved the DLT's precision by exploit-
ing the fact that when a series of pictures is taken by the same camera, e.g. during a
motion video, several camera parameters remain constant and do not have to be
recalculated every frame. This way - if e.g. the camera's position is known - fewer
pass points are required and the precision of the measured data is improved.
This advance could be achieved by transforming the 11 DLT parameters describ-
ing the camera projection into a new set of parameters which can be interpreted in a
more useful way, namely the camera's position and orientation as well as zoom and
distortion factors. This way situation-specific equations could be derived that can be
used to improve the accuracy ofthe measurement in many cases (Schiestl 2005).
2.3 Reconstruction of the Trajectory

The positions of the left and right ski binding were reconstructed on large parts of the
surveyed track. The images taken at a frame rate of 25 pictures per second allowed a
sufficiently precise reconstruction of the skier' s motion during a time period of 10
seconds on the whole. The mean value of the left and right ski was taken as an ap-
proximation of skier's center of mass. We received the trajectory by smoothing the
data during three different steps:
• The image coordinates of the pass points and ski bindings were slightly
smoothed.
• The reconstructed positions x(t;), y(t;) and z(t;) were smoothed as a function of
time. Cubic spline functions were used to approximate the data points.
• For each data point calculated in step 2 the curve length l; was determined. Then
the set of data points (t;, l;) was taken to evaluate smoothed curve lengths L; at the
same times !;. Finally, the data points (X(t;), Y(t;),Z(!;)) having the smoothed curve
length L; were calculated. Any further analysis resorts to the curve fitting these
positions.
Calculation ofFriction and Reaction Forces During an Alpine Downhill Race 271
Note that the third step does not change the shape of the trajectory, however it
smoothes the accelerations along the path.
2.4 Calculation of Friction and Reaction Forces
Let the measured trajectory be a curve r(t) := (x(t ),y(t ),z(t )Y

parameterized by the
timet. From Newton's Law the skier's equation ofmotion can be written as
mF(t) = tg + + f,. i\ (1)
with m being the skier's mass, F, := {o, o, - mg Y the gravity force, ~- being any
energy conservative force and F1 representing non conservative resistance forces
such as kinetic and air friction as weil as the resistance in tangential direction due to
the snow spray of the carving edges (Hirano and Tada 1996). Note that ~- arises
from the constraint of r(t) being restricted to values on the measured trajectory - in
other words ~- can be regarded as the reaction force.
By definition ~- does not perform work, hence ~- must be orthogonal to the
skier's velocity vector i=, in other words fa. ~- = o. Analogously, the friction force is
parallel to the velocity: lltr II= Ff ·i= tlli=II .

Let D := m F- Fg . Then dot-multiplying Eq. l with fa gives:
Ff =(iJ.fa)fa1lli=f ~- =D-Fr (2)
3 Results
In the following we will present the results for Stefan Eberharter's run during the
Alpine World Cup downhill race in Kitzbühel 2002. Three parts of the section
known as "Mausefalle" as weil as the "Steilhang" were successfully reconstructed
and analyzed, in Fig. 2-5 the second and third part ofthe Mausefalle are shown.
The precision of the measured trajectory strongly depends on the number and the
quality of the pass points. In favorable parts of the track the accuracy can be ex-
pected to be just a few centimeters and less. The accuracy of the reaction and friction
forces is worse (approximately +/- 20 %) due to the second derivatives involved in
the calculation. Also keep in mind that the reaction force includes the skier's body
weight of 100 daN.
r-:--ong,nal data ]
L-~ ~
__ .... .....4.. ... . .. . .·---:·

-.=·~.
Xlm)
Fig. 1. The triangulated starting area Fig. 2. The reconstructed trajectory

of the Streif Mausefalle part 2
250.----,-----.---.----,----- .---.----,---,
""i'.--- -.....- - •• 52
Tlfflel{MC)
~ ..
•• .
Fig. 3. Results Mausefalle part 2
Calculation ofFriction and Reaction Forces During an Alpine Downhill Race 273
70
/ /
60
1 t
: : : :..... 1 //
# ,
/
50
<IO
30
I
>- 20
10
- 10
- 20
60 20 0 - 20 --40 - 60
X [m)
Fig. 4. The reconstructed trajectory Mausefalle part 3
250
200 reacllon force [daN)
150
100
50
- 50
- 100
- 150
- 200 .... _.._ ____._

85 g 85 10 105 11 11 5 12 125 13
T1met[sec[
Fig. 5. Results Mausefalle part 3

4 Discussion
The reaction force curve assumes values up to 240 daN. Thus, the forces acting on
the leg extension muscles of a skier can reach critical values during short periods of
time. Friction values of 50 daN and )arger are fairly high and might be reduced by
better equipment and skiing technique.
The calculated friction consists of the air resistance and the tangential resistance
between ski and snow. For straight running or traversing the snow friction can be
modeled as Coulomb friction (Kaps et al 1996). Y et, it seems that during tums other
friction mechanisms have to be considered, e.g. the tangential resistance due to the
snow spray of the carving edges. Attempts to calculate drag and friction coefficients
cJA and µ without such effects failed. However, the development of more advanced
friction models might require the measurement of parameters such as the angle be-
tween ski axis and ski velocity which plays an important role for skidding. To deter-
mine this angle the accuracy of the measurements has to be improved.
5 Conclusion
This study shows that the friction and reaction forces acting on a skier can be meas-
ured during a competition without interfering with the athlete. By using better cam-
eras and more pass points one could improve the accuracy of the measurement
greatly allowing a refined analysis of friction models in future studies using this
method.
References
Hirano, Y., Tada, N. (1996), Numerical simulation ola turning alpine ski during recreational
skiing, Medicine and Science in Sportsand Exereise, Vol. 28, pp. 1209-1213.
Hugentobler M. (2004), Terrain Modelling with Triangle Based Free-Form Surfaces, PhD
Thesis, University of Zürich, Mathematical Faculty.
Kaps P., Nachbauer W., and Mössner M. ( 1996), Determination of Kinetic Friction and Drag
Area in Alpine Skiing, Skiing Trauma and Safety: Tenth Yolume, ASTM STP 1266, Mote
C.D., et al., American Society for Testing and Materials, pp. 165-177.
Kwon Y.H. (1998), DLT Method, http://kwon3d.com/theory/dlt/dlt.html.
Schiestl M. (2005), DLT Data Reconstruction on Clough-Tocher lnterpolated G 1-Swfacesfor
Simulation in Alpine Skiing, Diploma Thesis, University of Innsbruck, Mathematical Fae-
ulty, http://Michael.Schiestl.name/publications.
Measurement of Jumper's Body Motion in Ski Jumping
Yuji Ohgi1, Kazuya Seo2 , Nobuyuki Hirai3, Masahide Murakami3
1 Graduale School ofMedia and Govemance, Keio University, Japan, ohgi@sfc.keio.ac.jp

2 Faculty ofEducation, Art and Science, Yamagata University, Yamagata, Japan
3 Graduate School ofSystem and Information Engineering, University ofTsukuba,
Ibaraki, Japan
Abstract. The authors propose motion analysis of ski jumping by data logging method using
inertia sensors. A tri-axial accelerometcr and a single axial gyroscope were attached onto the
jumper's body. Simultaneously, high speed cameras were equipped for the image capturing in
order to obtain the jumper's motion from the take-off and thc middle phase of the flight. The
characteristics of the ski jumper during whole phases werc identified by using the inertia
sensors.
1 Introduction
Flight dynamics of ski jumping through the whole phases has not been explained. As
for the biomechanical aspect, kinematical and kinetical methods were applieci in the
field studies (Babiel, Hartmann, Spitzenpfeil and Mester 1997; Kaps, Schwameder
and Engstler 1997; Komi and Virmavirta 1997; Yamanobe and Watanabe 1999).
Aerodynamic approaches were also examined for the steady flight phase (Seo, Wa-
tanabe and Murkami 2004a ; Seo, Murakami and Yoshida 2004b). Most of these
studies were conducted for the motion either on the slope or in the air. Total kinema-
tical data measurement from the start, inrun, take-off, steady flight to landing was
not reported.
The purpose of this study was a kinematical measurement of ski jumping through all
phases by using inertia sensors and high speed video cameras. Then authors intended
to analyse a relationship between the sensor information and the jumper's motion
based on the video analysis.
276 Yuji Ohgi et. al.
2 Methods
2.1 Field Measurement

Figure 1 shows the schematic illustration of our experimental equipment. The au-
thors applied a tri-axial accelerometer (Hitachi Metals Inc., H480A-2G5VR) and a
single axial gyroscope sensor (Murata Corp., ENC-031). For the data acquisition, a
PDA (TOSHIBA GENIO, WindowsCE 2003) and compact flush A/0 converter (C-
Cubed Ltd., Dataq CF2) were used. The sensor unit and PDA were located on
jumper's low back at L5. Data sampling was taken on 256Hz.
A local coordinate system was defined on the jumper's trunk. As for the gyro-
scope sensor, it rotated around y-axis of the local coordinate system (Fig. 2). Positive
sign of the angular velocity means that the jumper takes forward tilting motion and
negative corresponds to the backward tilting, vice versa.
In the field measurement at Hakuba Ski Jumping Stadium, a high speed video-
graphy was also taken on the !arge hill slope. Three high speed video cameras were
equipped from both side of the !arge hill slope. Two of those cameras (Photron lnc.,
Fastcam-R2, 250Hz) were configured from the left side at a distance of 15m to take
both pre and post take-off phases. From the apposite right side, at a distance of 40m,
another camera (Photron Inc., Fastcam-1024PCI, 500Hz) was set up for the wide
range of the flight phase approximately 1.8 s.
In this study, the ski jump is assumed to be performed on the sagittal XZ plane on
the global coordinate system, which was located on the edge of the slope. Subject
was a high school ski jumper (171.0cm, 55Kg, four years career). Jumper's head
(helmet), hip, knee, ankle, toe, top and tail of the ski were digitized for analysis.
Those digitized coordinates were once approximated to the 5th ordered polynomial
equations, then reconstructed as the horizontal (X) and vertical (Z) coordinates.
2.2 Flight Dynamics Model of Ski Jumping

In Fig. 2, U indicates a velocity vector of the center of gravity of the jumper-ski
model. For our convenience, we assumed the hip joint tobe a center of gravity in this
study. And our sensor unit was also assumed tobe attached on this location. In this
figure, the jumper's trunk and lower extremities were combined as a single segment
and its angle 0, o. represented his flight motion in the air. 0 is the trunk angle to the
horizontal plane, o. is angle between trunk and the velocity vector U.
Mcasurcment of Jumper's Body Motion in Ski Jumping 277
Fig. 1 A schematic illustration ofthe sensor unit and the logging unit
Horizontal Line ·-··
x-axial component
L eo (7t/2 - 0) - Deo o: - mg sin 0
z-axial component
L sin(7t/2 - 0) - D in o: - mg eo 0
Fig. 2 Local coordinate system of the ski jumper. The gravity and thc aerodynamic force
excrted to the jumper-ski system were illustrated in figurc. Drag and lift forces were disas-
semblc to thc scnsor axis dircction.
3 Results
An example ofresults was shown in Fig. 3. The acccleration and the angular velocity
have two different frequency components around the take-off. For that reason, the
authors applied 7th ordered adaptive smoothing filter (Kawata-Minami filter), which
is applicable to time series with spike peaks or instantaneous transient local peaks.
Acceleration on x and z-axis of the sensor unit has large and high frequent vibration
at the take-off instance. Similarly, the angular velocity steeply changed negative,
then tumed into positive again after take-off. After that, you can see that the angular
velocity indicated positive, which means jumper's forward tilting motion. The result
shows that subject tried to do his forward tilting motion at least three times. Because
the angular velocity Wr has three local maximums after his take-off. This multiple
tilting attempts of jumper were stable within subject through his all jump trials. Bar
graph in Fig. 4 shows that duration time of the forward tilting (FT), steady flight (SF)
and backward tilting (BT) phases among successful seven jumps. Forward tilting
phase was determined by the interval between take-off transient acceleration peak
and the moment when the angular velocity reached O deg/s. After forward tilting
motion, the jumper kept steady flight phase. Finally, he draws up his upper body to
the upright position to prepare his landing. So, backward tilting motion was distin-
guished by the angular velocity when it became negative. Mean and standard devia-
tion of total flight time were 3.225±0.369s. The initial angle between jumper's trunk
and horizontal line 00 and the angle between trunk and the velocity vector a0 were
obtained from the video analysis.
. lO
~
10
! 10
~
] r
,10 Tm1e(•l
~
< ·lO
JO
-';.
"'I
10
"'
r-
! 10
y
~
i 10
~
<
-:1
!
~
=1
10
,
~
t
1
-10 TmlC( l
-:1
,10
-1!
f' 1 (l)y
__
!~ .~: 1 -~
i,
o,i-06 ~
Tun<(,)
100 T,mc()
T L <
Fig. 3 Accclcration and angular velocity ofthe ski jumpcr during wholc jump (right), and pre
and post takc-off phase (lcft). Start (ST), Inrun (IR), Take-Off (TO) and Landing (LA) arc
indicatcd.
Mcasurcment of Jumper's Body Motion in Ski Jumping 279
'ÖD
2,l
<:5 14
12
l
Landing
D Back ward tilting
tcady night
■ Fol"\ ard tihing
2 3 4 5 6 7 Takc-ofT
Trial
Fig. 4 Initial trunk angle (0 0 ), initial vclocity (v 0) ofthe jumper and his flight phasc time dis-
tribution.
4 Discussions
On the basis of simple model in Fig. 2, the authors assumed the center of gravity of
the jumper-ski system to be located on his hip joint. And the upper trunk and the
lower extremities were assumed a straight line. In addition, the ski is also assumed to
be same segment with jumper. lt is quite simple model of the jumper-ski system.
Usually, the aerodynamic forces were taken into account that those act to the CG.
Then drag and lift are decomposed to the orthogonal components on parallel and
vertical direction to the velocity. However, in this figure, those decomposed forccs
were decomposed again with respected to the sensor sensitivity axis. Thus, compos-
ite forces act on the x and z-axis direction of the local coordinate system, which
correspond to the sensor axis are forrnulated as below.
L cos (;r/2-0) - D cosa. - m g sin0 (1)
L sin (;r/2-0) - D sina. - m g cos0 (2)

Since our used accelerometer is capable to measure static acceleration, so that, when
we hold the sensor unit, it detects Jg. However, in case of free fall, it detects no
gravitational acceleration on any axis. Therefore, we have to consider that no gravi-
tational components, as terrns mgsin0, mgcos0, would be observed. For the calcula-
tion of the drag and the lift, we need the time series of both 0and a.. As for 0, it can
be computed by the integration of the angular velocity w,.
(Fig.5). On the other hand,
a. requires the tracked coordinates of CG and its velocity direction. First of all, we
have to prove the integration of the gyroscope data for0. Then, we can carry on the
calculation of the drag and lift force. An examination of those important aerody-
namic forces is now being investigated.
50·
40! FT SF
,-. i
~301
:320i
a,
10[
+--~~-~~2==---3~_i____i_
TO Time(s)
Fig.5 Jumper's trunk angle 0, which was calculated by the integration ofthe gyroscope data.
Initial angle 00 was determined by the video analysis at the take-off. Forward tilting (FT) ,
steady flight (SF) and backward tilting (BT) intervals were indicated in the figure.
5 Conclusions
The authors conducted a measurement of jumper's motion in ski jumping by using

inertia sensors with the high speed videography. According to the obtained inertia
sensors' data, we could derive jumper's trunk motion during whole jumping phases.
Also at the conference, the estimation of the aerodynamics will be presented.
References
Babicl, S., Hartmann, U., Spitzenpfeil, P. and Mcster, J. (1997) Ground reaction forces in
alpine skiing, cross-country skiing and ski jumping., Science and Skiing, Müller E. et al.
(Eds) 200-207.
Kaps, P., Schwameder, H. and Engstler, G. (1997) Inverse dynamic analysis oftake-offin ski
jumping, Science and Skiing, Science and Skiing, Müller E. et al. (Eds) 72-83.
Komi, P.V. and Virmavirta, M. (1997), Ski-jumping take-offperformance: Determining
factors and methodological advances. Science and Skiing, Müller E. et al. (Eds), 3-26.
Sco, K., Watanabe, I. and Murakami, M. (2004a) Aerodynamic force data for a V-style ski
jumping flight. Sports Engineering, 7, 31-39.
Seo, K., Murakami, M. and Yoshida, K. (2004b) Optimal !light technique for a V-style ski
jumping. Sports Engineering, 7, 97-104.
Yamanobc, K. and Watanabe, K. (1999) Measurement oftake-offforces in skijumping com-
pctition, Japanese Journal of Biomechanics in Sportsand Exercise, 3, (4 ), 277-286.
Checkfor
updates
RETRACTED CHAPTER: Riding on Air: A New Theory

for Lift Mechanics of Downhill Skiing and Snowboarding
Qianhong Wu 1, Yesim Igci2, Yiannis Andreopoulos 3 and Sheldon Weinbaum 3
ER
1 Villanova University, qianhong.wu@villanova.edu
2 Princeton University
3 The City College ofNew York
Abstract. A simplified mathematical model is derived to describe the lift mechanics of down-
hill skiing and snowboarding, where the lift contributions due to both the transiently trapped
PT
air and the compressed solid phase (snow crystals) are deterrnined. To our knowledge, this is
the first time that anyone has attempted to realistically estimate the relative contribution of the
pore air pressure to the total lift in skiing and snowboarding. The model uses Shimizu's em-
pirical relation to predict the local variation in Darcy perrneability due to the snow compres-
sion. The forces and moments on the skier or snowboarder are used to predict the angle of
HA
attack of the planing surface, the penetration depth at the leading cdge and thc shift in thc
center of pressure for two typical snow types, fresh and wind-packed snow. Our model pre-
dicts that, when there are no edging or tuming maneuvers and the velocity of the snowboarder
or skier, V= 20 m/s, for fine-grained, wind-packed snow approximatcly 50% of the total lift
force is generated by the trapped air for snowboarding and 40% for skiing. For highly perme-
able fresh powder snow the lift contribution from the pore air pressure drops substantially. The
C
force and moment balance on the planing surface due to the trapped air and the snow crystals
are then used to develop a theory for control and stability in rcsponse to changes in thc center
ofmass as the individual shifts his/her weight.
('
1 Introduction
Downhill skiing or snowboarding, in its simplest form, refers to the motion of a

human sliding down an inclined plane on a porous medium. The extensive classic
literature treating the science of skiing and snowboarding is summarized in (Lind
&7
and Sanders 1996). A major contribution in this literature is the pioneering work of
(Colbeck 1991; 1992; 1994a; 1994b; 1995) on the reduced frictional force that re-
sults from the micron-thick fluid film that forms on the underside of the ski or snow-
board due to frictional heating. More recently, Feng and Weinbaum (2000), hereafter
referred to as F&W, have developed a new lubrication theory for highly compressi-
$
ble porous media which shows a remarkable dynamic similarity between the motion
of red blood cells (RBCs) on the endothelial surface layer (ESL) that lines our mi-
crovessels and that of a human skier or snowboarder skiing on soft snow powder
75
even though their difference in mass is of order 10 15 • F&W predict that the excess
pore pressure generated by a planing surface moving on any compressible porous
media scales as d = h2/K, where h is the layer thickness and K is the permeability,
The original version of this chapter was revised:
The Editors have retracted this Conference Paper because the content overlaps significantly with
5(
a previously published article (Wu et al., 2006). Qianhong Wu does not agree with this retraction.
Yiannis Andreopoulos and Sheldon Weinbaum have not responded to correspondence about this
retraction. The Editors could not obtain a current email address for Yesim lgci. Wu, Q., lgci, Y.,
Andreopoulos, Y., & Weinbaum, S. (2006) Lift mechanics of downhill skiing and snowboarding.
Medicine and Science in Sportsand Exercise, 38(6), 1132-1146.
This has been added in the affected chapters. A correction of this chapter can be found at
https://doi.org/10.1007/978-0-387-46050-S_77.
282 Qianhong Wu et al.
ER
PT
Fig. 1. (a). Schematic illustration of a snowboard or ski comprcssing a layer of snow powder. (b). Forces
upon a skier as he/she descends down the fall linc.
HA
and that a is of order 10 2 or !arger for both RBCs gliding on the ESL and humans
skiing. Thus, the lift forces generated can be four or more orders of magnitude
greater than the classical lubrication theory. Thc enhancement in lift arises from the
fact that, as the porous media compresses, there is a dramatic increase in the lubrica-
tion pressure, because of the marked increase in the hydraulic resistance of the fluid
C
as it tries to escape from the confining boundaries of the planing surface through the
compressed porous layer. The principal difference between the tightly fitting RBC in
a capillary and a human skier or snowboarder is the leakage of the excess pressure at
the lateral edges ofthc skis, as can be seen from Fig. la.
('
F&W theory was first verified qualitatively by the snow compaction experiments
in (Wu, Andreopoulos, and Weinbaum 2004; Wu, Andrcopoulos, Xanthos and
Weinbaum 2005b; Wu 2005). However, this thcory has the following major limita-
tions for skiing or snowboarding: (i) the 2-O-fiber-array modcl in F&W is not appro-
&7
priate for predicting the variation of snow perrneability due to the wide variety of
shapes and sizes of ice crystals (Arons and Colbeck 1995); (ii) the assumption that
the ice crystals' contribution to the total lift is negligible in F&W is not appropriate
based on the experiments in (Wu, et al. 2005b; Wu 2005); (iii) the F&W analysis
emphasizes lift generation in porous media, but does not treat the other forces and
$
moments acting on the skier or snowboarder as shown in Fig. 1b. In the present
study, we shall develop a new theoretical approach to treat these limitations and
provide a realistic model for lift mechanics of downhill skiing and snowboarding.
75
2 Methods
Theories to predict the snow perrneability have had Iimited success (Jordan, Hardy,
5(
Perron and Fisk 1999) and the most widely used forrnula is Shimizu's ( 1970) empiri-
cal expression, K=0.077exp(-0.0078pJa'2, where p, is the density of snow and d the
mean diameter of the ice crystals. In the current application shown in Fig. 1a, p,
changes due to the motion of the ski/snowboard, thus, K changes as a function of x:
K = 0 .077cxp[(1-(I-l / k )·(1-x l L )' )tn( K 2 ! 0.077d 2 )]d 2 (!)
(Wu, lgci, Andreopoulos and Weinbaum 2006), where L is the length of the
ski/snowboard; k = h2/h 1, h 2 and h 1 are the local thickness of the snow layer at the
Riding on Air: A New Theory for Lift Mechanics ofDownhill Skiing and Snowboarding 283
leading (x = L) and trailing (x = 0) edges of the planing surface, respectively; and K 2

is the permeability beneath the leading edge, K 2 = 0.077exp[(holh 2)1n(Kof0.077
d2)]d2, where h0 and K 0 are the undeformed thickness and permeability of the snow
layer, respectively.
ER
For a skilsnowboard gliding with velocity U = ( Un U„ UJ over a snow layer as
shown in Fig. la, the trapped air's velocity in the x direction satisfies Darcy's law
(Feng and Weinbaum 2000), u = -(Klµ)(öPlöx) where µ is the viscosity of the air.
The flow in the transverse plane can be viewed, to a first approximation, as a stagna-
tion-point flow in a porous medium (Wu, et al. 2005a), v = -(Klµ)(öPlöy), w=-Az
PT
where A is the vertical velocity gradient, A=-U2 (x, y, h)lh. Combining with the con-
tinuity equation, au1ax+av1ay+aw1az = 0, we obtain
V 2 P = -µAl K -(II K)VP-VK. (2)
Equation 2 is the new simplified governing equation for the pore pressure in skiing
HA
or snowboarding. The two terms on the right hand side have a simple interpretation.
The first term proportional to A is the forcing due to the angle of attack and the lat-
eral edging of the skilsnowboard. The second term arises from the variation in per-
meability due to the compression of the snow layer. If both of them vanish one ob-
tains the potential equation for Darcy flow without any forcing. In the present study,
C
we shall only examine the simple case where there is no lateral tilt, thus, h = h(x), K
= K(x), A = (Ulh)dhldx where V= U 0 and the pressure distribution in the y direction
is parabolic (Feng and Weinbaum 2000; Wu, et al. 2005a). Introducing dimen-
('
sionless variables, p'= (P-P 0)IP0 , p/= (Pc-Po)IP0 , x'= xll, y'= yl(Wl2), h'= hlh 2 and
K'= KIK2, where Pc(x) is the centerline pressure corresponding to the cross section at
the location x, P 0 is the air pressure at the edge of the planing surface which is very
close to the atmospheric pressure, one obtains
d 2 p;(x') +-1_dK'(x') dp;(x') _~ , x' + 0L =0 ( 3)
&7
dx 12 K'(x') dx' dx' E P,( ) K'(x')h'(x') '

where E: = (WIL) 2, 01, =(µI P0)(Ulh 2)(dhldx)(L 2IK2). Equation 3 subject to the bound-
ary conditions, p/(0) = p,.'( 1) =0, can be solved numerically and the 2-D pressure
distribution beneath a snowboard/ski surface is determined, p'( x', y') = (1- y') p/(x').
$
The solid phase (ice crystals) lift force is obtained by measuring the quasi-steady
force generated when the snow is subject to incrementally increasing compressive
forces (Wu, et al. 2005b; Wu 2005). This force in its dimensionless form is given by
75
p, (x') =
1
~oliJ (x) I ~ = ( ~ng IP)· f ([ 1-A + A( 1- 1I k} x'] I ( q> 0 - 0.06)) (4)
(Wu, et al. 2006), where Pwlid (x) is the local solid phase pressure, Pmg =
mgcosahlLW where ah is the angle of the inclined slope,f is the empirical relation
obtained in (Wu, et al. 2005b; Wu 2005), <l>o is the undeformed porosity of the snow
5(
layer, and /4 = hi/h 0 is the compression ratio at the leading edge.

Figure I b shows the representative forces acting on a skier gliding on an inclined
snow slope. The weight mg is resolved into two forces, Fs parallel to the slope, and
FN normal to the slope. The lift force, which refers to the total reaction force of the
snow in the skiing community, is in the present analysis the sum of the distributed
forces due to the pore air pressure, N., and the ice crystals' reaction force, N,. The
skier gliding down the slope has a snow friction force Fr= T/Ns where 17 is the coeffi-
cient of friction, and a wind resistance or aerodynamic drag force F0 , which are
directed up the slope. FL is the aerodynamic lift force which is negligible compared
ER
with N. and N, (Perla and Glenne 1981). In this figure, the forces are shown at the
points at which they act. When a skier has achieved terminal velocity, there is no
acceleration and, thus, there is no inertial force, so all the forces as well as the
torques shown must sum to zero:
mgcosah = Na +NS, Na(xc -x.)+F1( = NJxs -xJ,
r r
(5a, b)
PT
where Na= c:JP-Pa)dxdy, NS= ~o/iix)Wdx' lc is the normal distance of
the CM from the ski surface, Xe, Xa and Xs are the x coordinates of CM, center of N.
and center ofN„ respectively. Equation 5a can also be written asfa;r +!solid =l where
HA
fair= N)mgcosah and holid = Nsfmgcosah.
3. Results and Discussion

C
When a skier/snowboarder (m = 80 kg) glides down a slope (ab= 15°, 17 = 0.04) at
velocity U, over an undeformed snow layer of thickness, h 0 = 10 cm, and permeabil-
ity K0, without changing the location of CM (for skiing, xc'= 0.40; for snowboarding
xc'= 0.45), one has to adjust the compression ratios k = hi/h 1 and ;i = hi/h0, to satisfy
('
the force and moment balance Eqs. 5a and 5b. The lift distribution between the
trapped air and the ice crystals strongly depends on the geometry of the planing
surface, W/L, the velocity U, and the properties of the snow layer, K 0 and q>0 • This is
reflected on the two dimensionless parameters, t: and 0L in Eq. 3. In this paper, we
&7
consider two typical snow types, wind-packed (K0 = 5.0x10- 10 m2, <l>o = 0.6, d = 0.42
mm) and fresh snow (K0 = l.7xl0-8 m2 , <l>o = 0.8, d = 1.0 mm) which bracket the
range of permeability for most skiing conditions (Wu, et al. 2005b; Wu 2005). Be-
cause the permeability of fresh snow is roughly 34 times larger than the wind-packed
snow, air can not be trapped efficiently. When snowboarding at a given speed (U =
$
20 m/s), one needs a much larger compression of the fresh snow layer (h 2/h 0 = 0.38,
hi/h 0=0.31) and a larger contribution from the solid phase ifsoud = 82%) to generate
the required lift compared with the case of wind-packed snow (h 2/h 0 = 0.70,
75
h1/h 0=0.65, !solid= 46%), see Fig. 2a. Figure 2a also shows that for a given snow
type, an increase in velocity leads to an increase in the trapped air's contribution to
the total lift and a decrease in the compression of the snow layer. This is because as
one increases his/her velocity, the contact time of the planing surface with the snow
layer decreases, the trapped air has less time to escape before the pore pressure de-
5(
cays and thus, the required snow compression is smaller. Since t: for a ski (L = 1. 7 m,
W= 0.1 m) is 1/16 that for a snowboard (L =1.16 m, W= 0.27 m), solutions ofEq. 3
for skiing differ greatly from those for snowboarding. In general, due to the large
increase in the pore pressure relaxation at the lateral edges, the required snow com-
pression is larger in skiing than in snowboarding. As shown in Fig. 2b, as one glides
Riding on Air: A New Thcory for Lift Mechanics of Downhill Skiing and Snowboarding 285
over a wind-packed snow layer at U =20 m/s, for skiing k = h2/h 1 = 1.315,J;,;, = 42%,
while for snowboarding, k = 1.072,f,;, = 54%.
(a) (b)
0.024 003
ER
0 02
p'
0.01
PT
0.004 -.....,.·-~1o·mts·.fr:est1·snow- ··
- U=15 m's,fresh soow
_0 001 .=.. J.F20 mls•. Ji:esh.soow,....... . 0.00
- 20 m/s, \tt'ird- ack d soow
0.0 0.2 0.4 ,o 6 08 10 O.O 02 04 0.6 08 10
Fig. 2. Centerline pore pres~ure distribution for (a). snowboardi~g on wind-packed or fresh snow.
(b).snowboarding or skiing on wind-packed snow.
HA
(a) (b)
- x/=0373
-o- xf=0.40
0 05 ....,=, .~,'=OA29
x,=0.452 0.01
~- x;-=o.465
p' . 7.3 p' '
0 03
osoo ~--~-o·.':".b„37~....~
C
--6-· xc'=0.40
-xc"=0.429
0 01 -0.0l -.O.-:-.. .x/=.0-452
- •c"=Ö.465
--0- •c'=Ö.4 73
('
02 o4 x' o s 0.8 , 0
Fig. 3 (a) Centerline pore pressure (b) solid phase lift pressure distribution beneath a snowboard surface
as one glides over a wind-packed snow layer at U = 20 m/s and shift their position ofthe CM. The dashed
lines crossing the pressure profiles show this shift in x,'.
&7
Figure 3 provides the critical insights for snowboard control and stability. A
snowboarder can alter his/her CM by shifting their weight from the front to the rear
foot (x/ decreases). This change is accompanied by a transfer of lift forces from the
air to the solid phase and a change in the angle of attack of the snowboard. The
$
dashed lines in Fig. 3a and b crossing the pressure profiles show this shift in xc'. The
curves in Fig. 3 apply to a neutral stability condition in which the sum of moments
75
about the CM vanishes. If one shifts their weight (changes x/) without changing
their angle of attack (trajectory a-b in Fig. 3), the initial neutral moment balance is
broken and an unbalanced pitching moment is generated. In order to maintain stabil-
ity, one has to input a muscular moment or change the compression ratios of the
snow layer (trajectory b-c in Fig. 3) to get back to a new neutral moment balance
5(
position. The latter requires no muscular input, and is accompanied by a transfer of

lift forces between the trapped air and the solid ice crystals as weil as changes of
snow compression at the leading and trailing edges.
In summary, we have developed a new theoretical analysis of the lift forces gen-
erated during downhill skiing or snowboarding, which incorporates the lift contribu-
tion from both the transiently trapped air and the compressed ice crystals. This study
is an important practical application and extension of the F&W lubrication theory.

The results presented herein agree with the more qualitative predictions in (Wu, et al.
2004; 2005b) where the pore pressures generated in snow were measured for the first
time using a porous-cylinder-piston apparatus. This new theory of lift mechanics of
ER
downhill skiing and snowboarding would be invaluable for future snowboard design.
References
PT
Arons, E. M., and Colbeck, S. C. ( 1995) Geometry of heat and mass transfer in dry
snow: a review oftheory and experiment. Reviews of Geophysics 33, 463-492,
Colbeck, Samuel C., and Warren, G. C. (1991) The thermal response of downhill
skis. Journal of Glaciology 37(126), 228-235.
Colbeck, Samuel C. ( 1994a) A review of the friction of snow skis. Journal of Sports
HA
Sciences 12, 285-295.
Colbeck, Samuel C. Bottom temperatures of skating ski on snow. Medicine and
Science in Sportsand Exercise 26(2): 258-262, 1994b.
Colbeck, Samuel C. (1995) Electrical charging of skis gliding on snow. Medicine
and Science in sports and exercise 27( 1), 136-141.
C
Feng, J., and Weinbaum, S. (2000) Lubrication theory in highly compressible porous
media: the mechanics of skiing, from red cells to humans. J. Fluid Mech. 422,
282-317.
('
Jordan, R. E., Hardy, J. P., Perron, Jr, F. E., and Fisk, D. J. (1999) Air pcrmeability
and capillary rise as measures of the pore structure of snow: an experimental and
theoretical study. Hydro/. Process. 13, 1733-1753.
Lind, D., and Sanders, S. P. ( 1996) The Physics of Skiing -Skiing at The Triple
Point, Woodbury, New York, pp. 1-268.
&7
Perla, R., and Glenne, B. (1981) Skiing In: D. M. Gray and D. H. Male (Eds), Hand-
book of Snow, Pergamon, Toronto, pp. 725.
Shimizu, H. ( 1970) Air permeability of deposited snow. Institute of Low Tempera-
ture Science: Sappora, Japan: Contribution No.! 053, English Translation.
Wu, Q., Andreopoulos, Y., and Weinbaum, S. (2004) From red cells to snowboard-
$
ing: A new concept for a train track. Physical Review letters 93( 19), 194501-
194504.
Wu, Q., Weinbaum, S., and Andreopoulos, Y. (2005a) Stagnation point flow in a
75
porous medium. Chemical Engineering Sciences 60, 123-134.

Wu, Q., Andreopoulos, Y., Xanthos, S., and Weinbaum, S. (2005b) Dynamic com-
pression of highly compressible porous media with application to snow compac-
tion. Journal of Fluid Mechanics 542, 281-304.
Wu, Q. (2005) Lift generation in soft porous media; from red cells to skiing to a new
5(
concept for a train track. Doctoral Dissertation. City University of New York,
New York, NY, May, 2005.
Wu, Q. lgci, Y., Andreopoulos, Y. and Weinbaum, S. (2006) Lift mechanics of
downhill skiing or snowboarding. Medicine and Science in Sports and Exercise,
to appear in June.
Subjective Evaluation of the Performance of Alpine Skis
and Correlations with Mechanical Ski Properties
Peter Federolf1.2, Mirco Auer2, Mathieu Fauve2, Anton Lüthi2, Hansueli Rhyner2
1 Christian-Doppler-Laboratmy Biomechanics in Skiing, Department of Sport Science and

Kinesiology, University ofSalzburg, Austria.peter.federolf@sbg.ac.at
WSL Swiss Federal Institute for Snow and Avalanche Research SLF, Davos, Switzerland.
Abstract. The competition betwccn ski and binding manufacturers is very strong. The pur-
chase decision of customcrs is occasionally bascd on ski testing by thc customer himself, but
often thc purchase decision mainly rclies on published results of commcrcial ski tcsts. In this
study wc analysed evaluation methods and results of five important ski tests for thc winter
season 2004/2005, whose results wcrc published in skiing related media. All of those tests are
vcry extensive, but thcy differ strongly in their evaluation methods, in the skiing skill of their
tcstcrs, and, hencc, in thc results and purchase recommendations they give. From our point of
view it is a shortcoming of all of thesc ski tests that while thcy established a very sophisticated
testing procedure, thcy neither evaluated mechanical properties of the tested skis, nor did they
record the snow conditions during their tests. Whether a ski exhibits a good performance or
not, depends not cxclusively on the properties of the ski, but rather on thc properties of the
whole system athlctc-bindin·g-ski and the intcraction of this system with the type of snow
prescnt during thc test. To analyse these intcrrelations we conductcd a ski test with five testers
and five pairs of skis. Thc ski testers wcrc all expcrienced, sport-orientatcd skiers. The snow
conditions during thc tests were hard snow with a good grip. Subsequent to thc subjective
evaluation of the ski pcrformance in thc field tcsts all ski-binding combinations were tcstcd in
the laboratory for thcir bending and torsional propertics. On the onc hand, the results of our
study underline thc strong differenccs in the subjectivc assessment of the ski performancc. On
thc othcr hand indications for corrclations betwcen bcnding and torsional sti ffncss of the skis
and the grades they achievcd in the subjcctive asscssment for the specific conditions werc
found. We also analyscd that changing extemal conditions affect strongly the outcome of the
subjectivc ratings.
1 Introduction
The development of new skiing equipment is a fast process and the competition
between ski and binding manufacturers is very strong. For most customers it is
nowadays virtually impossible to keep an up-to-date overview of progress and trends
in the skiing industry. Only a few customers have the opportunity to test a broad
range of skis themselves before they buy a new ski. Hence, the purchase decision of
skiers who buy a new ski is strongly influenced by reports and promotions published
in skiing-related media. The various ski tests annually conducted and published by
these media have a very significant impact ofthe purchase decision.
In this study we compared evaluation methods and results of five important "com-
mercial" ski tests published in German-speaking media, which were conducted for
288 Peter Federolf et al.
the winter season 2004/2005. All of these ski tests assess test skis by analyzing the
subjective ratings of individual ski testers. All tests have developed a highly sophis-
ticated test procedure and rely on a !arge number of ski testers and test runs. How-
ever, they differ in the skiing skill of their testers, in the task performed by the test-
ers, in their evaluation criterions, and in the evaluation method. As a result, they vary
substantially in their ratings of the skis and in the purchase recommendations they
give. (for a direct comparison please refer to www.carving-ski.de.)
All of the reviewed ski tests strongly concentrated on the skis, paying little atten-
tion to other factors, which might have an impact on the results. Whether a ski exhib-
its a good performance or not, depends not exclusively on the properties of the ski,
but rather on the properties of the whole system athlete-binding-ski and the interac-
tion of this system with the type of snow present during the test.
Another issue is that skiing equipment comprises not only the ski itself, but also
the binding, often an additional riser plate, and the ski boot. lt is unquestionable that
the ski is the most important component, however, it is obvious that the properties of
the other components also have some impact on the performance of the whole sys-
tem (e.g. Nigg et al. 200 I; Lüthi et al 2006). Hence, a different rating of the same ski
in different ski tests might also be caused by different equipment components.
The purpose of this study was to get a general idea to which extent mechanical
properties affect the performance characteristics of skis. A second aim was to deter-
mine and analyse which factors influence the results of ski tests and might be re-
sponsible for the observed discrepancies.
2 Methods
Table 1 lists the selected ski-binding combinations. The Stöckli skis were chosen
because they were not evaluated in the cited commercial ski tests, which reduced a
potential preoccupation of the ski testers' opinion. The fifth ski, the Atomic SL IIM,
was selected because it had achieved top grades in all commercial ski tests. Unlike
the other skis it was used in a rent ski configuration including binding plate.
test ski Ski Binding

number manufacturer Product length Manufacturer product
name [cm] name
1 Stöckli Spirit 170 Fritschi Diamir
2 Stöckli Spirit 170 Atomic Race
3 Stöckli Spirit 170 Atomic Race
4 Stöckli Laser SL I 71 Atomic Race
5 Atomic SLIIM 160 Atomic rent binding
Tab. 1. Ski cquipment sclcctcd for the test.
Subjective Evaluation of Skisand Correlations with Mechanical Properties 289
For the purpose of this study, the bending and the torsional stiffness of the front
and rear sections of the ski-binding combinations were determined subsequent to the
ski tests. The bending stiffness was characterized by a characteristic bending value,
defined as the load exerted to the ski-binding system divided by the ski 's deflection
(ÖNORM 1977). The ski 's torsion value was calculated by dividing the torque by
the deflection angle (ÖNORM 1977). A detailed description of the measurement
devices can be found in Lüthi et al. (this conference).
Five experienced skiers, which have a similar, sport orientated skiing style, but
differed in body proportions (body masses between 72 and 95 kg) have conducted
the field tests.
The evaluation criterions used in our rating of the test skis were defined in a dis-
cussion with all testers to ensure that the definitions were clear and equally applied.
The definitions contained a detailed description, two extremes between which a ski
has to be classified, and suitable test tums. The used evaluation criterions were easi-
ness of turning, self steering, edge grip, stability, and overa/1 impression.
The ski test was conducted on 24 th March 2005 between 9.00 and 11.20 in the Ja-
kobshom ski area in Davos, Switzerland. For each ski two runs were performed by
the testers, during which they individually performed the test tums. Successively,
they had to rate the ski on a one-to-five scale for each evaluation criterion. Five rep-
resented the optimal, one the worst extreme defined for each test criterion. On the
early hours of the test the snow was dry and very hard, but due to machine grooming
on the evening before the snow still allowed a good grip. During the course of the
tests the snow surface was exposed to the sun and became softer, giving the skis a
slightly better grip. To ensure an individual rating, the testers were not allowed to
talk about the skis during the test, because an optical neutralization of the test skis
was not possible due to the unequal ski-binding combinations. Prior to the test all
skis were base and edge grinded and waxed by a local sports store.
The mean rating given by the five testers is shown in Fig. 1. Error bars indicate
the standard deviation. Not surprisingly, test ski 5, which consisted of matched ski,
binding plate, and binding, was rated best in most of the criterions. lt performed
especially well in the criterion easiness of turning, while it rated less well in the
criterion stability. This can be explained by the fact that ski 5 was 10 cm shorter than
the other skis in this test. Among the results for the other test skis, which all had a
similar ski length, the rating of ski 4 differed significantly from the rating of skis I to
3. Obviously, the ski model had a stronger impact on the result of this ski test than
the binding model. Ski 1, equipped with a different binding mounted on the same ski
model, rated similarly as skis 2 and 3. Skis 2 and 3 were similar in construction, but
did not obtain the same rating. This might be explained by the fact that, due to dif-
ferent earlier usage, they slightly differed in their mechanical properties (Fig. 2 and
3).
□ Ski 1 □ ski 2 □ Ski 3 ■ Ski 4 EI ski 5
easiness of seif steering edge grip stability overall

turning irrpression
Fig. 1. Mean rating of the test skis.
Figures 2 and 3 compare the mean rating of the test skis to the bending and tor-
sion characteristics of the skis. In the specific test conditions (hard snow surface)
stiffer skis tended to perform better than softer skis (see Fig. 2). Higher torsional
stiffness of the ski shovel seemed to affect the ski negatively, whereas the torsional
properties ofthe ski end did not correlate with the rating (see Fig. 3).
5
skl3
! •
4
• • •
Cl
C
+l 3
...
«!
C
....'
•
i
: 2 skl 5
E
f l
■ easiness of turning
1
_. seif sleering
skl2 ♦ edge grip
skl1 skl4 • stability

■ overall i ression
0
4,5 5 5,5 6 6,5 7
bending value [Nimm]
Fig. 2. Comparison ofthe skis' bending values with their mean rating in the test.
Subjective Evaluation of Skisand Correlations with Mechanical Properties 291
„.... •
■
4 4
■ ■
Cl Cl
0
.!: 3 § 3
...
i6 ...
III
easiness of turning
:2
E
C:
■ easiness of turning
• seif steering • seif steering
♦ edge grip 1 ♦ edge grip
o stability • stability
■ overall i ression ■ overall lrrpression
0 +------====.====;::===:!....l 0
1,5 2 2,5 3 3 3,5 4 4,5 5
torslon value shovel [ttn /°] torslon value skl end [ttn /°]
Fig. 3. Comparison ofthe ski shovels' and the ski ends' torsion values with the mean rating.
The high variances in the individual ratings indicate a limitation of the validity of
the results and ask for a closer analysis of equipment-independent factors, which
could possibly affect the testers' rating. One such factor are changing extemal condi-
tions, e.g. changing snow properties. Therefore, a strikt test procedure had been set
up, which ensured, that the ski test could be completed within two hours.
4,5
III 4
32
III
-n,
0
C)
C:
3,5
""~ 3
C:
III easiness of turning seif steering
Q)
E 2,5 edge grip stability
---- CNerall irrpression
2
9:20 AM 9:40AM 10:00 AM 10:25 AM 10:50 AM
approximative time ofthe evaluation
Fig. 4. Overall rating of all testers and all test skis in dependence of time.
An indication for changing extemal conditions affecting the skis' rating is a gen-
eral trend in the mean rating of all testers for all skis. In fact, a clear trend is visible
in the overall rating of the criterion edge grip, which, probably due to the softening
snow surface, rated more than half a grade better at the end of the test (see Fig. 4).
The rating for easiness of turning declined strongly during the first hour, but in-
creased again in the second hour. The rating for stability also showed strong tluctua-
tions. Unfortunately an explanation for the last two effects could not be found.
5 Conclusions
In this study the subjective field performance was correlated to mechanical proper-
ties of different ski-binding systems. In fact, correlations of bending stiffness and
torsional stiffness of the ski shovel to the performance ratings are indicated by the
results of our study. Within the scope of this project the geometrical properties of
the skis, e.g. the side cut, were not considered. The individual ratings of different ski
testers tluctuate strongly. lt could be shown, that they are affected by changing ex-
temal conditions, such as the snow hardness. Hence, further studies with more test
skis and testers are needed to obtain a good overall picture of the interrelation be-
tween ski properties and performance. Based on the results of this study, we would
like to suggest that additionally to the evaluation of the skis in field tests the me-
chanical and geometrical properties of the test skis and the extemal conditions
should also be recorded. Correlations between performance and properties of the skis
offer the opportunity to give more specific recommendations for customers. They
would also provide valuable data for the ski manufacturers to potentially improve
their ski models or adapt them for the specific needs of individual target groups of
skiers.
References
Howe, J. ( 1983) Skiing Mechanics. Poudre, LaPorte, CO, USA.

Howe, J. (2001) The New Skiing Mechanics. Mclntire Publishing, Waterford, ME, USA.
Lind D. and Sanders S. (1996) The Physics ofSkiing. Springer-Verlag New York, NY, USA.
Lüthi, A., Federolf, P., Fauve, M., Rhyner H.U. (2006) Effect of Bindings and Plates on Ski
Mechanical Properties and Carving Performance. This conferencc.
Nigg, 8. M., Schwameder, H., Stcfanyshyn D. and Tschamer v. V. (2001) The Effect ofSki
Binding Position on Performance and Comfort in Skiing. In E. Müller, H. Schwameder,
C. Raschncr, S. Lindinger and E. Komexl (Eds.), Science and Skiing II, Verlag Dr.
Kovac, Hamburg, Gcrmany, pp. 3-13.
Önorm ( 1977) Alpinski, Elastische Eigenschaften, Labormessverfahren. Austrian Standart
Organisation (Österreichisches Normungsinstitut), Vienna, Austria.
Timing of Force Application and Joint Angles
During a Long Ski Turn
Takeshi Yoneyama 1, Nathan Scott2 and Hiroyuki Kagawa 1
1 Kanazawa University, yoneyama@t.kanazawa-u.ac.jp

2 The University ofWestern Australia
Abstract. Using a measuring system which is described in detail in another paper in this
conference, the load on the ski, sole pressure, leg joint motion, and turn direction have been
measurcd during a long turn of an expert skier. The instant of turn change was associated with
a change in the sign of the force moment about the ski direction. The total force on the outside
ski was generally about double that on the inside ski, while both loads instantly decreased at
the turn change. Foot pressure increased at the heel area during the steering process. The
center of the pressure was always kept in the rear part near the heel, but it moved forward at
the turn change. The main motion of the leg was a combination of flexion-extension of the hip
joint, knee joint and ankle joint. The outside leg was kept extended angle during the steering
process, while the inside leg gradually flexed and extended. The trajectory of the body was
estimated from the data of a magnetic compass at the backpack. The forces, foot pressure,
joint motion and body trajectory were compared with the video image of the skier. This com-
parison showed that the skier made the turn change earlier than the centerline of the turn tra-
jectory. During the turn change process, the skier first extended the previous inside leg without
flexing the outside leg. Next, he shifted the main load from the previous outside leg to the
other leg; at this time the force moment also changed. Then he flexed the new inside leg. We
think that the timing of these motions is the main factor determining the downhill speed
achieved.
1 Measurement
The measurements reported here were made using equipment that is described in
another paper in these proceedings (Scott et al 2006). An expert skier who was a test
player for a ski company wore the measuring apparatus as shown in Fig.1. The
equipment consisted of load cells between the binding plate and the ski, foot pressure
pad between the foot and the inside ofthe boot, Measurand ShapeTape™ to measure
leg joint angles, a magnetic compass and mechanical gyrocompass to measure the
backpack angle, and a data Jogger in the backpack. The side curve of the ski was
21 m. The athlete was 185cm tall and weighed about 90kg.
The experiments were done at Shiga-kogen, Nagano, in February 2005. The ski field
was planar with slope 20° to the horizontal. The athlete performed four long turn
294 Takeshi Yoneyama, Nathan Scott and Hiroyuki Kagawa
cycles after an initial

straight descent. Each cycle
took 4 seconds and the
distance travelled was 60m
in the plane of the snow: the Datn Jogger m
skiing speed was about 1h ba lpa k
60km/hr.
Although the forces and
foot sole pressures were
measured on both legs, the
leg angles were only
measured on the left leg.
This report is thus mainly to
do with the left leg.
Fig. 1. Measurement system

2 Forces on the Ski
U'ft 1in R1pt. n.im tdt. .wa Rtpl 11r11

"'""" o,...... ln..S."1 ,,..••
Täe,s
lotl'lhn Rlfhlt11n1 lcflha"ft AapllMIII
1 ,.....,, "1 1-.lt "1 o,• ..., ...
MIii , - - - - , - - - - - ~ - ~ MIii , - , - - - - . , . . . . . . - , - -.......
100 BL/L am
z ,.,., z ,.,.,
i..., i-
... ""' -'----'h----,f---e...""-"--...j/,.j
D
... a:111
·a:111 _ _ _ _............c.-_..u.,......"--1
~ u ~ P w ~ ~ n n n ~ u ~ P w ~ ~ n a n
Tllo,1 T ■ e,s
Fig. 2. Forces at the four points ofthe ski
The load cells detected upward forces at 4 points of the left ski: front right side, front
left side, back right side and back left side. By "upward" we mean normal to the
plane of the ski. The front points were 180mm in front of the boot center and the
back points were 140mm behind it. The transverse distance between the measure-
ment points was 54mm. The forces measured during two turn cycles are shown in
Fig 2. The vertical line marking the turn change instant will be explained below.
Note that during the right turn, the right side forces were !arge while the left side
front force was nearly zero and back force was a little bit negative. In the left turn,
the left side back force was !arger than the front side force while the right side force
was nearly zero.
The total upward force on the left ski is shown in Fig. 3. Comparing the loads for the
inside and outside period, the load on the outside ski was generally about double that
on the inside. However at the instant of the turn change both the insidc and outside
total forces were small.
Left turn Right turn Left turn Right turn
z Inside leg Outside leg Inside leg Outside leg
:si1500~
~1□□□~--
- - - - ---
:65□□ - - - - -- --
c:
0
~ □-----...._ _______________
3 14 15 16 17 lll 19 aJ 21 22 23
Tine,s
Fig.3. Total load applied on the left ski
Force moments around the ski direction (Side moment) and around the boot center-
line (Forward and backward moment) are shown in Fig. 4. The turn change instant
was considered to be the time of the sign change of the moment around the ski direc-
tion, i.e. the change of the running edge of the ski. The moment around the boot
centerline varied rapidly, which we take tobe due to the athlete's effort to keep bal-
ance in the skiing direction.
u~
Left turn Right turn Left turn Right turn
Inside leg Outside leg Inside leg Outside leg
Vl -40
~ ~ ~ ~ lll ~ ~ ~ ~ ~
Left ski
T:ine-s
(a) Force moment around the ski center axis
Tine,s
(b) Force moment around the boot centerlinc
Fig. 4. Force moments around the ski dircction and boot center
3 Foot Pressure
Fig. 5 shows "images" of the foot pressure during two turn cyclcs. In the images a
darker shade shows a higher pressure. The angled lines show the ski direction and
the boot centerline. A circle shows the position of the center of pressure. During the
part of the ski cycle where the left leg was outside, the pressure was mainly in the
heel area so the center of pressure is also towards the rear. At the turn change the
total pressure was quite small and the center of pressure was central. When the left
lieg \ a · n ihe ·insidc. ihc cntcr f prc ur " a again 1
•
16.5s 17s 17.5$ 18$ 185s 19s 19.5s 20s 205s
+-- Right turn : uhidc leg ----+ Lcfi turn: Inside leg ---+
Fig.5. Foot pressurc distribution and

the center of the pressure
4 Leg Joint Angles
The definition of the leg joint angles

is shown in Fig. 6. Leg joint motion \nkl, nc I n f
consists of hip joint flexion, abduc-

tion-adduction, thigh rotation, knee Fig.6. Definition of joint anglcs
joint flexion, lower leg rotation and
ankle flexion. In our experiments we recorded the signal from a ShapeTape.
l"l U 1,,._ U W l't dl il c. t:!l

~ ll ~ n ~ ~ ~ a ~ a
T • ••
Tlt.e•
("1) lr,,;IX'(' Joml 11..: 11W1 f #
l lD
f
~
s
T ••
Ce} 1"'"er k n,1 ttun f ,
1:
• aJ
...
• u
1D
ll D ll n
,
~ U ~ D ~ tt ~ H ll ~ ~ ~ ~ ~ aJ ~
Fig.7. Leg joint angles motion

This signal was post-processed to produce vectors tangential and nonnal to the tape.
The vectors at the tape attachment points on the left leg were used to estimate the
joint angles; the result is shown in Fig.7. The largest angle changes were for hip joint
flexion and knee joint flexion. The data of the flexion of the ankle looks too small to
fit the hip joint and knee joint motion. Some abduction-adduction, thigh rotation and
lower leg motion was also observed.
Comparing the motion of hip joint flexion with the turn change time, it can be seen
that the hip joint is still extended at the turn change. This means that the skier first
extended the previous inside leg while keeping the extension of the outside leg at the
turn change. Next, he shifted the
main load from the previous outside
leg to the other leg and there was a
change in the sign of the force mo-
ment. After that, he flexed the new 16.5s 17s 17.5a 18s llUs 19s 19.5s 20s 20.5s
inside leg. - Right turn --+ - Left turn -
Outside leg Inside leg
The flexion and extension posture
of the left leg, and upper body angle, Fig.8. Leg posture during the turn
are shown using stick figures in
Fig.8. Knee joint angles were assigned so that the upper body inclination looked
reasonable compared with the video image of the skier. The posture at the turn
change was intennediate between the most flexed posture and most extended posture
5 Trajectory
...... ,,.
A trajectory for the upper body was estimated using the
magnetic compass in the backpack and the assumption of
constant 60km/hr speed; see Fig. 9. The positions at the
turn change times are marked with circles. The radius of ,.,.
the first half part of the turn arc was larger than that of
the second half. The turn change point was a little bit
,.,
before the centerline of the turn curve. From this we infer
that the athlete anticipates the turn change as early as
possible for the next first half part turn.
""
6 Comparison Among the Data
Foot pressure, forces at 4 points, leg joint posture and

video image of the skier are compared in Fig. 10. Forces
at four points are expressed as vectors. The timing of
..
these motions must be critical factors in detennining the
overall downhill speed achieved.
We appreciate the cooperation of Ogasaka company. Fig.9 Body trajcctory
~ ~
• •
17
•
•
17 . .
h 1
r4 1
1 .5\
~ ~
~ ~
19
~ ~
19.5,
20s
;
20. a
L d on thc
1
Joint mouon idc-o 1m.1sc of
Pn:,,sure of thc tdl foo1 tdl 1., of thc lefl leg lhc ,l.,cr
Fig. l 0. Comparison among foot pressurc, forces and leg motions
References
Scott N.W., Yoneyama, T.,Kagawa,H. & Takahashi, M., Mcasuremcnt of joint motion and
acting forccs on a top athlcte skiing„ Proc. 5th International Engineering of Sport
Conf.,Davis, California, 13-16 Sept. 2004, Vol. 1 pp 394-400.
Scott N, Yoneyama T & Kagawa H, "A Unificd, Custom-Built Mcasuring System for a Ski
Athletc", Proc. 6th International Conferencc on thc Engineering of Sport, Munich, Aug
2006 (these procccdings).
Effect of Biodings and Plates on Ski Mechanical
Properties and Carving Performance
Anton Lüthi 1, Peter Federotr1-2, Mathieu Fauve 1 and Hansueti Rhyner 1
1 WSL Swiss Federal Institute for Snow and Avalanche Research SLF, Davos, Switzerland,
tuethi@stf.ch
2 Christian-Doppler-Laboratory Biomechanics in Skiing, Department of Sport Science and
Kinesiology, Univcrsity of Salzburg, Austria.
Abstract. The carving performance of modern ski equipmcnt is determined by the overall
mechanical properties of thc whole system composcd of ski, binding and binding plate. In the
current study the effect of different bindings and plates on the mechanical properties was
determined in the laboratory using different devices. Measured were the shape of the ski when
bent, bending and torsional stiffncss, as weil as the force distribution on the running surface of
the ski when prcssed against an array of 26 x 6 strain gauge sensors. The statistical uncertainty
on all quantitics determined is lower than 2 %. The performance in carving was evaluated by
test skiers according to thc criteria: easiness for tuming, stability and edge grip. Bind-
ing/plate/ski-combinations with high ranking in thc field evaluation were found to havc similar
mechanical charactcristics. This correlation between ficld tests and laboratory characterisation
allowed to givc indications on the ideal overall mcchanical propcrties of ski equipment for
carving.
1 lntroduction
Bindings and ptates ptay an increasingty important rote in ski equipment design.
They do not onty significantty affect various ski properties generally discussed as
being important for performance such as bending and torsionat stiffness, as well as
vibrationat and damping characteristics (Schuttes 1978; Lind and Sanders 1996;
Howe 200 l) but also affect how the force is transmitted from the skier to the ski.
This transmission, together with the ski geometry, the overall bending and torsional
properties of the binding/plate/ski system and extemal parameters such as ski load-
ing, edging angle and snow mechanical properties define how the ski is deformed
while skiing, what forces act on it and thus how the ski performs.
The purpose of the current study was twofold: to determine the effect of different
bindings and plates on the ski mechanicat properties and on the performance in carv-
ing, and to get insight on the ideal properties of ski equipment for carving through
the corretation of laboratory and fietd tests. Such information is scarce in the litera-
ture because on the one side, ski manufacturers rarely publish their knowtedge and
300 Lüthi et al.
on the other side, the definition of ideal properties is hindered by the fact that they
depend on such diverse parameters as ski technique, skiing ability and snow condi-
tions.
2 Measurements
The laboratory and field test were carried out for three binding/plate systems (Fig. 1)
which were all mounted on the same ski model, namely a 161 cm long slalom ski of
the manufacturer Stöckli. The first two systems (A, 8) are commercially available,
the third system (C) is a prototype design. One of its particularities is that its stiffness
can be altered. lt can also be fixed to the ski using various screw positions. For the
current investigations it was used in two different configurations, namely a stiff one
with screw positions 1-3 (C 1) and a soft one with screw positions 1-2 (C2).
Fig. 1. Tested binding/plate systems. A: Atomic binding with Hang! plate, 8: Atomic binding
with VIST plate, C 1: Prototype binding/plate stiff, with screw positions 1-3, C2: Prototype
binding/plate soft, with scrcw positions 1-2.
For the laboratory measurements the bending and torsion apparatus shown in
Fig. 2 and the force distribution device shown in Fig. 3. were used. For the bending
tests, the ski was supported 5 cm from the tail and 23 cm from the front and loaded
by a force piston at the center of the binding (located 69. 7 cm from the ski end) via a
phantom boot ISO 9338 Typ A. During Ioading, applied load and ski deflection at
the location of force application were continuously recorded using a strain gauge
force sensor and a distance transducer. From this recording, the bending stiffness
was derived. The distance transducer can also be automatically moved along the ski
axis and was employed to determine the shape of the ski before and after loading.
The difference of the two curves yields the effective ski deformation curve. For the
torsion tests, the ski was rigidly fixed at its center via the ISO phantom boot inserted
into the binding (Fig. 2) and a torsion moment applied 23 cm from the ski tail. This
set up was primarily aimed at determining the torsion stiffness of the binding/plate
link between boot and ski. From the continuous recording of the applied torsion
Effect ofBindings and Plates on Ski Mechanical Properties 301
moment and the torsion angle the torsion stiffness was derived. The force distribu-
tion device consists of an array of 26x6 strain gauge force sensors and a force piston
to press the ski equipment against it. For the current measurements, the readings of
the six force sensors in each array row were summed up to get a two dimensional
force distribution along the ski axis. For all binding/plate systems tested all men-
tioned measurements were repeated at least three times and then averaged, the statis-
tical variation on all quantities determined being lower than 2 %.
The performance in carving of the different plate/binding/ski combinations was
evaluated in free skiing by five experienced skiers having a similar, sport orientated
skiing style. The evaluation criteria were easiness for tuming, stability and edge grip.
During the tests the snow was hard but not icy. Each tester did two runs with a given
combination before noting down his comments and changing to the next one. After
the tests, a 1 to 4 ranking of the four evaluated systems was derived for each tester
from his subjective judgements ofthen the results averaged over all testers.
Fig. 2. Test apparatus in bending (A) and torsion (B) mode.
Fig. 3. Device for measuring the force transmission through the ski to the ground.
The shapes of the bent binding/plate/ski combinations when loaded with 320 N are
shown in Fig. 4. The corresponding bending stiffness values are listed in Table 1,
which also shows the torsion values and the mean ranking in the field tests. Figure 5
shows the force distribution on the running surface when loaded with 400 N. Con-
302 Lüthi et al.
ceming the field tests, all testers feit distinct differences between the tested systems.
All testers ranked system A and B superior to systems CI and C2, with system A
being the favourite one for four out of five testers.
- A AIOm1c bond,ng + H ngl plat 1

0 s AIOmlC bond,ng + v,st p1 t 1
E .,o
.s ._ __ C 1 Prototype sbfl. with screw po 1t,ons 1-3
C2 Prototype soft, woth scn,w po ,t,on 1-2
j 20
..
"C
~
>
-30
-40
"'
~
w -50
-10
0 200 ,<QO l!OO 1100 1000 1200 1,00
D1s1Bnce lrom end of sk• (mm]
Fig. 4. Effective ski deflection of same ski with different bindings and plates.
E
0
,o
- A A1om1c bond,ng + H ngl plate
B Atom1c bond,ng + Vist plale
C 1• Prototype still, woth scn,w pos,toons 1-3
C2 Prototype soft. w,th screw poS<toons 1-2
l
.s
1. 20
~
.. ~
-40
.,.
~
~
w 50
eo
70
0 200 "lO 600 1100 1000 1200 1,00
O, IBnce from end of sk, (mm]
Fig. 5. Force distribution on running surface of same ski with different bindings and plates.
Binding / Plate Bending Stiffness Torsion Stiffness Average ranking in

system [Nimm] [Nm/ 0 ] field tests
Atomic / Hang! (A) 6.89 3.77 1.2
Effcct of Bindings and Plates on Ski Mechanical Properties 303
Atomic / VIST (8) 6.61 3.56 1.8

Prototype sti ff (C 1) 7.21 2.86 3.4
Prototype soft (C2) 5.26 2.64 3.6
Table 1. Binding and torsion stiffness as weil as avcrage ranking in ficld lest of samc ski with
different bindings and plates.
The results show that plate and binding significantly affect the mechanical prop-
erties of ski equipment. The resulting differences can not only be measured in labo-
ratory but also considerably affect the perfonnance in field tests.
When comparing the results of the laboratory and the field tests, it is striking that
the two systems supcrior in the field test, namely A and B, show salient common
characteristics. These are a flattening of the bent ski shape in the ski center, a more
elongated force distribution on the running ski surface with the force maximum
being located in the front part of the ski, and a high torsional rigidity between boot
and ski. In general, systems performed the better, the flatter the bending curve, the
higher the torsional stiffness, the wider the force distribution and, with the exception
of system 8, the higher the overall bending stiffness was. The advantagc of a high
bending stiffness is consistent with the findings of Federolf, Auer, Fauve, Lüthi and
Rhyner 2006. A flat bending curve translates itself to an increased difference be-
tween maximal flexural stiffness at the ski center and minimal stiffness at the ski
ends, a finding consistent with Nachbauer, Rainer and Schindelwig 2005. The advan-
tage of a high torsional rigidity between ski boot and ski seems obvious from the
viewpoint of a finner fixation of the ski to the athlete. A more elongated force distri-
bution on the running surface seems to be advantageous from at least two points of
view: first the skier supporting portion on the running surface is wider, which de-
creases the local loading of the snow and hencc the risk of snow failure. Thereto, it
should be noted that a elongated distribution may be disadvantageous on very hard
snow where a very well centred distribution may be required to build a supporting
track into snow at all. Second, the force on the ski tail and specially on the ski front
is higher which makes the ski morc easily controllable.
The fact that systems A and B show a rather similar bending stiffness when com-
pared to system CI but perform significantly better in the field tests suggests that the
resulting force distribution on the running surface is an as important systcm charac-
teristic as the bending stiffness when regarding carving perfonnance. This distribu-
tion is affected by the system stiffness but probably to a !arger extend by the manner
in which binding and plate transmit the high loading forces in carving to the ski. This
suggestion is supported by Finite Element calculations of the prototype system (C 1,
C2) in which the calculated ski shape in the carving situation was found to be more
affected by the screw positions (1-2 vs. 1-3) than by its stiffness. Another possible
explanation for the performance difference between systems A, 8 and CI would be
the large difference in binding/plate stiffness.
304 Lüthi et al.
4 Conclusion
The current study shows that bindings and plates significantly affect the overall
mechanical properties of ski equipment. The resulting differences can not only be
measured in the laboratory but also considerably affect the performance in carving.
For carving on hard (but not icy) snow, it was found that ski, plate and binding
should be combined in such a way as to obtain a flat bending curve, a wide force
distribution on the running surface and a high torsional rigidity of the binding/plate
link between ski and ski boot. The results further suggest that more attention should
be paid to the force distribution on the running surface as an important parameter
when trying to define optimal system properties for carving. For this reason, a new
measuring device is currently being put in service which allows to determine this
force distribution in the bent ski situation.
References
Federolf P., Auer M., Fauve M., Lüthi A. and Rhyner, H. (2006) Subjective evaluation ofthe
perfonnance of alpine skis and correlations with mechanical ski properties. This confer-
ence.
Howe, J. (2001) The new Skiing Mechanics. Mclntire, Waterford
Lind, D. and Sanders, S.P. (1996) The Physics of Skiing. Springer, New York
Nachbauer W., Rainer F. and Schindelwig K. (2004) Effects of ski stiffness on ski perfonn-
ance. In: M. Hubbard, R.D. Metha and J.M.Pallis (Eds.), The Engineering of Sport 5 Vol.1.
Central Plains Book, Winfield, pp. 472-478.
Schuhes, H. ( 1978) Der Alpinski. Haller & Jenzer, Burgdorf
Development of a Prototype that Measures
the Coefficient of Friction Between Skis and Snow
Paul Miller 1, Andy Hytjan 1, Matthew Weber1, Miles Wheeler1,

Jack Zahle 1, Andy Walshe2, Alan Ashley2
1 Dept. ofMechanical Engineering, University ofColorado at Boulder,

paul.miller@colorado.edu
2 United Statcs Ski Association
Abstract. This paper discusscs the developmcnt a system for mcasuring the kinetic coefficient
of friction between thc bottom surface of a ski and snow. The prototype for taking these fric-
tion mcasurements has been built and tested. This mcasuremcnt system contains load cells,
accelerometers, data acquisition hardware and software, a laptop, skis, and connccting hard-
ware. Thc system developed is portable, accurate and can opcrate at a varicty of skiing condi-
tions and skier wcights. The kinctic coefficient of friction is calculated by measuring the
friction force on the ski and dctermining the normal force acting on the ski. Data was taken
betwcen speeds of 0 to 4 m/s with normal forces of 22 to 82 kgs on an icy surface as weil as
on surfaces with much higher coefficients of friction. The kinetic coefficient of friction be-
tween skis and an icy surface was calculated to be 0.041. The data obtained agreed with
previously publishcd results and was quite repcatable.
1 Introduction
A prototype device that measures the kinetic coefficient of friction between skis and
snow was built and tested to help determine which type of skis, ski waxes and wax-
ing techniques would produce the lowest coefficient of friction on certain snow con-
ditions, including an icy surface. The coefficient of friction between skis and snow
can vary significantly depending on the type of ski and wax that is used for a certain
type of snow. The temperature and humidity of snow has also been shown to affect
the coefficient of friction as weil (Spring 1989). Accurate and repeatable measure-
ments have been taken using the current prototype at speeds up to 4 m/s. The proto-
type was designed such that measurements can be taken on site, on actual ski
courses, as opposed to using devices where snow has to be packed on to a testing
surface (Buhl 2001 and Keinonen 1987). The device takes advantage of commercial
ski bindings and can be easily interchanged with different types of skis. The meas-
urement system is basically composed of off-the-shelf equipment. The device is
portable, easy to setup, and is designed such that a range of ski velocities can be
obtained from 0 to 13 .5 m/s. This paper will primarily focus on the proof of concept
306 Paul Miller et al
of the system to accurately determine the coefficient of friction on icy surfaces. A

picture ofthe measurement system is shown in Fig. 1
T wing
direction
Load eil
containing a
eromete , laptop,
data a qui iti n
Fig. 1 Ski Module with added weight simulating a skier.
2 Experimental Setup
The ski module contains all the components of the data acquisition system. lt con-
sists of a protective case with a set of cut off ski boots attached to the bottom of the
case. These modified ski boots fit into regular commercial ski bindings. Tension in
the two opposing cables were measured by load cells (Omega LC201-100 and
Omega 2511 LCEB-25) that were connected between the respective towing and
restraining cables and the protective case shown in Fig 1.
Inside the black protective case shown in Fig. 1 are the data acquisition compo-
nents including a laptop computer which ran a LabView data acquisition program.
Development of a Prototype that Measures the Coefficient ofFriction 307
Additionally, the case contains a 3 axis accelerometer (Three Analog Devices

ADXLO5 uni-axial accelerometers) that measured accelerations in the towing and
normal directions of movement. Movement of the skis was controlled by manually
pulling the ski module via a tow cable. The protective case is rigidly connected into
the bindings of the skis.
The coefficient of friction was measured/ calculated on two different surfaces to

show proof of concept for the system. The surfaces chosen were such that they rep-
resent the extremes for the expected coefficients of friction. The primary goal was to
prove that the system was capable of repeatable and accurate measurements. The
first set of tests measured the coefficient of friction between the protective case and a
polymeric tile floor, or plastic on plastic. The second set of tests was done to calcu-
late the coefficient of friction between skis and ice, with the weight of the module
being the variable. Each test was conducted for approximately 30 seconds. Tests
were performed by pulling the ski module from speeds of 0 to 4 m/s, while approxi-
mately maintaining a zero degree angle with respect to the horizontal for the two
opposing cables. Typical force data is shown in Fig. 2.
Towing ablc
... Lrainang hie
Fig. 2 Typical force data for towing and restraining cables.
4 Theoretical Analysis
The friction force acting on the skis was calculated by using the output of the load
cells to determine the tension in both the tow cable that pulls the ski module across
the surface and the opposing restraining cable. Fig. 3 shows a free body diagram of
the forces acting on the ski module.
Velocity
a force "'
Restraining Cab e T2 TowCableT
m
Friction Force ,.,._ __

Normal Force
Weight
Fig. 3 Free body diagram of forces acting on ski module.
All measurements were made on horizontal surfaces, therefore the angle with respect
to the horizontal was assumed to be zero degrees. Since measurements were made at
speeds below 4 m/s, the aerodynamic drag force was considered negligible and was
not included in the calculation of the coefficient of friction. While the towing veloc-
ity is relatively constant, the towing acceleration is not zero, and is measured by the
accelerometer oriented in that direction. Tue normal force is just the weight of the
ski module, including the weight of skis and bindings. Tue equation of motion for
the ski module in the towing direction is,
Wa
-=T; -T2 -µW (l)
g
where a= towing acceleration ofski module, W= weight ofski module, T 1=Tension
in tow cable, T2= tension in restraining cable, g= acceleration of gravity, µ= coeffi-
cient ofkinetic friction. Tue normal acceleration is assumed tobe zero. Thus, only
that segment of the data where the normal acceleration is zero is analyzed to deter-
mine the coefficient of friction. Thus the coefficient of friction is Eq 2.
(2)
5 Testing Results
5.1 Plastic on Plastic Friction Testing Results

Plastic on plastic friction testing was done by pulling the protective plastic case
along a horizontal plastic surface. The case was dragged at a constant speed of ap-
proximately 1.5 m/s. After four consecutive tests the coefficient of friction was
calculated to be 0.479. The standard deviation for this coefficient of friction was
calculated to 0.023, which shows the systems ability to make repeatable accurate
Developmcnt of a Prototype that Measurcs the Coefficient of Friction 309
measurements. This coefficient of friction falls within the plastic on plastic values
found on the MatWeb.com online material property database (MatWeb 2005).
5.2 Skis on lce Friction Testing Results

Friction testing was done between skis and a smooth ice surface which was at 17.5°
C. The ski module was dragged at speeds up to 4 rn/s. Weight was added on top of
the ski module to simulate the weights of different skiers. The results for the testing
are shown in Table 1.
Table 1 Results for Testing on Ice
Load Weight (kg) Coefficient of Friction

41 0.0422
61 0.0415
82 0.0418
Average Coefficient of friction
0.0412
between skis and ice
As expected the results adhere to Coulomb-Amonton's first law of friction and

the coefficient of friction did not significantly change by increasing the weight of the
ski module (Persson 1989). Strausky et al had calculated a kinetic coefficient of
friction between skis and ice tobe around 0.03; however, this was at speeds between
0.005 to 0.01 rn/s (Strausky 1998). Spring found that the kinetic coefficient of fric-
tion increases as the speed increases (Spring 1988). Thus, the kinetic coefficient of
friction obtained around .041 is reasonable, since it was taken at speeds up to 4 rn/s.
The coefficient between skis and ice also falls near the coefficient of friction for skis
on snow (Buhl 2001 ). The testing results for all ice tests produced a standard devia-
tion of 0.0019, again showing the good repeatability for this system.
6 Conclusions
The calculation method of the coefficient of friction discussed has been proven to be
accurate and consistent. Testing results for plastic on plastic and skis on ice, com-
pared well to published results. The method has been shown to work at speeds up to
4 rn/s, however higher speeds will be tested in the future. This method of calculating
the coefficient of friction is advantageous over other methods because it can be used
at the actual ski locations as opposed to !ab testing. This ability to test at race loca-
tions can lead to better predictions of what type of ski and ski wax should be used for
a particular race.
7 Future Work
The current prototype will continually be refined to improve measurement accuracy

and precision. Future testing will be done at higher speeds as weil as on inclined
surfaces; therefore calculated coefficient of friction values will be more closely re-
lated to actual racing conditions.
One improvement to the current prototype system is to automate the system to
allow for testing at higher speeds. The current prototype is manually pulled across
the surface while the rear axially opposing cable is also manually pulled on. This
significantly limits the possible testing speeds. Design plans are underway to create
a motorized system that would pull the ski module to higher speeds. A variable
speed motor will be used to allow measurements to be obtained at speeds up to 20
m/s.
The ability to test on inclined surfaces, will allow more flexibility of ski race
locations. An inclinometer would be then incorporated into the ski module.
Acknowledgements
The team would like to thank Puneet Parish and Brad Dunkin of the University of
Colorado lntegrated Teaching and Leaming Laboratory for all their help and advice
with developing our measurement system.
References
Buhl D, Fauve M, Rhyncr H (2001) The kinetic friction ofpolyethylen on snow: thc intluence
of thc snow tcmpcraturc and thc load. Cold Regions Scicncc and Tcchnology. 22, 133-
140.
Keinonen J. ( 1978) An experimental dcvicc for mcasuring friction betwcen ski and snow.
Acta Polytcchnica Sacndinavica. n PH 123, 6-11.
MatWeb, http://www.matwcb.com, 2005
Personn 8, Tosatti E ( 1989) Physics of Sliding Friction. Vol E3 1 1. NATO AS! Scrics, Bos-
ton
Spring E. ( 1988) A method for testing the gliding quality of skis. Tribologia. 7 n 1, 9-14.
Spring E. (1989) Ski Tribology. Acta Polytechnica Sacndinavica, Elcctrical Engineering
Series. N E 164, 108-119.
Strausky H, Krenn J.R., Lcitncr A, Aussennegg F.R. ( 1998) Sliding plastics on icc: fluorcs-
cence spectroscopic studies on intcrfacial watcr laycrs in um thickness regime. Applied
Physics 8 (Lasersand Optics). v866 n5, 599-602.
8 Football
Synopsis of Current Developments: Soccer
Matt Carre
Sports Engineering Research Group, University of Sheffield, m.j.carre@shef.ac.uk
Soccer is arguably the biggest sport in the world, in terms of the impact on everyday
Jives, either through spectating or participation. Indeed, as the conference to which
these proceedings are linked begins, we will all still be recovering from the excite-
ment of the 2006 World Cup final that occurred in Berlin a few days previously. A
sport that has so much impact, unsurprisingly also generates great research interest,
reflected in the number of papers in the following section as weil as in other sections
of these proceedings. In sports engineering this research is fairly weil distributed·
between four key components of the game: the ball, the pitch, the footwear used and
the players themselves.
The soccer ball has been re-designed and improved many times over the years
and current designs employ a combination of careful material selection and sophisti-
cated manufacturing techniques to produce balls which are impressive in terms of
performance, durability and consistency. Ronkainen and Harland describe a tech-
nique using laser Doppler vibrometry to carry out modal analysis of a soccer ball.
This will be of great use in determining the mechanical response of a ball due to
dynamic loading and aid in the further improvement of ball design.
Asai et al. concentrate their efforts on the flight of the ball and combine some
traditional wind tunnel experiments for spinning and non-spinning soccer balls with
a truly innovative method of obtaining flow visualisations for real, kicked balls. The
data obtained and findings observed will undoubtedly be of great use to future stud-
ies of soccer ball aerodynamics.
Two papers in this section are concemed with the performance of the pitch. Al-
catara et al. examine synthetic soccer surfaces, in particular those that incorporate an
in-fill of sand and rubber granules (known in the industry as 'third-generation
pitches'). Their paper compares pitch installations that have a varying size of rubber
granule with various measurement indices, to study the effect of rubber infill mor-
phology on mechanical performance. This information can be utilised in the future
design of these types of surface.
The paper by Jennings-Temple et al. looks at the more traditional surface used in
soccer, natural turf, and they investigate the link between physical soil properties and
playing quality for player-surface interaction. In many countries, soccer is played
during the winter season and the pitch is exposed to harsh weather conditions, which
can have a detrimental effect on the game. The findings from this study can be used
to improve pitch-management and maintenance strategies, improving the quality of
play for players and spectators alike.
314 Matt Carre
A very important topic of study related to soccer is traction of studded footwear,

where there is a need for high player performance, whilst preventing injury. Both
these factors become increasingly important at the professional level of the game.
Grund and Senner describe the design and development of a new device that can be
used to simulate the interaction between a studded boot and a surface, whilst measur-
ing the forces that are produced for different movements. This data can then be fed
into a multi-body computer model of the human knee to calculate the resulting ante-
rior cruciate ligament (ACL) tensile forces. This study, combined with the planned
future work, has the potential to gain a real understanding of the cause of ACL inju-
ries, with respect to boot-surface interaction, which would be of huge benefit to the
health of soccer players.
The three remaining papers are related to how players perform under different
conditions. Work carried out by Marcolin et al. is also concemed with boot-surface
interaction, but in this case the focus is on how the placement of the support foot
affects shooting accuracy for different types of instep kick. Findings from this study
will be of great interest to coaches and is directly applicable to a training environ-
ment.
An intriguing study by Hiramatsu and Ohshima analyses how the position of a
gaze point (e.g. knee, waist or ankle) is used by defensive players to predict the ensu-
ing movement of attacking players, even in situations where the attacking player is
trying to trick the defender by use of 'feint' moves. The accuracy of this prediction
for different situations, as analysed by a brain-mimicking function could provide
some incredibly useful advice for players and coaches to improve defensive ability.
To address the balance, Kerwin and Bray's study could ultimately improve the
attacking nature of a side through more successful penalty kicks. Their study first
involves a simulated penalty shoot-out between a ball firing machine and real goal
keepers, whose movements are monitored using a force platform and video analysis.
This information is fed into a computer model which can predict the likelihood of
certain shots being successful, with reference to a 'diving envelope' within which the
keeper is able to make a save. At the point of writing this, I am hoping that Kerwin
and Bray have passed their findings onto the England Football team, who, on past
experience, often rely on penalty shoot-outs for the success (or otherwise) of their
international matches!
Soccer research is clearly moving forward in many areas, leading to greater
safety and enj oyment for players, easier monitoring of the rules by FIFA, and better
entertainment for spectators.
An Investigation into the Link Between
Soil Physical Conditions and the Playing Quality
of Winter Sports Pitch Rootzones
Marke Jennings-Temple, Peter Leeds-Harrison, Iain James

Cranfield University at Silsoe, m.a.jennings-temple.s0 l@cranfield.ac. uk
Abstract. Playing quality standard devclopment failed to demonstrate how set targets could be
achieved, resulting in Groundsmen being unable to manage pitches to optimize playing qual-
ity. This research linked easy-to-measure pitch parameters to the outcome of tests for player-
surface interaction quality, to enable this to be monitored in real-time and ensure appropriate
pitch management options are selected. 25 pitches were tested three times over an 18-month
period. Tests for surface traction and hardness were conducted, along with a range of soil and
grass factors; multiple linear regression was used to generate prediction models. R2 values
varied with soil type and weather conditions, although increased sand content generally re-
duced the reliability of the prediction equations. lt was concluded that top-dressing may have
skewed the data; suggesting morc sand-based pitches than actually existed, or that sand-
dominated rootzoncs varied littlc in playing quality. The production of significant regression
equations has dcmonstrated which easily-influenccd pitch factors can be manipulated to alter
playcr-surface interaction quality and ultimately, lower the risk of injury.
1 Introduction
Shildrick and Dye ( 1983) identified the need for a code of practice to assist
Groundsmen to make management decisions which would aid the provision of high
quality sports pitches. Playing quality standards were first proposed by Bell, Baker
and Canaway ( 1985) and the final report from the Natural Turf Pitch Prototypes
Advisory Panel was published in 1992 (Adams, Gibbs, Baker and Lance 1992). The
research, sponsored by the Sports Council and performed by the Sports Turf Re-
search Institute (STRI) failed to produce a code of practice, or management plan.
The Performance Quality Standards (PQS) document published by the Institute of
Groundsmanship (IOG; IOG 2001) contains targets for particular tests for pitch qual-
ity. Absent are guidelines on how to achieve these numbers.
The Jack of coherent guidelines to achieve the recommended targets for pitch
quality has contributed to football (soccer) being identified as 1000 times more dan-
gerous than high-risk industrial occupations (Hawkins and Fuller 1999). A recent
report from the Department for Culture, Media and Sport (DCMS 2002) detailed the
aims of the British Govemment to use increased sport participation to combat in-
316 Marke Jennings-Temple
creasing levels of obesity and associated illnesses. In the document, it is explicitly

stated that increased participation rates must not be mirrored by increased injury
rates.
The aim of this research was to establish a link between easily measured pitch
parameters with the outcome of the two tests for player-surface interaction quality.
This would enable Groundsmen to monitor pitches in real-time and eventually facili-
tate the selection of optimum pitch management tools for the production of a safe
surface.
Twenty five pitches at fifteen football clubs in England and Scotland were visited
three times during January and May 2004 and April 2005. The range of clubs in-
cluded those in the Premiership and Scottish Premiership, lower league clubs and
private schools. Each had a dedicated Groundsman.
2.1 Soil Physical Tests
Tests determined grass cover (%) using a point quadrat, Grass length (mm) using a
rising disk, surface evenness (mm) measured using a graded wedge and a 2m straight
edge and volumetric moisture content (%) was determined in the laboratory from soil
cores removed from the pitch. Bulk density (Mg/m3) was determined for the 0-50mm
and 50-l00mm core sections. Soil water tension (matric potential; kPa) was deter-
mined using a bespoke tensiometer. Each of these tests was conducted at 5 locations
on each pitch. A füll particle size analysis was performed on the combined soil from
each of the five cores.
2.2 Pitch Quality Tests
Player-surface interaction quality was assessed using two tests; surface hardness and
traction. The hardness test used a Clegg hammer, first developed by Clegg (1976) for
evaluating pavement base course layers. A 0.5kg hammer was dropped 0.55m onto
the surface. A single drop method was used and the reading in g was recorded. Three
tests were performed at each location.
Traction test equipment was manufactured and operated in accordance with the
guidelines presented by Canaway and Bell (1986) and BS 7044 (BSi 1990). A metal
disk encompassing six 15mm studs equally spaced at 46mm radii from the centre
was loaded with 41kg of steel weights. With the studs fully engaged into the surface,
a torque wrench, attached to the disk via a meta! shaft was tumed and the maximum
force at failure was recorded in Nm. Total equipment weight was 45.5 kg. This ·test
was performed once at each location.
An Investigation into the Link Betwcen Soil Physical Conditions and thc Playing Quality 317
2.3 Pitch test locations
Test locations have varied in previous studies, in general ranging from twelve
(Holmes and Bell 1986) to six (Holmes and Bell 1987; Baker and Gibbs 1989). The
five test locations displayed in Fig. 1 included high and low wear areas.
QJ
Fig. 1. Pitch test locations. This figure was present on data collection sheets and was
annotated to ensure each visit used the same test locations.
2.4 Statistical analysis
Using the statistical package Statistica™ (Statsoft lnc., USA), backward stepwise
regression eliminated non-significant variables and generated equations to enable the
prediction of playing surface quality from a combination of soil and grass factors; all
ofwhich can be managed or influenced by the Groundsman.
The factors entered into the regression analysis were: Grass length, grass cover,
surface evenness, volumetric moisture content, 0-50mm bulk density, 50-100mm
bulk density, Log clay content, Log silt content, medium sand content and Log fine
sand content.
This study avoided using the results of one quality test to predict another and
only a regression statistic (R 2) >0.5 was deemed acceptable.
Non- normality in the data was treated by Log transformation. This was necessary
for the Clegg hammer readings and the soil constituents: clay, silt, fine sand and
coarse sand.
3.1 Soil types
Six soil types were identified: clay (n=30), clay loam (n=55), loamy sand (n=35),
sand (n=205), sandy loam (n= 15) and sandy silt loam (n= 15). Tabs. 1-2 demonstrate
the regression models for two contrasting soil types; sand and clay loam.
318 Marke Jcnnings-Temple
3.2 Regression models
Regression models for sand and clay loam pitches under wet and dry conditions are
shown in Tabs. 1-2. 'Wet' indicates rain or irrigation within 2 days ofthe tests.
St. error R
Sand Wet Traction = 0.31 (grass cover) - 0.29 5.57 0.55 p<0.01
(grass length) - 9.43 (50-l00mm bulk
density) + 37.80
Dry Traction = 0.23 (grass cover) - 0.22 5.64 0.38 p<0.01
(moisture content) + 1.98 (Log silt
content) + 28. 79
Clay Wet Traction = 0.17 (grass cover) - 0.54 4.56 0.89 p<0.01
Loam (grass length) + 0.86 (evenness) +
31.66
Dry Traction = -0.94 (moisture content) + 7.19 0.68 p<0.01
74.68
Table 1. Traction regression cquations for sand and clay loam soil types
St. error R-
Sand Wet Hardness = 0.002 (grass cover) + 0.56 0.21 0.13 p<0.01
(0-50mm bulk density) + 4.34
Dry Hardness = -0.01 (moisture content) + 0.15 0.39 p<0.01
0.23 (50-100mm bulk density) + 4.95
Clay Wet Hardness = -0.01 (grass length) + 0.04 0.13 0.76 p<0.01
Loam (evenness) + 0.71 (0-50mm bulk
density) -0.03 (medium sand) - 0.27
(Log fine sand) - 0.98 (Log silt) +
8.64
Dry Hardness = 0.004 (grass length) - 0.16 0.91 p<0.01
0.04 (moisture content) - 0.01 (me-
dium sand) + 6.58
Table 2. Hardness (Log g) regrcssion cquations for sand and clay loam soil typcs
Unsurprisingly increased grass cover increased traction on wet and dry sand
while increased bulk density increased hardness on wet sand. This relationship was
particularly weak however. On wet clay loam soil, grass length reduced traction
possibly through reduced friction between the surface and the meta! plate, while
longer grass may have cushioned the blow of the hammer on the surface. Impor-
tantly, in each of the regression equations are soil and / or moisture content factors
demonstrating that player-surface interaction quality could be manipulated through
comprehensive soil management. The weak R2 values for sand were likely to be the
result of limited variability within the sand data results. Grass cover was a predictor
An Investigation into the Link Bctwecn Soil Physical Conditions and the Playing Quality 319
variable but this may be a poor reflection of the below ground biomass present.
Roots add strength to the soil matrix, but even after grass cover reduced, van Wijk
( 1980) recorded sustained soil strength because of roots retained in the soil.
Furthermore, under conditions of drying, matric forces within the soil may add to
strength but this would be less successful in coarse-grained material. The depth and
spacing of the drains and time since rainfall or irrigation event may be more critical
in coarse-grained (sand-based) rootzones when measuring strength.
The issue of top-dressing was also highlighted; the underlying soil type may be dif-
ferent to that recorded in the upper 50mm. Deeper sampling and particle size analy-
sis using soil from a range of depths may have shown a different soil type. The result
ofthis would be to include pitches that were not sand in the 'sand' category, limiting
the production of reliable prediction equations.
3.3 Predicted v Actual

Sand, wet conditilro ClayLoom, ,,.., conditions
:§
:::so
:::'.00
. 200~-----~
.
150
lll
j 150 ♦
100
.,,
jl
100
50
'B~ 50
p..
0 50 100 1Sl 100 150 50 100 150 200
Qb.,,rved Hanlness (g)
Fig. 2. Predicted v obscrved hardncss (untransfonned data) for sand and clay loam pitchcs
undcr wet lest conditions, using rcgrcssion cquations from Tabs. 1-2.
Figure 2 demonstrates that for the sand based rootzones, predicted values were
overestimated, limiting the reliability of the regression model. Although for a smaller
data set, the proximity of the data points to the I: 1 line demonstrates that hardness
values can be reliably predicted on clay Ioam soils.
4 Conclusions
A link between tests for player-surface interaction quality and components of the
pitch that can be influenced by a Groundsman has emerged. Further research should
undertake more frequent testing, on pitches of known soil type to verify the regres-
sion models discovered.
Commercial applicability has been suggested. Bespoke software would enable a
Groundsman to monitor player-surface interaction quality in real-time and periodic
soil testing, would ensure model parameters are up to date.
320 Marke Jennings-Temple
Poor regression results in some instances may have been the result of top-dressed
pitches skewing particle size classification towards sand, or simply that sand-based
pitches are less variable and regression equations could not be established.
Finally, player-surface interaction tests are only one aspect of playing quality but are
directly relevant to player safety. The facilitation of pitch management to maximize
player-surface interaction quality should help minimize injury rates and associated
costs.
5 Acknowledgements
This research was sponsored by Turffrax, Oakley, Bedfordshire, UK and the Engi-
neering and Physical Sciences Research Council (EPSRC) under the Engineering
Doctorate (EngD) program.
References
Adams, W.A., Gibbs, R.J., Baker, S., & Lance, D. ( 1992) Making the most of natural turf
pitches. A national survey of winter games pitches with high quality drainage designs.
Natural Turf Pitch Prototypes Advisory Panel Report No. 10, Sports Council.
Baker, S.W. & Gibbs, R.J. (1989) Levels of use and the playing quality of winter games
pitches of different construction types: case studies at Nottingham and Warrington. Jour-
nal of the Sports Turf Research Institute, 65, 9-33.
Bell, M.J., Baker, S.W., & Canaway, P.M. (1985) Playing quality of sports surfaces: a review.
Journal of the Sports Turf Research Institute, 61, 26-44.
BSi ( 1990) Artificial Sports Surfaces. 7044-2.2: 1990, London: British Standards Institute.
Canaway, P.M. & Bell, M.J. ( 1986) Technical note: an apparatus for measuring traction and
friction on natural and artificial playing surfaces. Journal of the Sports Turf Research In-
stitute. 62, 211-214.
Clegg, B. ( 1976) An impact testing device for in situ base course evaluation. Proceedings of
the Australian Road Research Bureau, 8, 1-6.
DCMS/Strategy Unit (2002) Game plan: a strategy for delivering Govemments sport and
physical activity objectives. London: British Govemment.
Hawkins, R.D. & Fuller, C. W. ( 1999) A prospective epidemiological study of injuries in four
English profcssional football clubs. British Journal of Sports Medicine, 33, 196-203.
Holmes, G. & Bell, M.J. ( 1986) A pilot study ofthe playing quality of football pitches. Jour-
nal of the Sports Turf Research Institute, 62, 74-90.
Holmes, G. & Bell, M.K. ( 1987) Standards ofplaying quality for natural turf Project Re-
port, Sports Council/STRI.
IOG (2001) Guidelines for perforrnance quality standards part one; sports surfaces natural and
non turf. Milton Keynes, UK: The Institute ofGroundsmanship.
Shildrick, J.P. & Dye, A.L. (1983) A review ofplaying surfaces research. London: National
Turfgrass Council.
van Wijk, A.L.M. ( 1980) Playing conditions ofgrass sportsfields. Wageningen: Centre for
Agricultural Publishing and Documentation.
Measuring and Modelling the Goalkeeper's Diving
Envelope in a Penalty Kick
David G Kerwin 1 and Ken Bray2
1 UWIC, CardiffSchool ofSport, Wales, dkerwin@uwic.ac.uk

2 University ofBath, Sport and Exercise Science, k.bray@bath.ac.uk
Abstract. The penalty kick is a key set play in competitive soccer which often determines the
outcome of a match, and when used in its 'shoot-out' format, resolves tied games during the
knock-out stages of international toumaments. 85% of penalty kicks in open play, and 75% in
penalty 'shoot-outs', result in a goal being scored. The purpose of this study was to establish
the area of a soccer goal - the diving envelope - which could not be defended by a goalkeeper
when facing a penalty kick. Video analysis of the 2004 European championship penalties
revealed that the time between ball strike and arrival at the goal line ranged from 500-700 ms.
Officially the goalkeeper cannot move towards the penalty taker until the ball has been struck
and the area that he has to defend is 7.32 m x 2.44 m (24 ft x 8 ft) with the ball having to travel
from a fixed spot, 10.97 m (12 yd) from the centre ofthe goal line. An electrically driven ball
launcher (BOLA, Bristol, UK) was used to deliver two sets of 20 ball launches at two experi-
enced goalkeepers in a simulated penalty shoot-out. The goalkeepers took up position on a
force plate located at the centre of a soccer goal. All trials were simultaneously recorded using
three digital video cameras within an indoor sports facility which had previously been cali-
brated to facilitate 3D DL T reconstruction of the video data. The launch velocity remained
constant at 21 m·s· 1 (47 mph) whilst the direction and height of each trajectory was varied to
include left and right shots directed to low, mid and high levels. A goalkeeper model was
developed based on jumping ability, derived from ground reaction force data, and anthro-
pometry. Based on the ball flight time and the goalkeeper model a region of the goal was
identified comprising 28% of the area which cannot be protected by a goalkeeper, meaning
that players could guarantee scoring if they shoot into this 'unsaveable zone'. Observations
from the two penalty shoot-outs at the 2004 European Championships revealed that many
players do not take advantage of this zone and continue to gamble on the goalkeeper guessing
incorrectly when taking a penalty.
1 lntroduction
Penalty kicks in elite level soccer toumaments including W orld Cups from 1982-
1998 and the European Nations Cup (Euro) 1996 included 81 penalties in open play
of which 85% were successfully converted and 176 in penalty shoot-outs of which
75% were goals (McGarry and Franks 2000). The same authors reported that at least
one of the finalists in each of these toumaments had won at least one match, includ-
ing the 1994 World Cup Final, via a penalty shoot-out. Rule changes now allow a
goalkeeper to move sideways prior to the ball being struck. As a result it is possible
for a goalkeeper to move in anticipation of the kick being taken and in so doing,
decrease the area of the goal to one side that he needs to defend. Obviously if he
322 David Kerwin and Ken Bray
anticipates incorrectly, the unprotected area to the opposite side increases. The ball
flight time for a typical penalty is around 600 ms (McMorris, Copeman, Corcoran,
Saunders and Potter 1993) compared to the goalkeepers' diving motion which takes
from 500-700 ms (McGarry and Franks 2003). In a previous study of goalkeepers'
diving motion, two semi-professional and two college goalkeepers were required to
<live towards a football suspended ±2.5 m from a centre line at varying heights (Su-
zuki, Togari, Isokawa, Ohashi and Ohgushi 1988). The two semi-professionals re-
corded takeoff velocities close to 5 m·s- 1, with the horizontal component being ap-
proximately four times that of the vertical component. These results do not reflect a
normal penalty environment, where the goalkeeper is required to judge the flight of
the ball and anticipate its arrival location and to adjust the width as weil as the height
ofhis <live.
The purpose of this study was to measure maximum jumps and dives in a simu-
lated penalty shoot-out to provide data to develop a model of a goalkeeper's maxi-
mum diving envelope.
2 Methods
2.1 Data Collection

Two experienced goalkeepers (21 yrs, 95 kg, 1.83m; 20 yrs, 76 kg, 1.84 m) gave
written informed consent. A soccer goal was located in a !arge indoor sports hall
with an electrically driven ball launcher (BOLA, Bristol, UK) located 10.97 m (12
yards) from the goal line. Safety mattresses (crash mats) were placed around the goal
mouth to protect the goalkeepers when landing. Three digital video cameras (DCR-
TRV-900E, Sony, Japan) were sct at distances of 11 m to 25 m from the centre ofthe
goal. One was located at the side to record the whole of the flight. A second was
located at a height of 3.5m and angled to include the goal mouth and the majority of
the ball flight. The third camera, directly behind and above the ball launcher, viewed
the goal mouth (Fig. 1). All cameras operated at 50 Hz with the shutter speed of
1/1000 s. A common trigger initiated three banks of 20 LEDs which illuminated in
sequence at I ms intervals enabling the shutter opening times to be synchronized to
1 ms (see Kerwin and Bray 2004 for details). A 3D volume containing the goal
mouth and the penalty spot was calibrated using a single pole with four 0.10 m
spherical markers at 1 m vertical intervals. Video images were recorded whilst the
pole was moved in sequence around 18 locations (Fig. 1).
A simulated penalty shoot-out was conducted with cach goalkeeper. The launch
speed was maintained at 21 m·s- 1 (47 mph). Ten 'shots' were made to each goal-
keeper's right with the launch angle being altered to produce a range of low, mid and
high shots. The sequence was repeated to the goalkeepers' left. Video images from
the camera behind the ball launcher were made of the goalkeepers performing a
series of four maximal effort counter movement jumps (cmj) from a force plate (Ki-
stler 9287BA, Switzerland) located at the centre of the goal. Force data ( 1000 Hz),
Measuring and Modeling the Goalkeeper's Diving Envelope in a Penalty Kick 323
--
I
-1
,
Fig. 1. Equipment layout showing the calibration pole and camera locations, goal, crash mats
and trigger unit linked to LEDs and force plate.
triggered with the same LED system, were numerically integrated to determine take-
off kinematics.
2.2 Data Analysis

All digitizing was completed using the Target high resolution system (Kerwin 1995).
Following six repeats of the calibration images, sequences of the ball flight were
digitized from two views and reconstructed using 3D DLT techniques to determine
the ball's trajectory. Images of each goalkeeper's dives were digitized from the in-
stant that the ball left the launcher until his arm contacted the crash mat. The left and
right toes and fingers, an estimated mass centre (-navel) and head centre were digi-
tized. The planar motion of the goalkeepers was reconstructed using 2D DLT tech-
niques (Kwon 1999). Images ofthe vertical jumps were used to determine maximum
stretch (toes to fingers) and estimated arm span for each ofthe goalkeepers.
Digitizing ofthe ball's trajectory commenced when the ball first appeared in both
views. This instant varied from trial to trial depending on the launch angle. Extrapo-
lated data for the unseen part of the ball flight were generated using a linear fit for
the motion towards the goal (y). The missing lateral (x) and vertical (z) data were
generated using Burg's prediction method (Mathcad 13™, Adeptscientific). Trials
resulted in a 'goal', a 'save' or a 'miss' and therefore the ball did not always cross
the goal line between the posts. To determine the theoretical flight time for each
launch, the linear fit to the 'y' data from the ball's first appearance until it was 2.5 m
from the goal line was used to generate the whole flight from y = 10.97 m (12 yards)
to y = 0 m (goal line). The actual time at which the ball crossed the line (or until the
minimum value ofy in a 'save' or a fingertip contact with the ball) provided an esti-
mate of the time available to the goalkeeper to make a save.
For each goalkeeper's <live, the nearest finger location to the ball at the time at
which the ball crossed the goal line was determined. In addition, the closest a finger
came at any time in the goalkeeper's <live to the position where the ball crossed the
line was determined. For those trials where the goalkeeper made contact with the
ball, the minimum value was at ball contact. Planar plots of the paths traced out by
the goalkeeper's limbs (Fig. 2) during the <live represent the goal area which could
be protected.
A model of the goalkeeper's maximum diving range was developed from the
force plate and video images of the goalkeepers' jumps and the penalty dives. These
values, together with the fingertip to ball distances, radii and angles, and mass centre
position and takeoff velocity in all dives, provided data to inform the model used to
the determine the goalkeeper's maximum diving envelope.
Of the 40 trials recorded, 32 were suitable for analysis. One was off target, eight
were saved by the goalkeeper, ·18 resulted in a goal. In the remaining trials, the ball
struck either the bar or post. Thus a similar percentage of 'shots' in the trials, as have
been reported in toumament soccer, resulted in a save (8/32). A plot of all 'shots'
was expanded by assuming left/right symmetry so that 64 ball positions were gener-
ated (Fig. 3). The time for the ball to travel from the penalty spot to the goal line was
0.642 ± 0.014 s. This represents the time that the goalkeeper has to make a save for a
penalty kick taken at 21 m·s- 1• When a 'save' was made, contact with the ball oc-
curred approximately 0.75 min front ofthe goal line and-600 ms from launch.
The maximum vertical takeoff velocity achieved (cmj) was 2.9 m-s- 1 and the
maximum reach 2.84 m. The highest takeoff velocity (4.13 m·s- 1) was found in the
penalty dives when there was a large horizontal velocity component.
The images in Fig. 2 show four dives. When the ball was close to the goalkeeper
(±2 m) from the centre ofthe goal (Fig. 2, upper images), the goalkeeper dived from
his initial standing position with minimal prior foot movements. When the ball was
aimed just inside the goal post or to a high central position, the goalkeeper stepped
sideways by about 1.1 m before commencing his <live (Fig. 2, lower images). The
time to make the side step constituted 66% of the total takeoff time, and so whilst
increasing his total diving range, his diving 'reach' prior to the ball's arrival was
restricted. By combining the data from all trials it was possible to map out the total
area of the goal which the goalkeeper was able to defend.
Based on the time of ball flight, the takeoff position and velocity, and the maxi-
mum reach of each goalkeeper, a model of the diving envelope which the goalkeeper
could describe was developed. Initially the vertical takeoff velocity and reach (cmj)
were used to predict the area that the goalkeeper could cover in a <live. The peak of
this area was higher than the goal (2.44 m) in the centre, but narrower than the goal
(7.32 m), and covered 65% of the goal area (Fig. 3 grey arch). If allowance was
made for the goalkeeper stepping to the side before diving, and the time needed was
ignored, he could cover almost all ofthe goal area (Fig. 3 dash and dot-dash arches).
When the goalkeeper moved sideways before diving, his resultant velocity increased
Measuring and Modeling the Goalkeeper's Diving Envelope in a Penalty Kick 325
"
es
Fig. 2. Images ofthe goalkeeper's movement in trials (B1,4,5,8). Black lines indicate the
position of the goalkeeper as the ball crossed the goal line, grey lines track the complete <live
with the locus ofthe closest finger tip traced throughout. The ball is plotted at the time when it
crossed the goal line. Image dimensions = ½ the goalmouth (ie 3.66 m x 2.44 m).
but, in the time of ball flight to the goal line, the maximum area of the goal that he
would be able to cover was 76% (Fig. 3 black arch). When the calculations were
repeated with the goalkeeper reaching forwards by 0.75 m as seen in the 'saved'
trials, the goal area and the reach radius were reduced, producing a revised area
which could be covered of 72%.
·'
.
•
••• • ••
• ••
••••
'•• •••
•• ..
Fig. 3. Left image = zones of the goal which the model predict can be covered. Grey rectangle:
goal area; dotted grey rectangle: 0.75 m goal area; black diamonds: limit ofthe goalkeeper's
<live; grey triangles: cmj reach; dash and dot-dash arches: cmj arch ±1 m step prior to takeoff.
Right image = the recorded ball locations in the pseudo penalty shoot-out (grey circles =
actual shots and white circles = reflections about a central vertical line, ie symmetry assumed).
4 Conclusions
By modeling a goalkeeper's <live, it is possible to predict the area of the goal that he
can defend in a penalty. There remains a region, around 28% of the goal, which
cannot be covered. This zone comprises the corners of the goal. A tall goalkeeper
may be able to move towards one post and save a wide-low shot aimed just inside
the post. This was seen in the European Cup Final 2005 penalty shoot-out in Turkey.
Even allowing for a goalkeeper's movements prior to ball strike, the two areas close
to the top corners of the goal remain exposed.
Data from a simulated penalty shoot-out resulted in a 'score to save' ratio similar
to that previously reported for international penalty shoot-outs. The data confirmed
that there is an area ofthe goal that is unreachable in the time available.
Based on the ball flight time and the goalkeeper model, a region of the goal was
identified comprising approximately 28% of the area which could not be protected,
meaning that players could guarantee scoring if they shoot into this 'unsaveable
zone'.
Acknowledgements
The authors gratefully acknowledge Toru Nakagami's assistance in the field trails
andin the data collection and data reduction phases of this study.
References
Kerwin, D. G. ( 1995) Apex/Target high resolution video digitising system. In Proceedings of
the Biomechanics Section of the British Association of Sport and Exercise Science (edited
by J. Watkins), pp. 1-4. Glasgow: BASES.
Kerwin, D.G. and Bray, K. (2004) Quantifying the trajectory ofthe long soccer throw-in.
Engineering in Sport 5. University ofDavis California. (Eds. M. Hubbard, R.D.Metha and
J.M. Pallis). International Sport Engineering Association publication, 63-69. ISBN 0-
9547861-0-6.
Kwon, Y.H. (1999) 2D Object plane deformation due to refraction in !wo-dimensional under-
water motion analysis. Journal of Applied Biomechanics. 15, 396-403.
McGarry, T. and Franks, I. M. (2000) On winning the penalty shoot-out in soccer. Journal of
Sports Science, 18, 401-409.
McGarry, T. and Franks, I.M. (2003) The scicnce ofmatch analysis. In Science and Soccer
(edited by Reilly, T. and Williams, A. M.), pp. 265-275. London: Routledge.
McMorris, T., Copeman, R., Corcoran, D., Saunders, G. and Potter, S. (1993) Anticipation of
soccer goalkeepers facing penalty kicks. In Science and Football II (edited by T. Reilly, J.
Clarys and A. Stibbe), pp. 250-253. London: E & FN Spon.
Suzuki, S., Togari, H., Isokawa, M., Ohashi, J. and Ohgushi, T. (1988) Analysis ofthe goal-
kcepcr's diving motion. In Science and Football (edited by T. Reilly, A. Lees, K. Davids
and W.J. Murphy), pp. 468-475. London: E & FN Spon.
Flow Visualization on a Real Flight Non-spinning
and Spinning Soccer Ball
Takeshi Asai 1, Kazuya Seo2, Osamu Kobayashi3 and Reiko Sakashita4
1 Inst.of Health and Sports Science, Tsukuba Univ., Tsukuba, Japan,

asai@kescriv.kj.yamagata-u.ac.jp
2 Fac. ofEduc. Art and Science, Yamagata Univ., Yamagata, Japan
3 Department of Aeronautics and Astronautics, Tokai Univ., Hiratsuka, Japan
4 Fac. ofEduc., Kumamoto Univ., Kumamoto, Japan
Abstract. Wind tunnel tcst with a füll size non-spinning soccer ball was carried out to meas-
ure the aerodynamic forces (drag, lift and side forcc). The drag coefficient of a non-spinning
soccer ball in the subcritical regime was nearly 0.43 and that of in the supercritical regime was
ncarly 0. 15. lt is estimatcd that the critical Reynolds numbcr of a soccer ball is about 2.2x 10 5•
In the visualization experiment using titanium tetrachloride, we compared the flow around the
ball during a non-spinning, low-spccd kick (5 m/s) and a high-spced kick (29 m/s). During the
low-spccd kick, the boundary layer separation point was about 90 degrces from the front
stagnation point, while during the high-speed kick the separation point had recedcd to about
120 dcgrecs from thc front stagnation point.
1 Introduction
Many studies have been conducted on the aerodynamics of sports balls including
golf (Bearman and Harvey, 1976; Smits and Ogg, 2004), cricket (Mehta, 1983),
tennis (Haake et al. 2000; Mehta and Pallis, 2001), and baseball (Watts and Ferrer,
1987). Mehta ( 1985) carried out the first proper review of sports ball aerodynamics
by covering in depth the sports of cricket, golf and baseball. Mehta ( 1985) discussed
that the drag on a ball was dominated by the size and deflection of the wake. How-
ever, there have been few such studies on soccer balls (Carre and Asai, 2004; Carre
et al. 2004).
Contemporary soccer requires a player to be able to control the amount of spin on
the ball when kicking it and to be able to intentionally generate a curve ball. lt is thus
extremely important to elucidate the aerodynamic properties involved in soccer. We
therefore analyzed using a wind tunnel experiment the aerodynamic properties of a
soccer ball during non-rotation. We also compared the flow properties surrounding
spinning and non-spinning soccer balls in actual flight by a visualization experiment
using titanium tetrachloride.
328 Takeshi Asai, Kazuya Sco, Osamu Kobayashi and Reiko Sakashita
2 Methods
The wind tunnel used in the experiment was a low-speed circulating wind tunnel
(with six-component wind-tunnel balance; maximum wind velocity: 40 m/s; measur-
ing section: 1.5 x I m; turbulence level: ~! %) at the Department of Aeronautics and
Astronautics, Tokai University. In the wind tunnel experiment we conducted meas-
urements on a hand-stitched soccer ball (Size 5 Fevemova), and a thermally bonded
type of soccer ball (Size 5 Roteiro ), both of which were officially approved for inter-
national games. Each ball was supported at the rear, with the tip fitted to the six-
component wind-tunnel balance (Fig. 1). The ball was stabilized to the shaft with an
adhesive so it could not rotate.
Fig. 1. Photo ofthc wind tunnel experimcnt sct-up
Each of the following coefficients were calculated: wind velocity ( V), force acting in
the direction of the wind (drag (D)), force acting perpendicular to the wind direction
(lift (L)), and force acting sideways based on a frontal view (S). The aerodynamic
forces acquired in the experiment wcrc converted to the drag coefficient (C0 ), lift
coefficient (Cd, and lateral force coefficient (Cs), as shown in equations 2-1 to 2-3:
C = D (2-1)
v lj2pV 2 A
C = L (2-2)
'- lj2pV 2 A
C = S (2-3)
5 1/ 2pV 2 A
where p is the density of air (p= 1.2 kglm\ V is the flow velocity, and A is the pro-
jected area of the soccer ball (A = rr x 0.11 2 = 0.038 m2) . The wind velocity used in
the tunnel experiment was set within a range of7 m/s to 35 m/s.
A visualisation experiment using titanium tetrachloride was also conducted in or-
der to visualise the flow around the soccer ball during flight. A soccer ball was
placed directly in front of a soccer goal 15 m away and we had a subject perform a
straight kick that involved virtually no rotation and a side-spinning curve kick. Both
kicks were placement kicks delivered at the same velocity, as would occur in a real
Flow Visualization on a Real Flight Non-Spinning and Spinning Soccer Ball 329
game. A high-speed VTR camera (Photron Ultima; Photron Limited) was set up at a
midpoint between where the ball was placed and the soccer goal, and photographs
were taken at 4,500 fps. The wake angle (W) was defined as a parameter expressing
the size of the vortex area trailing the ball.
W= 360 ~ 2 x SP (deg.) (2-4)
Here, W is the wake angle, and SP is the angle made by the boundary layer separa-
tion point and front stagnation point.
In the tunnel experiment, observation of the changes in drag in relation to wind ve-
locity revealed that drag increased from 0.5 to 3.8 N as the wind velocity increased
from 7 to 35 m/s, although the changes were not uniform. At a wind velocity of 12 to
15 m/s a drag crisis was observed, whereby the drag coefficient temporarily de-
creased. Examination of the relationship between the Reynolds number and drag
coefficient revealed that the drag coefficient was about 0.43 at a high range and
about 0.15 at a low range. This meant that the critical Reynolds number for the non-
spinning soccer ball in the present experiment was about 2.2 x 105 (Fig. 2).
0.5
_._ Cd (Fevemova)
0.4 --0-- Cd (Rotelro)
0.3
0
" 0.2
0.1
0
100000 1000000
Re
Fig. 2. Drag coefficient versus Reynolds number for a non-spinning soccer ball
There was little difference in the drag properties between the hand-stitched soccer
ball and the thermally bonded one, although further research will be required to es-
tablish the details of variations between the two makes. Carre and Asai (2004) re-
ported a wind tunnel experiment using a l /4 miniature soccer ball whereby the criti-
cal Reynolds number for a non-spinning soccer ball was about 1.3 x 105. The critical
Reynolds number in the present study was found tobe higher, however. This might
be because the small size of a 1/4 miniature soccer ball comparcd with a full-sized
ball means that the effect of stitching on the ball panelling would be comparatively
!arge. In any case, the drag properties of this miniature ball were thought to be
somewhere between thosc of a smooth ball and a golf ball. When a player performs a
powerful shot or takes a free kick during an actual game of soccer, the initial velocity
is commonly considered to be about 25 to 35 m/s. In that case, the ball will reach the
330 Takeshi Asai, Kazuya Seo, Osamu Kobayashi and Reiko Sakashita
goal at the previous supercritical region related to the critical Reynolds velocity
region.
Images of the visualization experiment, which depicted the smoke trails produced
by the titanium tetrachloride, were examined based on observations and the angle
from the front stagnation point of the boundary layer separation point (tip of ball).
First, we compared the flow around the ball between a non-spinning low-velocity
kick (5 m/s) and a high-velocity kick (29 m/s). lt can be seen during the low-velocity
kick that the boundary layer separation point was about 90 degrees from the front
stagnation point and the vortex region was comparatively !arge, while during a high-
velocity kick the separation point had receded to about 120 degrees from the front
stagnation point and the vortex region had also shrunk (Fig. 3).
(a) (b)
Fig. 3. Flow visualization of a non-spinning soccer ball from above (ball velocity. (a) 5 m/s,
(b) 29 m/s) (flow is from right to left)
Taneda (1978) reported that the critical Reynolds number during transition of a non-
spinning smooth ball is about 3 x 10 5 and that separation at no more than the critical
Reynolds number occurs at a position of about 75 degrees from the front stagnation
point, while separation above the critical Reynolds number occurs at about 135 de-
grees from the front stagnation point. In the case of a soccer ball, the difference be-
tween before and after transition was not as great as the smooth ball, but there was a
difference up to about 30 degrees, which was also accompanied by a difference in
the vortex region. In the present study, we compared the wake angle (W) between
the subcritical time and supercritical time, and found a difference between the two at
about 60 degrees, but there was no notable difference within each respective region
(Fig. 4).
200
180 ,,.. 1 • Soccer ball 1
_160
,,:140
.
-120
m100
.,
.„
C
80
:J 60
:;: 40
20
o+-~~~-~~~
0 5 10 15 20 25 30
Ball YOlocity (ml s)
Fig. 4. Relationship between ball velocity and wake angle (360-2 * the separation angle from
the front stagnation point)
Flow Visualization on a Real Flight Non-Spinning and Spinning Soccer Ball 331
Examination of lateral images during high-speed kicking of a spinning curve ball (26
m/s; 8 rps) showed that the separation point was positioned about 120 degrees from
the front stagnation point, and the layer between the top and bottom surface of the
ball became turbulent (Fig. 5a-c).
(a) (b) (c)

Fig. 5. Flow visualization of a spinning soccer ball from side view (ball velocity: 26 m/s, spin
rate: 8 rps) (flow is from right to left and the ball is spinning from left to right)
Parietal images also showed that, as with the above, during high-speed kicking (29
m/s) of a non-spinning ball the separation point receded to about 120 degrees from
the front stagnation point, and the vortex region also shrunk. On the other band,
during high-speed kicking (27 m/s; 7 rps) of the spinning curve ball the vortex was
seen to deviate, due to the effects of a relative difference in fluid velocity caused by
the rotation (Fig. 6a-c). The curvature of the ball's trajectory was attributed to the
generation of a lateral force caused by the vortex counteraction. However, although
the separation point was not symmetrical on the left and right in relation to the direc-
tion of travel (top and bottom symmetry on the images) it was similar to about 124
degrees from the front stagnation point and the angle attained during high-speed
kicking of a non-spinning ball. During a high-speed kick of a spinning ball also, the
boundary layer became turbulent and the vortex region shrunk. The above findings
suggest that, during high-speed kicking of a non-spinning ball, the boundary layer
becomes turbulent. However, during high-speed kicking of a spinning curve ball it
can also be assumed that the boundary layer is a turbulent layer, based on the posi-
tion of the separation point and vortex region.
(a) (b) (c)

Fig. 6. Flow visualization of a spinning soccer baU from above (ball velocity: 27 m/s, spin
rate: 7 rps) (flow is from right to left and the ball is spinning in a counter-clockwise direction)
332 Takeshi Asai, Kazuya Seo, Osamu Kobayashi and Reiko Sakashita
4 Conclusion
Examination of the relationship between the Reynolds number and drag coefficient
for a non-spinning ball in the wind tunnel experiment revealed that the drag coeffi-
cient was about 0.43 at a subcritical region and about 0.15 at a supercritical region.
This suggests that the critical Reynolds number was about 2.2 x 105. In the visualiza-
tion experiment using titanium tetrachloride, we compared thc flow around the ball
during a non-spinning, low-speed kick (5 m/s) and a high-specd kick (29 m/s). Dur-
ing the low-speed kick, the boundary layer separation point was about 90 degrees
from the front stagnation point, while during the high-speed kick the separation point
had receded to about 120 degrees from the front stagnation point.
References
Bearman, P. W. and Harvey, J. K. ( 1976) Golf ball aerodynamics. Aerona11ticul Q11arter(r, 27,
112-122.
Carre M. J. and Asai T. (2004) Biomcchanics and Aerodynamics in soccer, Biomedical Engi-
neering Princip/es in Sports, G. K. Hung and J. M. Pallis Eds., Kluwer Academic plenum
Publishcrs, pp 333-364.
Carre M. J., Goodwill, S. R., Haake, S. J., Hanna, R. K. and Wilms, J. (2004) Understanding
the aerodynamics of a spinning soccer ball. The Engineering o/Sport 5 (Eds. M. Hubbard,
R.D. Mehta and J.M. Pallis). Pub. The International Sports Engineering Association,
Sheffield, UK, 70-76.
Haake, S. J., Chadwick, S. G., Dignall, R. J., Goodwill, S. R, and Rose, P. (2000) Engineering
tennis - slowing the game down. Sports Engineering, 3 (2), 131-143.
Mehta, R.D., Bentley, K .. Proudlove, M. and Varty, P. ( 1983) Factors affecting cricket ball
swing. Nature, 303, 787-788.
Mehta R. D. ( 1985) Aerodynamics of Sports Balls, Annual Review of F/11id Mechanic:s, 17,
151-189.
Mehta, R. and Pallis, J. (200 l) The aerodynamics of a tcnnis ball, Sports Engineering 4( 4 ),
177-189.
Smits, A.J. and Ogg, S. (2004) Golfball aerodynamics. The Engineering o/Sport 5 (Eds. M.
Hubbard, R.D. Mchta and J.M. Pallis). Pub. The International Sports Engineering Asso-
ciation, Sheffield, UK, 3-12.
Taneda, S. ( 1978) Visual Observations of the Flow past a Sphere at Reynolds Numbers Be-
tween 10"4 and 10"6. Jnl. F/11id Mech., 85, 187.
Watts, R.G. and Ferrer, R. ( 1987) The Lateral Force on a Spinning Sphere: Aerodynamics of a
Curvebal. Am. J Phrs., 55, 40-44.
Gaze Point Analysis in Movement Prediction of Soccer
Players by Image Processing
Yuusuke Hiramatsu 1, Shigemichi Ohshima 1 and Atsumi Ohtsuki 1
1 Department ofMeehanical Engineering, Meijo University, Nagoya, Japan,

m0553502@ccmailg.meijo-u.ac.jp
Abstract. This paper describes the effect of positioning the gaze point of a defensive player to
predict the movement of an offensive player in a soccer game by using the recognition
algorithm. In this research, positioning of thc gazc point is systcmatically varied to prcdict the
movcment of the offensive player, target. Hcrc, pattem images of the target moving to the
right and left side (dircctional movcment), and swccping thc left or right foot in front of the
ball and moving to the right and lcft sidc (feint movemcnts) arc rccorded by a digital camera.
Thc stored feints movements are thcn matchcd with the dircctional movcments to predict the
direction of movement of thc target. The matching systcm is developcd by mimicking brain
function for image rccognition. Thcrefore, in using the matching systcm, we can analyze the
gazc point as a system that is similar to thc visual information exchange ofthe retina and brain.
The optimum position of the gaze point is then cvaluatcd bascd on the differcnce of the
matching dcgree outcomc in thc opposite directions.
1 lntroduction
In the soccer game, defenders basically try to predict/anticipate what offensive

player is going to do with the ball, which direction offensive player is going to go.
The offensive player uses "the feint" a part of a movement designed to trick/fake to
distract the defenders. Defenders should not be cheated and react to the movement of
the ball and the body language ofthe offensive player with the ball.
According to the past research, in one-to-one in soccer game, the accuracy of
prediction was influcnced by the positioning of the gaze point of a defensive player.
The effect of the gaze point positioning pattems was analyzed by dividing expert's
group and beginner's group. The prediction ability of the expert's group was highcr
than that of the beginner's group. Therefore, we thought that the gaze point selection
is crucial for the accuracy of prediction and that it is possible the effect of
positioning the gazc point of a defensive player to predict the movement of an
offensive player by using the matching system. The matching system is developed by
mimicking brain function for image recognition. In using the matching system, we
can analyze the gaze point as a system that is similar to the visual information
exchange of the retina and brain. The optimum position of the gaze point is then
evaluated based on the difference of the matching degree outcome in the opposite
directions.
334 Yuusuke Hiramatsu, Shigemichi Ohshima and Atsumi Ohtsuki
Tlieoffe11ce Tlrecomero
Tl,e hegi11i11g poilll o( tlre feilll '}\ The 1-ir/110/ defe11se po itio11 (stoy)
Tl, last 1011 Ir poi111 );., '
i ... ....._·....
Im Im 4
Fig. 1. Schematic illustration of experiment
2 Experimental Conditions
In one-to-one in soccer, as for relations between the offensive player and the
defensive player, facing from the front, running in a row, and contact play are given.
In this experiment, the situation of facing from the front is chosen from them. In this
situation, the defensive player should predict whether the offensive player moves to
the right or left.
Figure 1 shows the schematic illustration of experiment. The feint is begun at a
position away from the defensive player by two metres, and last touch is clone at a
position away from the defensive player by one metre in the front side of the
defensive player. Sight of the defensive player is obtained from the camera put on a
position away from the defensive player by four metres in the rear side of the
defensive player. The feint images are obtained using a camera running at thirty
frames per second, from the beginning of the feint to the last touch. Then these
frames are converted into the one equal with sight of the defensive player. The set of
the still pictures that consists of these converted frames is used for the numerical
simulation to predict the movement ofthe offensive player.
3 Prediction System Developed by Mimicking Brain Function
Figure 2 shows the variation matrix of the recorded images. Figure 2 (a) shows the
four directional image pattems under consideration. Here, offensive player moves to
the right or left by hooking the ball using the outside or inside of the foot without any
feint. For example, d0 r is the case in which the offensive player moves towards the
left by hooking the ball using the outside of the left foot. And d;R is the case in which
the offensive player moves towards the right by hooking the ball using the inside of
the left foot. Figure 2 (b) shows four feint image pattems under consideration the
player's feint is clone by sweeping the left or right foot in front ofthe ball and then as
it lands on the ground use it to pivot and turn using the right or left foot to take the
ball to the right or left. For example, F 0 r is the case in which offensive player moves
towards the left after sweeping the right foot to take the ball to the left using the
outside of the left foot. And, FiR is the case in which offensive player moves towards
the right after sweeping the right foot to take the ball to the right using the inside of
the left foot.
Gaze Point Analysis in Movemcnt Prediction ofSoccer Players by Image Processing 335
In this system, the first phase is image translation from the log-polar coordinate
that is mimicking the translation from the retina to the brain (Jurie 1999). Here, the
origin of the log-polar coordinate is taken to be the gaze point under consideration.
Figure 3 (a) shows the target image in the log-polar coordinate system. Gaze
points on the target are selected to be on the waist, ancle and knee. Figure 3 (b)
shows the image in the translated coordinate system. The images on the left and right
are images produced when the origin is taken at gaze points I and 2, respectively.
Figure 3 (c) shows the target in the translated coordinate system without the
background. The dark portion is the target. lt should be noted that the shape of the
target in the coordinatc system changes with the selected gaze point. The gaze points
shown in the translated coordinate system include characteristic information that can
be used to indicate the predicted movement. The amount of information varies with
the position of the gaze point. Therefore, gaze point selection is crucial for the
accuracy of prediction.
Figure 4 shows matching of images produced by differential of Gaussian filter
with four direction and five frequencies. These extracting features by DOG are
mimicking the function of the visual processing at the primary visual cortex of the
brain (Nakano, Morie and lwata 2003). Figure 4 (a) shows the filtered directional
images of one directional image depicted Fig.2 (a). Figure 4 (b) shows the filtered
feint images of one feint image depicted Fig. 2 (b ). One set of filtered feint image is
matched against 40 directional images by the dynamic link matching method
(DLMM) that is flexible template matching method.(Lades, Vorbruggen and Konen
1993).
The right matching degree (ER 1) of one feint image (Fj) is computed by summing
the outcome of the matching to the 20 directional images towards the right (d;R and
d0 R) by DLMM. The left matching degree (E 1J) of one feint image (Fj) is computed
the same as thc right. The prcdiction parameter (Ej) is defined the difference of the
matching degrce outcome in the opposite directions (ERi and EL1). These parameters
are defined as:
I " I " (!)
E,, =-
/1
Ir F, ®d.,. J.
L·I
E,, = -Ir
/1
F, ®d.,, J.
L-1
E r =E R1 -E l1
(®: matching operation hy DLMM, *: or outside) and ir inside ),j. Flame numherof Feint image)
Here, j (i= 1 ➔ 31 frame) is frame number of feint image and n ( =20 frame) is
number of directional images each of left and right directions. If E1<-0. I then the
defensive playcr predicts that the offensive player, target, moves to the left direction.
If EpO. I then the defensive player predicts that the offensive player moves to the
right direction. lf -0.1 < =E1< =0.1 then the defensive player doesn 't predicts whether
the offence moves to the left or right. When the absolute value of E1 is the )arger, the
confidence of the defense for prediction is the !arger.
4 Results
Figure 5 (a) shows the nine gaze points based on the past research that the expert
defensive player's gaze points are tend to be put on knee, waist and ankle in feint
section.
k. 1 •·· 6 7 8 9 10
(a) D1rec11011lll image
Fig. 2. Variation matrix ofthc rccorded imagcs
Ga=ept1111t J
p Ga:~pnmt2
Gu:~po•m 1
Xn X
•
(a) Log-polar coordinate srstem
8 Ga=e point 1 Gaze point 2 Ga::.e point 1 Ga::.e point 2
.............
(h) Trcmsluted coordinate .,:rstem (c) Witholll hackground
Fig. 3. Log-polar transformation
7 9 10
mutd1i11g
X 10/ramrX-1 (d~.,l.t..d,1.d11)
(o/ D1rectionol ,muge (1,) Femt imuge
Fig. 4. Matching of imagcs by the differential of Gaussian

Gaze Point Analysis in Movement Prediction ofSoccer Players by Image Processing 337
Thc simulation results are shown in Fig.5 (b ), (c) those show the prediction
parameter E1 when the gaze point is put on each part of the knees of thc offensive
player who does F;1,. The vertical axis in graphs is the frame number j and the second
that is spent to last touch. The horizontal axis is the values of prediction parameter E1.
The predictions are done at the same time as behavior of the offensive player. The
defensive player deeides own behavior bascd on Ei since the 22nd frame to whieh the
dot line is pulled. The defensive player reacts to the offensive player, when the seven
frames with E1 of the same direction are accumulated between from the 22nd frame
to the 31 st frame. Ei before the 21 st frame are used only for the gaze point movement
of defensive player. The gaze point movements of two pattems are shown by arrows
and double circles in Fig.4 (b), (c). Gaze point movement has delay ofabout 133[ms]
that is initiation of regular saccade (Dorris, Pare and Munoz 1997). When the
movement of the gaze point happens since the 22nd frame, the prediction that begins
to move is accumulated as the four frames with E of the same direction during the
movement.
Figure 5 (b) shows the gaze point pattem I to Fil. that the gaze point is moved to
the predicted direction. Figure 5 (c) shows the gaze point pattem2 to F,L that the gaze
point is moved to the opposite of the predicted direction. In Fig.5 (b ), the
accumulation of E1 become four frames on the left direction, and become six frames
on the right direction. Because t1 of the same direction more than seven framcs are
not accumulated, the defensive playcr cannot react to stop offensive player. In Fig.5
(c ), the defensive player can react to stop offensive player by moving to the right,
because the accumulation of the E1 becomes seven frames on the right direction at
the 28th frame.
Figure 6 show thc success rate of the defensive player to four subjects of the
offensive player who does feint. The vertical axis in graphs is the kind of gaze point
of seven pattems. The horizontal axis is the success rate that the defensive player
stops the offensive player. There are the pattems of 100% in the fixed gaze point on
the each part of knees and the moved gaze point respectively. The gaze pattem 1 that
the gaze point is moved to the predicted direction become 100% in Fig.6 (a), and it is
corresponding to the past re search. When all gaze point pattem are evaluated at the
same time, the gaze point pattem with 100% doesn't exist. Figure 6 (c) shows the
average of Fig.6 (a) and Fig.6 (b). The gaze point pattem2 is most high rate in Fig6
(c) among all gaze point pattem.
5 Conclusion
The system to evaluate the relation between thc feints and the gaze points pattems
was proposed in the numerical simulation analysis. Then the effect of the gaze point
pattems to predict the feints was researched and proposcd by using the matching
system. In the future, it will have to do that the result is fed back to the training site.
,,,, -../,, 11,m /lflldnu l1, / pr, lu ,u,,, JStuc1m, h, /

r1>,:_h1 /1 Jt u-.:111 I, /1 r1,;l1t /, /1 11x_h1 I, lt "1-:''' lt/t 11,-:lrt /1 /1
1 II I I O 1 1 IJ I 1 II I I II 1 1 II 1
, ,~ht S..111, ,md,lt..· of lt Jl l,.m, 11~lt1 A.nn m,d,11, 41 1 lt-/1 J.m

l..llttl 1..nu..·~
(d) Gu~t p,11111 puu,..,-11, (/>) (u,_·,· ,,,,,,,, /1tllf<'rll 1 (I') (,cC(' /HI/III p<lll<'/"11-:J.
Fig. 5. The nine gaze point and two gaze point movements based on E for subject3 in F;r
/.-/1411«•
e::
...
nuddh· o/ 4,1n-,
rithl#..m·,
A.m.·c ,m (o-ot o/ ,u. cpm~ o,-.,
Am.t. mt«.U\ foot "ht1J ,uupm~n, ,
Cia:1.: /lfllnt />am,.,, 1
Citet' ,,,,,,,,, /Jilllt.. n,~
II :o .JH 0 ~(I U ,... 0 100, .,
(ht S11,uHr,1h h:) \ucu \\ rdh m

mf unJF /· l 1- .,,,JF
Fig.6. The success rate for defensive player to stop the four subjects of offensive player
Acknowledgment
The authors would like to thank Mr. K. Inatomi at Meijo University for data
processing.
References
Jurie, F. (1999) A new log-polar mapping for space variant imaging. Application to face
detection and tracking. Journal ofthe Pattern Recognition Society. 32, 865-875.
Nakano, T., Marie, T. and Iwata ,A. (2003) A Face/Object Recognition System Using FPGA
Implementation ofCoarse Region Segmentation. SICE Annual Conf, Fukui, pp. 1418-
1423.
Dorris, C. Micheal., Pare, Martin., Munoz, P. Douglas. (1997) Neuronal Activity in Monkey
Superior Colliculus Related to the Initiation of Saccadic Eye Movements. J N eurosci 17,
8566-8579.
Lades, M., Vorbruggen, C. J., Buhmann, J., Lange, J., Maisburg, v.d. C., Wurtz, P. R., and
Konen, W. (1993) Distortion Invariant Object Recognition in the Dynamic Link
Architecture. IEEE Trans. Campt., vol. 42, no. 3, pp. 300-311.
Traction Testing of Soccer Boots Under Game Relevant
Loading Conditions
Thomas Grund' and Veit Senner'
1 TU München, Faculty of Sports Science, Department of Sport Equipment and Materials,

grund@sp.tum.de
Abstract. The final goal of this research is to determine anterior cruciate ligament (ACL)
tensile forces under game relevant loading conditions with respect to different typcs of soccer
boots and different stud design. To realize the given target a combination of experimental
testing and computer simulation was chosen. Based on the findings of previous studies and the
analysis of video-recorded injury situations in soccer, a new traction test device ('Trak-
Tester") has becn developed. The distinguishing feature ofthis device is to simulate and meas-
ure not only pure traction force betwcen shoe and surface but also to simulate high risk load-
ing situations such as thc "plant-and-cut-maneuvcr". Thc resulting forces and torque
transmittcd to the tibia will be measurcd using a 6-component-load-ccll. Comparative testing
conceming the amount of traction and the extent of force and torque produced with different
sole designs will follow. For calculation of ACL tensile forces the cxperimentally determined
loads transmitted to thc lower leg will latcr be used as input data for a multi-body computer
model of the human knee. Thc current paper will demonstrate the necd for a new traction test
device and illustratc the ongoing first steps of its developmcnt.
1 Introduction
Knee injuries are one of the most frequent injuries in soccer (Dvorak and Junge
2000). Espccially ruptures of the anterior cruciate ligament (ACL) are very severe,
causing at least six months of rehabilitation and some cases the end of the active
career of the injurcd playcr. Numerous studies showed a prevalence of non-contact
ACL-injuries compared to contact ACL-injuries in team-sports (Myklebust, Enge-
bretsen, Braekken, Skjolberg, Olsen and Bahr 2003).
The "plant-and-cut maneuver" (Besier, Lloyd, Cochrane and Ackland 2000) and
the "one-leg jump-landing" (Olscn, Myklebust, Engebretsen and Bahr 2004) have
been identified as the major non-contact injury mechanisms. Both mechanisms are
characterized by a planted (fixed) foot, slight knee flexion, valgus-position of the
tibia and intemal rotation ofthe body.
lnvestigators analyzed more in detail the kinematics of the plant-and-cut maneu-
ver (McLean, Myers, Neal and Walters 1998; Sigward and Powers 2005). They all
agree that the movements observed during a typical plant-and-cut seem not tobe in a
340 Thomas Grund
critical range for suffering an ACL-rupture. However the large variation between the
subjects makes general statements impossible.
Non-contact ACL-injuries were related to high traction forces between the sole
and the playing surface (Heidt, Dormer, Cawley, Scranton, Losse and Howard 1996).
They conclude that excessive traction increases the risk of foot fixation. A number of
scientists (Heidt et al. 1996; Haake, Carre, Kirk and Senior 2004) then investigated
thc translational traction behavior of different boot types and different stud-design.
Other authors conccntrated on the rotational traction component (Valiant 1989; Heidt
et al. 1996; Barry and Kummer 2000; Cawley et al. 2003). However no information
is available concerning the intluence of the stud-design during combined loading
situations such as cutting maneuvers.
Therefore we attempt a new approach to obtain ACL tensile forces in situations
which have the potential to cause injuries. The principle method is a combination of
experimental traction measurements (using a mechanical test device) with a com-
puter simulation. The focus of the presented paper is on the development of this
traction test device.
The determination of the unknown ACL-tensile force will be done by means of a

computer simulation using a complex model of the human body and of the knee
joint.
For running this simulation realistic joint kinematics as input data are necessary.
To achieve this data from TV recordings showing ACL-injuries in soccer we have
chosen the computer based reconstruction method described by Krosshaug
(Krosshaug and Bahr 2005). This data will then be used as kinematical target crite-
rion in the optimization process of a forward dynamics problcm using an existing
computer model of the human body (Grund, Wallrapp and Senner 2004). lt will be
combined with a dctailed representation of the human knec joint (Lehner, Eichhorn
and Senner 2005) allowing to determine the unknown ACL tensile force.
A major difficulty in the development of this model is how to describe the con-
tact between the playing surface and the studded sole of the soccer boot. To avoid
thc complex shoe-surface contact modeling wc plan to use experimentally deter-
mined contact loads instead. These will be taken from simulations with a physical
model, a traction tcst device, able to reconstruct ACL injury situations. As several
assumptions and simplifications will have to be made in the development of this
traction tester ("TrakTcster"), it is clear that this physical model can only determine
thc resulting loads between boot and ground with limited accuracy. However we
consider the measured contact load histories which will then be used as input for the
computer model to be a good estimate of the situation to analyze.
2.1 Design Requirements

Literature describes some state-of-the-art traction-tcst devices (Barry et al. 2000;
Cawley et al. 2003). Both allow mcasurement of translational as weil as rotational
traction. The device of Cawley et al. furthermore allows the adjustment of plantar-
and dorsi-flexion and setting both, inversion and eversion up to 25 degrees. But there
are some more demands on the test device that need to be fulfilled for a realistic
simulation of an injury-situation and for the acquisition of the necessary data for the
simulation:
Replication of the füll range of motion of the ankle joint in order to secure the
realistic position of the tibia in reference to the boot and the playing surface,
respectively (especially varus and valgus position of the tibia).
Application of 3-component angular and 3-component linear momentum on the
soccer boot.
Appropriate dynamics of load application.
Measurement of resultant forces and torques at the tibia (transmittable intemal
force).
Portable design of the test device in order to accomplish measurements on different
playing surfaces and under various weather conditions.
Comparable to the test devices mentioned above our TrakTester is controlled via
software and a computer. In contrast to other devices we have chosen not to fix sam-
ple boxes of natural grass on a force plate. Instead we want to bring the device to the
lawn (or other spart grounds) and measure the force and moment components with a
built-in six-component load cell.
2.2 Some Design Details

A stable frame supports an artificial lower leg and a simplified model of the foot.
The anatomic shape of the foot is given through a silicone cast of a volume model of
a 50th percentile male. Inside the silicone a meta! structure simulates major behav-
iour of the foot's bones. The lower leg consists of a shaft of steel with an adapter for
load transmission. The connection between lower leg and foot is a mechanical repli-
cation of the upper and lower ankle joint that assures the position of the joint axis
and the range of motion based on the findings of lnman (lnman 1976). This artificial
joint is shown in Fig. 1 and Fig. 2 shows the complete unit of the artificial leg (Ebert,
Knauer and Senner 2005).
Fig. 1 Ankle Joint Axis in the CAD-Model Fig. 2 Artificial Lower Leg
342 Thomas Grund
The load application will be challenging due to high dynamic movements to

realize. To solve this problem we will either build a mechanism with a pre-stressed
spring, a pneumatic solution or a combination ofboth.
For recording the joint angles of the artificial ankle joint, two goniometers are
implemented. The measurement of the transmittable intemal force is accomplished
by a 6-component load cell, located on the distal end of the surrogate tibia, just
above the ankle joint. Data acquisition will be conducted using a NI-DAQ-Card
(National Instruments, Austin, Texas), a standard laptop with PCMCIA-Slot and the
software NI LabVIEW (National Instruments, Austin, Texas, USA).
A modular construction will allow portability of the TrakTester by separating
single components for purpose of transportation.
3 Current Status
The artificial foot, the lower leg and the artificial ankle joint are ready for use. Also
the construction of the main frame has been completed. At the moment we are
searching for the best technical solution for realizing the linear and rotational impact.
The two models (three dimensional muscle-skeleton model of the player and the
knee joint model) have both been implemented in SIMPACK™ (Intec GmbH,
Wessling, Germany), which is a commercially available multi-body-system software
package. Fig. 3 and Fig 4 illustrate the two models.
Fig. 3 computer model ofthe human body Fig. 4 Knee Joint Model (Lehner et al. 2005)
The human body is represented by 37 rigid segments. Additional masses are con-
nected to 13 of them, representing wobbling masses (Zaun er 2002).
The knee model by Lehner has been developed within the past 10 years (Lehner
1995; Lehner et al. 2005) and is now being validated by experiments with human
cadaver knees. Representation of the major anatomic structures and the most impor-
tant characteristics ifthe human knee joint is realized.
4 Discussion
N igg ( 1990) questioned the relevance of laboratory-based test devices in general

because these tests only provide information on the surface-shoe-interaction. His
conclusion was that the combination of laboratory material-tests and subject-tests
would be the best to get meaningful information on what shoe-surface combination
has the highest potential for reducing the risk of injury.
To our knowledge the presented study is one of the few studies which try to
simulate a real injury-situation using a mechanical test device. Our aim is to increase
our knowledge on what influence different stud-designs may have on the forces
generated during a sport-specific movement. By comparative measurement of differ-
ent soccer boots with the TrakTester it should be possible to obtain adequate input
data for our models and to draw the correct conclusions from the simulations hereof.
The main difference compared to known traction test devices (Heidt et al. 1996;
Barry et al. 2000) is that we will test soccer boots under more realistic loading condi-
tions. This means first testing under a combined situation instead of pure transla-
tional and/or pure rotational traction testing. Secondly this requires the application of
the loads with respect to physiologic axis instead of a straight perpendicular force
application.
Although our TrakTester should be a step towards meaningful traction testing
several limitations will have to be considered. The influence of the stud-design will
only be investigated for one certain movement (the one for which we will determine
the kinematic data from the TV recordings). Therefore generalisation of the results
on any cutting manoeuvre will be critical.
Further it is almost impossible to adequately model the complex muscle interac-
tion during a plant-and-cut manoeuvre. Therefore the simulation results will only be
able to show tendencies, but nevertheless they will improve our understanding of
what happens inside the knee in such situations.
5 Future Prospects
Having once realized the entire procedure combining the TrakTester measurements
with our computer model, the systematic development of an ACL-friendly stud de-
sign is the next logical step. Basing on extensive measurements with commercially
available soccer boots we will focus on the question, if there are differences regard-
ing the resulting ACL tension force for boots with cylindrical studs and those boots
with studs formed in blade-style. Also prototype-boots with alternative sole layouts
and a completely different alignment and shape of the studs will be analyzed.
6 Acknowledgement
This study is supported by the "Verwaltungs-Berufsgenossenschaft" (VBG). We

would also like to thank the "Deutsche Fußball Liga GmbH" for the free of charge
access to TV recordings.
344 Thomas Grund
References
Barry, E.B., Kummer, R. and Milburn, P.D. (2000) The design of a traction-measuring device
for footware. In: Subic, A. and Haake, S. (Eds.) The Engineering ofSports - Research,
Development and Innovation. Blackwell Science, Oxford, pp. 103-111.
Besier, T.F., Lloyd, D.G., Cochrane, J.L. and Ackland T.R. (2000) External loading ofthe
knee joint during running and cutting maneuvers. Med&Sci Sports&Exc. 33, 1168-1175.
Cawley, P.W., Heidt, R.S., Scranton, P.E., Losse, G.M. and Howard M.E. (2003) Physiologie
Axial Load, Frictional Resistance, and the Football Shoe-Surface Interface. Foot&Ankle,
Int 24, 551-556.
Dvorak, J. and Junge, A. (2000) Football Injuries and Physical Symptoms - A Review ofthe
Literature. Am J Sports Med 28, S3-S9.
Ebert, C., Knauer, C. and Senner V. (2005) Objectivating Safety Aspects of Ski and Snow
board Boots by a New Testrigg and Optimized Prosthesis. Proceedings of 16th Interna-
tional Symposium on Ski Trauma and Skiing Safety. Niigata, Japan. 31.
Grund, T., Wallrapp, 0. and Senner V. (2004) Simulations in Human Movement. Proceedings
ofSIMPACK User-Meeting 2004, Wartburg, Eisenach, Germany.
Haake, S.J., Carre, M.J., Kirk, R.F. and Senior T (2004) Traction of studded boots on turf. In:
Hubbard, M., Metha, R.D. and Pallis, J.M. (Eds.) The Engineering of Sports 5, Volume 2.
International Sports Engineering Association. Pp.544-551.
Heidt, R.S., Donner, S.G., Cawley, P.W., Scranton, P.E., Losse, G. and Howard, M. (1996)
Differences in Friction and Torsional Resistance in Athletic Shoe-Turf Surface Interfaces.
Am J Sports Med 24, 834-842.
Inman, T.V. (1976). The joints ofthe Ankle. 1-117, Baltimore: Williams and Wilkins.
Krosshaug, T. and Bahr R. (2005) A Model-based image-matching technique for three-
dimensional reconstruction ofhuman motion from uncalibrated video sequences. J Bio-
mech 38, 919-929.
Lehner, S. (1995) 3D Simulation des menschlichen Kniegelenks. Diploma-thesis Munich
University of applied sciences, Department of Precision- and Micro-Engineering.
Lehner, S., Eichhorn, S., Senner, V. (2005) Towards an asymmetric ski binding release: ACL
versus binding forces with the rotated knee in deep flexion. In: Proceedings: 1st World
Congress on Sports Injury Prevention, Oslo, Norway.
McLean, S.G., Myers, P.T., Neal, R.J. and Walters, M.R. (1998) A Quantitative Analysis of
Knee Joint Kinematics During the Sidestep Cutting Maneuver. Bulletin Hospital for Joint
Diseases 57, 30-38.
Myklebust, G., Engebretsen, L., Braekken, I.H., Skjolberg, A., Olsen O.E. and Bahr, R. (2003)
Prevention of Anterior Cruciate Ligament Injuries in Female Team Handball Players: A
Prospective Intervention Study Over Three Seasons. Clin J Sport Med. 13, 71-78.
Nigg, B.M. (1990) Tge validity and relevance of tests used for the assessment of sports sur-
faces. Med&Sci Sport&Exc 22, 131-139.
Olsen, O.E., Myklebust, G., Engebretsen, L. and Rahr, R. (2004) Injury Mechanisms for Ante-
rior Cruciate Ligament Injuries in Team Handball - A Systematic Video Analysis. Am J
Sports Med 32, 1002-1012.
Sigward, S.M. and Powers, C.M. (2006) The influence of gender on knee kinematics, kinetics
and muscle activation patterns during side-step cutting. Clin Biomech 21, 41-48.
Valiant, G.A. (1989) A Method ofMeasuring Translational and Rotational Traction Character-
istics ofFootware. J biomech 22, 1091.
Zauner, C. (2002) Integration ofthe man model Ramsis into MBS software Simpack-model
comparison and biomechanical simulation for verification. Master-thesis, Technische Uni-
versität München.
Correlation Between Support Foot Placement
and Goal Accuracy for Instep Kicks in the Soccer Field
Giuseppe Marcolin, Nicola Petrone and Claudio Robazza

University of Padova, Italy, nicola.petrone@unipd.it
Abstract. The aim of this work is to evaluate the correlation between the placement of the
support foot on the ground and the precision of the shot: the placement of the support foot
relative to the ball was supposed to have an effect on the shot precision for a given target. The
goal was divided with a visual grid that enabled to identify six possible targets of the shot,
high/low for vertical placement and left/center/right for horizontal placement. Skilied players
were asked to perform repetitive penalty instep kicks in the soccer field with a defined target
such as high/left or low/right. Ground reaction forces were recorded at the supporting leg by
means of a force platform installed in a suitable board and the player's movement was video
recorded. The position of COP at the kick instant with respect to the ball and the average
trajectory of COP on the platform tumed out tobe correlated with successful targeting in such
a way that training procedures can be defined to improve the player coordination for precision
shooting. Finally, differences between free kicking and kicking with a goalkeeper were re-
corded, thus enabling us to analyze a realistic player behavior in the study.
1 Introduction
Soccer is one of the most popular sports in Europe and in the world: the application
of motion analysis techniques, EMG recording and force acquisition showed great
advances in the last few years.
The ball speed is correlated with the Ground Reaction Forces (GRFs) and the
values that skilled players exhibit vertically, anteriorly-posteriorly and laterally are
greater than among unskilled players (Barfield, 1998). The direction of approach to
the ball has an influence on ground force components (Kellis, Katis and Gissis,
2004); lastly the ball speed in different ways of kicking (instep and side-foot) has
also been investigated (Nunome, Asai, Ikegami and Sakurai, 2002).
The aim of the present work is not only to study force magnitude in relation with
ball speed, as known from previous works, but also to evaluate the correlation be-
tween the placement of the support foot on the ground and the precision of the shot
in the soccer field in order to closely simulate whathappens during a football match.
The X-Y distance of the support foot from the ball and the foot orientation at the
kick instant were supposed to have a correlation with the goal target sector.
For this reason, the goal was divided with a visual grid enabling the subject to
identify six possible targets of the shot, high/low for vertical placement and
left/center/right for horizontal placement.
Skilled players were asked to perform repetitive penalty instep kicks in the soccer
field with a given target such as high/right or low/left: ground reaction forces were
346 Nicola Petrone, Giuseppe Marcolin and Claudia Robazza
recorded by means of a force platform installed in a suitable board together with the
kick instant.
2.1 Instrumentation
The Ground Reaction Forces (GRF) at the support foot and the trajectory of the
Centre of Pressure were measured by means of a Bertec force platform (600x400
mm, Model 4060, Bertec Corporation, U.S.A.). Signals were recorded on a National
Instruments NI-PXI-1042 at 1000 Hz per channel.
The force platform was fixed inside a wooden module (dimensions lm x lm,
height 120mm) and placed on the football ground together with a set of 6 equal
modules to create a suitable and safe runway for the player at a constant distance
from the ground.
The goal was divided into six equal rectangular sectors (2,44 m x 1,22 m) and a
wooden rectangular panel was hanged in correspondence of the target sector.
The instant ofball kicking and the average horizontal ball speed were detected by
means of a suitable electric circuit including a support trigger under the ball, a trigger
at the target panel, a LED at the force platform and a 3.6 V battery. Two triggers,
normally open, were connected in parallel to the NI-PXI-1042 and recorded syn-
chronously with the force signals. The support trigger was applied on a hollow grid
placed laterally to the force platform allowing the tester to adjust the position of the
ball. The placement of the ball onto the support trigger closed the circuit closure and
switched on the LED; at the kick instant, the circuit opened and the LED switched
off until the ball hit the target, closing again the circuit and switching the LED on
again. The trigger signal enabled to individuate the kick instant and to estimate the
ball average speed on the basis of the ball flight time. Each test was filmed via two
digital video cameras NC GR-DVL357, one from the back and the other from the
left side ofthe tester, both viewing the LED for image/force synchronization.
2.2 Test Protocol

Tests were performed on a regular football field: the ball was placed centrally with
respect to the goal, at a distance of 16m from the goal line. The balls were nr. 5 FIFA
approved.
Four non-professional players with experience in the secondary leagues were in-
volved in the study: the data ofthe testers are summarized in Table 1.
Corrclation Between Support Foot Placement and Goal Accuracy 347
(a) (b)
Fig. 1. (a) Force Platforrn embedded in the wooden nmway. (b) Detail ofthe ball trigger.
Fig. 2. Rear view ofthe experimental setup, for a right foot player.
TESTER AGE HEIGHT lcml MASS IK2:I FOOT ROLE

1 (M.C.) 29 183 84 RIGHT Midfielder
2 (M.S.) 29 186 80 RIGHT Midfielder
3 (E.Z.) 32 168 67 RIGHT Midfielder
4 (G.M.) 27 184 77 LEFT Forward
Tablc 1. Testers involvcd in the study.
Once in thc football ground, the placemcnt of the ball on the support grid and of
the running pathway were chosen by each tester in order to follow his natural habit
and to ensure the placement of the whole support foot on the force platform. All
players were wearing standardized shoes for indoor soccer and were requested to
read and sign an informed consensus about the tests.
348 Nicola Petrone, Giuseppe Marcolin and Claudio Robazza
The test protocol was subdivided into several stages:

Stage 1: warm-up. Before the kicking session, each tester performed a warm up
conceming 5 minutes slow running, 6 series of 10m skip and rear kick running, 4
stretching exercises for the quadriceps, knee flexors, gastro and soleus.
Stage 2: pilot shots for ball and runway placements. After 5 shots, the tester was
requested to define the most suitable placement for the ball with respect to the plat-
form and for the runway modules. This was done to reduce the influence that the use
of a wooden surface, rather than a grass surface, might have on the kicking action.
Stage 3. precision shots. Two series of 18 shots without goalkeeper were per-
formed with a given target: for right-footed players, the first target was High-Left
(HL); the second target was Low Right (LR) (the left footed targets were High Right
and Low Left).
Stage 4. power and precision shots with goalkeeper. A series of 16 shots with a
goalkeeper, with the aim of scoring in the given target sector of the goal. Four possi-
ble targets, Low Right, Low Left, High Right, High Left were randomly mixed and
communicated to the player immediately before each shot, with the request of shoot-
ing with the maximum power and aiming to the given target. The goalkeeper was
asked to wait for the ball kicking before moving. Only successful shots (even if
caught by the goalkeeper) aiming to the given target were analysed.
2.3 Data Analysis

For each shot, the three components of Ground Reaction Forces Fx (anterior-
posterior), Fy (medio-lateral) and Fz (vertical) were plotted synchronously with the
trigger as in Figure 3. Peak values of Fz were normalized to the body weight and
evaluated; the average horizontal speed of the ball was also calculated as the ratio
between the goal distance and the flight time.
The COP of each shot was plotted in the platform system of reference within the
interval between the instant of contact and 50 ms after the ball kick. The first part of
the contact, from the Heel-Strike to the Foot Fiat, corresponded to the raising of
force Fz from zero to a peak and in,volved the settling of the ankle. The following
part, from the Fz peak to the kick, corresponded to the swing phase before the ball
kicking. For this reason, the COP trajectory occurring between the Fz peak and 50
ms after the kick was linearly approximated: the slope coefficient m was evaluated as
shown in Figure 3, together with the coordinates of the Impact Point (IP), that is the
position ofthe COP when the ball impacted.
For each series with a given target, the different values of m and of IP coordi-
nates were averaged: the IP scatter was expressed as an ellipse with semi axes equal
to the x and y standard deviations. The slope m scatter was expressed as three lines
pointing to the ellipse centre and defining the upper and lower standard deviation
bands around the mean value of slope m, as reported in Figure 4.
The coordinates of the impact point IP, measured in the platform reference sys-
tem 0xyz as described in Fig. 1.a, were transformed in the ball coordinate system
OXY thanks to the knowledge of the placement of the ball trigger relative to the
platform in the suitable grid shown in Fig. 1.b.
Correlation Between Support Foot Placement and Goal Accuracy 349
-300 -200 -100 0 100 200 300

t
:f
100
- -------
f ___
50 -
(a
-150
-200
(b)
Tlme [sec)
-1000 ~
Fig. 3. (a) COP trajectory for a right footcd player (grey line) HL, linear approximation ofthe
kick phase (black line) and COP at the impact (black points: smallcr points for other success-
ful shots) referred to the force platform system. (b) Corresponding GRF plots and triggcr.
Lo, Right tar et 600 y Hi h eft target

y
550
( R) (H
500
500
450
450
3
2
3 4
- 350--;- ~ 300-t
-300 -200 -100
-300 -200 -100
Fig. 4. Comparativc diagram between the different players in the ball reference system.
350 Nieola Petronc, Giuseppe Mareolin and Claudia Robazza
The goal accuracies ofthe four testers were similar (1-27%, 2-36%, 3-28%, 4-25%),
but lower than the values that may be expected with professional players.
The values of normalized forces Fz were generally higher than values presentcd
in the literature (Kellis et al. 2004 ): this can be due to the fact that in the present
study the tests were performed in a football court and not indoor.
The ball speed resulted lowcr than the values reported by Nunome et al. (2002):
differently from that work, in this study the players had been asked to be precise,
aiming to the given target. Speed was generally higher in the scries of shots with
goalkeeper.
The final rcsults of the four testers under invcstigation are compared in Figure 4,
where left-footed player 4 has been reported by symmetry. Differences between HL
and LR series can be appreciated.
For each tcster, paired t- test statistical analysis between the two series of free
kicks with different targets (e.g. HL-LR) was carried out on the following parame-
ters: the anterior-posterior coordinate of IP (xfP), the slope (m) of the COP trajectory
trend line, thc average ball speed (vm) and the normalized peak vertical force
(Fz!BW). The anterior-posterior placement of the foot, coordinate xfP, proved tobe
significant for testers 3 and 4. The direction of the foot sole, slopc m, was significant
for testers I and 2. The ball spced vm was significant for tester 4 only; the normal-
ized peak forces Fz/BW tumed out to differ significantly (p<0.05) for testers 2, 3, 4,
with a tendency for tester 1 (p=0.071 ).
Further tests with professional players are supposed to confirm the method and
hypothesis of the study.
4 Conclusions
The anterior-posterior position of COP at the kick instant with respect to the ball and
the direction of the support foot on the platform recorded in a soccer field tumed out
to be correlated for two testers out of four with different goal sector targets in such a
way that training procedures can be defined to improve the player's coordination for
precision shooting.
References
Barficld W.R. ( 1998) The biomeehanies of kieking in soeeer. Clin. Sports Med.; 17 (4 ): 711-
28.
Kcllis E., Katis A. and Gissis 1. (2004) Knee biomcehanies of the support leg in soeecr kieks
from three angles of approaeh. Med. Sei. Sports Exere.; 36 (6 ): 101 7-28.
Nunome H., Asai T., Ikcgami Y. and Sakurai S. (2002) Three dimensional kinetie analysis of
side-foot and instcp socccr kieks. Mcd. Sei. Sports Exere.; 34 ( 12): 2028-36.
Analysis of the Influence of Rubber Infill on the
Mechanical Performance of Artificial Turf Surfaces
for Soccer
Enrique Alcantara, David Rosa, Javier Gamez, Antonio Martinez, Mario Comin,
Maria Jose Such, Pedro Vera and Jaime Prat
Instituto de Biomecanica de Valencia (IBV), enrique.alcantara@ibv.upv.es
Abstract. Artificial turf is increasingly being used in the construction of football pitches.
Characteristic for this product is the infill, usually consisting of sand and rubber granulates. A
significant role is attributed to it in the perfonnance of the surface. At present, different mate-
rials and thicknesses, as weil as grain sizes are used with little scientific support about their
influence in mechanical and biomechanical properties. However, knowledge from materials
sciencc makes reasonable to think that grain morphology will also have a great influence in
the field perfonnance. This paper presents a rescarch conducted to assess the influence of
different parametcrs related to infill grain morphology on the mechanical properties of artifi-
cial turf, as weil as on their wear with use.
1 Introduction
Installation of artificial turf for football pitches has experieJ?ced a huge increase in
last years and a further increase is expected in next years after UEF A and FIFA ar-
rived to an agreement for the rules to allow the use of artificial turf pitches in their
competitions.
Nowadays 'third generation' artificial turf field is used for soccer. Characteristic
for this field is the infill consisting of sand and rubber granulate. A significant role is
attributed to the infill in the performance of the surface at both mechanical and bio-
mechanical lcvel; mainly in relation to impact forces reduction, vertical deformation
and ball bounce, which have been suggested as injury factors (Dura et al. 1999; Dura
et al. 2002; Shorten and Himmelsbach 2002; Baroud, Nigg and Stefanyshyn 1999;
Ekstrand and Nigg 1989). However, there are not scientific works published with
this respect. Another point that needs clarification is about the influence in mechani-
cal properties of infill compacting due to use of the field along time.
Nowadays, the discussion is mainly on the type and thickness of rubber and sand
tobe used (Walker 1996) and the only parameter used to make a choice is the granu-
late size. Standards only consider the minimum size for the rubber, and the morphol-
ogy of the grain (roundness) for the sand (FIFA 2005). However, from materials
science knowledge, it is reasonable to think that grain morphology has a great influ-
ence in the mechanical performance of the field and thus, it could be used in the
dcsign of fields.
352 Enrique Alcantara et al.
The regulation bodies (FIFA, UEF A), in collaboration with several laboratories
and research centers, have developed standard methods to measure the performance
of football pitches (Socraturf 1999), which can be divided into technical and sports
performance properties. This paper presents a research conducted to assess the influ-
ence of the type, grain morphology and size of rubber infill on artificial turf proper-
ties and their modification with use.
2 Material and Methods

The work consisted in mechanical testing of a series of prototypes of artificial turf
built using different infill in a same reference turf: FJELD TURF INC. FTOS 1S
(Field Turfüutdoor Series 1 for Soccer) and pile height 60 mm.
Two different granulometries of cryogenic recycled rubber were used in the study;
the particles of the first one are comprised between 1.4 mm and 2 mm of size (here-
inafter "thick infill") and the second in the of 0.6 mm to 1.4 mm ("fine infill").
The following parameters were used to describe the grain morphology:
-<l>max: the !arger particle diameter in the turf rubber filling. In this study, two rub-
bers were studied: fine rubber ( 1.4 mm) and thick rubber (2 mm).
-Icor: this is the index of coordination and corresponds to the value obtained from
the division of the smallest diameter of the particles present in the infill by the largest
diameter.
Icor=--<jlmin
(1)
<jlmax
-Mg: it corresponds to the granulometric mean of the infill rubber. lt is obtained with
the following equation:
M = A% ·(<j>max+<j>min)+ B% ·(<j>max+<j>min)+ ...
g 100 2 100 2 (2 )
The equation has as many terms as different types of rubber of different granu-
lometries are included in the turf infill. The value of the percentage in weight being
added to the infill will be entered as A %, B% ...
-Dg: it is a measure of the degree of structural "disorder" measured in terms of in-
formation and corresponds to the number of parameters needed to characterise the
granulometries and the ordering of the rubber infill. This parameter can take three
values:
Dg = 1: infill is made of a single type of rubber.
Dg = 2: infill is made of various types of rubber, all mixed in the same propor-
tions in weight.
Dg = 3: infill is made of various types of rubber and in different proportions in
weight.
-Ncycles: number ofwear cycles by LISPORT.
Analysis of the Influence of Rubber Infill 353
Ten prototypes were deterrnined by statistical design of experiments using as vari-

ables the presence or not of sand, the thickness of rubber infill and the morphology
parameters (Table 1).
Prototype Sand (mm) Rubber thickness <l>max (mm) Icor Mg Dg

1 20 25 mm (thick) 2 0,7 1,7 1
2 20 25 mm (fine) 1,4 1 1 1
25 mm (50% thick-
3 20 2 1,35 1,35 2
50% fine)
25 mm (80% fine-
4 20 2 1,14 1,14 3
20% thick)
25 mm (40% fine-
5 20 2 1,42 1,42 3
60% thick)
9 NO 45 mm (thick) 2 1,7 1,7 1

10 NO 45 mm (fine) 1,4 1 1 1
Table 1. Prototypes description.
2.1 Mechanical Testing
The mechanical properties of vertical ball bounce, force reduction and vertical de-
forrnation were tested according to FIFA standards (FIFA QUALITY CONCEPT).
The tests were done on the prototypes with no wear (cero cycles) and after prototype
wearing of 200 cycles and 2000 cycles using the LI SPORT machine as described by
FIFA. (FIFA 2005). Test were done at three points on each prototype.
The parameters of study were Force reduction (FR), Standard vertical deforrnation
(St V), Maximum deforrnation (Dmax) and Bounce height (H).
2.2 Data Analysis

Data analysis consisted first in descriptive statistics of all parameters for comparison
against FIFA requirements. Then an analysis of variance (p ::c 0.05) was done to
assess the influence of the type of infill and of the presence of sand. Finally, a linear
regression analysis was done using Force reduction, standard vertical deforrnation.
Maximum deforrnation and bounce height as dependent variables and all morpho-
logical parameters and number of cycles as independent variables. The regression
equation was obtained for each dependent variable.

The results of descriptive statistics showed that not all prototypes met FIFA require-
ments showing a clear influence of the infill in field perforrnance.
The results of ANOV A presented statistically significant differences due to sand

presence (Figure 1), in a way that FR (force reduction) and StV (standard deforma-
tion), are higher in the prototypes mounted without sand (prototypes 9 and l 0) to the
extent that both are outside FIFA specifications. There were no differenccs in bounce
height and all prototypes met FIFA specifications. However, the variation of this
property with use was higher without sand.
75
70 ......................... !!! ..... ~... .
65 1111 111 1lL

~
u. 60- ~
55- -· N cycles
.................. llll o
0.6
0200
50- ~ 2000
1 1 1 1 1 1
2 10 10 1 2 9 10
Prototype Prototype Prototype
Fig. 1 Comparison FR(%), StV (mm) and ball bounce (m) with (prototipes I an 2) and with-
out sand (prototipes 9 and 10). FIFA requircmentes (horizontal lincs).
Analysis of variance also reflected significant differences for morphology parameters

(Figure 2). The compacting suffered by the prototypes with a mixture of granu-
lometries in different percentages (prototypes 4 and 5) seems to be grcater than for
the other configuration showing a greater reduction if force reduction. and vertical
deformation with wearing. In some cases, StV became lower than FIFA thresholds.
Moreover, it should be mentioned that the most stable behavior in tcnns of force
reduction and standard deformation parameters are achieved by prototypcs 2 and 3,
corresponding to the fine rubber granulate and to the 50% mixture of granulometries
respectively. The bounce height did not show differences and for the fivc prototypcs,
remains stable with wear.
75 10
lil
70
II
lll
55 N cycles
llll o
0 200
50- ra 2000
1 1 1 l 1 1 1 1 1 1
1 2 3 4 5 1 2 3 4 2 3 4 5
Prototype Prototype Prototype
Fig. 2 Comparison of prototypes with different granulometrics The three abovc graphics
compare thc parameters obtained for the prototypes made with sand and Recipncu material
using different granulometries
Analysis ofthe lnfluence ofRubber Infill 355
The results of linear regression for each mechanical property are showed below
in the form of mathematical equations:
FORCE REDUCTION (FR): The following equation predicts the value ofthe FR
parameter from four input variables with an 84, 4% ofthe variance explained. A
uniform size of rubber granules gives a high force reduction while a range of rubber
granules sizes gives low force reduction. This is probably due to the compaction of
the smaller granules between the !arger granules giving a more resilient surface. This
is confirmed by the fact that force reduction reduces with average granule size im-
plyin that smaller ranules increase the hardness ofthe surface.
FR= 72,950-(11,744· Icor)-(6,258-Mg)-(5,702·<1>max)-(0,001 · Ncycle$ (3)
STANDARD VERTICAL DEFORMATION (StV): this equation predicts the value

of the StV parameter from four input variables with an 88% of the variance ex-
plained. Increasing the disorder ofthe structure (Dg) will reduce StV without signifi-
cant influence in Force Reduction. The effect in energy restitution should be studied
as a greater disorder is ex ected to increase ener dissi ation b reor anisation.
StV= 3,663-(6,323· Icor)-(0,00012· Ncycle$-(l,625· Dg)+ (3,713· <!>max) (4)
MAXIMUM DEFORMATION (Dmax): the following equation predicts the value of

the Dmax parameter based on four input variables with an 86, 7 % of the variance
explained. The influence of disorder is similar as for STv.
1Dmax = 3,233-(5,217 · Icor)-(0,00013 · Ncycles) + (3,414 · <l>max)-(1,397 · Dg) 1 (5)
BOUNCE HEIGHT (H): The following equation predicts the value ofthe "H" pa-
rameter based on three input variables with a 71,4% ofthe variance explained. In-
creasing Disorder or the range of rubber granules sizes will increase ball bounce
height.
1H = 1,0446 + (2,48 · Icor)-(0,18 · <l> max)+ (0,059 · Dg) 1 (6)
Values of force reduction and standard deformation parameters of the prototypes
without sand under the rubber layer are outside the specifications determined by
FIFA, whereby this composition of materials requires alternative constructions or
materials if to be used to assemble artificial turf fields.
With respect to the influence of parameters which define the granulometric com-
position of the infill on FIFA variables, Anova showed significant influence of gran-
ule size in all parameters but vertical bounce whereas linear regression analysis is-
sued equation in which values of all mechanical properties can be predicted from
those parameters.
On the other hand, the mean granule size (Mg) showed a linear direct relationship
with FR value in a way that as it increases the FR is reduced without altering another
parameter. That means increasing either the size in general or the percentage of big-
ger grains.
Parameters like StV and Dmax have an inverse relationship with Dg; thus as
Disorder increases, they decrease. However, when Dg increases, vertical bounce ball
increases too. Finally, this study has showed the effect of wear in surfaces. All prop-
erties decreased when the number of cycles increased.
4 Conclusions
The results of this work provide a clear evidence that infill influences mechanical
performance of artificial turf for soccer, in fact not all prototypes met FIFA require-
ments. Wear by use (number of cycles) appears to significantly reduce the value of
all parameters that define deformation and force reduction, but not to vertical ball
bounce. The value of variables such as FR, St V and Dmax showed an inverse linear
relationship with kor, whereas it was direct with bounce height. The variation of the
kor had more influence on the StV parameter. That means that as the ratio between
the minimum and maximum diameter increases, bounce height increases and force
reduction, StV and Dmax decrease. The maximum diameter of the infill particles
(<l>max) influences the value of parameters studied, specially the value of Dmax and
StV. Thus, if it increases, StV and Dmax increase, whereas the values of FR and
bounce height reduce. Concluding, this work has showed that infill effectively influ-
ence performance of artificial turf pitches and this influence depends of granule size
to a great extent. Results show that grain morphology can be used in the design of
artificial turf infill and future studies about the role of infill "disorder" in artificial
turf performance are deemed very interesting.
Acknowledgments
The authors wish to thank the collaboration in this study of the cryogenic recycled
rubber supplier Recipneu Lda.
References
Dura, J. V., Hoyos, J. V. and Martinez, A. ( 1999) The effect of shock absorbing sports sur-
faces in jumping. Sports Engineering, 2 (2), pp. 97-102.
Dura, J. V., Martinez, A. C. and Solaz, J. (2002) Testing absorbing materials: the application
of viscoelastic linear model. Sports Engineering, 5, pp. 9-14.
Shorten, M. R. and Himmelsbach, J. A. (2002) Shock attenuation of sports surfaces. The
engineering of Sports 4. S. Uj ihashi and S. J. Haake (Eds. ), pp. 152-159.
8aroud, G., Nigg, 8. M. and Stefanyshyn, D. (1999) Energy storage and retum in spart sur-
faces. Sports Engineering, 2, pp. 173-180.
Ekstrand, J. and Nigg, 8. M. (1989) Surface-related injuries in soccer. Sports Medicine, 8 (1 ),
pp. 56-62.
Walker, C. A. ( 1996) Experimental mechanics and artificial turf. The engineering of Sports.
Haake (Eds.), pp. 239-242.
FIFA quality concept. Handbook of test methods and requirements for arti ficial turf football
soccer (2005).
Socraturf (1999). Development of improved artificial turf fields; CRAF-1999-70805.
Soccer Ball Modal Analysis Using a Scanning Laser
Doppler Vibrometer (SLDV)
Jouni Ronkainen and Andy Harland

Loughborough University, j.a.ronkainen@lboro.ac.uk
Abstract. Soccer manufacturers are investing significant amounts of time and money re-
searching and developing soccer balls, using advanced materials and constructions in an at-
tempt to create a ball that has better flight and impact characteristics. An important considera-
tion in any structure subject to dynamic or impact loading is its mcchanical response. The
recent development of non-contact optical vibration measurement tools such as thc SLDV
have made the measurcment of such rcsponses a more realistic possibility. The technique of
vibrometry utilises thc Doppler principle to provide a measure of the surface vclocity at the
point at which a laser beam is incident. The SLDV benefits from its non-contact and non-
marking method and thc speed and ease with which measurements can be recorded. This paper
reports the method and results from a study aimed at dctermining the dynamic responses of
two different soccer balls. The balls are excited using either an acoustic source or mechanical
shaker and the velocity of each ball's surface at a series of points is recorded. The natural
frequencies and vibration mode shapes are identified and a comparison is made between the
responses of each ball. Significant mode shapes were observed between 50 Hz and 1400 Hz.
At the lower frequencies the mode shapes were observed to be independent of the outer pan-
els, based more on the structure of the ball as a whole, at the higher frequencies the mode
shapes were centred on individual panel oscillations. The soccer balls tested show some no-
ticeable differences in mode shapes.
1 lntroduction
In 2006 the second largest sporting goods manufacturer, adidas, have predicted that
they will seil 10 million replica world cup soccer balls worldwide (Reuters 2005). lt
is clear to see how important it is to manufacture and market balls effectively for the
mass market as weil as achieve acceptance at the professional level.
Soccer is a global game and arguably the most popular sport on the planet. For
the 2005/06 soccer season it is forecast that within the UK the English professional
soccer clubs will have an income of approximately fl .935 billion (Mintel 2004), this
figure is huge and illustrates the vast sums of money involved within the game. Due
to the vast interest and money involved, manufacturers are spending significant
amounts of time and money researching and developing soccer balls, using advanced
materials and constructions in an attempt to create a ball that has better flight and
358 Jouni Ronkainen and Andy Harland
impact characteristics. Therefore this paper explores an alternative method for soccer
ball analysis through modal testing.
An important consideration in any structure subject to dynamic or impact loading
is its mechanical response. Traditionally these tests have been carried out by attach-
ing accelerometers onto the target structure at locations that are predicted to yield
results. However problems arise from this technique since attaching an accelerometer
onto a structure will result in a mass loading effect, which results in the Frequency
Response Functions (FRFs) being altered for the test structure. The actual attach-
ment/detachment ofthe accelerometers can cause problems and the location at which
they are attached onto the structure is crucial.
The recent development of the optical vibration measurement tool, the laser
Doppler vibrometer (LDV) allows the vibrational engineer to measure the mechani-
cal response of structures in a completely non-contact method unlike the accelerome-
ter approach. This technique ofvibrometry utilises the Doppler principle to provide a
measure of the surface velocity at the point at which the laser beam is incident. Lit-
erature explains that LDV has seen great benefit in structures that are hot, light or
rotating (Ewins 2000). A FIFA 'inspected' soccer ball weighs between 410 to 450
grams, which undoubtedly would be affected by the mass loading effect of acceler-
ometers. The SLDV allows a very high spatial density for the measured points on the
target structure and therefore removes the need for predicting mode shapes for struc-
tures. The scanning capability for LDV is achieved by utilising two orthogonal mir-
rors to direct the measurement laser beam. Generally laser measurements have re-
quired the use of retroreflective marker tape attached onto the target structure in
order for the laser interferometer to detect a signal back from the target structure.
However preliminary testing using a Polytec PSV-300 SLDV on a plain white soccer
ball, approximately 80% of the ball could be measured using the laser without ret-
roreflective marker tape, see "Fig. l ". Therefore for 80% of the ball a 'strong' signal
to noise ratio (SNR) is achieved for data analysis.
Streng Poor
SNR SNR
Fig.1. Signal back from plain white soccer ball

This study compares the mechanical response functions of two different soccer
balls in order to ascertain whether it is feasible to utilise a SLDV in order to assess
the FRFs of soccer balls and to highlight if there are any differences
2 Method
Two soccer balls were tested; a plain white 32 pane) 71anually stitched puma ball and
an adidas 32 panel thermally bonded ball. This choice allowed a typical match ball to
be compared with a novel championship ball.
The setup for the testing is pictured in "Fig.2". The ball is suspended by fishing
line attached at the valve, which allowed the known out of balance points, valve and
counterbalance, to be removed from the measurement grid. A grid of orthogonal,
equispaced points was projected onto the surface of the ball, which were used by the
Polytec software to mark the grid points for measurement by the laser. This allowed
any number of balls to be measured with exactly the same number and location of
measurement points.
B&K Conditioning Amplif1er Ban hanging by Fishing Wire
Speaker
Power
Anl>lifiar Laplop
Vibrometer
Controller
SLDV Scanner Head
Fig.2. Modal analysis setup

The puma ball was initially excited acoustically and mechanically using a shaker
via a stinger attachment. Both methods have their advantages and disadvantages the
main positive with the acoustic excitation source is that like the SLDV it is a non-
contact method and therefore avoids mass loading the structure.
The mechanical response of both the puma and adidas ball was then determined.
In order to achieve this, the soccer ball was excited by white noise in a 'check out
run' whereby the number of measurements at each point was kept to a minimum,
with an average of three measurements taken at each grid point. Once the 'check out
run' was completed the FRFs were analysed and the resonant frequencies of the
soccer ball extracted. The next stage of the testing was to pinpoint the resonant fre-
quencies andin turn excite the soccer ball at a given resonant frequency. The excita-
tion was inputted through a continuous sinusoidal wave. The Polytec software was
used to carry out a Fast Fourier Transform (FFT) working at a bandwidth of20 kHz,
measuring 6400 FFT lines, taking 320 ms to record each individual measurement
having a frequency resolution of 3.125 Hz. The measurement was carried out on a
5% rising trigger level in order to obtain useful phase information. lmportantly for
the measurements at resonant frequencies an average of ten measurements were
taken at each grid point and the measurement consisted of 283 grid points.
Since the known out of balance points on the ball were purposefully removed
from the measurement grid it was interesting to see if they actually affect the meas-
ured values. Therefore a brieftest was carried out whereby the valve was positioned
at the centre of the measurement grid.
The initial test where the puma ball was excited acoustically as well as mechanically
showed very similar results are obtained through both excitation methods therefore
to avoid duplicating work, the acoustic excitation method was used for further test-
ing. The acoustic method was chosen because the setup time is quicker and less
apparatus is required.
To ascertain the resonant frequencies the FFT graph is analysed for white noise
excitation, the graph is shown in "Fig.3." for the adidas ball. A comparison between
the puma and adidas balls showed similar resonant excitation frequencies and are
listed in Table. l.
<O
Magnitude, µm/s
Frequency, Hz
500 1000 1500
Fig.3. Surface velocity FRF for white noise excitation of adidas ball
Table.. . fiorpuma an a 1 as soccer baIIs

1 Resonant firequenc1es
Puma Resonant Frequencies adidas Resonant Frequencies
<100 Hz <100 Hz
150Hz 150Hz
284Hz, 334 Hz 278 Hz, 309 Hz, 334 Hz, 362 Hz
950 Hz, 981 Hz 950 Hz, 981 Hz
1000Hz, 1156Hz, 1175Hz 1000 Hz, 11 00Hz
1206 Hz 1359 Hz
As a general result for both soccer balls, at the lower frequencies of excitation
the mode shapes were observed to be independent of the outer panels, based more on
the structure of the ball as a whole as shown in "Fig.4." At higher frequencies of
excitation, above 500 Hz, the puma ball mode shapes were centred on individual
panel oscillations shown in "Fig.5." whereas the adidas ball was not observed to
exhibit this behaviour, presumably because ofthe difference in panel technology.
lt was observed that the stitched seams of the puma ball act as a skeleton, con-
straining the vibrations at high frequency. This is consistent with the findings of
(Price 2005), who reported the structural rigidity given by the stiff seams of a manu-
ally stitched ball.
Fig.4. Puma ball excited at 334 Hz
Fig.5. Puma ball excited at 981 Hz

Both the puma and adidas ball exhibited the same behaviour when the valve was
measured. The valve is much stiffer and therefore the maximum displacement ob-
served at this panel is much less, shown in "Fig.6.", the central panel at maximum
displaces 86 µm whereas the outer panels exhibit maximum displacement of 210 µm.
This result is anticipated, but highlights the accuracy ofthe SLDV.
Fig.6. Puma ball excitation at 981 Hz, valve located on central panel
5 Concluding Remarks
The approach utilised allowed mode shapes to be viewed graphically in three-

dimensional animations, showing interesting mode shapes for two different soccer
balls. Different responses at high frequencies were observed which are attributed to
the differences in manufacturing methods. Soccer balls are hit in transient dynamic
fashion therefore it is assumed that the observed mode shapes would be excited.
How these results correlate to match play performance would require further work.
The approach utilised would be a valuable tool in acoustics research, validation
of computer models, player safety research particularly focusing on head impacts
and as shown to test differences in behaviour due to design modifications.
Acknowledgements
The authors wish to thank the EPSRC for funding this work and the assistance of Mr.
A. Hallam and Mr. S. Carr for their help with experimental instrumentation.
References
Ewins, D.J. (2000) Modal testing: theory, practice and application, 2nd ed. research studies
press, Baldock, England, pp.194-287.
Mintel (2004), Mintel international group ltd, "The football business, UK, December 2004"
http://reports.mintel.com/sinatra/reports/display/id=68234#about, acc. 6th January 2006
Price, D.S. (2005), PhD Thesis, Loughborough, "Advanced modelling of soccer balls."
Reuters (2005), Posted 10th Dec "World Cup ball back in traditional black and white"
http://today.reuters.com/news/newsChannel.aspx?type=sportsNews, acc. 6th Jan 2006
9 Tennis
Synopsis of Current Developments: Tennis
Stuart Miller
ITF, London, UK, stuart.miller@itftennis.com
There is a substantial body of scientific research in tennis, the major growth of which
began in the late 1970s, around the same time as the introduction of the oversize
racket and the subsequent change in playing style that it facilitated. As such, rackets
constitute a significant part of the body of research. A second key element of tennis
is the ball, which, although narrowly defined by the Rules of Tennis, is sufficiently
important to warrant attention.
The Racket
A seminal comparisons of the vibration and power characteristics of conventional

and oversize racket was made by Elliott et al. ( 1980). This was followed by a com-
plete description ofthe 'sweet spot' ofthe racket by Brody (1987a). The modern
racket, being built out of aluminium, then graphite, has an inherently higher natural
vibration frequency, and there was speculation that this was linked to upper limb
injury. A series ofresearch papers have subsequently been produced that examine
the characteristics of racket vibration, including those of Tomosue et al. ( 1992) and
Hennig et al. (1992). Recently, models ofthe tennis racket have been developed,
such as those by Goodwill and Haake (2001), which have allowed the modification
of its physical parameters and, thus, its performance, on a computer. The paper by
Goodwill et al. will add to this body ofknowledge by providing the first known data
of the spin-generating characteristics of a substantial number of racket/string combi-
nations. Surprisingly little is known about the orientation of rackets at impact, a gap
in knowledge that will be addressed at ISEA by Choppin, using an advanced 30
analysis method. Rackets continue to become more sophisticated, and several con-
tain methods by which the vibrations described above can be attenuated. One such
method will be considered by Cottey and co-workers.
The Ball
Ball-related research has often centred around an understanding of its impact with
the racket (e.g., Grabiner et al., 1983), which is understandable, given the obvious
importance of this event. This line of research has been continued by authors such as
Brody ( 1987) and further developed by Haake et al. (2003 ).
366 Stuart Miller
The tennis ball has been further analysed from the perspectives of its impacts
with the court surface (e.g., Brody, 1987), dynamic rebound properties (Mill er and
Messner, 2003) and its aerodynamics (e.g., Goodwill et al., 2004). The latter area is
particularly relevant today given the increascd speed with which elite players hit the
ball and the considerable topspin applied in the majority of groundstrokes. The
power that modern players generate means that the durability of tennis balls is a key
issue, a detailed assessment of which will be given at ISEA 2006 by Steele, which
includes the consideration of factors not previously reported in the literature. Mean-
while, Honda will provide an analysis of the force and displacement histories of
dynamic impacts between a ball and a flat surface.
References
Brody, H. ( 1984). That's how the ball bounccs. The Physics Teacher, 494-497, November.
Brody, H. ( 1987a). Tennis SciencefiJr Tennis Players. Philadelphia: The Univcrsity of Penn-
sylvania Press.
Brody, H. ( 1987b). Models oftcnnis rackct impacts. lnternational Journal o/Sports Biome-
chanics, 3, 293-296.
Elliott, B.C., Blanksby, B.A. & Ellis, R., (1980). Vibration and rebound velocity characteris-
tics of convcntional and ovcrsize tcnnis rackets. Research QuarterlyfiJr Exercise and
Sport, 51, 608-615.
Goodwill, S.R. and Haake, S.J. (2001) Comparison of flexible and rigid body modelling of a
tennis racket. In: Proceedings olthe Materials and Science in Sports Symposium, San
Diego, Califomia, 223-236.
Goodwill, S.R., Chin, S.B. & Haakc, S.J. (2004) Wind lunncl tcsling ofspinning and non-
spinning tcnnis balls. Journal ol Wind Engineering & lndustrial Aerodynamics, 92, 935-
958.
Grabiner, M., Groppe!, J.L. & Campbcll, K.R. ( 1983). Resultant tennis ball veloeity as a fune-
tion of off-center impact and grip firmness. Medicine and Science in Sport and Exercise,
15, 542-544.
Haake, S.J., Carrc, M.J. & Goodwill, S.R. (2003) The dynamic impact characteristics oftcnnis
balls with tennis rackcts, Journal o/Sports Sciences, 21, 839-850.
Hennig, E.M., Rosenbaum, D. & Milani, T.L. (1992). Transfer oftcnnis racket vibrations onto
the human forearm. Medicine and Science in Sport and Exercise, 24, 1134-1140.
Millcr, S. and Messner, S. (2003). On the dynamie coefficient ofrestitution oftennis balls. In
Tennis Science and Technology 2 (edited by S. Miller), pp. 97-104. Toronto: Webcom.
Tomosue, R., Mutoh, Y., Yoshinari, K. & Kawazoe, Y. ( 1992). Vibrations of a racket handle
and the wrist joint in the tennis forehand drive. Journal ol Biomechanics, 25, 719.
Normal Impact of Hollow Balls on Fiat Surfaces
Yoshihisa Honda
Kinki University, Japan, honda@waka.kindai.ac.jp
Abstract. Normal impact of hollow balls on flat surfaces is theoretically analyzed and the
dynamic deformation of hollow balls is derived as an axisymmetric motion of an elastic
spherical shell by using modal expansion method, where variation of contact area is taken into
account. Numerical calculations were conducted and the dynamic characteristics are dis-
cussed. lt is found that the center displacement of the ball possesses nearly a half-sine wave
form while the reaction force has a triangular time history. lt is also shown that the collision
time decreases with increasing initial velocity ofthe ball.
1 Introduction
Many researchers have studied the impact phenomenon in ball games, such as golf,
tennis, baseball and so on. Stronge (2000) has mentioned that collisions in a system
of flexible structures will involve dynamic deformation of the structures. In most of
the published analytical works, clubs, racquets and bats are treated as a continuous
elastic bodies and their structural dynamics is discussed with emphasis on their vi-
bratory modes. A hollow ball is also flexible and, therefore, collisions may generate
vibratory motion in it. Hubbard and Stronge (2001) studied the bounce of a hollow
ball on a hard flat surface by a quasi-static lumped model and a finite element model.
The author examined the dynamic behavior of a hollow ball impacting at a point
against a linear spring (Honda 2004).
In this paper, the dynamic deformation of a hollow elastic ball on impact is theo-
retically studied as a continuous body. Axisymmetric motion of an elastic spherical
shell with uniform thickness is adopted as a physical model, and is analyzed by using
the modal expansion method, in which variation of contact area is taken into account.
Numerical calculations were conducted and the dynamic characteristics ofthe impact
phenomenon are discussed.
2 Theoretical Formulation
2.1 Analytical Model

lt is assumed that a non-spinning hollow ball impacts normally on a flat surface of an
elastic body as shown in Fig. 1. Therefore, the excited motion of the ball is axisym-
metric about the diameter through the initial contact point. An elastic spherical shell
368 Yoshihisa Honda
Elastic Body
w1th Fiat Surface
Elasti
Spherical
Shell
Fig. 1. Analytical modcl of a hollow elastic ball
with uniform thickness is adopted as a physical model of the ball. Both extensional
and bending deformations are considered, but damping, nonlinearity and inner pres-
sure are ignored. The density, Young's modulus, Poisson's ratio, the thickness, and
the radius ofthe shell are denoted by p, E, v, h, and a, respectively.
For convenience, the following expressions are written in a nondimensional form
on the bases of length a, mass pa2h, time a(p!E/1 2• Therefore, the mass of the ball,
which is 47Cpa2h in dimensional form, is 47C in nondimensional form. The dimen-
sionless displacement components in the normal and meridian directions are denoted
by w(~, tJ and u(~, tJ, respectively, where ris the dimensionless time.
2.2 Equations of a Spherical Shell in Axisymmetric Motion
Equations of axisymmetric motion of a spherical shell can be written as

a (Q sm~ . . . aw
. ) - (N.+NB )sm~+p„sm~=sm~ar2, 2
(1)
a~
d (N . ) Q . . . d2u (2)
a~ .sm~-NBcos~+ sm~+p„sm~=sm~ar2'
where p„ and p 11 are the extemal distributed forces in the normal and meridian direc-
tions (Kraus, 1967; Soedel, 1993). Expressions for the intemal forces denoted by N
and Q can be found in the refcrences.
The displacement components can be expressed as
~ ~
w(~,r)= Lq}(r)w)~), u(~,r)= Lq , (r:X,,)~), (3)

/' l
where W;{ ~) and U;( ~) are the natural mode components. These mode components
can be given by
(4)
Normal Impact of Hollow Balls on Fiat Surfaces 369
where the nonnegative integer n depends on modal number j, and Pn( ) is the Legen-
dre function of ordern. The expressions for coefficients Cj and other parameters are
given in the reference (Honda 2004).
The modal participation factors q/ r) satisfy the following differential equations.
d2
M ~+K q = F (r) (5)
' dr 2 1 .1 1 '
F,(r)= 2,r f(p"W' + p,,U)in(/xl</J, (6)

where ~ and Kj are the equivalent mass and the equivalent stiffness of the j-th mode,
respectively, given by
M J =N }2 ' K=N'Q'
J j } J
(7)
N2 = 21rf(w2 + U2 )sin</Xf</J = _!!!_[1 + C2 n(n + !)} (8)

' i i 2n + 1 '
2.3 Impact on a Fiat Surface
The impact force is the interaction between the ball and the impacting or impacted
body, and depends on their dynamic characteristics. In this analysis, simplified char-
acteristics are assumed for the impacted body with a flat surface. The axis of symme-
try of the ball is selected as the x-axis of a Cartesian coordinate system and that the
flat surface is located at x= 1. The x coordinate of the surface of the ball at </! can be
expressed by
x = (w+ i)cos(/J-usin </J, (9)
lt is assumed that the restoring force applied by the impacted body on the ball is
directed along the x-axis and that the magnitude of this force is proportional to the
positive compression x-1. The following equations for the restoring force hold during
the collision.
p" =-k(x-l)cos(!?, p,, =k(x-l)sin(,?, (10)
where k is constant, which represents the simplified elasticity of the impacted con-
servative body.
3 Numerical Results and Discussion
Numerical calculation was carried out for a shell of h/a=0.02, v=0.3. The infinite
series of Eq. (3) were truncated to include only the modes whose frequencies are
lower than 50. A pair of modes with n=I have same equivalent mass 47fand one of
the pair which has zcro frequency is the translational mode. Below 50 of frequencics,
there are 89 bending modes and 48 stretching modes. The lowest nonzero frequency
is 0. 73508 (n=2; bending). The initial conditions are that the shell
370 Yoshihisa Honda
0 10
t 008
c.,
006
"'~
::ä
'-
0 0.04
c.,
E
...g
Q.
002
"' 000
i5
-002
0 10 20 30 40 50
20E-3 ---
.,u
15E-3 J
5
g10E-3
C:
0
50E-4
00E+0
0 10 20 30 40 50
20
15
~J
~
.,
eo
C: 10
<
J;l
c0
5
0
0 10 20 30 40 50
Time
Fig. 2. Time histories (h/a=0.02, v=0.3, k=I00)

Normal Impact ofHollow Balls on Fiat Surfaces 371
0.10 2.5E-3 20 - -.---,---r-------;r-------;, - , 50
00.08 2.0E-3 ~
ei,
"'"' "'
f
'--- .,
"'O
"; 15
ci)
~ 0.06 1.5E-3 e < "'
.-.S
C:
5 C
"- ti
E ,:: ö
.g
.!E
u C: "'
'Ei
"
~0.04 1.0E-3 ~ u
0
u
0
0 :ä = ~ - f---1------+l--'ll-lf-----+----+-30
E
~10
:;)
E
·x
·;;; 0.02 5.0E-4 ~"'
~
"'
5 ..................._ _ _ _ _ _ _ _ 20
0.00 ....,......,.....,.....,.....,....,.....,.....,.....,..._._ O_OE+O
0.000 0.002 0.004 0.006 0.008 0.010 0.000 0.002 0.004 0.006 0.008 0.010
Initial elocity
Fig. 3. Dependence on initial velocity (h/a=O.ü2, v=0.3, k= 100)
contacts the flat surface at the point <fr=O with a uniform velocity v0 and that the shell
and the surface are in a free state. The calculated coefficient of restitution (COR) is
greater than 0.997 for v0<0.01 and k= 100. Under these conditions, the residual vibra-
tion of the shell after impact is negligibly small and the COR is almost equal to
unity.
Figure 2 shows the time histories of the displacement of the mass center of the
ball, the impact force and the contact angle for various initial velocities. By compari-
son of the time histories, The displacement of the ball possesses a half-sine form
while the force possesses a more triangular form. The contact time is much longer
than the longest natural period of the shell's modes (8.5477 for n=2; bending). The
contact area grows rapidly in the initial stage of the impact.
As the initial velocity increases, the displacement, the force and the angle become
!arger, and the contact time becomcs shortcr. The incremental or decremental rates of
the displacement, the angle and the contact time to the initial velocity gradually de-
crease. On the other hand, the incremental rate of the force to the initial velocity
gradually increases. Figure 3 shows the dependence of those values on the initial
velocity. While the initial velocity rises from 0.001 to 0.01, the maximum values of
displacement of the mass center of the shell, the contact force and the contact angle
are magnified by 5.58, 21.8 and 3.15 respectively, and the contact time is shortened
to 54.3%.
Figure 4 shows an example of the relation between the displacement of the mass
center of the shell and the contact force during impact. As the disp\acement in-
creases, the incremental rate of the force grows up. The relationship between them
represents nonlinearity.
372 Yoshihisa Honda
O.OE+OO ..,.c;;;;,a,._ _1,_,_....__ ___.i...._..,__ ___._ _....__...,1__..,__ __,

0 0.02 0.04 0.06 0.08 0.1
Displac ment of enter
Fig. 4. Relation between displacement and force (hla=0.02, v=0.3, k=l00, v0=0.01)
4 Conclusions
In this paper, the dynamic behavior of a hollow ball during normal impact against a
flat surface was analysed as an axisymmetric motion of an elastic spherical shell
using a modal expansion method. lt is found that the center displacement of the ball
possesses nearly a half-sine form while the reaction force possesses triangular form
in time histories. lt is also shown that the collision time decreases with increasing
initial velocity of the ball.
References
Honda, Y. (2004) Impact Analysis ofHollow Elastic Balls. In: M. Hubbard, R. D. Mehta and
J. M. Pallis (Eds.), Engineering of Sports 5, 1. ISEA, Sheffield, UK, pp. 344-350.
Hubbard, M. and Stronge, W. J. (2001) Bounce ofHollow Balls on Fiat Surfaces. Sports
Engineering, 4, 49-61.
Kraus H. (1967) Thin Elastic Shells. Wiley, New York.
Soedel W. (1993) Vibration of Shells and Plates (2nd ed.). Marcel Dekker, Inc., New York.
Stronge, W. J. (2000) Impact Mechanics. Cambridge University Press, Cambridge.
Factors in Tennis Ball Wear
Carolyn Steele, Roy Jones, and Paul Leaney
Loughborough University, C.Stcele@lboro.ac.uk
Abstract. Recent rcsearch has shown wcar to affect ball perforrnance during play, but con-
tributing factors to wear differences have not been weil identified. A framework has been
developed to aide a systematic investigation into these factors. Several of these features are
investigated in this paper, including repeated impacts, racket impact conditions, precipitation,
and natural pressure loss. The results from these investigations are intended not only to give
players a better understanding of what occurs during play, but also enable manufacturers and
goveming bodies to address specific areas of importancc in ball wear and perforrnance.
1 Introduction
Ball wear is becoming an increasingly popular area of tennis research as the proper-
ties of new balls are weil documented. Recent research has investigated the effects
of ball surface condition on aerodynamics (Mehta and Pallis 2001 ), repeated impacts
on ball mass loss (Capel-Davis and Miller 2003), and the coefficient of restitution
(Miller and Messner 2003). While various stages of ball wear can be observed
through play and simulated in laboratory experiments, these have yet to be incorpo-
rated into a structured study identifying the individual causes of differences in ball
condition and performance.
This paper presents a framework for identifying key areas contributing to wear
differences with the intent to aide manufacturers and goveming bodies in improving
ball wear characteristics and monitoring the effect of wear on the game. Factors
such as racket impact conditions, repeated impacts, precipitation, cloth construction,
and natural pressure loss are discussed in this paper.
2 Framework Development
A framework for investigating the causes of wear in tennis has been developed
through extensive interviews with elite players. These interviews encompassed both
the causes and effects of ball wear, as weil as implications to ball performance, and
were conducted at several professional toumaments where players were asked to
consider not only their current training and playing environment, but also experi-
ences throughout their career. These interviews were combined with a review of
literature on ball characteristics and the game of tennis to provide a structure for the
causes ofball wear (Steele 2005). This framework is presented in Fig. 1.
374 Carolyn Steele, Roy Jones, and Paul Leaney
/lJlitude
Preapitati cn
Tem perature
Fig. 1. Framework used to assess factors in ball wear.
3 Play
Wear through play is perhaps the most obvious cause to changes in ball properties.
Two components of this factor are discussed in this paper: racket impacts and re-
peated impacts.
3.1 Racket Impacts
Stringbed properties and impact speed were identified as important characteristics of

racket impacts. Fora given racket, string type and tension are variables players can
modify to suit their preferences. Elite players are known to use high stringing ten-
sions, favoring increased control over the power found with lower tensioned strings
(Brody, Cross, and Lindsey 2002). Players also employ a range of string types with
differing properties, such as monofilament polyester and natural gut. Additionally,
racket impacts have been influenced by recent improvements in racket technology
and improved player fitness contributing to increased ball speed during play.
These three factors were investigated using an air cannon to fire two premium
grade, sateen weave balls (for each impact condition) at a rigidly clamped, oblique
racket head. This method was used to isolate differences in racket properties as the
air canno~ itself causes no significant changes to ball mass during testing and the
ball was not allowed to impact another oblique, abrasive surface that may have also
affected mass loss characteristics.
Figure 2 indicates the results of this work, with ball speed causing the most no-
ticeable effect on mass loss. Stringing tensions did not produce mass loss differ-
ences, though natural gut caused higher mass losses than synthctic gut, while polyes-
ter caused less.
Role of impact speed on ball mass loss Role of string type on ball mass lass
----- 25m/s
1.2 - ♦ - 35m/s 1.2
§ 0.8
]
1 0.6
0.4 0.4
--e-- Synthetic
0.2 - • - Natural gut
♦ Polyester
o~--~-~---~-~--~-~
100 150 200 250 300 350 400 450 500 50 100 150 200 250 300 350 400 450 500
Number of impacts Number of impacts
Role ot string tension on ball mass lass
1.2
0.4
0.2 ·
o~-~----~-~-----~~
50 100 150 200 250 300 350 400 450 500
Number of impacts
Fig. 2. Effects of racket properties on changes in ball mass.
Ball Core COA with Repeated Impacts Dtfferences in mass 109S for wet samples
-+-water Resistent Treated Cloth

7 - • - Regular Cloth
0.53
----•---
0.52
f
0.51
a:
8
0.5
0.49
0.48
100 150 200 250 300 350 400 450 500

Number of impacis Number of Impacts
Fig. 3. Effect of repeated impacts on Fig. 4. Effect of cloth treatment on

core COR. mass changes during wet testing.
3.2 Repeated Impacts

Players often comment on the 'loosening up' of the ball, or feeling of increased ball
speed and lower stiffness, during the early stages of play as rubber is known to
stiffen during extended rest. A wear rig was designed to cyclically wear the balls at
30m/s with two oblique impacts (a racket and steel plate). To eliminate changes to
ball fuzziness and cloth stiffness during testing, ball cores were used to isolate
changes in the rubber. The coefficient of restitution (COR) for the balls was evalu-
ated for sixteen impacts at 30± 1m/s off a steel plate using a high speed camera run-
ning at 2,000 fps.
Figure 3 shows the mean COR, plotted with one standard error. While there
appears to be slight increase in ball COR through 40 impacts, the variation in stan-
dard error makes it difficult to form any definitive conclusions. The minimal differ-
ences also suggest that while the ball may not travel significantly faster, changes to
ball stiffness may improve ball 'feel' for players.
4 Natural and Environmental Factors
Additional components of ball degradation are those occurring due to innate differ-
ences in ball construction, such as core and cloth differences, and environmental
conditions such as a temperature and precipitation. A cyclical wear machine with a
rotating racket was used to project balls towards two oblique surfaces covered in ball
feit. The system averaged 33.9 (a=l.76) ball impacts per minute with a racket head
velocity of 14.63m/s (and approximate ball speed of 55m/s).
4.1 Precipitation
Given no current methodology exists for evaluating wet tennis balls, a standardized
process from a ball cloth manufacturer was used. Balls were submerged in water for
five minutes, bounced ten times to release excess water, and then placed in the wear
machine.
Three balls were evaluated at each impact lcvcl. The first ball tested was a pres-
surized, sateen weave ball, while the second was the same ball produced with a water
resistant dye. Figure 4 shows ball mass loss during testing, with the original weight
being that of the ball immediately before it was placed in the machine. While both
balls show similar rates of mass loss with repeated impacts, the untreated ball shows
significantly higher initial losses due to the !arge amount of water absorbed by the
ball while it was submerged. This figure also shows the success of the water resis-
tant dye at minimizing ball weight increases due to moisture.
4.2 Cloth Construction

A similar investigation was perforrned with a needle feit ball and a premium grade
sateen weave ball. The needle feit ball showed significantly lower mass losses
through extended testing, as shown in Fig. 5. These results differ from those pre-
sented by Capel-Davies and Miller (2003) that show greater losses from a needle feit
ball and comment on the continued linear trend in mass loss through 1000 impacts,
but the slowed rate of loss over extended impacts for a sateen weave ball. These
differences could be due to the wearing mechanism, impact speed, or ball brand and
highlight the need for standardized wearing procedures and methods.
___
-
l.oN
~
,..
2s;,--:::-::=--==---=-=-:===:;-------i
SaJNll'l--ft
---c.--
-- --
„ 15
!
1
,•
.• ,,
1
••
•• 100 MO
11'tUfof~
300 400 500
Fig. 5. Effect of cloth type on changes in ball mass.
,_
---1M4K
........
-• -
130 - - 3WNkl
130 - t2moAlhl
- - - punctured
125
,20
115
____________
110 a:,;
.
,••• ...__
0 10
Sr.andard Almoapherc Pinsure
30
_,
03
,~~.-~20-~..--,~,--:,;-~••-~.,--'~
~ ~ ~ ~ ~ ~ ~ llnOaaSOMdlmlSl
o-,s rrom opening
Fig. 6. Interna! ball pressure for aged balls. Fig. 7. Role ofnatural pressure loss on
the coefficient of restitution
and intemal ball pressure.
4.3 Natural Pressure Loss

Interviews and testing with elite players has highlighted their preference for pressur-
ized balls, citing improved feel and performance (Davies 2005). Previous work has
used punctured balls as an approximation of the changes to ball characteristics, but
Fig. 6 indicates that through three months of natural ageing, intemal pressures are
still noticeably higher than standard atmospheric pressure. The high pressure gradi-
ent present when the balls are first opened is responsible for the high rate of initial
pressure losses that slows with increased time.
With regards to the dynamic COR for aged balls against a rigid surface, Fig. 7
indicates two groupings of aged balls - those under three months in age and those six
months and older (including the punctured sample). These results also indicate that
punctured balls approximate some aged balls in dynamic impact testing.
5 Conclusions
The framework developed in this paper provides a structured method to examine and
classify differing contributions to ball wear. Initial experimental work indicates that
some factors play prominent roles in ball wear, with future work planned to investi-
gate remaining areas of the framework such as court surface interactions. The fol-
lowing conclusions can be drawn from the completed studies:
• Impact speed affects ball mass loss, making wear more noticeable to elite
players using higher ball speeds. Mass loss is affected by string type but
not string tension.
• Core COR did not conclusively indicate repeated impacts may cause the
ball to play faster. Future work is planned to investigate any changes to ball
stiffness that may occur and the subsequent relationship to player feel.
• Needle feit balls showed lower mass losses than sateen weave balls.
• Water resistant ball feit can significantly reduce initial water absorption,
though the rate of loss with impacts is consistent.
Acknowledgements
The authors would like to thank Dunlop Slazenger for supporting this work and the
Wolfson School ofMechanical and Manufacturing Engineering for providing labora-
tory facilities.
References
Brody, H., Cross, R., and Lindscy, C. (2002) The Physics and Technology ofTennis. Racquet
Tech Publishing, Solana Beach.
Capel-Davies, J. and Miller, S. (2003) Durability oftennis balls worn in a test rig. In: S. Miller
(Ed), Tennis Science & Technology 2. International Tennis Federation, Roehampton, pp.
113-122.
Davies, G. (2005) Determination and analysis of dimensions of 'feel' in tennis ball impacts.
PhD Thesis, Loughborough University.
Mehta, R. and Pallis, J. (2001) The aerodynamics ofa tennis ball. Sports Eng., 4, 177-189.
Miller, S. and Messner, S. (2003) On the dynamic coefficient ofrestitution oftennis balls. In:
S. Miller (Ed), Tennis Science & Technology 2. International Tennis Federation, Roe-
hampton, pp. 97-104.
Steele, C. (2005) Tennis ball degradation. Interna! Research Report, Loughborough Univer-
sity.
Measuring Ball Spin Off a Tennis Racket
Simon Goodwill 1, Jamie Douglas 2, Stuart Millet\ Stephen Haake 1
1 Sports Engineering, CSES, Sheffield Hallam University, s.r.goodwill@sheffield.ac.uk

2 International Tennis Federation
Abstract. A series of experiments were carried out in which a spinning tennis ball was pro-
jected obliquely at a head clamped racket. In each experiment a different string was used in
the racket frame. An automated image analysis algorithm was used to measure the velocity
and spin of the ball. lt has been found that, for the majority of impacts, the magnitude of the
ball rebound spin was dependent on string stiffness. lt has been concluded that the string
stiffness influences the magnitude of stringbed lateral movement during impact. This differ-
ing magnitude oflateral string movement, in turn, influences the amount ofball rebound spin.
1 Introduction
The research presented in this paper has been conducted in collaboration with the
International Tennis Federation (ITF). The ITF are responsible for the rules of ten-
nis, and there is currently no ruling which specifies the magnitude of spin which a
tennis racket can impart on the ball.
The absence of such a ruling could potentially lead to a situation similar to that in
1977 when a novel 'spaghetti' stringing system was introduced (Fischer 1977). This
racket had the inherent ability to apply considerably more spin onto the ball com-
pared to a conventionally strung racket. The receiver could not predict the flight or
bounce of the ball, and so gave the player who was hitting the ball an unfair advan-
tage over the receiver. In 1978, the ITF banned this stringing system.
In this paper, an experimental investigation of the oblique impact between a
tennis ball and head clamped tennis racket is described. The aim of this study is to
determine whether the string type (material and stiffness) affects how much spin is
applied to the ball. Previous published research (Goodwill and Haake 2004a) has
suggested that the string type does not affect the rebound spin. However, it should
be noted that Goodwill and Haake only tested a small sample of strings. In this cur-
rent study a significantly larger sample size was used in an attempt to identify a
relationship between string type and ball rebound spin.
2 Methodology
Identical racket frames were strung at a tension of 270 N (60 lbs) for each of the
thirty tennis strings that were tested. Each racket was strung 24 hours prior to being
380 Simon Goodwill
tested. All the strings that were tested in this study were manufactured from either
polyester or nylon (or some composite based on one ofthese materials).
2.1 Impact Test Experiment

In this experiment, new pressurized tennis balls were projected at a nominal velocity
of 25 m/s (56 mph) on to a head clamped tennis racket, as shown in Fig. I (a). Tests
were conducted for a range of inbound ball spins between zero and 400 rad/s (3820
rpm). The ball impacted with back spin and rebounded with top spin. This simulates
the change in spin which occurs in a typical top spin ground stroke.
The racket clamp was designed such that the racket could be rotated to give any
relative impact angle The clamp pivots asymmetrically to ensure that the ball impact
is always centered on the longitudinal axis of the tennis racket. In this experiment,
the ball was projected at both 40 and 60 degrees to the normal, to simulate two dif-
ferent top spin shots. These angles were based on work by Choppin (2005) who
measured maximum relative impact angles of38° for good leisure standard players.
A Phantom v4.0 high speed camera, operating at 1000 fps, was automatically
triggered every time a ball was projected using the bowling machine. The camera
was located 8 meters from the impact, and a typical image is shown in Fig. I (b ). The
camera automaticall fbitma file on to a la to P .
Fig. 1. (a) Test rig setup showing the two different racket clamp settings used and (b) typical
high speed video image.
2.2 Automated Image Processing

Bespoke image analysis software has been written which automatically calculates the
velocity and spin of the ball, before and after impact. For each impact, the camera
captured a blank image before the tennis ball entered the frame. This was defined as
the base image. Each subsequent image was then subtracted from the base image
using a standard image subtraction algorithm in Matlab v7. This image matrix was
then filtered to remove noise, and processed to detect the position of the tennis ball.
Three mutually perpendicular lines were drawn on the ball, to allow the spin to
be measured. The algorithm which was used to determine the ball spin is more com-
plex and a füll description of this method is beyond the scope of this paper. The
algorithm was written in Matlab v7, and based on Hall (2002). lt is assumed that the
Measuring Ball Spin Offa Tennis Racket 381
ball rotates purely around an axis that is perpendicular to the image plane. The basic
function of the algorithm is to determine how much the ball rotates between a pair of
consecutive images. Pattern matching is used to rotate the second ball image in the
sequence so that it has the same orientation as the first image. The average rotation
of each pair of images is used to determine the ball spin. The uncertainty in the
measured velocity and spin values has been determined to be less than 2%.
2.3 String stiffness

The main aim of this study is to determine whether string type affects how much spin
is applied to the ball. However, defining a string solely by its material type does not
differentiate between the wide range of, for example, polyester strings available.
One parameter of a string that is commonly used to characterize its properties is the
string stiffness (Lindsey 2002, 2004). The stiffness of the strings used in this study
has already been obtained by Lindsey, and those values are used here. The string
stiffness values quoted by Lindsey were measured on a single string, using the
method described in Cross (2001).
3.1 Results
400 400
-;;;-
i (a)
'i:I
g,300
(b)
~300
3 Polyest r 8
%200
Nylon
4 ·[200
r:.
et, Polyest r
., ✓,# .,
,, ,:, 4°0
C:
5 100 ,oo1j~w C:
5 100 x/-Nyton
.,
D .,
D
a:: a::
0 0
0 100 200 300 400 0 100 200 300 400
String stiffness (lbf111) Strilg stiffness (lbf111)
-.,
'i:I
400
(c) x"(fl
• ,:olyester -;;;-
'i:I
~300
400
(d)
~300 Nylon
3 3
%200 t200
,,.,
.,
C:
,,
C:
5 100 5 100
.,
D .,
D
a:: a::
0 0
0 100 200 300 400 0 100 200 300 400
Strilg stiffness (lb/in) Strilg stiffness (lbf111)
Fig. 2. Test rig setup showing the two different racket clamp settings used.
382 Simon Goodwill
Tests were perfonned using inbound spins that ranged from zero to 400 rad/s. How-
ever, for simplicity, only the rebound spin results for inbound spins of I 00 and 400
rad/s are presented here. These rebound spin values are plotted against the appropri-
ate string stiffness values. Figure 2 (a)-(b) show the results for the ball impacting at
40° to the nonnal. lt can be seen that all the polyester strings were stiffer than the
nylon strings .. In both Figs. 2 (a) and (b) it can be seen that the ball rebounds with
highest spin from the rackets strung with polyester material (inbound angle 40°).
However, in Fig. 2 (c) the rebound spin is essentially independent ofthe string stiff-
ness (inbound angle 60°, inbound spin 100 rad/s ). Furthennore, in Fig. 2 (d) it can
be seen that the rackets strung with the polyester strings give the lowest rebound
-~
spins (inbound spin = 400 rad/s, inbound angle= 60°).
2·5 100 rod/1 0

0
2 ~
0 O
.2
oi 1.5 o PolyHter
Polyesw xK,l-
a:: Nylon
NytonK*;
%1
Cl)
... ··········••··· ... ·1 c ~· --~ ·- "'·· ... ...............................x···"··"····· .......................... .
0.5 0.5
0 +---.----.------,..------l 0 +---.----.----..----
0 100 200 300 400 0 100 200 300 400
String stiffness (IMn) Strng stiffness (lb/in)
2.5 ~ - - - ~ ------~ 2.5 ..-:-:,:,--,---.....,-.,,-------,

(1)
2 2 ' ~ (/)
,
.2
oi 1.5 !1 .5
a:: Nylon ~ Polyester a:: Nylon ( ~
%1
Cl)
·······•·•··············•······ .......... ..""............... . %1 ...................................">f"-••••••••.••••••••••••••••••••••••
Cl) XX
0.5 0.5
0 +---.------,..---~..---- 0+---..---..---,----
0 100 200 300 400 0 100 200 300 400
Strng stiffness (lb/in) Strng stiffness (lb/in)
Fig. 3. Test rig setup showing the two different racket clamp settings used.
lt is difficult to identify the reason for the varying dependency between the string
stiffness and the rebound spin using only the data in Fig. 2. However, the ball re-
bound velocity and rebound ·angle were also measured in this study. The actual
values of the ball rebound velocity and angle are not presented here, but the results
can still be used to further our understanding of the impact mechanism. This is done
by considering the mode of the ball immediately prior to it leaving the surface. The
ball can either be in sliding or rolling mode, and the spin ratio is used to define
which mode the ball is in (Goodwill and Haake (2004a)). Thc spin ratio (SR) pa-
rameter is defined as,
SR= m.r/V(x) (1)
Mcasuring Ball Spin Off a Tennis Racket 3 83
where OJ is the rebound spin, r is the ball radius (0.033 m) and V(x) is the horizontal
component ofthe rebound velocity.
Previous researchers have shown that SR can be,
( 1) Less than unity - the ball slides throughout impact.
(2) Equal to unity - the ball rolls off the surface.
(3) Greater than unity - the ball rolls during contact, and then leaves spinning
faster than rolling.
The spin ratio values for the impact tests in this study are shown in Fig. 3. The main
point to note from this figure is that, for the majority of the impacts, the spin ratio is
greater than unity. lt can therefore be concluded that there is sufficient friction be-
tween the ball and stringbed to cause the ball to roll at some point during the impact
which is consistent with the findings of Goodwill and Haake (2004). The only ex-
ception to this finding is shown in Fig. 3 (d). This figure shows that, for impacts at
60° and 400 rad/s, the spin ratio for the polyester strings is significantly lower than
unity. This is interesting as it implies that the ball slides throughout impact on the
stiffer strings.
3.2 Discussion
There are two major findings in this work which are,
1. For all impacts at 40°, the stiffer polyester strings impart more spin com-
pared to the nylon strings.
2. For impacts at 60° and 400 rad/s, the polyester strings impart less spin.
For all the impacts at 40°, Figs. 2 (a)-(b) clearly show that the balls rebounding from
the polyester strings have a higher spin. Intuitively it might be concluded that the
higher spin attained by the polyester strings might be due to a higher friction coeffi-
cient. However, the friction coefficient for virtually all the strings is high enough to
induce rolling, and any value higher than this critical value does not increase the
magnitude of the rebound spin (Daish 1972).
The reason for the higher spins off the polyester strings is likely to be linked to
the magnitude of the lateral stringbed deformation. This lateral deformation occurs
because, during the compression phase of the impact, the ball grips the stringbed and
causes the strings to deform laterally in the direction of the ball motion. The impor-
tance of this deformation in terms of the spin generation, is the recovery of the
stringbed during the restitution phase. The stringbed will attempt to recover in some
part to its original position. This causes the stringbed to move in an opposite direc-
tion to the motion of the ball. This stringbed motion acts to increase the relative ball-
surface velocity. If the string is gripping the surface, then this motion will accelerate
the ball rebound spin. The results in Figs. 2 (a)-(b) appear to suggest that the stiffer
polyester strings are able to utilize this mechanism more efficiently. This may be
because the strings are stiffer and therefore more able to recover. Altematively, it
may be because the stiffness of the strings will determine how fast the string can
recover, and the stiffer strings recover in the most efficient time period.
The second finding (based on the results in Fig. 2 (d)) appears to contradict the
conclusion discussed above. This is because the balls rebound off the polyester
string with less spin. However, the results in Fig. 3 (d) highlight the point that the
384 Simon Goodwill
ball is sliding off the polyester strings (SR < 1) and therefore the impact mechanism
is different for these impacts, compared with those described above.
The lateral motion of the strings will now be reconsidered in an attempt to under-
stand why the stiffer polyester strings now give less spin. In the impacts presented in
Fig. 2 (d) the ball will have a tendency to want to travel a long lateral distance (due
to the shallow impact angle) and want to slip (due to the high inbound spin). In this
extreme case the ball will attempt to deform the stringbed laterally, but the stiffer
polyester strings will not be able to deform as much as the nylon strings. Therefore
the impact on the stiffer strings will be more analogous to an impact on a rigid sur-
face where the ball will be subjected to more sliding during the impact. So instead of
the polyester strings gripping the ball and deforming laterally during impact, their
high stiffness prevents this lateral deformation, and the ball has no option but to slide
across the surface. The consequence of this is that, unlike in the other cases, any
recovery of the lateral deformation during the restitution phase is less significant.
The nylon strings have a greater ability to deform during the restitution phase, and
therefore the ball is subject to less sliding on these strings.
4 Conclusions
lt has been found that, for balls inbound at 40 degrees to the normal of the string
plane, the stiffer (polyester) strings give more spin. However, for balls inbound at 60
degrees, the stiffer strings generally give less spin. There is a considerable amount
of anecdotal evidence that top professional players are using stiff polyester strings.
In this study it has been shown that they will achieve more spin with polyester string,
if the relative ball-racket impact angle is 40 degrees. This extra spin allows them to
hit the ball harder, whilst keeping the ball in play, thus increasing their chance of
winning the point.
References
Chop~in, S., Goodwill, S.R. and Haake, S.J. (2005) 3D Player testing. In: Procecdings ofthe
61 International Conference on the Engineering ofSport (ed F. Moritz & S.J. Haakc).
Cross, R. (2001) Stretch tests on strings, Racquet Tech, September, 12-18.
Daish, C.B. ( 1972) The Physics of Ball Games. English Universities Press, London.
Fischer W. (1977) Tennis Racket, US Patent 4273331, 8th December 1977.
Goodwill, S.R. and Haake, S.J. (2004a) Ballspin generation for oblique impacts with a tennis
racket, Experimental Mechanics, 44(2), 195-206.
Goodwill, S.R. and Haake, S.J., (2004b) Effect ofstring tension on thc impact betwcen a
tennis ball and racket. In: Procecdings ofthe 5th International Conference on the Engineer-
ing ofSport (ed M. Hubbard, R.D. Mehta & J.M. Pallis), 2, 3-9.
Hall, D. (2002) Three-dimensiona/ reconstruction_ji-om p/anar s/ices,
http://www.mathworks.com/matlabcentral/fileexchange/loadAuthor.do?objectType=autho
r&objcctld=982 l 74
Lindsey, C. (2002) String Selector Map, Racquet Tech, February, 4-8.
Lindsey, C. (2004) String Selector Map 2004, Racquet Sports lndustry,1(7), 24-29.
3D Player Testing in Tennis
Simon Choppin 1, Simon Goodwill 2, Stephen Haake 2
1 Sports Engineering Research Group, University of Sheffield,

mep04sbc@sheffield.ac.uk
2 Sports Engineering, CSES, Sheffield Hallam University
Abstract. Although qualitative shot analysis and rudimentary 2D player testing has been
performed in thc past, a comprehensive 3D study has yet to be done. This paper outlines a
method that has been used to record player baseline shots and serves in 3D. The method al-
lows accurate tracking of rackct velocity (any point on racquet), ball vclocity, impact instant,
impact position, and all associated angular velocities. Details of the methodology used in
obtaining recorded shots are described, as weil as the planar/vector calculations used to obtain
the required information from thc recordings. The movement of racket and ball were consid-
ered just prior to, and post impact, but testing is not limited to this case. Two Phantom high
speed cameras were uscd in the analysis at l000 frames per sccond. To date, testing has been
performed on recreational, to county level players with a mind to extend thc testing in the
future to world ranked professional players.
1 Introduction
Player testing is an important tool, and has its place in sports science, engineering
and biomechanics. To date, and with tennis analysis firmly in mind, photogrammet-
ric player testing has generally been performed in 2D at low (<200fps) frame rates
with a specific aim, whether this be some definition of player accuracy (Blievemicht,
1968), or more recently, studying advanced player kinematics (Knudson and Black-
well 2005). There has also been some notable 3D work performed using the DLT
method on serve (Elliott, Marsh and Blanksby 1986) and backhand (Elliott, Marsh
and Overhue 1989) strokes. This work is biomechanics based and is limited due to
the technology and frame rate used at the time.
The method proposed in this paper focuses primarily on the impact; the racket
and ball movement just prior to, and post impact. Analysis is concentrated on the
movements of the racket and ball only, biomechanical movements are not consid-
ered. This method varies from previous work, in that instead of obtaining quantitive
measurement from the photos directly, specifically marked points on the racket are
used to set-up a plane, and a point in space in the case of the ball. With this informa-
tion it is possible to accurately track racquet velocity (any point on racquet), ball
velocity, impact instant, impact position, and all associated angular velocities. The
386 Simon Choppin, Simon Goodwill, Stephen Haake
advantage of this method, is that unlike previous methods, velocities are not limited
to a single tracked point, or singularly considered axis of rotation. lt also allows
further in-depth 3D analysis should the need arise (an example being the instantane-
ous rotation matrix and helical axis of rotation (Spoor and Veldpaus, 1980) of racket
movement) the key being that this method does not limit the target objective of the
testing, as long as it is grounded in racket/ball dynamics.
The testing performed to date has used recreational to county level players to
refine and test the methodology, with a mind to move on to professional, ranked
players in the future. The testing has been developed as a validation exercise to de-
termine typical racket head speeds, impact angles and impact positions, for use in
future testing.
2 Methodology
All testing was performed using stereo videogrammetric methods and on a standard
size, outdoor tennis court (although it should be noted that provisions for indoor
testing have been made). Players were situated at the baseline in a calibrated 2x2x2m
volume and recorded performing a variety of shots, the balls were fired from a re-
peatable air-cannon into the control volume, and all shots that landed within the
court boundary were recorded.
A checkerboard calibration technique was used to define 3D space (Choppin,
Whyld, Goodwill and Haake 2005), with a set of global axes defined as in fig.1
Because this method used only two cameras, placement of the cameras, type of
markers used (5 markers are needed on the racket), operating speed of the cameras,
and the markers position on the racket are all vital aspects to this method, and a füll
methodology review was carried out to ensure the correct choices were made.
: ' :. : .·.: 0 1 0 1 1 1 1 ·: I •
... ..... , .. ...... ·. : ..

0
.-
y
Baseline
Fig. 1 Diagram of global axes in relation to court

30 Player Testing in Tennis 387
2.1 Markers
Set-up
The purpose ofthe markers is not only to define the racket face as a plane, but also to
define a set of local co-ordinate axes at each instant, for these reasons, it is vital that
a minimum of three markers are visible to define the plane(two of which must be
used for axes co-ordination)
A reflective tape is used to create markers 20-25mm wide at five points on the
racket face, markers 1-3 can be used to create the local axes set (figure 2.1.1 ), mark-
ers 4 or 5 are used if one of the markers 1-3 is not visible at any one instant.
A tape type marker is used so that it is visible from both sides of the racket. Ideally,
spherical markers surrounding the frame of the racket would allow the markers to be
seen from most orientations and also allow accurate tracking of the markers (a
spherical object's centre can always be found in 20 image tracking). lt was decided
that tape markers - because of the minimum alterations to the appearance, and weight
of the racket - would be the best in terms of gaining accurate results from player
testing, and the one or two mm discrepancy in tracking accuracy was a worthy trade
off.
Plane and local axes generation

The plane ofthe racket is defined algebraically as:
ax + by + cz + d =0 (1)
Where (x,y,z) is any point on the surface of the plane, and the perpendicular vector
to the plane is given as (a,b,c). Each vector is defined in terms ofthe global axes set
set-up at calibration. The plane equation can be obtained from three points which
exist in that plane. If three of the markers global co-ordinates can be given as:
(2)
P2(X2,Y2, 2 2) (3)
P3(X3 y3,Z3) (4)

The coefficients can be found quite simply from the knowledge of these three
points(Anton)
Local co-ordinates are set-up using by defining the axes as unit vectors from the
position oftwo ofthe three markers (figure 2.) the z-axis is always normal to the face
of the racket, hence:
ai + bj + ck
z=----- (5)
a2+b2+c2
Error in axes generation:
Error testing on both the data collection and calculation stages of analysis shows the
planar/vector model for racket and ball analysis to be robust and sufficiently accurate
as to be insignificant in terms of error. For example, a point tracked on the racket
directly from the images is typically around ±0.25mm from the same point tracked
via the planar model using initial position and velocity vectors. The accuracy of the
calibration method concurs with previous findings (Choppin, Whyld, Goodwill and
Haake 2005).
1
Fig. 2 Local racket axes and their relation to the racket markers
2.2 Cameras
The cameras were placed at either end of the net and focused on the centroid of a
2x2x2m volume situated at the baseline, from this orientation all the markers are
visible at the point of impact and for 30 or so frames either side. A camera situated at
the side of the player may have a clearer indication of the balls position in the global
x-direction, but most of the markers are hidden around the time of impact, making it
impossible to define the racket plane. The cameras were run at 1000 fps, at racket
speeds of around 30 m/s, the racket is moving around 30mm between frames and this
was found to give an accurate analysis. Daylight provided sufficient light such that
shutter speeds of 100µs produced no blurring or distortion of the images.
2.3 Analysis - Examples

Using the method described in section 2.1, the plane and local co-ordinate set is
defined for each instant as weil as the point co-ordinates of the ball in each of these
instants. Any point on the racket can be recreated at each instant (for example, the
centre of mass, or the tip of the racket) assuming the racket travels linearly in the
period of analysis the velocity vector of that point on the racket can be calculated.
The balls velocity is assumed linear in x and z directions, but 2 nd order polynomial in
the vertical. Separate tests checking the validity of these assumptions, showed them
tobe valid.
The impact point of the racket can be calculated by using a bisection method to cal-
culate the time at which the perpendicular distance between the ball and plane is a
minimum. If a set of points are carefully selected on the racket, it is also possible to
ascertain the rotational speeds around certain axes with simple rigid body dynamics,
3D Player Testing in Tennis 389
for these to have any particular meaning, velocities must be transformed into the
local co-ordinate set. If the ball is clearly marked, it is also possible to calculate 3D
ball spins (Tamaki, Sugino Yamamoto 2004), with translations and rotations for ball
and racket, a füll 6 degree of freedom model can be generated from which the possi-
ble racket/ball observations possible becomes vast.
3 Results
Analysis shows that, a local axes set produced from markers 1 and 2 is typically less
then 0.1 ° misaligned from a set produced from markers 2 and 3. Racket markers can
be repeated to within 2mm, and the ball within 1mm, meaning at l000fps velocity
uncertainties of±2ms· 1 and ±lms· 1 respectively.
Single Player Analysis:
The results (shown in Table 1) for a single player are given below as an example of
typical data collected. (Player plays ortce a week at recreational level)
6 Shots in total
Playing Angle Racket Speed (ms- 1) Impact accuracy(mm)
Average Max Average Max Average StD

26.6° 38° 24.45 29.59 43.0 15.8
Table 1 Results of a single player analysis
Playing angle:
The angle at which the racket impacts the ball, taking into account racket and ball
velocities and the vertical angle of the racket at impact.
Racket speed:
The speed of the centre of mass of the racket immediately prior to impact.
Impact accuracy:
Distance in millimetres of the impact point from the stringbed centre.
4 Conclusions
4.1 Analysis and metnodology

Thorough testing of both the method and accompanying analysis method has proven
them to be a reliable, accurate and versatile player testing procedure. The apparatus
used in testing does not present any intrusions or distractions for the players, who are
able to play on a standard tennis court indoors or outdoors. The testing can be per-
formed in a controlled 'set-shot' way, or used to record particular shots during a

game. The errors measured suggest that the model assumptions are fair for this ap-
plication.
4.2 Player Testing

To date, the player testing has adequately shown the effectiveness of the methodol-
ogy, as well as helped refine certain aspects of its execution. The results obtained so
far help to assess the characteristics of play for lower level players and perform sepa-
rate analysis involving (for example) grip characteristics during impact, and valida-
tion of existing racket models.
4.3 Further Development

The methodology and analysis method itself is at a usable stage of refinement and
any further work would involve specialising the calculations for a specific purpose.
In terms of player testing, to build a thorough cache of players' shots, up to high
level professional players, will provide a !arge amount of highly relevant, useful
data.
Acknowledgements
I would like to thank the ITF for continued support throughout this project, and the
players which have given their time and skill towards the continued understanding of
the game of tennis.
References
Anton. Lines and Planes in 3 space. Elementary Linear Algebra, (2000), Wiley. 8th Ed: 149-
151.
Blievernicht.J.G. (1968). "Accuracy in the Tennis Forehand Drive: Cinematographic Analy-
sis." Res. Q. Ex. Sport 39(3): 776-779.
Choppin.S.B., Whyld.N.M., et al. (2005). "3D Impact Analysis in Tennis." The Impact of
Technology on Sport 1(1): 373-378.
Elliott.B.C, Marsh.A.P, et al. (1986). "A Three-Dimensional Cinematographic Analysis ofthe
Tennis Serve." Internationaljournal ofSport Biomechanics 2(4): 260-271.
Elliott.B.C, Marsh.A.P, et al. (1989). "The Topspin Backhand Drive in Tennis: A Biome-
chanical Analysis." The Journal ofHuman Movement Studies(l6): 1-16.
Knudson.Duane.V. and Blackwell.John.R. (2005). "Variability ofimpact kinematics and
margin for error in the tennis forehand of advanced players." Sports Engineering 8(2): 75-
80.
Spoor.C.W. and Veldpaus.F.E. (1980). "Rigid Body Motion Calculated From Spatial Co-
Ordinates ofMarkers." Journal ofBiomechanics 13: 391-393.
Tamaki.T, Sugino.T, et al. (2004). "Measuring Ball Spin by Image Registration." The 10th
Korea-Japan Joint Workshop on Frontiers ofComputer Vision: 269-274.
An Extended Study Investigating the Effects of Tennis
Rackets with Active Damping Technology on the
Symptoms of Tennis Elbow
Robert Cottey1, Johan Kotze 1, Herfried Lammer 1 and Werner Zimgibi2
1 HEAD Sport AG, Kennelbach, Austria, r.cottey@head.com

2 Praxisklinik für Orthopädie und Sportmedizin, München
Abstract. The aim of this research was to determine what effect an active damping tennis
racket technology had on players suffering with symptoms of tennis clbow. The study was
conducted to verify findings of prcvious research, which concluded that the symptoms of
tennis elbow had been dramatically reduced by playing with a Head rackets containing the
Head Chipsystem™ (Kotze et al. 2003). A similar study over an extended period was com-
pleted to further substantiate these findings and to test the improved generation of 'Chip'
rackets. This study used two versions of the Head Protector Oversize tennis rackets; both
containing piezo ceramic fibres integrated with the elcctronic Chip system™, but only half
with the chip "active", thus providing a control. The subjects were male and female experi-
enced tennis players diagnosed with either acute or chronic tennis elbow. They were given
unspecified rackets to facilitate a blind study, and the subjects' elbow condition was medically
assessed and recorded over an extended period of time. Results of the study indicated that for
the players who were initially diagnosed with acute tennis elbow, a !arge improvement in their
condition was recorded for those using the active rackets, whilst the players with the control
rackets showed little improvement in their condition. Similar results were found for the play-
ers diagnosed with chronic tennis clbow although to a lesser extent; those using rackets with
active chips showed an overall improvement, whilst the players with the control rackets again
showed little sign of improvement. The results of the study have shown that an active damp-
ing technology, when applied to a tennis racket, can reduce the symptoms of both acute and
chronic tennis elbow.
1 Introduction
Tennis elbow (Lateral Epicondylitis) is the most common (Pluim and Safran 2004)
and investigated injuries in tennis and although different forms of the injury also
occur in other walks of life, it has predominantly been connected to tennis. Tennis
elbow is defined as the occurrence of micro-fractures in the tendon attaching to the
lateral epicondile of thc forcarm. The fractures cause swelling of the joint and inten-
sive pain to the player, resulting in anything from slight discomfort to complete
debilitation (Roetert et al 1995). In an extensive investigation, by Cooke et al (2002),
into the current knowledge related to the injury, it was concluded that the nature of
the injury is fairly weil understood and thoug there were still many uncertainties
regarding the exact cause the gencral consensus was that it is caused by repetitive
392 Robert Cottey et al
impact loading. Hence manufacuturers have developed various technologies focused

on reducing the effect of impact loading on the arm.
These systems have traditionally all been passive damping systems, which means
the mechanical vibration energy is dissipated via some damping material or mecha-
nism in the form of material friction or heat and have never been proven to have any
real effect on the actual injury. Head sport therefore went a step further in developing
rackets with an active damping system in order to find a more affective damping
solution. In an active damping system the mechanical deformation energy is con-
verted into electrical energy, which is stored and released back into the material such
that it actively damps the vibration, without using any extemal energy. The first
series of rackets were tested during an independent study, during which carefully
selected subjects, suffering from tennis elbow, were given the Intelligence i.X16 and
i.Sl8 rackets (Fig. la and Fig. lb) to play with as per usual, while being examined at
the start and the end of a six week research period and their progress recorded (Kotze
et al. 2003). Results from the tests were reassuringly positive. As a result, the design
of the rackets was further improved in order to develop a racket specifically aimed at
players with tennis elbow, which resulted in the new HEAD Protector series. From
this new series the Protector OS (Fig. lc) was selected for another stringent set of
medical testing. As before, the players diagnosed with either acute or chronic tennis
elbow participated in the testing but this time in a more conclusive random, double
blind, placebo controlled study to determine the effectiveness of the new rackets on
their symptoms.
(a) (b) (c)

Fig. 1. Active damping piezo rackets from HEAD Sport: (a) i.Sl8, (b) i.Xl6, (b) Protector OS.
An Extendcd Study Investigating the Effccts ofTennis Rackets with Active Damping 393
2 Experimental Method
Strict selection criteria were used to detennine which of the possible candidates were
used in the study; these comprised a preliminary medical examination, being within a
certain age bracket, and play tennis at a predetennined ability level.
The preliminary mcdical examination was conducted and candidates clinically
diagnosed with epicondylitis humeri radialis and had the condition for at least 6
weeks were suitable for the study. lt was also specified that the candidates had to
have been free of any treatment for a minimum of I 2 weeks prior to the start of the
testing, and could not receive any subsequent treatment for their symptoms; this
treatment included any injections or other therapeutic remedies. Candidates who
were diagnosed with nerve compression syndrome, instabilities of the shoulder /
elbow or if they suffered from arthrosis of the humeroulnar joint, were not allowed to
participate in the study. Candidates had tobe between the ages of 18 and 70 to pre-
vent some degenerative factors influencing the rcsults. The candidates also had to
have at least 3 years tennis playing experience.
Male and female players were randomly selected from thosc who met thc prereq-
ms1te requirements. A total of 102 experienced tennis players from southem Ger-
many, 58 men and 44 women were chosen to complete the study. The group had an
average age of 51.2 years and the average playing experience was described as
"Good".
At the start of the investigation each participant was 'prescribed' a Head Protec-
tor Oversize tennis racket as the sole treatment for their condition. They were re-
quested to reduce activities such as gardening, handiwork, or other manual and el-
bow-burdening activities. They werc also asked to retain their usual tennis schedule
using only the racket assigned to them at the start of the testing. Any other treatment
of their conditions such as the use of injections or other therapeutic medicine was
restricted. All activitics that the participant completed were documented for the
testing period.
The sclected participants werc dividcd randomly into two groups, and a clinical
investigation of these two groups was completed using an x-ray and an ultrasound
test. The medical examinations were supervised by Dr. med, F.Söller, a scientific
employee of the Orthopädie Klinikum Großhadem at the Ludwig-Maximilians Uni-
versität München. The medical examination enabled further classification ofthe two
groups into sub-groups detennined by the diagnosed condition of the participants,
either acute or chronic tennis elbow. Acute epicondylitis symptoms was the classifi-
cation for those who had the painful symptoms for less than 3 months, and chronic
epicondylitis defined as those who had suffered with the symptoms for more than 3
months.
Two methods used to evaluate the severity of the pain are presented in this paper,
one using the scoring method of Broberg and Morrey ( 1986) and the other using the
classification of the Mayo Elbow Perfonnance Index (in Morrey 1993). Both tests
comprise of an objective and subjective part and have been used to compare the
participants' symptoms before and after completion of the test period. The objective
comprises of motion, stability and strength or ability to perfonn tasks, and the sub-
jective part comprises of an evaluation based on an interview with the patient. For
the purpose of evaluating the participants elbow condition prior to and after the test-
ing period these methods are suitable.
A statistical analysis of the results was completed to determine if the difference
between evaluations made before and after the completion of study were statistically
significant. A two sample t-test was completed allowing the spread of data to be
evaluated and to determine if there was a statistically significant improvement of the
patients' condition.
Doktoranden Herrn F. Dörfler defined the protocol for the participant evaluation,
which consisted of an intemal examination of the affected elbow using Sonography
(with a 7.5MHz linear sound head) and radiography. The examinations were com-
pleted at the Orthopädischen Klinik und Poliklinik Großhadem which has a !arge
orthopaedic department. Follow up examinations were at 6 and 12 weeks. The
results taken were recorded and scored on a scale indicated by using the Borberg and
Morrey and the Mayo Clinic Performance Index for the elbow. After each examina-
tion the tennis rackets were re-strung to the same tension, providing further control
of the testing conditions.
The Head Protector rackets were divided according to the original random group-
ing, both groups were given identical looking rackets, however one set of rackets did
not contain the active chip system, and so acted as a control group. The participants
were not aware of the type of racket they were given for the testing. Only Dr Söller
had prior knowledge of which type of racket each participant received and had no
personal contact with any of the participants at any time during the testing.
The careful organisation of the groups and distribution of the rackets ensured that
this was a double blind study and would show any effect that playing with a racket
with the active chip system would have on the symptoms oftennis elbow.
3.1 Mayo Elbow Performance Index

The group that was diagnosed with acute epicondylitis showed an improvement in
their condition for both the participants using the control rackets and the participants
using the active rackets. Both distributions of the scores after 12 weeks showed a
statistically significant increase, however the difference between the test results
showed that for the participants using the rackets with the active chip system the
increase was almost double, from an initial rating of 'poor' with an average score of
51. 7 to a final rating of 'excellent' average score of 97. 7 compared with the control
group rising from 'poor' with an average of 46.0 to a final rating of 'fair' and a score
of 68.2 (Fig. 2a).
The group diagnosed with Chronic epicondylitis increased from 'poor' 51. 7 to
'good' 87.5 for the participants with the active chip system compared with an initial
rating of 'poor' average of 47.8 rising to a final rating of 'poor' average of 55.6 for
the participants using the control rackets (Fig. 2b ). For the chronic test results the
rise in the score for the two groups were both statistically significant but the differ-
An Extended Study Investigating the Effects ofTennis Rackets with Active Damping 395
ence between the rate of change was much greater for the participants that used the
,~r •
Boxplot of Initial, and score after 12 w ks Boxplot of Initial, and sc:ore after 12 weeks
uslng the Broberg and Money system using the Broberg and Morrey system
Partlopents dlagnosed wrth Aalte Ep,a,ndyltlls Particpents d,aono.ed wrth Olrornc Ep,a,ndyl
1
lnlbal 12 weelcs lnlbal12 w
+ 1,~1____
+ _,____..
1
In 1 12 weelcs lnrtlal 12 wee1cs

1
With Chip Without Chip With Chip Wlthout Chip
~ (b)
Fig. 2. The scoring distributions for the Mayo scoring for the participants diagnosed with
acute (a) and chronic (b) epicondylitis.
3.2 Broberg and Morrey Scoring System

The examination of the participants and the scoring of their condition according to
the Broberg and Morrey scale showed similar results. The participants diagnosed
with acute epicondylitis using the rackets with the active chip system scores in-
creased from an average rating of 'fair', score of 62.6 to a final rating at 12 weeks of
'good' 88.7, there was no change in category for the participants with the control
rackets, starting at 'fair' with an average of 61.4 the final score after 12 weeks re-
mained at 'fair' with an average of 69.0 (Fig. 3a). Statistically the results for the
rackets with the active chip was significantly different and there was no difference in
the condition for those patients with the control rackets.
Similar results were observed for the participants diagnosed with chronic epi-
condylitis. The group using the rackets with the active chip scores rose, although the
categories did not change, starting at 'fair' with an average of 60.4 remained so with
an increased score of 72.6 after 12 weeks. The control racket participants also
showed no change in the category 'fair' but a smaller increase in the average scores
was recorded from 62.6 to a final score of 65.3 (Fig. 3b). A statistical analysis of
these scores for the participants diagnosed with chronic epicondylitis showed that
only the group that used the rackets with the active chip system had a statistically
significant increase in their average scores even though the categorical ranking did
not change.
Boxplot of Initial, and score after 12 w ks Boxplot of Initial, and score after 12 weeks
-_1 ·~1
uslng the Mayo Elbow Performance Index uslng the Mayo Elbow Performance Index
Partlapenl:$ d,agnosed wrth Acute Epccndyl,ns Partiapenb d1agnosed wm, Olrcn,c Epccndylltls
1·~~.____ 1
lnrbal 12 weeks IMJBI 12 wedcs

1
lnrbal 12 weeb lnrbal
~
12 wedcs
With Chip Without Chip Witt! Chip Wiölout Chip
(a) (b)
Fig. 3. The scoring distributions for the Broberg and Morrey scoring for the participants diag-
nosed with acute (a) and chronic (b) epicondylitis.
4 Conclusion
By using a random, placebo controlled, double blind study the true effects of using a
HEAD tennis racket with an active damping EDS cari be accurately assessed. These
results confirm the previous study (Kotze et al., 2003) that the symptoms of tennis
elbow can be statistically and significantly reduced with the use of this specially
designed tennis racket. Participants who used the HEAD Protector tennis rackets
equipped with the EDS showed a significant improvement in their diagnosed condi-
tions for both the acute and chronic form of tennis elbow.
The results for this extended study have yielded statistically significant results
with a clear reduction of the complaint of tennis elbow for those participants using
the rackets with the chip system compared to those using the control rackets.
References
Broberg, M. A., Morrey, B. F. (1986) Results of delayed excision ofthe radial head after
fracture. J Bone Joint Surg Am, 68 pp 669-74.
Cooke A. (2000). An overview ofracket technology. In: Tennis Science & Technology (Ed.
by S.J. Haake & A.O. Coe), pp. 43-48. Blackwell Science Ltd, Oxford, UK.
Kotze, J., Lammer, H., Cottey, R., Zimgibl, W. (2003) The Effects ofactive piezo fibre rack-
ets on tennis elbow. Tennis Science and Technology 2. Edited by S. Miller. Published by
the ITF.
Morrey, B. F. (1993) The elbow and its disorders, 2nd edition, (edited by Morrey) Philadel-
phia: Saunders.
Pluim, B., Safran, M. (2004) From Breakpoint to Advantage. A Practical Guide to Optimal
Tennis Health and Performance. Racquet Tech Publishing, USA.
Roetert E.P., Brody H., Dillman C.J., Groppe! J.L., and Schultheis J.M. (1995). The biome-
chanics oftennis elbow: an integrated approach, Clinics in Sports Medicine, 14 (1), 47-57.
10 Watersports
Synopsis of Current Developments: Water Sports
Jani Macari Pallis, Ph.D.

Cislunar Aerospace, Inc., deke@cislunar.com
Water sports continue to provide an arena of innovation for the sports engineer. The
mixed fluid medium of air and water and the often harsh operating environment
provide challenges in the design, instrumentation, technology implementation and
material science for these sports.
Water Sports and the Sports Engineer
Engineers in sports, recreation and fitness have the same goals as other sports profes-
sionals: enhance performance; prevent injury; assure safety; increase enjoyment and
health benefits; support longevity, accessibility and diversity (to participate through-
out the human life cycle regardless of physical challenge). Clearly, these papers
presented on water sports exemplify each ofthese objectives.
This series of work covers a wide range of water sports: water skiing, rowing,
kayaking, surfing, swimming and white water rafting. Equally diverse is the athletic
ability and situation of the sportsperson studied: world ranked and Olympic athletes
(both male and female), individuals retuming to or beginning a sport after a serious
physical injury and those being rescued during water sport participation. The tech-
niques utilized and the technical subject matter of each paper is likewise distinctive:
füll body scanning and computational fluid dynamics, testing of equipment for
strength and durability, modeling of forces, instrumentation for performance moni-
toring, development of software analysis tools and use of optimization methods in
design.
In the following paragraphs the works of these authors (several just beginning
their careers in the field) are summarized.
The indoor paddling biomechanics of the 2004 Italian women's Olympic kayak
team was analyzed by integrating data from both a motion capture system and in-
strumented footpads, seats and paddles. Paddle trajectory, trunk and limb motions
(shoulder, pelvis, trochanters excursions, range of knee flexion) as weil as force
measurements from a dynamometric footpad were extracted. Motion symmetry and
regularity were correlated to an athletic ranking system and demonstrated that the
"best" athletes performed regular paddle trajectories and steady trunk motion - in-
formation which can be key criteria for an evaluation system for trainers.
To provide an alternative to on-water performance analysis systems for rowing,
an on-land feedback device using rowing ergometers was developed and tested by a
400 Jani Macari Pallis, Ph.D.
member of the Austrian national rowing team. Portable units using load cells and
strain gages wcre developed to measure reaction forces at the foot stretcher. On-land
and on-water results were correlated to validate that similar reaction forces were
indeed observcd. Additionally, a monitor which displayed a history of key kinetic
parameters provided feedback to the rower during on-land training.
Computational Fluid Dynamic software was used to analyze the aerodynamics of
a water ski jump for British Water Ski. Using the geometry of a ski jumper, water
skis, fins, bindings, helmet and tow handle, sevcn key, characteristic positions were
simulated to obtain changes in lift and drag throughout the jump on the components
modeled. Mid-flight results dcmonstrated that the ski jumper's body accounts for
about 33% of the total lift generated, stressing thc important of proper position.
Recent equipment failures during simulated white-water rescues have demon-
strated the need for detailed mcasurement of loads created and which can be safely
sustainable by equipment currently used for white water rescue. Areas explored
included the potential forces involved in a white-water rescue, the forces a three-
person rescue team generates, an analysis of suitable ropes for white-water rescue
and an analysis of the current mechanical advantage rescue techniques. Conclusions
in each of these areas are dtawn including information on the types of rope fiber
materials that should be utilized.
As part of a !arger project directed at facilitating design of fins and surfboards for
manufacture, a Computer Aided Design tool has been developed to facilitate 3-
dimensional design of surfboard fins. The tool also provides thc basis for in-depth
studies through the use of stress analysis and Computational Fluid Dynamics (CFD)
software, to provide insight into potential design and material modifications. Drag
and lift forces predicted by the CFD were fed into a Finite Element Analysis (FEA)
to obtain displacements of the fin undergoing these hydrodynamic forces.
A swimming aid for individuals with upper-arm amputation was developed using
optimization methods. The effect on the front crawl was analyzed by simulation
since the upper limb motion generates the most thrust. Researchers developed a
swimming prosthetic to compensate for the body imbalance created by the missing
limb. An optimization method was used in the design and an initial trial test was
confirmed by an experiment.
The Future of Sports Engineering in Water Sports
Water sports have the added complexity of a mixed fluid medium (air and water)
which raises the bar in terms of engineering degree of problem difficulty. The envi-
ronment of thcse sports can be quite harsh. However, water sports continue to attract
both individuals and families and are enjoyed through the entire life cycle of a per-
son even as their bodies age or physical ailments or disabilities develop. The sports
engineer will continue to develop or utilize the newest innovations in computing
technologies (both hardware and software), material science, emerging technologies
(wireless and nanotechnology) and MEMS (Micro-Electro-Mechanical systems). As
our population ages and to increase market share efforts will continue to attract
broader populations to the sport through engineering innovation.
Computational Fluid Dynamic Analysis
of a Water Ski Jumper
John Hart, David Curtis, and Stephen Haake

Sheffield Hallam University, Sports Engineering, CSES, John.Hart@shu.ac.uk
Abstract. Water ski jumping is one of the oldest disciplines in water skiing. The first jump
was performed by Ralph Samuelson of Minnesota (US) in 1925, three years after he had in-
vented water skiing. Samuelson jumped 18 m off the end of a greased ramp. Today water ski
jumping is an international sport with elite male athlctes jumping distances in excess of 70 m.
The Sports Engineering Research Group (SERG) at the University of Sheffield have con-
ducted a Computational Fluid Dynamic (CFD) analysis of the aerodynamic system of a water
ski jumper for British Water Ski (BWS) in support of their 2005 World Championship cam-
paign in China. The geometries of the water ski jumper and associated equipment were created
using SERG's in-house non-contact laser scanning facilities. Seven characteristic positional
stages were analysed over the ski jump to obtain information on the fluctuations in lift and
drag force acting upon the water ski jumper. The individual contribution of lift and drag, to the
overall aerodynamic system of the water ski jumper, from each modeled component could be
determined by the use of CFD. This indicated that the skis generate an average of 65 % of the
entire system lift and drag, with the front third ofthe ski's creating up to 50 % ofthese forces.
1 Introduction
SERG (Sports Engineering Research Group) were approached by British Water Ski
in the Autumn of 2004 to investigate the aerodynamic system of a water ski jumper.
This was tobe conducted in advance ofthe water ski 2005 World Championships in
Tianjin, China. The objective of the investigation was to identify areas of the current
system where improvement was achievable, with the provision of possible solutions.
Although the aerodynamics of Nordic ski jumping has been investigated, (Seo,
Watanabe, and Murakami 2004; Virmavirta, Kivekas, and Komi 2001 ), the aerody-
namics of water ski jumping is an area devoid of research. This is despite the fact
that water ski jumping was invented nearly 80 years ago when Ralph Samuelson of
Minnesota (US) performed the first jump. Similarities however can be drawn be-
tween the two disciplines. Water ski jumpers try to manipulate their skis into a V
flight style as used by their winter counterparts. However due to the dynamics ex-
perienced as a water ski jumper leaves the top of the jump ramp and climbs through
the jump this is much harder to achieve. Water ski bindings are also rigidly fixed to
the ski with no articulation, meaning that the jumper can not Jean out as far over the
skis as a Nordic ski jumper.
Differences exist in the ski design between the two disciplines. Whereas Nordic
jump skis have a fixed maximum width of 11.5 cm, and a maximum length set as a
function of jumper height, water skis have no maximum length, however width is
402 John Hart, David Curtis, and Stephen Haake
limited to 30% ofthe ski length. Water jump skis commonly do not have a constant
curvature (rocker) from the tip to the tail. The front third of the skis are angled up-
wards and tumed out. This design feature of the ski is known as the Stokes tip. The
ski also has a short fin attached to the underside at the tail.
The aerodynamics of Nordic ski jumpers has also been investigated using CFD
(Asai, Kaga, and Akatsuka 1997). However these studies have used simplified hu-
man geometry. In this current study SERO intended not only to investigate the aero-
dynamic performance of the ski, but provide as detailed a description as possible of
the aerodynamic system around and over the ski jumper. SERO therefore used non-
contact laser scanning techniques to capture as realistic a human geometrical form as
possible.
2 Geometrie Model
The basic modeled geometry, (Fig. 1), consists of ski jumper, "Connelly" water skis
and appropriate fins, bindings, a "Jofa" sky diving helmet, and tow handle. The ge-
ometry was acquired using SERG's in-house non-contact laser scanning facilities.
Scanning was performed using a ModelMaker X70 scanning system, to generate an
initial point cloud representation of the geometric components. The point cloud data
was then converted to a fully water tight NURBS (Non-Uniform Rational B-Splines)
model, using Raindrop Geomagic Studio, that could be manipulated within a com-
mercially available CAD (Computational Aided Design) package.
Fig. 1 Modeled ski jumper geometry showing both handle grip styles
lt was not possible to use an actual ski jumper to provide the human geometry
used in this investigation due to time constraints. Instead an anatomically correct
flexible mannequin was used which could be manipulated into a characteristic jump
position. The mannequin upper torso was scanned twice, to obtain two different hand
holds of the tow handle. During the jump flight the ski jumper alters the hand hold,
from a double grip to single hand grip. The tow handle geometry was created using
CAD.
The individual scanned geometries were assembled in to a single model within
the Fluent pre-processor Gambit. The identities of each component were preserved
Computational Fluid Dynamic Analysis ofa Water Ski Jumper 403
however to enable the individual contribution to drag and lift of each component to
be determined.
3 Computational Model
Computational mesh were constructed using Fluent's Gambit and TGrid mesh gen-
erators. Each mesh consisted of approximately seven million tetrahedral, prismatic,
and hexahedral cells, concentrated in regions of detailed geometrical interest. The
prismatic cells were constructed over the entire surface of the modeled geometry to
ensure that surface boundary layers were adequately captured.
The CFD code Fluent 6.1 was used to perform the simulation. This solved the
goveming equations of fluid motion sequentially, with turbulence closure provided
by the realisable k-E turbulence model used in conjunction with a non-equilibrium
wall function model. All goveming equations were discretised with 2nd order interpo-
lation schemes.
The modeled jump velocity and directional components were determined using
the trajectory model, as detailed in the next section, and applied appropriately. Simu-
lations were performed using 8 processors on a custom built Linux cluster, and con-
verged results were obtained after a runtime of approximately 12 hrs. Post-
processing was performed using Fluent to obtain lift/drag forces, and all graphical
output using CEI's Ensight8 software.
4 Modeled Jump
The jump has been modeled in stages as a selection of snapshots in time as agreed
with BWS. lt is not yet possible to model a water ski jump from start to finish in a
smooth transition using CFD, due to the geometrical changes that take place over the
jump. BWS providcd data of a jump performed by Jason Sccls (2004 & 2005 Euro-
pean Jump Champion), detailing key gcometrical angles of the ski jumper and skis
over time, (Fig. 2).
Fig. 2 Modeled jump angles

Using the BWS data, seven characteristic positions were chosen, including key
reference positions; off the ramp, change of handle grip, and final position prior to
landing. The ski jumper adjusts the tow handle grip from both hands to a single hand
hold after approximately one second in the modeled jump. The scanned geometries
were adjusted accordingly to the required angles using CAD, (Table 1).
Stage Time Jump Angles (°) Vx Vv X y

(s) a p 'Y 15 (m/s) (m/s) (m) (m)
1 0.00 29.40 85.10 54.10 15.50 29.01 11.72 0.00 1.80
2 0.45 37.80 79.70 38.50 35.65 28.22 7.31 12.87 6.08
3 1.00 28.40 79.70 30.80 55.80 27.24 1.91 28.11 8.62
4 1.33 22.90 74.60 23.40 55.80 26.67 -1.28 36.86 8.72
5 1.78 19.20 64.20 19.10 55.80 25.87 -5.69 48.67 7.15
6 1.98 13.20 62.00 10.30 24.00 25.52 -7.65 53.81 5.82
7 2.45 25.80 95.00 58.30 12.00 24.68 -12.31 65.72 1.08
Table. 1 Modeledjump conditions
4.1 Trajectory
Detailed data of a water ski jump flight trajectory does not currently exist. In
particular there is no data conceming the velocity vector of the ski jumper at each
stage of the flight. To obtain this information a standard projectile motion model has
been used. A constant horizontal deceleration has been assumed over the jump based
on existing jump data. This was determined from the known take off velocity, the
length of the jump, and the final velocity required to cover the distance in the time
jumped. During a jump a ski jumper not only travels forward but also to the side due
to the approach path made towards the jump ramp. This side ways movement has
been omitted. The water ski jumper hits the ramp with an initial velocity of 70 mph,
and the distance jumped is 67 m over a time of 2.55 seconds. The ramp has a slope
of22° and a takeoffheight of 1.8 m. The flight conditions, as shown in Table 1. were
applied to the simulation. Where V x and Yv are respectively the horizontal and
vertical velocity components, X and Y is the horizontal and vertical displacement of
the ski jumper.
5 Results
The predicted ratio of lift/drag force (L/D) for the aerodynamic system of the water
ski jumper is shown in Fig. 3. LID is seen to increase rapidly between stages 3 and 4
as the jumper maneuvers the skis into the characteristic V flight style. Maximum LID
Computational Fluid Dynamic Analysis ofa Water Ski Jumper 405
is found to occur at stage 5 when the ski angle 8 is a maximum and the jumpers body
is leaning out over the skis. This position can only be maintained for a short duration
however as the jumper is already realigning their skis by stage 6 in preparation of
landing. LID therefore decreases rapidly.
-• M
•
l'I
• •• lfl
• •
,0 ...•
2.1
l l l l l l l 10.0
,.o
2.0 ....,,,,...
_
I.O
•.. ..·" 7.0 I

1.1 -- - 7
1•
.
I.O
! ...•♦•• 1.0 z
... •
1.0 ,.o
I.O
0.1 2.0
-UD 1.0
0.0 0.0
0.0 0.1 1.0 1.1 2.0 2.1 a.o
FllgM Tlllle(Nc)
Fig. 3 Predicted LID, and vertical height of ski jumper, over jump
100
..
~
0
:J
90
80
70
■ Drag Sk is ~ Drag Jumper ■ Lift Skis Lift Jumper
60
{!.
ij 50
40
~
·c 30
5 20
,,.
u 10
0
S1age 1 Slage 2 Slage 3 Stage 4 Sl119e 5 Stege 6 Siege 7 Average
Fig. 4 % Contribution to lift and drag at each jump stage
Fig. 4. shows the contribution to lift and drag ofthe skis andjumper at eachjump
stage. lt was found that over the entire jump the skis account for 68 % of the lift
generated, with the Stokes tips accounting for 50 % ofthis force. The body ofthe ski
jumper generates an average of 31 % of total lift, clearly demonstrating the impor-
tance of the ski jumpers posture.
As would be expected regions of high pressure were observed to form on all
leading surfaces of the ski jumper. Regions of high pressure were also observed to
form on the inside edge of the ski, and the underside of the Stokes tips. This was due
to the inclined attack angle of the ski, causing the tips to strike the atmosphere
obliquely. Consequently a significant region of low pressure was formed behind the
Stokes tips, indicating the possibility of a large flow separation.
Flow was indeed observed to separate behind the Stokes tip, Fig 5, with the for-
mation of a large vortex core, from stage 4 of the flight. This vortex core separates
cleanly away from the ski due to the uptumed angle of the Stokes tips. Regions of
separating flow were also observed from the ski jumpers rear, limbs, and helmet.
Flow separation over the ski jumpers body can be seen in the oil flow plot in Fig.5.
This shows how air flow moves over the body, with the spiral pattems indicating
regions where vortex cores originate.
Fig. 5 Flow separation from skis & oil flow plot over ski jumpers body
6 Conclusions
SERG has conducted a CFD analysis of the aerodynamic system of a water ski
jumper for British Water Ski. Seven characteristic positions have been analysed over
the jump, to obtain information in the fluctuations of lift and drag. lt was found that
the skis generate an average of 69 % of the entire system lift, with the Stokes tips
creating up to 50 % of these forces. Large regions of flow separation are seen to form
behind the Stokes tips. The ski jumper's body generates an average of 31 % of the
total lift force, highlighting the importance ofbody posture.
References
Asai, T., Kaga, M., and Akatsuka, T. (1997) Computer Simulation ofthe V-style Technique in
Ski Jumping using CFD. Proc. 6th Int. Symp. Computer Simulation in Biomechanics. To-
kyo, Japan, pp.48-49
Seo, K., Watanabe, I., and Murakami, M. (2004) Aerodynamic Force Data for a V-Style Ski
Jumping Flight. Sports Engineering. 7, 31-39
Virmavirta, M., Kivekas, J., and Komi, P.V. (2001) Take-off Aerodynamics in Ski Jumping.
Jn. Biomechanics. 34, 465-470
Feedback Systems in Rowing
Arnold Baca, Philipp Kornfeind and Mario Heller

University of Vienna, arnold.baca@univie.ac.at
Abstract. On-land feedback devices using rowing ergometers provide an alternative for on-
water systems. In order not to draw incorrect conclusions it is essential to compare the rowers'
technique in the boat to that on thc ergometer. Units for mcasuring reaction forces in the boat
and at the crgometer havc bcen constructed. Similarities in the reaction forces at thc foot
stretcher could be found for elitc rowers.
1 Introduction
Technique analysis in rowing involves the consideration of fine details of the move-
ment of the rower with regard to the boat. In addition to kinematic analyses the study
of the kinetics of the boat-rower system provides valuable insights into strengths and
weaknesses (e.g. peculiarities in motion coupling) (Spinks and Smith 1994; Badouin
and Hawkins 2004). Feedback systems incorporated directly in the boat are used in
elite rowing (Smith and Loschner 2002). Data are processed on-board and may be
transmitted to a PC located on the coach's launch using wireless communication
technologies (Collins and Anderson 2004).
Analyses of the rowing technique in the boat are difficult to realize and are very
demanding in time and instrumentation. In many cases analyses are therefore based
on rowing simulators (ergometers) on land (Page and Hawkins 2003; Loh, Bull,
McGregor and Schroter 2004). In order not to draw incorrect conclusions from the
training sessions on land it is essential to compare the rowers' technique in the boat
to that on the ergometer (cf. Lamb 1989).
A specific setup has been developed to compare the dynamics. Units have been
constructed to measure reaction forces at the foot stretcher in two dimensions and
may be used in the boat as weil as at the ergometer (Concept 2 Indoor Rower Model
D) with or without slides (a construction that is attached to the legs of the ergometer,
allowing the ergometer to roll back and forth during the rowing stroke). Reaction
forces at both feet are acquired separately.
In addition to the forces at the foot stretcher the pulling forces also allow to draw
conclusions on the rowing tcchnique. In the case of ergometer measurements a force
transducer is connected to the chain attached at the handle. In the boat, dynamomet-
408 Arnold Baca, Philipp Kornfeind and Mario Heller
ric oarlocks are used for this purpose. Data measured in the boat are recorded using a
data Jogger or a Personal Digital Assistant (PDA).
A comparison of reaction forces at the foot stretcher has been performcd for elite
rowers. The methods applied and selected results (case study) are presented in the
sequel.
2 Methods
Reaction forces at the foot stretcher are measured using two identical constructions
(Fig. 1) based on load cells (HBM, type HLC220) and strain gages (HBM, type
XY91-6/120). The (portable) units may easily be attached to the foot stretcher ofthe
boat or of the ergometer. Forces are induced into a cover plate made of aluminum.
Components vertical (load cell) and parallel to the platform (strain gages) can be
acquired. From the data recorded the resulting force vector (magnitude, orientation)
is calculated. The load cell acts as double bending beam, the strain gages have been
applied to acquire parallel forces. To obtain an optimal position to mount the strain
gages, stress calculations have been performed utilizing the software Ansys™. A
CAD model of the load cell has been constructed in order to simulate the load cases
in longitudinal direction. The local maxima of the material tensions resulting from
these simulations were selected as positions for bonding the strain gages.
Fig. 1. Left: construction for measuring reaction forccs at the foot strctcher, right: modified
load ccll with strain gagcs
The strain gages (2 measuring grids configured in a T-rosette arranged perpen-

dicular to one another) have been configured as a füll bridge. Because of their orien-
tation in the circuit they compensate forces perpendicular to the load cell and simul-
taneously double the sensitivity in longitudinal direction. In order to condition and
amplify the bridge signals a dual stage amplifier circuit was dimensioned, manufac-
tured and integrated into the platform.
In the boat the platforms are mounted directly to the foot stretcher by screwed
connections, in the case of the ergometer quick clamps at the lower side as weil as
fastening angles at thc upper side are used for fixation (Fig. 2).
Fig. 2. Fixation ofthe platforms. Left: boat, right: ergometer
The linear relationship between force and output voltage was investigated by per-
forming a comprehensive calibration procedure with static loads in both force axes
(normal and parallel) in positive as well as negative direction. The measuring points
obtained by this procedure yield a nearly plane grid, showing a high linearity (Fig.
3).
I00
.
700
.!
-0
."'•
... ~
•oo
Fig. 3. Calibration. Left: perpendicular to platform (load cell - LC), right: parallel to platform
(strain gages - SG)
However, the linear relationship observed is not necessarily sufficient for the dy-
namic case, since inertia properties of the system and moreover the frequency behav-
ior may influence the signal. In order to analyze the dynamic behavior, comparative
measurements were performed using the ergometer equipped with the platforms. The
ergometer was put onto a force plate. A force transducer measuring the pulling force
was attached to the chain ofthe handle.
If horizontal force components are considered only, the following equation can
be set up
(1)
where Fs is the horizontal reaction force at the foot stretcher, Fr the pulling force and
FGRF the horizontal ground reaction force in the direction ofmotion. A comparison of
two different measuring systems is therefore possible.
A typical example for one stroke is shown in Fig. 4. Measured (Fs•) and calcu-
lated horizontal reaction forces (Fs) at the foot stretcher are shown. Note that Fs• is
the sum of components normal (load cell) and parallel (strain gages) to the foot
stretcher considering its angle with respect to the boat/ergometer. A correlation of
r=0.99 was calculated between the values (200 samples per second) ofFs and Fs• for
this stroke.
-FP
1000
-FQRF
z 500 --Fs
GI 1
C,) 0
0
II..
-500
1 F s - F QRF-F p 1
1
-1000
0. 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50
1000
z 500 1
.
GI
C,)
0
II..
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25 2.50
Time [s)
Fig. 4. Upper chart: pulling force (Fr), horizontal ground reaction force (FGRF) and calculated
horizontal reaction force at the foot stretcher (Fs). Lower chart: measured horizontal reaction
force at the foot stretcher (F s•)
3 Case Study
Horizontal reaction force curves at the foot stretcher of an Austrian elite rower (na-
tional team) are presented in Fig. 5. All measurements have been performed on the
same day. Remarkable asymmetries between left and right foot can be seen in all
situations. In particular, the amplitudes of the right foot are higher during the pulling
phase (negative forces). For the three· successive strokes presented the quotients of
the areas under the curves (negative parts only) of left and right foot are 0.68, 0.73
and 0.73 for the boat, 0.83, 0.82 and 0.80 for the ergometer with slides and 0.89, 0.91
and 0.90 for the ergometer without slides.
200
~
., 0
~
-200
0
u. -400
-600
0.0 0.5 1.0 1.5 2.0 2 .5 3.0 3.5 4,0 4.5 5 .0 5.5 6.0 6.5 7.0
200
~
.,
E L~
0
u -200
ö -400 · Rlgh1
u.
-600
0 .0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6 .0 6 .5 7.0
200
~
.,u 0
ö
u.
-200
-400 1- Left
:---· Rlght :
1
-600
0 .0 0 .5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4 ,5 5.0 5.5 6.0 6 .5 7.0
Time (s)
Fig. 5. Horizontal reaction forces at the foot stretcher from elite rower (3 strokes, 30 strokes
per minute ). Upper chart: ergometer without slides, middle chart: ergometer with slides, lower
chart: boat
Moreover, a specific irregularity marking the start of the pulling phase (denoted
by little arrows in Fig. 5) can bee seen in both ergometer conditions as well as in the
boat. One possible reason for the strong occurrence of this irregularity might be that
the upper body of the rower under investigation straightens up too early at the start of
the pulling phase.
The rower may therefore benefit from feedback sessions on the ergometer. Dur-
ing these sessions knowledge-of-performance feedback is given. The time histories
of the relevant kinetic parameters are displayed on a monitor in view of the rower
during motion execution. The rower is thereby able to discover how changes in the
movement pattem alter the shape of the curves.
4 Conclusions and Perspectives
Case studies indicate that peculiarities in the rowing pattem may also be observable
when using rowing simulators. In these cases benefits are expected from the use of
feedback systems based on ergometers.
The results also indicate that the ergometer with slides compares better to on-
water rowing. This is not surprising, since on the ergometer with slides the rower has
to accelerate/decelerate the ergometer whereas on the ergometer without slides
he/she has to accelerate/decelerate his/her body. lt should, however, be considered
that many (elite) rowers are not used to exercise on ergometers with slides.
Upcoming experiments will also consider pulling forces for the comparisons.
A variant of the on-land feedback system for use in the boat assisting both coa-
ches and athletes and a cascaded double ergometer system are under development.
Acknowledgements
We thank Armin Blaha and Emanuel Preuschl for their valuable contribution in
constructing the foot stretcher dynamometer.
References
Baudouin, A. and Hawkins, D. (2004) Jnvcstigation ofbiomechanical factors affecting rowing

perfonnance. J. Biomech. 37, 7, 969-976.
Collins, D. J. and Anderson, R. (2004) Thc use of a wireless network to provide real-time
augmented feedback for on-water rowing. In: Proceedings of the American Society of
Biomechanics 28th Annual Conference, Portland, USA, pp. 590-591.
Lamb, D. H. ( 1989) A kinematic comparison of ergometer and on-water rowing. Am. J.
Sports Med. 17, 3, 367-373.
Loh, J. M. H., Bull, A. M. J., McGregor, A. H. and Schroter, R. C. (2004) Instrumentation of a
Concept II rowing ergometer for kinetic and kinematic data acquisition, In: M. Hubbard,
R. D. Mchta and J. M. Pallis (Eds.), The Engineering of Sport 5, Volume 2. isea, Sheffield,
pp. 173-179.
Page, P. N. and Hawkins, D. A. (2003) Areal-time biomechanical feedback system for train-
ing rowers. Sports Eng. 6, 67-79.
Smith, R. M. and Loschner, C. (2002) Biomcchanics feedback for rowing. J. Sport Sei. 20,
783-791.
Spinks, W. L. and Smith, R. M. ( 1994) The effects ofkinematic infonnation feedback on
maximal rowing perfonnance. J. Hum. Mov. Stud. 27, 17-35.
Biomechanical Analysis of Olympic Kayak Athletes
DuringlndoorPaddling
Nicola Petrone 1, Andrea Isotti I and Guglielmo Guerrini 2

1 University of Padova, nicola.petrone@unipd.it
2 Italian Kayak Federation
Abstract. The aim of the work was the biomechanical analysis of elite female olympic kayak
athletes during indoor paddling on ergometers. An integrated motion capture system was used
for the acquisition of paddle trajcctory, trunk and limbs motion and for thc mcasurc of forccs
devclopcd by the athlctes on a dynamometric footpad. Sevcral quantitative biomechanical
parameters were correspondently defined. Symmetry and regularity criteria wcrc adoptcd to
define a skill classification for each analysis paramcter and a final correlation with sport
results classification was invcstigatcd. This will help trainers to dcfinc suitable functional
evaluation methods.
1 Introduction
Olympic kayak is a spectacular discipline where technique, fitness and race strategy
are essential factors for winning. Different athletes can develop different techniques
and usually require specially designed equipment such as paddles and seats to give
top performances: on the other side, new shapes or devices required adapted
techniques.
The work reports the experience developed with the ltalian Female Nationalteam
during the preparation of the past 2004 olympic games. By means of instrumented
footpads, seats and paddles the motion pattems and the forces developed by the
athletes during laboratory testing were recorded and compared with sport ranking.
2.1 Instrumentation and Testers

The study focused on the measurement of kinematic and kinetic data during indoor
paddling sessions performed on a rope paddling ergometer as shown in Fig. l .a.
To measure kinematics, a motion analysis system Smart® (BTS - ltaly) was
used: 6 infrared coaxial cameras working at 50 Hz were placed around the ergometer
and a volume of about 4x4x3 m 3 was calibrated. The testers were prepared with
adhesive passive markers following the marker protocol presented in Fig. l .b and
414 Nicola Petrone, Andrea Isotti and Guglielmo Guerrini
1.c. In particular, 3 markers were placed on a special wireframe helmet, 3 markers

were applied to the ergometer frame and 3 non-aligned markers were placed on the
paddle as shown in the detail of Fig. l .b. Bone landmarks markers were
symmetrically placed on the limbs with additional markers over C7, T12 vertebrae
and the two PSIS (Posterior Superior Iliac Spine).
Two rotational potentiometers were applied to the knees with straps, to measure
the two flexion angles <1>L and <1>R as shown in Fig. l .b, zeroed with legs extended.
The measure of forces applied by the feet was performed by means of a
dynamometric footpad developed in a previous work (Petrone, Quaresimin and
Spina, 1998) and fixed to the ergometer. Two force components, the forces normal to
the footpad at the left (FNL) and right (FNR) foot were measured as shown in Fig.
l .a by a synchronous data acquisition system at 200 Hz.
Five elite female athletes volounteered for the study. All testers were taking part
in the Italian National Team training and had obtained good results at the national
level either in single (Kl) or double kayak (K2). On the basis ofrecent performance
results, the National Team Trainer ranked the testers in descending order. This
ranking, antropometric data as well each athlete's best result are noted in Table 1.
TESTER PERSONAL BEST RESULT AGE H [cm] m lk!!]

1 IJ Olvmpic Gold Kl 500m 30 178 74
2 SF National Champion Kl 500m 23 174 73
3 RA National Champion Kl 1000m 28 172 86
4 FA National Champion K2 5000m 24 170 65
5 BT National Champion K2 5000m 23 175 73
Table. 1. Information about the five athletes involved in the study.
2.2 Test Protocol

Tests were performed on the same day after preparation in sequence of the 5 athletes.
Each tester was asked to perform four runs of paced paddling on the ergometer set at
constant resistance.
The runs were conducted at 70 and 90 strokes per minute (spm) with two
different types of seats: a normal fixed seat and a special rotating seat having a free
rotational degree of freedom along the vertical axis. For each run at least 50 strokes
were recorded. Each athlete was given a 5 minutes recovery period between each
test.
2.3 Data Analysis

Kinematic and kinetic data were analysed based on several criteria to determine
possible correlations between different quantitative biomechanical parameters and
the athlete's ranking. This can provide a useful tool to trainers in defining detailed
protocols of functional evaluation and monitoring the training state of athletes during
preparation.
The first analysis criteria was based on the paddle motion.
Biomechanical Analysis ofülympic Kayak Athletes During Indoor Paddling 415
Fig. 1. (a) Experimentalsetup with indication ofmeasured forces at feet, a detail ofthe paddle
markers and a detail the dynamometric footpad. (b) Stick model of an athlete after a motion
capture session, with indication ofthc global reference system XYZ, the knee flexion angles
<j)L - <j)R and the Great Trochanter motion range ~GT (c) Rear view of a tester with the set of
anatomical markers adopted.
Given the presence of a triad of non-aligned markers on the paddle, the

geometrical centre of the paddle was virtually reconstructed and its trajectory was
calculated via the motion analysis BTS ® software. The most significant pattems
416 Nicola Petrone, Andrea lsotti and Guglielmo Guerrini
occurred in the transverse plane XZ and in the frontal plane ZY reported in Fig. 2,
corresponding respectively to an upper and a posterior view. Ten cycles out ofthe 50
recorded were consistently extracted from each run and were projected on the
cartesian planes. The stroke repeatability, defined as the scatter of the 10 cycles
curves, the plot symmetry and the placement of curves centroid were considered as
quantitative parameters. Trajectories for testers 1 and 5 are compared in Fig. 2.
The second comparative parameter was the relative angle between shoulder and
pelvis, named 0sp. lt was defined through the introduction of two reference systems
sharing a common vertical axis y' as the angle between the vector defined by the two
PSIS at the pelvis and the vector defined by· the two Acromiom markers at the
shoulders. The two triads and the angle 0sp are sketched in Fig. 3.a at the end of a
left stroke. This angle 0sp was supposed to be preferably low, indicating a more
steady motion of the trunk with little rotation of the shoulders about the pelvis. In
any case the angle should be symmetric in weil balanced athletes with small
deviations from an average value. Angles were calculated from the motion captures
and positive and negative peaks of 10 significant cycles were compared, as reported
in Fig. 3.b, where the four athletes are compared for the 90 spm run with rotating
seat. Each line connects the two sides average values with Standard Deviation
superimposed.
The third data analysis concentrated on the motion of the Great Trochanters (GT)
in the sagittal plane XY. The motion L\GT of Left and Right GT were linearized
between the two extreme points and plotted in the XY plane. The symmetry of the
two slopes and the minor amount of elevation in the Y direction of the GT from the
seat correlate to a good stroke technique.
The fourth analysis focused on the range of flexion angles q>L and <!>R at the two
knees as indicated in Fig. l .b. The two paces and two type of seats runs were
compared in diagrams such as Fig. 4.a and symmetry between the two legs were
studied. Finally, the two normal forces applied to the footpad FNL and FNR were
investigated, again plotted as in Fig. 4.b and evaluated in terms of symmetry between
the two body sides.
The different analysis criteria supported athlete comparison and the ability to define
a relative classification of each tester for each criterion, as summarized in Table 2.
The paddle motion analysis as in Fig. 2 showed how the best athletes were able to
perform a symmetric, steady and compact trajectory, whereas the worst tests showed
asymmetric cycles with irregularities and open loops in the XZ plane.
The analysis of angle 0sp as shown in Fig. 3 revealed how testers I and 4 were
able to produce a compact symmetric trunk rotation, whereas testers 3 and 4 revealed
a much higher range ofrotation and tester's 5 results were strongly asymmetric.
Motion analysis of the GT at the two sides exhibited the best results for tester 1,
with the highest symmetry and lowest delta Y, and the worst results for tester 3. On
the contrary, tester 3 showed the best knee range of flexion where tester 1
demonstrated poor symmetry as can be seen in Fig. 4.a.
Biomechanical Analysis of Olympic Kayak Athletes During Indoor Paddling 417
-150 .75 0 75 150 ,so

.300 y
-225
75
-150
-75 z
0
0
X
75
-75
150
225 ~
300
_J -150
-300 -225 -150 -75 0 75 150 225 300
(b)
Fig. 2. Comparison ofpaddle motion between tester 1 (solid) and 5 (dotted). (a) Transverse
plane XZ trajectories (in mm). (b) frontal plane ZY trajectories (in mm).
37.91 °
5 BT
4 FA 36.42°
IJ 37.76°
44.21 °
47.06°
8sp
-45• .30• .,5• o· 15• 30' 45' 60'
(a) (b)
Fig. 3. (a) Definition ofrelative angle 0sp between shoulders and pelvis. (b) Comparison
between the five testers for the 90 spm tests with rotating seat.
80 DL 500 O FNL
(N) Tester 1_U
(')
72
400
70 66
Fig. 4. (a) Results ofknee range offlexion from angles <)>Land <)>R for tester 1 at 70 spm or 90
spm, with fixed (F) or rotating seat R. (b) Corresponding foot forces peak values.
418 Nicola Petrone, Andrea Isotti and Guglielmo Guerrini
The footpad force analysis showed the best symmetry for tester 1 and the worst
for tester 2. They were the only testers able to increase the forces at increased pace.
Tester 2 showed the highest absolute normal forces, even ifunbalanced.
The presence of a rotating seat generally induced an increase in the knee range of
motion and in the footpad forces, with the exception of tester 3 for which the seat
resulted tobe unsuitable.
From the analysis of classification of different testers in Table 2, some
biomechanical quantitative parameters like the shoulder-pelvis relative rotation 0sp,
the GT excursion and the footpad forces syrnmetry resulted in major interest in
defining a functional evaluation protocol for trainers. A regular paddle trajectory
should be considered a basic requirement.
In general, all athletes, despite their national or international ranking, showed
aspects of the paddling technique that may be improved and that were revealed by
the present analysis. Further development of this study will focus on the measure of
motion of these significant body angles during real sessions of paddling in the kayak.
For this, suitable lightweight data acquisition systems and instrumented devices
should be developed to better simulate the real boat dynamic behaviour and to give
füll freedom of motion to the athletes during the trial.
TESTER Paddle 0sp LiGTL/R Lid>L/R FNL/R

1 IJ 2nd 1st l st 4th 1st
2 SF 4th 4 tn 2nd 5th 5th
3 RA 3rd 3rc1 5th 1st 3rd

4 FA 1st 2na 4th 3,a 2na
5 BT 5 tn 5th 3rct 2"ct 4th
Table 2. Relative classification of different testers for the five quantitative parameters
introduced and analysed.
4 Conclusions
The relative range of motion of the shoulder and pelvis of five elite female athletes
was investigated, together with the trochanters excurions, the symmetry of the
paddling action, the knee range of flexion and the footpad forces. The athletes were
ranked based on recent event results. The different kinematic and kinetic parameters
were correlated with skill levels and showed how the best athletes were able to
perform regular paddle trajectories and steady trunk motion at the different trials.
References
Petrone N., Quaresimin M. and Spina S. (1998) A load acquisition device for the paddling
action on olympic kayak, 11th International Conference on Experimental Mechanics,
Oxford-UK, 24-28 August 1998, pp. 817-822.
So you think you know the ropes? White Water Rescue
Ropes and Techniques
Matt Barker
Auckland University of Technology, New Zealand, matt.barker@aut.ac.nz
Abstract. Recent equipment failures during simulated white-water rescues have shown the
need for detailed measurement of the loads created and the loads that can be safely sustained
by the equipment currently in use for white water rescue. This research experimentally finds
out; what are the potential forces involved in a white-water rescue? What forces can a three-
person rescue team generate? Which of the ropes on the market are suitable for the demands of
white-water rescue? Which of the current mechanical advantage rescue techniques are best
suited to the equipment available? Are there any experimental techniques that could improve
the force generation and safety of the rescue system? Weak points in the complete rescue
system are highlighted.
1 Introduction
White-water rescue equipment can be made to fail using conventional rescue prac-
tices. If failures of this type were to occur in an actual rescue environment, catastro-
phic results would ensue for both rescuer and rescuee. The current investigation set
out to try to answer these questions,
1. Which of the ropes on the market were suitable to the demands of white-
water rescue.
2. Which of the current mechanical advantage rescue techniques are best
suited to the equipment available.
3. What force can a 3 man team generate
4. Are there any experimental techniques that could improve the force genera-
tion and safety of the rescue system?
2 Methods
2.1 Buoyancy Aid Testing

On the Roberts Testing Equipment Serial number 97041 hydraulic test bed at One-
hunga Chain and Rigging, the strength of the various components in the rescue har-
ness and auxiliary clip in points of two manufacturers buoyancy aids were tested.
2.2 Rope Testing

The following criteria was used to select ropes for the tests, they had to have a den-
sity of less than 1 so that they tloated (Polypropylene, Polyethylene and Spectra
420 Matt Barker
fibres have a density of less than 1), they had to be between 6 and 8mm in diameter
(below 6 mm they are hard to hold and over 8mm it is hard to coil them into useable
throw bags) and the materials used in construction had to be resistant to bacterial
decay if left wet for extended periods. Ropes obtained were largely manufactured for
use in yachting but two ropes are specially manufactured for use in aquatic rescue,
PMI Water Rescue and Esprit Swiftwater Line. These ropes were tested on the Rob-
erts Testing Equipment Serial number 97041 hydraulic test bed at Onehunga Chain
and Rigging, when knotted and in the various 3: 1 mechanical advantage systems.
2.3 Haul System Testing

A Tedea Huntleigh model no. 619 load cell was placed at the terminal end of a hau!
system where the forces generated by a 3 man team with a combined mass of 230kg
could be measured. The haulers wore normal training shoes and stood on a flat grass
surface. The variables applied were, merely pulling the rope with hands (Armstrong
method), 3 to 1 pulley system, 6 to 1 pulley system, and 9 to 1 pulley system. To
each of these variables were added, using or not using Petz! Ultralegere pulleys to
reduce frictional losses, and pulling with hands only or attaching slings to the hau!
line with a marlinspike hitch and wrapping these around the haulers back so they
could all pull like a 'tug o war' anchorman. A vector pull was also tested where a 2-
man team tensioned a 3 to 1 pulley system while the third locked the system off. A
sling was clipped to the middle of the system at 90 degrees to the rope line and then
all three pulled with their hands on the sling.
3 Results
Test Force at failure Type of failure
Hydraulics ehest strap plastic buckle 4.1 kN Webbing broke

Webbing pulled through
Macpac ehest strap plastic buckle l .6kN buckle
Hydraulics ehest strap plastic and meta] Pulled the central pin out of
buckle 6. 7kN alloy buckle
Macpac ehest strap plastic and meta] buckle 9.8kN Stitching on buckle broke
Hydraulics 2004 cowstail ring 2kN Weid broke
Hydraulics 2005 cowstail ring llkN Bent so as to trap webbing
Macpac 2005 cowstail ring 2kN Ring broke
Hydraulics 2004 cowstail 2.7kN Webbing broke
Macpac 2004 cowstail 6.3kN Stitching broke
Hydraulics 2005 Buoyancy aid shoulder
strap 4.3kN Stitching broke
Hydraulics 2005 Buoyancy aid waist strap l.5kN Fastex buckle broke
Table.l. Buoyancy aid testing

So you think you know the ropes? 421
Table 1 shows the forces at failure for the components of the rescue hamess incorpo-
rated into white water buoyancy aids as weil as auxiliary attachment points that
could be used to clip a person to a rope in a rescue. This testing highlighted some
surprising mismatches in components and materials. Components of the same system
varied in strength from 2kN to 9.8kN. The ring that attaches the cows tail to the ehest
strap was found to be a weak link in the both manufacturers system, failing at 2kN
and the cows tail webbing and buckles used on the ehest strap failed at as low as
2.7kN and l .6kN respectively.
Rope Construction Fig 8 Overhand

Spectralite 8mm Braided Polypropylene sheath, Spectra core l 5.4kN 11.6kN
Southem light Braided Spectra and Polypropylene blend l l.3kN l l.3kN
Robline extreme Braided Polypropylene sheath, Spectra core 9.lkN
Spectralite 6mm Braided Polypropylene sheath, Spectra core 8.5kN 7.6kN
Swiftwater line Braided Polyester sheath, Polypropylene core 7.lkN 7.2kN
Econobraid 8mm Braided Polypropylene sheath and core 6.4kN
PMI Water rescue 7mm Braided Nylon sheath, Polypropylcne core 5.6kN
Robline Albatross Braided Polypropylcne 4.3kN
Econobraid 6mm Braided Polypropylene sheath and core 2.9kN
Table.2.Tenninal knot strength
Table 2 shows the strength of the ropes using Overhand and Figure of Eight knots.
The Figure of Eight knot with Donaghys' Spectralite rope and Southem Ocean's
Southem light provided the strongest terminal knots failing at 15.4kN and l l.3kN
respectively. The ropes specifically manufactured as water rescue ropes, PMI's Wa-
ter rescue rope and Esprit's Swiftwater line, failed at 7.lkN and 5.6kN respectively.
These ropes displayed separate sheath and core failure, probably due to the sheath
and core materials not having similar stretch characteristics, therefore not loading
across all fibres, leading to progressive failure at relatively low loads.
Marlin- Clove Truckers 5mm Tape

Rope Inline 8 spike hitch Tibloc hitch prussik prussik
Spectralite 8mm 15.3kN 13.4kN 13.9kN I0.7kN 10.SkN 9.4kN
Southem light 9.8kN I0.7kN 8.7kN 2.3kN
Robline extreme 9.lkN
Spectralite 6mm 8.SkN 8.5kN
Esprit Swiftwater line 7.1 kN 7.2kN 6.3kN
Econobraid 8mm 5.3kN 7.lkN 6.6kN 6.SkN 3.5kN
PMI Waier rescue 4.2kN 4.4kN
Robline Albatross 4.3kN 4.lkN
Table. 3. 3 to I pulley strength
422 Matt Barker
Table 3 shows the force at failure for the various 3 to 1 pulley systems. The best
performing knot used to create the 3 to 1 mechanical advantage system was a
Marlinspike hitch (Pawson 2001 ), although the Inline figure of Eight knot and Clove
hitch sometimes were stronger than the Marlinspike hitch, they tightened up to such
an extent during loading that they became impossible to untie thus leaving a knot, or
knot and karabiner, in the rope rendering it useless as a clean throw or rescue rope
later. Prussik knots although widely used and recommended by various authors (Fer-
rero 1998; Walbridge and Sundmacher 1995; Ray 1997), are in fact quite inadequate
to handle the maximum forces able to be applied without slippage.
3 man Force Generation
12
10
□ arm5tn>n9
kN 6
■2 to 1
■3 to 1
■ 6to 1
■ 9 to 1
0
Pulllnv w1lh Pulllng wilh Pull,ng With Pulllng ,.,th
hands sllngs hands end slings end
pullf!yS pulleyS
Fig 1. 3-man force generation
Figure 1 shows the force able to be applied by a 3-man rescue team. With the inclu-
sion of small karabiner pulleys, the force able to be applied was increased by 30%.
The inclusion of slings in the haul system to allow the haulers to pull with the sling
round their backs rather than holding the rope in their hands only, yielded a 12%
increase in force. The haulers stated that the limiting factor when just hands were
applied was their grip on the rope and that with the sling it was much easier to create
a higher force but the limiting factor then became their traction. On the riverbank,
there may well be boulders, tree roots or stumps that may allow greater forces to be
applied with the sling system. The most significant increase in force was with a vec-
tor pull, using no additional equipment the force generated by our 3-man team was
increased by 37%. Even more surprising is the fact that the original tensioning was
achieved by a 2-man team, as one was locking the system off, so the actual increase
from a 2-man 3: 1 pulley to 3-man vector pull was 107%, making this a very efficient
use of limited resources.
So you think you know the ropes? 423
4 Conclusions
The force applied by moving water has an exponential relationship to the speed of
flow. However the speed of water is not the only factor that rescuers have to over-
come, as there is usually a !arge component of friction adding to the force required to
free an object. Friction has a relationship to pressure and also rises exponentially, but
is not calculated here, as friction is dependent on the way objects are trapped and the
surface they are held against. The forces created by drag are l 2kN for a swamped
kayak and over 2kN for a body in water flowing at 5.3 metres per second (Bechdel
and Ray 1985; Ray 1997). These figures will increase in faster flows and including
variables based on friction, could lead to somewhat !arger forces in actual rescues.
Thoracic injury is avoided at loads of 4kN and 50% will receive injury of broken
ribs on one side when well-distributed loads of 6.9kN are applied to the shoulders
and ehest, with a maximum survivable force of 8kN (Cavanaugh 2000). Forces re-
quired to damage the vertebrae in adults is between 6 and 8kN (Cassillo 2005).
Given the above, rescue harnesses in buoyancy aids should be manufactured to
withstand forces of between 6 and 8kN as a minimum as it is better to rescue some-
one injured than leave them to drown because the equipment failed at loads as low as
2kN as is presently the case. lt is of little use to have some parts of the same rescue
system built to take forces of 9.8kN when they are dependent on other components
that fail at as low as 2kN. The separate components should be matched in their
strength rating and an overall minimum rating should be formulated for these safety
items. At present some buoyancy aid rescue harnesses could potentially fail under
the forces generated by even simple rescues.
Hau! systems need to be able to produce loads of over 10kN in order to rescue
swamped kayaks in swift water (Bechdel and Ray 1985; Ray 1997). Using prussik
knots to form a 3 to 1 pulley system will give you a safety valve allowing slippage
before the rope breaks but will ultimately limit the forces able to be applied to any
system. lt is therefore counter productive to apply a mechanical advantage system
greater than 3 to I as the prussik will likely slip at greater loadings particularly when
used on slippery (Moyer, Trusting, and Harmston 2000) Spectra sheathed or blended
ropes. lt is recommended to use a device such as the Petz! Tibloc to create a me-
chanical advantage hau! system where it is necessary to have a traveling attachment
point on the main hau! line as this produces a stronger and more consistent traveling
attachment point.
The strongest hau! systems were created using Spectra cored Polypropylene
sheathed or Polypropylene and Spectra mixed braids of 8mm diameter using a
Marlinspike hitch and karabiner pulleys wherever the rope ran round a karabiner.
Indeed the reduction in friction with the addition of plastic karabiner pulleys added
the equivalent of an extra person to a 3-man tcam making them weil worth their 10
gram weight.
424 Matt Barker
Pulling using slings attached to the main hau! line also creates a useful boost to peak
force generation and may be even better in places where there is better traction.
Ropes for aquatic rescue should be manufactured containing a significant pro-
portion of floating fibres (Spectra, polypropylene and polyethylene). The ropes core
and sheath should have similar elongation characteristics, so that the sheath and core
share loads and are therefore stronger per cross sectional area than ropes where the
core and sheath behave differently. These ropes should be manufactured with a target
knotted tensile strength in excess of 10kN in 8mm diameter form, to be aligned with
the strength of attachment points on kayaks, the maximal survivable loads through
ehest hamesses and the likely forces required in rescue of swamped kayaks.
Acknowledgments
I would like to thank the following for their help in getting this research from the
ideas stage to completion.
Institute of Sport and Recreation Research New Zealand (Auckland University of
Technology) for grant aid.
Hydraulics (New Zealand) for supply ofbuoyancy aids and materials.
Donaghys Ropes (Australia) for supply ofropes.
Southem Ocean Ropes (New Zealand) for supply ofropes.
Onehunga Chain and Rigging, Auckland, for use of their test bed.
References
Bechdel, L. and Ray, S. (l 985)River Rescue. Appalachian Mountain Club Boston
Cassillo, F. (2005) Kinesiological and Anatomical approach to the Deadlift. Retrieved
14/ l 0/2005 from http://www.bodybuilding.com/fun/casi4.htm.
Cavanaugh, J. (2000) The Biomechanics ofthoracic trauma. BME 7160, Winter 2000. Re-
trieved 14/ l 0/2005 from http://ttb.eng.wayne.edu/-cavanau/ucsdout.html
Ferrero, F. ( 1998) White Water Safety and Rescue. Pesda Press, Wales
Hopkins, R. (2003) Knots. Thunder bay Press, San Diego
Moyer, T., Trusting,P. and Harmston,C. (2000) Comparative Testing of High Strength Cord.
Paper prescnted at 2000 International Technical Rescue Symposium
Pawson, D. (2001) Pocke/ guide to knots and splices. PRC publishing,
London
Ray, S. ( 1997) Swiftwater rescue. CFS Press, Ashville NC.
Walbridge, C. and Sundmacher, W.A. ( 1995) Whitewater rescue manual. Ragged Mountain
Press, Camden ME.
Computational Modelling of Surfboard Fins
for Enhanced Performance
Dave Carswell, Nicholas Lavery, and Steve Brown

University of Wales Swansea, N.P.Lavery@swansea.ac.uk
Abstract. A Computer Aided Design (CAD) tool called Fin Designer has been developed at
the University of Wales, Swansea for three-dimensional design of surfboard fins. This tool has
been developed as part of a )arger project aiming to facilitate design of fins and surfboards for
manufacturers with user-friendly software and direct linking into manufacturing processes
such as CNC machines and injection molding. This type of tool also provides the basis for in-
depth scientific and engincering studies, using engineering software for stress analysis and
Computational Fluid Dynamics (CFD), giving dceper insights into potential improvcments
from design and material modifications. Previous papers (Lavery, Fostcr, Carswell and Brown
2005; Carswell and Lavery 2006) have dealt with the description of the geometrical models
used in the software as weil as prcliminary computational results made possible by the soft-
ware. In this paper the emphasis is on the correlation between drag and lift forces for surf-
board fins as predicted and measured in a flow tank. Drag and lift forces predicted by thc CFD
were fed into a specially coded Finite Element Analysis (FEA) to obtain displacements of the
fin undergoing these hydrodynamic forces. This paper presents preliminary validation offluid-
solid coupling on the standard benchmark of a cylinder, as weil as some results for a single
fin. While further verfication of the models are required, the current results appear to suggest
that the displacements are of a couple of orders of magnitudc smallcr than those expccted by
current fin manufacturers, and hcnce that fin stiffness remains a strong candidate for fin design
improvements.
1 Background
A previous paper (Lavery, Foster et al. 2005) presented a detailed background to the
current project, with a description of surtboard evolution over the past three decades,
and the emphasis currently being placed on fin design as a means of increasing surf-
board performance, largely brought about by relatively new attachment systems
available which allow faster exchange of the fins to suit changing surfing conditions,
but which are also lighter and less intrusive than previous systems.
Whereas higher profile water sports such as sailing, power boat racing and
swimming enjoy a measure of engineering and scientific treatment in terms of tech-
nical understanding and product development, this is primarily driven by economics,
popularity and possibly their presence in Olympic and international arenas. Over the
426 Dave Carswell et al.
past decade, the number of surfers taking to the waters has seen a steady increase,
fueled by a media portrayal of the sport as being cool and trendy, and possibly more
importantly, its image has been replaced by professional surfers doing an exciting
and sometimes dangerous job.
The project currently underway at the Swansea University is applying rigorous
engineering treatment to surfboard and fin design in the following ar-
eas/developments: Computer Aided Design (CAD) software, Computer Aided Engi-
neering (CAE) and Experimental Validation of computational models.
To date, and using the methodology described above, the project has addressed
issues related to fin design, (Lavery, Foster et al. 2005; Carswell and Lavery 2006),
namely, as to whether glass-on fins induce less drag than fins fitted in boxes and
whether the use of alternative NACA foils could provide lower drag, hence promot-
ing a faster fin design. The answer to both these questions has been a cautious yes,
but further work still needs to be done to assess the extent to which this is true.
2 Computational Modelling
The progression to perforrn a CAE analysis, consisting of a CFD analysis followed

by a Finite Element Analysis (FEA) stress analysis, is shown in Fig. 1.
-----u~hlmmt~
AD ~ (FI;~ ~
.._ 1 :;: 1 ,J
\ t
1 ™ 1• 1 GAMBr[ 1
Fig. 1. Computational methodology used
2.1 Computer Aided Design for Surfboards and Fins

Two specific CAD packages are being developed based on NURBS surfaces, one for
fins (Fin Designer) and the other for surfboards (DAT Designer). Underlying geo-
metrical algorithms are described in (Carswell and Lavery 2006). The data exchange
forrnat presently used is the IGES (Nagel, Braithwaite and Kennicott 1980; Kenni-
cott 1995).
2.2 Computational Fluid Dynamics Analysis

CFD analysis is done using the commercial package Fluent, and fin data created in
Fin Designer is imported into Gambit (Fluent mesh generator), using the IGES for-
mat, as shown in Fig. 2.
New Computer Aided Design Software of Surfboard Fins for Enhanced Performance 427
Fig. 2. (a) CAD drawing to (b) CAE solid mesh for a typical fin
The fin is then enclosed within a fluid domain having inlet and outlet apertures
through which flow is passed over the fin on a flat base to simulate boundary condi-
tions on a surfboard. Four-noded tetrahedral elements are used for both solid and
fluid domain. The analysis can be done for various velocities and angles of attack
giving pressures on the fin surface and, hence, lift and drag forces.
The Reynolds number, Re, is a classical non-dimensional engineering number. For

cylinders it is determined, by the diameter, D, by
Re= pUDIµ (1)
Where, p is the density, U is the velocity of flow at the inlet and µ is the viscosity.
The pressure coefficient, Cp, is defined by,
Cp = (P-P 0) / ½pU/ (2)
Where, P is the pressure, U0 is the free stream velocity and P O is the static pressure.
All these values are obtained from the CFD results.
2.3 Finite Element Analysis for Fin Deformation

The deformation analysis uses the Finite Element Virtual Energy approach (Smith
and Griffiths 1998) for elastic stress, with pressure derived from the CFD as a sur-
face loading and performing a one-off stress analysis, i.e. assuming a steady state
flow. Input options are the Young's Modulus, E, and Poisson's ratio, typical values
being in the 10-20GPa and 0.29-0.32 ranges, respectively.
3 Verification of Deformation of a Solid Cylinder in Cross Flow
As a benchmark to verify the fluid/structure coupling, models were run for a vertical
solid cylinder positioned in a cross-flow and fixed at both ends. Reynolds numbers
of 20 and 40 were examined, based on cylinder diameter, with a laminar model. The
same mesh was used in both cases and was 150mm high and 4mm in diameter, con-
sisting of 12,000 nodes and 60,000 elements. All measurements of pressure and
displacement were taken from the middle of the cylinder. Although much greater
Reynolds numbers occur to those given here, the test case is still valid as a check on
the fluid-structure coupling. Fig. 3 (a) for Re=20 and Fig. 3 (b) for Re=40 show the
calculated streamfunction pattems. Cp is plotted around the circumference of the
cylinder and compared to previous numerical work by (Benson, Bellamy-Knights,
Gerrard and Gladwell 1989) in Fig. 4, and it can be seen to compare relatively well.
Benson et al. used a polar mesh for their computations and differs from the one used
here as this used Cartesian coordinates. Discrepancies may be due to domain extent,
mesh distributions and/or boundary conditions. The resulting deformed circumfer-
ences for various Young's moduli are shown in Fig. 5 and appear to be physically
reasonable. At the time of writing, no previous experimental work was available to
validate the deformation.
(a) Re=20
(b) Re=40
Fig. 3. Streamline pattems at Reynolds numbers of (a) 20 and (b) 40.
Fig. 4. Pressure distribution Cp around the cylinder for Re=20 and 40.
New Computer Aided Design Software of Surfboard Fins for Enhanced Performance 429
20 10 00 -10 -20 -30 20 10 00 10 -20 -30
Fig. 5. Original and deformed shapes of a cylinder for a Reynolds number of (a)
20 and (b) 40, note greater deformations for the latter. Points on the cylinders do not
represent actual mesh nodes but points of intersection between element edges and the
intersection algorithm.
4 Validation of Drag, Lift and FEA of Fins
An experimental rig has been constructed which used a flume with flow rates up to 1
ms· 1 together with load cells to measure drag and lift forces. Surfboard velocities
measured by (Walker 1972) and (Paine 1974) are strongly dependent on wave height
and can be in excess of 13 ms· 1• Thus, current full-scale fin experiments would only
be valid for a small-wave surfing.
~ O~ T
001&
GE
0014
l00,2
1oo,o 1
10008
!OOOI
••• 1
000<
0002
•2 ..___ _ _ _ _ _ _ _ _ __
0000
0 10 12
Fig. 6. (a) Numerical versus experimental lift-to-drag ratios for the middle XI fin and (b) tip
displacement for various angles of attack and upstream velocities.
Various fins provided by Daum Tooling Ltd. (RedX) have been tested against
angle of attack. Results are presented for the middle Xl fin only. The experimental
and numerical results for the drag to lift ratio are given in Fig. 6 (a). The numerical
results obtained used a k-c turbulence model with standard wall functions and the
Reynolds number based upon the inlet was about 50,000. The mesh contained 47,000
nodes and 220,000 tetrahedral elements with Young's Modulus being l0GPa and
Poisson's ratio being 0.3. The drag and lift forces predicted by the CFD for fins at
higher flow velocities leads to the prediction of very small tip displacements using
the linear elastic FE Stress model developed for this work, Fig 6(b ). These displace-
ments are smaller than those given by manufacturers by a few orders of magnitude.
5 Conclusions
The CAD software being developed at Swansea is providing clean CAD surfaces
suitable for engineering analysis in a manner much easier to use than commercial
CAD packages. The experimental data for the drag and lift for separate fins appear to
follow expected trends. The current CFD results, by taking the average error, are
within 10% for pressure and 30% of drag and lift forces of experimental tests and
previous CFD work. Model accuracy needs to be improved. The fluid-structure cou-
pling appears to be behaving as expected for low Reynolds numbers tested here but
further work needs to be done to verify the correctness of the models and indeed the
application to new designs with variable stiffness. This paper represents a first and
thorough set of steps in the verification and validation of the models, and certainly
preliminary results would suggest that there is a potential to modify stiffness proper-
ties of fins in ways which could have beneficial effects on drag/lift ratios for the fins.
Acknowledgments
The authors would like to thank the other members of the team without whose help
none of this would have been possible: Dr Ian Pearce, original developer of DAT98,
Dr Adam Bere, for structural analysis and Mr. Graham Foster of JF Design Ltd., who
designed the experimental rig. The authors would also like to thank Tom O'Keefe
from Daum Tooling Ltd for his invaluable technical insights and sponsorship in the
form of the numerous fin samples tested. Finally, the authors would also like to
thank the School of Engineering at the University of Wales Swansea for the Re-
search Incentive which has partly funded this work.
References
Benson, M. G., P. G. Bellamy-Knights, J. H. Gerrard and I. Gladwell (1989). "A Viscous
Splitting Algorithm applied to Low Reynolds Number Flows round a Circular Cyl-
inder." Journal ofFluids and Structures(3): 439-479.
Carswell, D. and N. Lavery (2006). 3D Solid Fin Model Construction from 2D Shapes Using
Non-Uniform Rational B-Spline Surfaces. Advances in Engineering Software, (Ac-
cepted for publication).
Kennicott, P. R. (1995). The Initial Graphics Exchange Specification (IGES) Version 5.3,
IGES/PDES Organisation.
Lavery, N., G. Foster, D. Carswell and S. Brown (2005). "Optimization ofSurfboard Fin
Design for Minimum Drag by Computational Fluid Dynamics (Do Glass-on fins in-
duce less drag than boxed fins?)." Natural and Artificial Surfing Reefs.
Nagel, R., W. Braithwaite and P. Kennicott (1980). Initial Graphics Exchange Specification
IGES (Version 1.0). Washington, DC, National Bureau ofStandards.
Paine, M. (1974). Hydrodynamics of Surfboards. Mechanical Engineering, University of
Sydney. Bachelor of Science.
Smith, I. and D. Griffiths (1998). Programming The Finite Element Method, Wiley.
Walker, J. (1972). Recreational Surfing on Hawaiian Reefs. 13th Coastal Engineering Confer-
ence, Vancouver.
Development of Swimming Prosthetic
for Physically Disabled
(Optimal Design for One Side
of Above-Elbow Amputation)
Keiko Yoneyama and Motomu Nakashima
Graduate School of Infonnation Science and Engineering,
Tokyo Institute ofTechnology, Tokyo, Japan
Abstract. Swimming is a suitable sport for the physically disabled, since the buoyancy pro-
vided by the water enables the whole body to be trained through it. However, physical disabil-
ity may worsen the swimming fonn and affect body balance. In this paper, the effect ofupper-
extremity amputation on the crawl swimming was firstly analyzed by simulation, since crawl
is the most common stroke, and its thrust is mainly generated by the upper limb motion. Next,
a swimming prosthetic which compensates for the amputated limb was designed using an
optimization method. Finally, a swimming prosthetic was manufactured based on the results of
the optimization, and the prosthetic' validity was confinned by a preliminary evaluation.
1 Introduction
Swimming is a suitable sport for the physically disabled since the buoyancy provided
by the water enables the whole body to be trained through it. However, the physi-
cally disabled may not be able to swim using a suitable swimming fonn, and their
body balance may worsen compared to the able-bodied. Especially for people who
have recently have been disabled, they may lose their interest in swimming since
they may not be able to swim as they used to do.
Therefore, a swimming prosthetic is needed which can decrease the influence of
the physical disability, but only few prosthetics have developed before (Sawamura
1999, http://www.oandp.com/products/trs/sports-recreation/swimming.asp, and
http://www.hanger.com). However, these studies have not clarified the influence of
amputation theoretically, and not designed optimally in order to decrease the influ-
ence of amputation.
The aim of this study is to develop a swimming prosthetic for the physically
disabled. In this paper, the case of a one-side, upper-extremity amputation, and front
crawl stroke is investigated, since the front crawl stroke is the most common in
swimming, and its thrust is mainly generated by the upper extremity motion.
In this paper, the influence of upper-extremity amputation on the front crawl
swimming stroke is analyzed first. Second, a prosthetic which decreases the influ-
432 Keiko Yoneyama and Motomu Nakashima
ence of the amputation is designed by an optimizing calculation. In addition, a pros-

thetic is manufactured based on the results of the optimization, and the validity is
examined by conducting a preliminary evaluation.
2 Simulation Analysis Qf the Influence of Upper-Extremity

Amputation on the Front Crawl
For the analysis in this chapter, the swimming human simulation model SWUM
developed by the authors (Nakashima, Miura, and Sato 2004) is used. In SWUM, the
human body is expressed as a rigid body which consists of 21 elliptic cylinders, and
the fluid force is calculated by modeling without solving the flow field. Inputting
relative motion as the joint angles into SWUM, absolute motion and the fluid force
acting on the human body are obtained as outputs. Details of SWUM are available at
the web site (http://www.swum.org/).
By changing the length of the upper arm, lower arm and hand, three kinds of
upper-extremity amputation (Sawamura 1999) are expressed. That is, ( 1): hand am-
putation (hand length 10%), (2): below-elbow amputation (lower arm length 50%),
and (3): above-elbow amputation (upper arm length 90%). In each case, only the left-
side was changed. These amputations are shown in Fig. 1.
Each case is analyzed for 20 cycles in the same swimming form as for able-
bodied swimmer. The simulation result of the locus of the human body for 20 cycles
is shown in Fig. 2. TheX-Yplane expresses water surface, and the Y-axis is the head-
ing of the able-bodied swimmer. In addition, X and Y are normalized by body height.
Stroke length L, which means the advanced length in one cycle, and 0, which means
the angle between the propulsive direction at the present and the previous cycles, are
shown in Table 1. Note that the value for the able-bodied swimmer in Table I is a
single value, whereas the values for amputations are in ranges. That is, while abso-
lute movement of the able-bodied swimmer converged after 3 cycles, the result for
amputations did not converge because of the unbalance caused by the amputation.
The results in Fig. 2 and Table 1 indicate that upper-extremity amputation has a
negative effect of swimming speed and direction. Also, the shorter the upper-
extremity is, the !arger the negative effect. Therefore, above-elbow amputation,
whose influence on swimming motion is the largest, was decided as the target of the
following part ofthis paper.
Shoulder load, for the case of the above-elbow amputation (3), was examined
next. The forcc F acting in the human body's longitudinal axis as a reaction to thrust,
and the torque T about the frontal-horizontal axis caused by this force, are considered
to be the main components of the load, and are calculated by the simulation. Fig. 3
shows the result of time history for Fand T for the 20th cycle. The values in Fig. 3
are normalized by body height, cycle, and water density. The right (non-amputated)
side is displayed with a delay of a half-cycle to be in phase with the left. The peak
=
visible in the figure (at t 19.4) is a result of the motion used to push the water. lt
can be seen that this type of amputation worsens the body balance, since the load on
the amputated shoulder is extremely lower than normal, whereas the load on the non-
amputated shoulder is !arger than normal.
Development ofSwimming Prosthetic for Physically Disabled 433
Amputation(3): 20 •
1
-~.l>ove elbow amputation 1
10 1
Amputation (2): 1
'' Below elbow amputation
,- - - -Amputation (1 ):
Hand amputation X
Fig. 1 Kinds of amputation Fig. 2 Locus ofhuman body for 20 cycles
Table 1 Stroke length L and swimming direction 0

Able-bodied Amputation(!) Amputation(2) Amputation(3)
L l.35 1.15-1.18 0.20- 1.08 0.08-1.05
0 0.0deg l.3-3.3deg 2.3-119deg 5.6-174deg
0.05
\mputatcd side
-0.05 ~ ~on-amputatcd side
-Able-bodied
19 19.5 20
!\londimcntional time t Nondimentional time t
(a) F (b) T
Fig. 3 Time history of the force in the longitudinal axis F and the torque about the
frontal-horizontal axis Ton shoulder
3 Optimal Design of Swimming Prosthetic
3.1 Optimization Procedure

A prosthetic for above-elbow amputation, whose influence on swimming stroke is
the largest as calculated in the previous analysis, was designed. A prosthetic which
has no moving parts was selected because of ease of water-proofing, cost and ease of
use in swimming. In addition, a cylinder-shaped prosthetic was chosen, since it is
bilaterally-symmetric around the longitudinal axis of the upper arm. The prosthetic
consists of 2 cylindrical units (Unit 1: Root, Unit 2: Tip) to give the degrees of free-
dom to the optimization, and 6 design variables were selected (radius, length and
relative density of each unit). Fig. 4 shows how the prosthetic fits an amputee.
In this paper, the 'optimal' prosthetic was defined as one which allows the ampu-
tee to swim with the swimming form closest to the able-bodied swimmer. In order to
realize this, load difference on the shoulders between the amputated and non-
amputated sides should be minimized. Based on this definition, the objective func-
tion was defined as a next formula:
f = (F - F ) 2 / F 2 + (T - Tl IT 2 (1)
r l r r l r
Where:
F: Force in human body's longitudinal axis on shoulder
T: Torque about frontal-horizontal axis in steady state on shoulder
r : Right (non-amputated) side
l : Left (amputated) side
2
( ) = Mean square for one cycle
That is, the objective function f represents the mean square load difference for one
cycle between right and left shoulders, divided by the mean square load on the non-
amputated shoulder. The optimum prosthetic is one which minimizes this function.
The SUBPLEX module of the OPT package in NETLIB (http://www.netlib.org/),
which is the non-linear optimization routine, was used for optimization.
3.2 Optimization Results

The optimized values of the design variables are shown in Table 2. The optimized
profiles of the prosthetic are shown in Fig.5. After optimization, unit 1 became
thicker and shorter, and its relative density became greater than unit 2.
'
Unit 2
Fig. 4 Fitting image of prosthetic Fig. 5 Optimized profile of prosthetic
Table 2 Optimized values of design variables
Radius [mm] Length [mm] Relative density

Unit 1 54.2 68.8 1.87
Unit2 29.5 373 0.099
The history offorce in the human body's longitudinal axis Fand torque about the
frontal-horizontal axis Ton the shoulder in steady state after optimization are shown
in Fig.6. Comparing Fig.6 with Fig.3, it can be seen that the load on the shoulder of
the amputated side became much greater, and that on the non-amputated side became
smaller. That is, the load on the shoulder of both the amputated and non-amputated
side became as great as that ofthe able-bodied model by optimization.
Also, the swimming speed improved to 91 % of that for the able-bodied model,
and the direction stabilized to a level of 3.5 deg curving in 1 cycle. From these re-
Development of Swimming Prosthetic for Physically Disabled 435
sults, it was found that swimming speed and direction were also improved in aiming
to match the load on the right and left shoulders.
o.os
....0.01
.002
----,.\ mputated Kle .003 - a \ ß1MJtalcd 11,dr

.0.05 - N n-ampu111od ide -----N on-■ mput.alcd fidc
---- bk-bodicd -0 049~ - - _ _ ;•~
-9i;:;"aib~k:!?l•!!!l!!!od'----,JI0
9 9.5 10
'\'.ondlfflt"nhonal 1tmc , 'ond.o:ncnuon■ I umc I
OOF ~T
Fig. 6 Time history of F and T after optimization
4 Manufacture of Swimming Prosthetic Based on Optimal Design

and Preliminary Evaluation
4.1 Manufactured Swimming Prosthetic and Experimental Method

A prosthetic was manufactured based on the optimal design found in the previous
chapter, and the validity of the optimization was examined by conducting a prelimi-
nary evaluation, in which several subjects, fitting on the prosthetic, evaluate the
effectiveness of the design.
In this experiment, since it was difficult to secure enough amputated subjects, a
prosthetic for the able-bodied, imitating the above-elbow amputation by flexing the
elbow joint, was manufactured. Size and relative density of this prosthetic are shown
in Table 3, and a fitting image of the prosthetic is shown in Fig.7. The prosthetic
consists of a semi-cylindrical socket, cushion, fixing belt, Unit 1, and Unit 2. The
size of the socket was chosen to match the diameters of the subjects' upper arm. In
addition, a cushion is used to fill an individual difference in diameter of the upper
arm. Material ofunit 1 is acrylic, and that ofunit 2 is polyethylene resin foam. These
materials were selected based on their optimized relative density, and all are rust-
proof and waterproof.
The subjects were five able-bodied healthy males. All of the subjects can swim
easily, despite no experience of serious training. There were 4 trials for each subject,
that is; trial 0: able-bodied, trial 1: imitating an amputee without prosthetic, trial 2:
imitating an amputee with an optimized prosthetic, and trial 3: imitating an amputee
with an unoptimized prosthetic. The optimized prosthetic used in trial 2 was manu-
factured based on the results of the optimization. Trial 1 did not involve any units of
prosthetic, and the unoptimized prosthetic of trial 3 consisted only unit 2. In each
trial, subjects swam at a self-determined appropriate speed.
The evaluation method was questionnaire, evaluated directly after each trial. The
evaluation questioned ease of swimming and load on the shoulders, as weil as other
areas which do not relate to the optimization, for example, the validity of the struc-
ture of the socket.
4.2 Experimental Results

The results of the evaluation for ease of swimming are shown in Fig.8. Scores are
normalized so that 100 corresponds to the level trial 0 (able-bodied), and 0 corre-
sponds to the level of trial 1 (without prosthetic). Since all evaluations of trial 2
(optimized prosthetic) are higher than those of trial 1, the effectiveness of the pros-
thetic is confirmed. In addition, since four subjects answered that trial 2 was better
than trial 3 (unoptimized prosthetic), the validity of the optimization can be also
confirmed.
Table 3 Size and relative density ofprosthetic
Length[ mm] Diameter[mm] Relative density
Unit 1 60.0 50.0 1.23
Unit 2 345 37.5 0.082
.•
0
0
60
~
-.
~
..;
40
0
20
0 •
0
Tnal numM'r
Fig. 7 Fitting image of prosthetic Fig. 8 Evaluation about ease of swimming
5 Conclusions
In this paper, the influence of an above-elbow amputation was analyzed using simu-
lation, and a swimming prosthetic which is able to decrease this influence was de-
signed by an optimization. A prosthetic was manufactured based on the optimization,
and a preliminary evaluation was performed.
Following an analysis of 3 kinds of upper-extremity amputation, the influence of
above-elbow amputation was found to be the greatest. Then, in the result for the
optimal design of a swimming prosthetic (consisting of 2 cylinder units without
moving parts), an optimized prosthetic can greatly improve the balance of load on
both right and left shoulders. Finally, the effectiveness of a prosthetic and the valid-
ity of the optimization were confirmed by a preliminary evaluation.
In this study, the prosthetic for the able-bodied was manufactured, because ofthe
difficulty to secure enough amputated subjects. Therefore, in the future, modification
based on the result of evaluation by amputee will be expected.
References
Nakashima M., Miura Y., and Sato K. (2004) Swimming Human Model "SWUM" to Analyze
Dynamical Swimming Problems. The Engineering of Sport 5, vol. l. 594-600
Sawamura S., (1999) Amputation and prosthetic 4th edition. Ishiyaku Publishers, Inc. Tokyo
(In Japanese)
Retraction Note to: Riding on Air: A New Theory for Lift
Mechanics of Downhill Skiing and Snowboarding
Qianhong Wu, Yesim Igci, Yiannis Andrepoulos and
Sheldon Weinbaum et al.
Retraction Note to: Qianhong Wu, Yesim Igci, Yiannis Andrepoulos and
Sheldon Weinbaum et al., Riding on Air: A New Theory for Lift
Mechanics of Downhill Skiing and Snowboarding,
DOI: 10.1007/978-0-387-46050-5_50
The Editors have retracted this Conference Paper because the content overlaps
significantly with a previously published article (Wu et al., 2006). Qianhong Wu
does not agree with this retraction. Yiannis Andreopoulos and Sheldon
Weinbaum have not responded to correspondence about this retraction. The
Editors could not obtain a current email address for Yesim Igci. Wu, Q., Igci,
Y., Andreopoulos, Y., & Weinbaum, S. (2006) Lift mechanics of downhill
skiing and snowboarding. Medicine and Science in Sports and Exercise, 38(6),
1132-1146.
__________________________
The updated online version of the original chapter can be found at
https://doi.org/10.1007/978-0-387-46050-5_50
E. Fozzy et al. (eds.), The Engineering of Sport 6,

DOI 10.1007/978-0-387-46050-5_77 C1
© Springer Science+Business Media LLC, 2021
Author Index
Ae M., 137 Fitzpatrick K., 155

Alam F., 127 Fleming P., 241
Alcantara E., 351 Fliege 1., 97
Anderson B.C., 167 Fuss F.K., 43, 51, 57
Anderson R., 155
Andreopoulos Y., 281 Gamez J., 351
Asai T., 327 Geraldy A., 97
Ashley A., 305 Goodwill S.R., 379, 385
AuerM., 287 Gotzhein R., 97
Grund T., 339
Baca A., 407 Guerrini G., 413
Barker M., 419
Bäurle L., 251 Haake S.J., 115, 235, 379,385,401
Bissuel M., 69 Harland A., 205, 211, 357
Blackbum K., 217 Hart J., 401
Bourban, P-E., 263 Heller M., 407
Brabant, J-D., 69 Hintzy F., 85
Bramley A.N., 63 Hirai N., 275
Bray K., 321 Hiramatsu Y., 333
Brown S., 425 Holmes C., 211
Honda Y., 367
Carre M.J., 115, 203, 235, 313 Hopkins J., 23
Carswell D., 425 Horvais N ., 85
Caton C., 79 Hoshino Y., 161
Chong L., 23 Hubbard M., 177, 183
Choppin S., 385 Hytjan A., 305
Combaz E., 263
Comin M., 351 Igci Y., 281
Cottey R., 391 Iida H., 137
Cox A., 205 Irwin G., 189, 195
Curtis D., 401 Ison A., 29
Custer D., 45 Isotti A., 413
Dixon N., 241 JaitnerT.,97, 103

Douglas J., 379 James D.A., 235
Drane P.J., 5 James 1., 217,229,315
Jenkins M., 79
Fauve M., 251,263,287,299 Jennings-Temple M., 315
Federolf P., 287, 299 Jones R., 211,373
Fischer C., 263 Justham L., 205
438 Author Index
Kaczmarski H., 23 Pallis J.M., 399

Kagawa H., 35, 293 Petrone N., 345,413
Kaps P., 269 Petzing J., 211
KerwinD.G., 175,189,195,321 Phillips A., 63
Kimme! W., 183 Plummer C.J.G., 263
Kobayashi 0., 223, 327 Prat J., 351
Kobayashi Y., 161
KochJ., 103 Redfield R., 109
Koike S., 137 Rhyner H., 251,263,287,299
Kornfeind P., 407 Richard M., 69
Kotze J., 391 Robazza C., 345
Kuhn T., 97 Ronkainen J., 357
Rosa D., 351
Lammer H., 391 Russell D., 11
Lavery N., 425
Leaney P., 373 Sajima T., 143
Leeds-Harrison P., 315 Sakashita R., 327
Lukes R., 115 Samozino P, 85
Lüthi A., 287,299 Schiestl M., 269
Schindelwig K., 257
Manin L., 69 Scott N., 293
Manson J.A.E., 263 Senner V., 121,249,339
Marcolin G., 345 Seo K., 223, 275, 327
Martinez A., 351 Shaw R.H., 5, 17
Mather S., 135 Sheets A.L., 177
McLeod A., 217 Sherwood J.A., 5, 17
Michaud V., 263 Shipton P., 229
Miller P., 305 ShirakiH., 137
Miller S., 365, 379 Smith L., 29
Mizota T., 149 Strangwood M., 77, 79
Mössner M., 269 Steele C., 373
Mueller M., 121 Stefanyshyn D.J., 167
Murakami M., 223, 275 SubicA.,91, 127
Such M.J ., 351
Nachbauer W., 257, 269 Sutela C., 109
Nakashima M., 431 Suzuki S., 161
Naruo T., 149
Nathan A., 3, 23 Takahashi M., 35
Niebhur D., 103 Tempia A., 91
Niegl G., 51, 57 Trapp M., 103
Tsunoda M., 143
Ohgi Y., 275
Ohtsuki A, 333 Vera P., 351
Oshima S., 333 Vickers A., 229
Vogwell J., 63
Pagliarella R.M., 91
Author Index 439
Walshe A., 305 Yabu M., 143

Watkins S., 127 Yamaguchi T., 143
Webei C., 97 Yoneyama T., 35, 293, 431
Weber M., 305 Y oung C., 241
Weinbaum S., 281
West A., 205 Zable J., 305
Wheeler M., 305 Zimgibl W., 391
Wimmer M., 121
Wood G., 217
Wright I.C., 167
Wu Q., 281
Subject Index
accelerometer, 155 tennis elbow, 391

aerodynamics
atmospheric boundary layer, 149 climbing
baseball, 23 dynamics of speed, 51
cycling, 115 equipment
helmets, 127 anchor, 45
golfball, 149 belay, 69
dimple design, 143 experimentation, 69
magnus force, 223 fall arrest, 45
rugby ball, 223 holds, 57
ski, 269 instrumentation, 69
soccer, 327 modelling, 69
spinning ball, 23 rock, 69
water ski jumping, 401 rope forces in a fall, 63
wind tunnel, 127, 149, 327 safety, 45, 63
air cannon, 17 sport climbing, 57
coefficient of restitution, 29
ball spin computational fluid dynamics (CFD)
baseball, 23 surfboard fins, 425
tennis, 379 water ski jumping, 401
baseball, 23 cricket
bat, 5, 11, 35 bowling delivery, 205
durability, 5 oblique ball impact on pitch, 235
holding methods, 35 pitch rolling, 229
properties, 11, 17 safety, 229
sensation of sting, 11 cycling
softball, 29 aerodynamics, 115, 127
biomechanics ambient intelligence system, 97
baseball, 35 biomechanics, 121
cycling, 85, 103 helmets, 127
total knee arthoplasty, 121 indoor simulation, 103
disability, 431 mountain bikes
golf, 155 rear linkage dynamics, 91
swing, 161, 167 rear shock modelling, 109
wrist release, 161 non-circular chain rings, 85
gymnastics physiological data, 103
musculoskeletal work, 195 racing wheels, 79
shoulder compliance, 177 rolling resistance, 115
kayak,413 team training, 97, 103
motion analysis, 413 total knee arthoplasty, 121
orthosis development, 431 track cycling, 115
ski, 257,293
swimming, 431 disability, 431
442 Subject Index
drag moment, 137

cycling helmets, 127 kinetic energy, 167
soccer ball, 327 real-time kinematic data, 155
water ski jumping, 401 robot, 161
dynamics swing, 161, 167
climbing wrist release, 161
climbing speed, 51 gymnastics
sport climbing, 57 balance strategies, 183
cycling, 91 dynamics, 189
gymnastics, 189 high bar, 177, 189, 195
mountain bikes, 91 musculoskeletal work, 195
ski, 263 shoulder compliance, 177
swmg
ergometer giant, 177
kayak,413 longswing, 189, 195
rowing, 407 surface compliance, 183
experimentation gyroscope, 155
baseball, 5, 17
climbing, 69 hydrodynamics, 425
golf, 149
traction testing, 339 ice
friction, 251
field hockey, 241 ski, 251
finite element analysis (FEA) impact
ski, 263 oblique cricket ball on pitch, 235
surfboard fins, 425 racket, 373
flow visualisation tennis balls, 367
titanium tetrachloride, 327 injuries
football tennis elbow, 391
(see also soccer) instrumentation
goal keepers diving envelope, 32 i accelerometer, 155
modelling, 321 climbing, 51, 57
penalty kick, 321 golf, 137
pitch playing quality, 315 gyroscope, 155
soil physical conditions, 315 kayak, 413
turf, 315 rowing, 407
friction, 251, 269, 305 ski, 257, 293
golf kayak,413
ball
aerodynamics, 143 lawn sport
dimple design, 143 cricket, 205, 229, 235
trajectory analysis, 149 field hockey, 241
clubs rugby, 211
instrumented grip handle, lift
137 baseball, 23
Subject Index 443
rugby ball, 223 climbing, 45, 63

soccer ball, 327 cricket, 229
water ski jumping, 401 white water rescue, 419
lift mechanics snowboarding, 281 scanning laser doppler vibrometer
(SLDV), 357
magnus force, 223 simulation
materials computational fluid dynamics
bicycle wheels structural foams, (CFD), 401,425
79 cycling, 103
ski, 251, 263 finite element analysis (FEA),
measurement 263,425
football, 321 ski
golf, 137 alpine skis, 263, 287
mechanical performance base materials, 251
artificial turf for soccer, 351 binding, 287
mechanical properties carving, 299
bindings, 287 downhill, 269,281
ski, 287 friction, 251, 269, 305
modelling ice, 251
climbing, 69 jumping, 257, 275
football, 321 leg joint motion, 293
mountain bike rear shock, l 09 mechanical properties, 287, 299
track cycling, 115 motion analysis, 275
moment of inertia, l 7 snow, 269,305
motion analysis, 23 snowboarding, 281
kayak,413 snow
ski, 275 friction, 269, 305
mountain bikes ski, 269, 305
rear linkage dynamics, 91 soccer
rear shock modelling, 109 (see also football)
artificial turf, 351
orthosis, 431 ball
flow visualisation, 327
rescue ropes, 419 modal analysis, 357
robot, 161 non-spinning, 327
rolling resistance, 115 spinning, 327
rowing ergometers, 407 boots, 339
rowing instrumentation, 407 gaze point analysis, 333
rugby instep kick, 345
ball launch characteristics, 211 movement prediction, 333
kick, 211 scanning laser doppler vibrometer
kick multi-optimisation, 223 (SLDV), 357
magnus force, 223 traction testing, 339
vibration, 357
safety softball, 29
buoyancy aids, 419 sport climbing, 57
444 Subject Index
swimming, 431 soil physical conditions, 315

surface traction, 315
tennis synthetic
3d player testing, 385 ball interaction, 241
active damping technology, 391 evaluation, 21 7, 241
ball wear, 217
hollow, 367
spin, 379 vibration
wear, 373 scanning laser doppler vibrometer
racket, 373, 379 (SLDV), 357
tennis elbow, 391 soccer, 357
thermal efficiency, 127
trajectory water
baseball, 23 kayak,413
golfball, 149 rowing, 407
kayak paddle, 413 ski jumping, 401
rugby ball, 211 surfing, 425
turf swimming, 431
artificial, 351 white water rescue, 419
ground reaction loads, 345 water ski jumping, 401
mechanical behaviour of soil, 229 wear
rootzones, 315 synthetic turf, 217
rubber infill morphology, 351 tennis ball, 373

2006 Book TheEngineeringOfSport6

Uploaded by

Copyright:

Available Formats

2006 Book TheEngineeringOfSport6

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2006 Book TheEngineeringOfSport6

Uploaded by

Copyright:

Available Formats

The Engineering of Sport 6

Eckehard Fozzy Moritz and Steve Haake (Eds.)

The Engineering of Sport 6

Library of Congress Control Numbcr: 2006927112

Printed on acid-free paper.

© 2006 Springer Science+Business Media LLC, corrected publication 2021

Printed in the United States of America. (EB)

Eckehard Fozzy Moritz

An Experimental lnvestigation of Baseball Bat Durability 5

Bending Modes, Damping, and the Sensation of Sting in Baseball Bats 11

Experimental Investigations of the Relationship of Baseball Bat Properties 17

The Effect of Spin on the Flight of a Baseball 23

Rigid Wall Effects on Softball Coefficient ofRestitution Measurements 29

The Effect of Holding Methods on a Baseball Bat Performance Estimation 35

2 Climbing - Instrumentation And Testing 0/ Equipment

An Estimation ofthe Load Rate Imparted to a Climbing Anchor During Fall 45

Dynamics of Speed Climbing 51

lnstrumented Climbing Holds and Dynamics of Sport Climbing 57

Forces Generated in a Climbing Rope During a Fall 63

Rock Climbing Belay Device Analysis, Experiments and Modelling 69

Thermo-mechanical Modification Techniques for Structural Foams used in 79

The Effect of a Non Circular Chainring on Cycling Performance 85

Dynamic Characteristics ofModem Mountain Bikes Rear Linkages 91

An Ambient Intelligence System to Assist Team Training and Competition in 97

Indoor-Simulation ofTeam Training in Cycling 103

A Bond Graph Model of a Full-Suspension Mountain Bicycle Rear Shock 109

Track Cycling: An Analytical Model 115

Forces During Cycling After Total Knee Arthoplasty 121

A Study of Aerodynamic Drag and Thermal Efficiency of a Series of Bicycle 127

The Aerodynamic Influence of Dimple Design on Flying Golf Ball 143

Experimental Verification of Trajectory Analysis of Golf Ball Under 149

Validation of Accelerometers And Gyroscopes to Provide Real-Time 155

Segmental Sequencing of Kinetic Energy in the Golf Swing 167

Effects of Horizontal Surface Compliance on Balance Strategies 183

Predicting High Bar Forces in the Longswing 189

Musculoskeletal Work in the Longswing on High Bar 195

Quantification of the Cricket Bowling Delivery; a Study of Elite Players to 205

Ball Launch Characteristics for Elite Rugby Union Players 211

A Novel Quantitative Method for the Determination of Wear in an Installed 217

Multi-Optimization of Three Kicks in Rugby 223

The Mechanical Behaviour of Cricket Soils During Preparation by Rolling 229

Studies on the Oblique Impact of a Cricket Ball on a Cricket Pitch 235

7 Skiing, Snowboarding and Ski Jumping

Biomechanical Instrumentation of the Bergisel Jumping Hili in Innsbruck 257

Dynamic Properties ofMaterials for Alpine Skis 263

Measurement of Jumper's Body Motion in Ski Jumping 275

Subjective Evaluation ofthe Performance of Alpine Skisand Correlations 287

Development of a Prototype that Measures the Coefficient of Friction 305

Measuring and Modelling the Goalkeeper's Diving Envelope in a Penalty 321

Gaze Point Analysis in Movement Prediction of Soccer Players By Image 333

Analysis of the Influence of Rubber lnfill on the Mechanical Performance of 351

Normal Impact ofHollow Balls on Fiat Surfaces 367

Factors in Tennis Ball Wear 373

Measuring Ball Spin off a Tennis Racket 379

30 Player Testing in Tennis 385

Computational Fluid Dynamic Analysis of a Water Ski Jumper 401

Feedback Systems in Rowing 407

Biomechanical Analysis of Olympic Kayak Athletes During Indoor Paddling 413

Computational Modelling of Surfboard Fins for Enhanced Performance 425

Development of Swimming Prosthetic for Physically Disabled (Optimal 431