2007:31

DOCTORA L T H E S I S

Plate Buckling Resistance

Patch Loading of Longitudinally Stiffened Webs and Local Buckling

Mattias Clarin

Luleå University of Technology

Department of Civil, Mining and Environmental Engineering
Division of Structural Engineering - Steel Structures

2007:31|: -1544|: - -- 07⁄31 -- 

 Doctoral Thesis 2007:31

Plate Buckling Resistance

- Patch Loading of Longitudinally Stiffened Webs and Local Buckling -

Mattias Clarin

Luleå University of Technology

Dept. of Civil, Mining and Environmental Engineering
Division of Structural Engineering - Steel Structures
Luleå, August 2007
 Preface

Preface

A couple of weeks ago, I spent some time in contemplation over how different time may be
experienced. More than five years have passed since the 1st of April 2002 which was the day I
took my first tremulous steps towards this thesis. That I actually was moving towards writing a
doctoral thesis wasn’t that obvious during that day, neither during many of the days to follow.
Nevertheless, that day was the beginning of a period which has contained so many things. A
period which will be ended by this thesis. Five years are a short period in some senses, a very
long in others. Isn’t it strange that end and beginning can be so close in some ways, distant in
others? Just as a reminder, this preface is a beginning, a beginning of an end which in fact this
thesis is...

During this period in my life, I have been fortunate to be aided by my supervisor Professor
Ove Lagerqvist. From the earlier days mainly investigating residual stresses and local buckling
to patch loading resistance in the end, your experience and knowledge in the field has been
invaluable. Though, working with you has also brought many memorable times regarding other
things. Reflecting over music, books and other small and large things in life has been delightful.
I thank you not only for drafting me to the Division of Steel Structures in my beginning, but also
for your time, support and friendship!

The first three years of the work resulting in this thesis I was favoured to get great assistance
by Dr. Eva Pètursson. As co-supervisor Eva was reading, correcting and questioning but maybe
more important, supporting and encouraging. When Eva engaged in new challenges outside the
university I was fortunate to get a strong “substitute” on the co-supervisor position; Professor
Bernt Johansson. Using his vast knowledge, calmly explaining and answering my questions has
been the very best support a Ph.D. aspirant can get. I thank you both and I am truly grateful for
assisting me during this period!

As always, research is maybe not impossible, though difficult to conduct without financial
support. The financial aid provided by Luleå University of Technology (LTU) and by RFCS -
Research Fund for Coal and Steel within the frame of the two projects LiftHigh - Efficient

I
Lifting Equipment with Extra High Strength Steel and ComBri - Competitive Steel and
Composite Bridges by Improved Steel Plated Structures are gratefully acknowledged.

I am also very grateful for the friendly support and help given the staff at Complab which
have helped out with huge effort during the experimental work. Special gratitude is paid towards
Lars Åström, Georg Danielsson and Claes Fahlesson for aid during the all the tests!

The immensely friendly and warm atmosphere at the Division of Structural Engineering has
been a great aid in the days starting not that productive. This especially regarding the research
group for Steel Structures with which I have shared many good times. Supporting late night and
week-end workers, coffee breaks, research discussions; the memorable occasions are so many...
I have enjoyed the period with you and will miss you all!

I can hardly imagine how this period would have been without my companion Jonas Gozzi.
Much has been going on during these years, work- and otherwise. The former stretch from the
beginning of office and computer sharing, via doctoral courses, laboratory work and assisting
guests researchers to the thesis discussions in the end. The latter stretches over an even wider
spectra of events; caravan customizing, skiing, transparent toilet doors, Sarek, the queues of
China, popcorn dinners and much much more. It has been a pure pleasure my friend!

Nonetheless, nothing of this would have been possible without the support, understanding
and love of my cherished Annica.You kept encouraging me with your hearty laughter and
glowing and kind spirit regardless how messy and absent-minded I was. As much as this is the
beginning of the end of this period, the end is the beginning of a new period for us. At the same
place, at the same time, how sweet it will be!

Consequently, all periods come to an end, also prefaces... However, this preface was just the
beginning of an end. Though, an end which is the beginning of something yet not written. Ergo,
time is a strange thing. Occasionally slow moving, usually fast. Aye, plainly strange it is...

Luleå, 25th of August, 2007

Mattias Clarin

II
 Abstract

Abstract

Incremental launching of steel bridges is a demanding undertaking, on the erection site as

well as on the designers desk. Not seldom, the structure itself is during the launching subjected
to high concentrated forces on the lower flange when passing over a launching shoe or an
intermediate support (e.g. column). These concentrated forces, commonly referred to as patch
loads, may be of such magnitude that it governs the thickness of the web in the bridge girder.
Though, a small increase in web thickness leads to a substantial gain of steel weight of the
bridge. Hence also a higher material cost.

One solution to this problem is to increase the buckling resistance of the web with the use of
a longitudinal stiffener of open (a plate) or closed type (closed profile of e.g. V-shape). The
improved patch load resistance is in the european design code EN 1993-1-5 nowadays
determined with the help of the yield resistance for the web and contributing parts of the loaded
flange reduced with a factor dependent of the slenderness of the web and the influence of one
or more longitudinal stiffeners. Parts in the expression for the yield resistance and the reduction
factor have been somewhat questioned and over the years a substantial amount of tests and FE
simulations of longitudinally stiffened webs has been carried out. This research work has
produced a large amount of test data which has been used herein to further improve the
prediction of the patch load resistance of longitudinally stiffened steel girder webs.

Based on the use of the gathered test data from the literature and previously done research, a
calibrated patch load resistance function was developed for both open and closed longitudinal
stiffeners. Furthermore, a partial safety factor for the proposal was determined according to the
guidelines in EN 1990 (2002). In all, the proposal was shown to clearly improve the accuracy
of resistance prediction when compared to other resistance models as well as the EN 1993-1-5.

Another questioned part in the commonly used design codes is the reduction function
regarding local buckling under uniform in-plane compression. The nowadays used function (the
Winter function) has been developed during the 1930’ies and was based on tests on cold formed
specimens. This reduction function has been criticized as being too optimistic regarding plates
with large welds. A series of tests on welded specimens made of high strength steel with large

III
welds was conducted to investigate the aforementioned concerns. Along with test data found in
literature survey, the Winter function was proven to be too optimistic regarding these heavily
welded plates. A new reduction function, based on the test data, was proposed and validated
through a comparison with the available experimental results.

IV
 Notations & Symbols

Notations & Symbols

The notations and symbols used in this thesis are described within this chapter. The notations
and symbols are listed in alphabetical order, roman and greek respectively.

Roman notations and symbols

a - Weld size, numerical coefficient or panel length

A - Area

A5 - Elongation measurement, 5%

Afl - Area of flange

Aw - Area of web

b - Correction factor

b - Width of plate

b1 - Depth / Height of upper panel

beff - Effective width

bf - Width of flange

bst - Width of longitudinal stiffener

cu - Half the length in the web which resists the applied force

Co - Parameter used for calculating the buckling coefficient of a

longitudinally stiffened web

d - Plate thickness

D - Flexural plate rigidity

V
 E - Modulus of elasticity, Youngs modulus

fu - Ultimate tensile strength

fue - Ultimate strength, electrode

fyk - Characteristic value of yield strength

fy - Yield strength

fye - Yield strength, electrode

fyf - Yield strength of flange

fyw - Yield strength of web

F - Force

Fcr - Elastic critical buckling load

Fcr1 - Elastic critical buckling load for the upper (directly loaded) panel,
patch loading

Fcr2 - Elastic critical buckling load for the whole web panel, patch
loading

Fexp - Ultimate load from tests

FE - Applied transverse load

FR - Predicted load resistance

FRd - Design resistance

FRl - Predicted resistance for a longitudinally stiffened web according

to an amplification factor model

Fu - Ultimate resistance

Fy - Yield resistance

grt(X) - Resistance function of basic variables in design model

h - Height / length of plate in specimen

h1 - Distance between upper flange and centre of gravity of

longitudinal stiffener

hst,o - Depth / Height of closed stiffener, outer dimension

VI
 Notations & Symbols

hst,w - Depth / Height of closed stiffener, dimension closest to web

hw - Depth / Height of web

If - Moment of inertia, flange

Ist - Moment of inertia, longitudinal stiffener

k - Coefficient

kc - Error term

kcr - Buckling load coefficient

kd,n - Design fractile factor

kF - Buckling load coefficient, patch loading

kF1 - Buckling load coefficient for the upper (directly loaded) panel,
patch loading

kF2 - Buckling load coefficient for the whole web panel, patch
loading

kn - Characteristic fractile factor

ksl - Buckling load coefficient addition for a longitudinally stiffened

web

kV - Buckling load coefficient according to EN 1993-1-5

L - Plate length

m, n - Number of half waves over plate

ME - Applied bending moment

Mi - Plastic moment resistance, inner plastic hinge in flange

Mo - Plastic moment resistance, outer plastic hinge in flange

Mpf - Plastic moment resistance, flange

Mpw - Plastic moment resistance, web

MR - Bending moment resistance according to EN 1993-1-5

N - Normal force

VII
 Ncr - Critical load

Nel - Buckling load

Nx, Ny - Normal forces per unit distance

Nxy - Shearing force per unit distance

r - Value of resistance

rd - Design value of the resistance

re - Experimental resistance

rk - Characteristic resistance value

rm - Predicted resistance by the resistance function using the mean

values of basic variables, i.e. grt(Xm)

rn - Nominal resistance value

rt - Resistance predicted by the resistance function grt(X)

Rm - Ultimate resistance

Rp0.2 - 0,2% Proof stress

s - Standard deviation

ss - Loaded length

sy - Distance between plastic hinges in loaded flange

t - Thickness

tf - Thickness of flange

ti - Flange thickness, idealized

tst - Thickness of longitudinal stiffener

tw - Thickness of web

T - External work

U - Internal work

VG - Coefficient of variation of the error term G

VIII
 Notations & Symbols

Vfy - Coefficient of variation of the yield resistance

Vrt - Coefficient of variation of the resistance function

w - Amplitude of lateral deflection

wo - Initial amplitude of lateral deflection

W - Section modulus

Weff - Effective section modulus according to EN 1993-1-5

x, y, z - Cartesian coordinates

X - Array of j basic variables X1, ..., Xj

Xm - Mean value of the basic variable

Greek notations and symbols

D - Angle

D - Distance between yield lines in web

D, DF - Imperfection factor, reduction function

E - Distance between plastic hinges

J - Boundary condition dependent parameter

JM - Partial factor for resistance

J M* - Corrected partial factor for resistance

JM1 - Partial factor for members susceptible to instability

Jst - Relative flexural rigidity of longitudinal stiffener

Jst,t - Relative flexural transition rigidity of longitudinal stiffener

G - Error term or deformation

Gw - In-plane deformation of web

' - Logarithm of the error term G

H - Strain or Material depentent parameter

K - Correction factor for bending moment or imperfection factor

IX
 T - Angle defining deformation of web with yield lines

O0, O0F - Plateau length, reduction function

OF - Plate slendernes parameter, patch loading

OP - Plate slendernes parameter, local buckling

Q - Poisson’s ratio, Q = 0,3 if nothing else is stated

V - Stress

Vc , Vrc - Compressive residual stress

Vcr - Critical stress

Vmax - Maximum stress

Vmin - Minimum stress

Vr - Residual stress

Vrs - Residual stress

Vu - Ultimate stress

Vw - Stress in web

Vx - Normal stress

Ist - Relative torsional rigidity of longitudinal stiffener

F - Reduction factor

FF - Reduction factor, patch loading

FP - Reduction factor, local buckling

\ - Stress ratio

Throughout the thesis mean values are marked overlined, e.g. fy represents the mean yield
strength.

X
 Table of Contents

Table of Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . III
Notations & Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . V
Chapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Local buckling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Patch loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Purpose and Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5.1 Effective cross-section of longitudinal stiffeners . . . . . . . . . . . . . . . . . . . . 5
1.5.2 Bending resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 Disposition of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 2: Plate Buckling - Theory . . . . . . . . . . . . . . . . . . . . 9
2.1 Plate buckling theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Elastic analysis / Calculation of critical load . . . . . . . . . . . . . . . . . . . . . . 11
2.1.2 Simply supported plates under uniform compression. . . . . . . . . . . . . . . . 13
2.1.3 Initial plate imperfections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.4 Geometric imperfections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.5 Residual stresses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2 The effective width concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.1 The von Kármán effective-width formula . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.2 The Winter function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 Patch loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3.1 Resistance for girders without longitudinal stiffeners. . . . . . . . . . . . . . . . 24
2.3.2 Resistance for girders with longitudinal stiffeners . . . . . . . . . . . . . . . . . . 33
2.3.3 Interaction with bending. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Summary of the theoretical review . . . . . . . . . . . . . . . . . . . . . . . 44
Chapter 3: Patch Loading - Test Results . . . . . . . . . . . . . . . . 45

XI
 3.1 Patch loading experiments on longitudinally stiffened girders . . . . 46
3.1.1 Rockey et. al (1978) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.1.2 Bergfelt (1979) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.1.3 Bergfelt (1983) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.1.4 Galea et. al (1987) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1.5 Shimizu et. al (1987) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.1.6 Janus et. al (1988) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.1.7 Dubas and Tschamper (1990). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.8 Dogaki et. al (1990). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.9 Carretero and Lebet (1998) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.1.10 Walbridge and Lebet (2001) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.1.11 Kuhlmann and Seitz (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2 Numerical simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2.1 Davaine (2005) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.3 Summary of the experimental review . . . . . . . . . . . . . . . . . . . . . 62
Chapter 4: Patch Loading - Design Proposal . . . . . . . . . . . . . 65
4.1 Yield resistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2 Elastic critical load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3 Reduction function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.4 Proposal of design approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.5 Validation of the design proposal. . . . . . . . . . . . . . . . . . . . . . . . . 70
4.6 Comparison with other models . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.7 Interaction with bending moment . . . . . . . . . . . . . . . . . . . . . . . . 75
4.8 Summary of the proposed design procedure. . . . . . . . . . . . . . . . . 77
4.9 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Chapter 5: Local Buckling - Test Results . . . . . . . . . . . . . . . . 83
5.1 Nishino et. al (1967). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1.1 Test Setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.1.2 Test results and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.2 Dwight et. al (1968) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3 Dwight and Moxham (1969). . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.1 Tests made by J.D. Harrison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.3.2 Tests made by K.E. Moxham . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Fukumoto and Itoh (1984) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.5 Rasmussen and Hancock (1992) . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.5.1 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.5.2 Residual stress measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.5.3 Test results and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.6 Möller and Johansson (1995). . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

XII
 Table of Contents

5.6.1 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.6.2 Test results and conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.7 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
Chapter 6: Local Buckling - Experimental Work . . . . . . . . . . 95
6.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Experimental investigation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.3 Uniaxial tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3.1 Specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3.2 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.3.3 Test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4 Buckling tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.4.1 Specimens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.4.2 Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .100
6.4.3 Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .101
6.4.4 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .102
6.5 Test evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Chapter 7: Local Buckling - Design Proposal . . . . . . . . . . . . . 107
7.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.2 Proposal and validation of new reduction function . . . . . . . . . . . 108
7.3 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Chapter 8: Discussion and Conclusions . . . . . . . . . . . . . . . . . 113
8.1 Patch loading - Discussion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8.2 Local buckling - Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
8.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.4 Proposals for future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Chapter 9: References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
APPENDIX A: Patch Loading - Specimen Data. . . . . . . . . . . . . 125
APPENDIX B: Patch Loading - Further Evaluation . . . . . . . . . . 143
APPENDIX C: Local Buckling - Further Evaluation . . . . . . . . . . 163

XIII
XIV
 Introduction

Chapter 1:
Introduction

The civil engineering of today is a demanding undertaking. A structural designer has not only
to guarantee that the structure to be built is safe to use, but also take economical, environmental
and architectural aspects into account. A part of this work is to decide what material to use, e.g.
a materially homogeneous composed structure or a composite creation? The early civil
engineers often used what was nearby, usually stone or timber. Today there are a multitude of
different materials available on the market. Concrete, timber, fibre reinforced polymers, glass
and steel are all examples of materials used in civil structures today.

When the structural steel entered the market, the civil engineers were provided with a
possibility to design more slender structures than before. However, making the structural
members more slender in order to minimize the use of material (dead weight and economy) the
designers also had to pay an increased attention to possible buckling related issues.

The designer has a couple of tools to use to make their structure as perfect as possible with
respect to the aspects of safety, economy, architecture and environment. The material was one
example of these, another is the design regulations which is a way for the designer to ensure the
safety of the structure. However, the design codes available for the designer has to be applicable
with respect to not only safety (yet being economically efficient), but also be kept up to date
with respect to advances by the steel industry and the production methods of civil structures.
Developing and up-dating the design codes are usually some of the work a structural researcher
is facing, e.g. through European research projects.

The work presented within this thesis is an example of some of the outcome of such projects.
The two RFCS (Research Fund for Coal and Steel) sponsored research projects LiftHigh -
“Efficient Lifting Equipment with Extra High Strength Steel” and ComBri - “Competitive Steel
and Composite Bridges by Improved Steel Plated Structures” were the frame within which the
herein presented research work was conducted.

The project LiftHigh was initiated in 2002 and under three years an investigation of how
using steels with a higher strength than commonly used (e.g. fy > 600 MPa) could benefit the

1
crane industry was carried out. This with respect to an increased lifting capacity and / or a
reduced dead weight of the products. The work in this thesis focused on investigating the
resistance of plates subjected to uniformly distributed compressive stresses, referred to as local
buckling, was conducted as a part of the LiftHigh project.

The other part of this thesis, focusing on the resistance of longitudinally stiffened plates
subjected to in-plane local compressive loads, referred to as patch loading, was conducted as a
part of the ComBri project. The ComBri project was a three year research activity, started in
2003. The main objectives of the research work was to promote the use of steel plated structures
mainly in bridge applications and to further improve the cross-sections of steel in both
composite and pure steel bridges. This with respect to design with respect to both final and
erection state.

1.1. Local buckling

As mentioned earlier, designing a structure of steel often includes slender members / cross-
sections which have to be treated safely and properly with respect to possible buckling
phenomena. Even though the presented work within this thesis only comprises plates subjected
to uniformly distributed compressive in-plane stresses, buckling of a plate is not out of
consideration if the stresses differ from being evenly distributed. Applying bending moments
and shear stresses also induces in-plane stresses, i.e. plate buckling has to be considered.

In the European design regulation used for design of plated steel structural elements, EN
1993-1-5, the method of taking local buckling into account is based on the effective width
concept, originated from the work of Theodor von Kármán and his colleagues in the 1930’ies.
Though, original concept by von Kármán was refined in the years to follow and in the end of
the 1940’ies George Winter presented a modified version of the effective width concept. The
work by Winter ended up in a reduction function validated with respect to a large quantity of
experiments, i.e. on plates with imperfections. This was the major difference between the work
of von Kármán and Winter, the former was derived with respect to a perfect plate without any
imperfections. However, the tests conducted by Winter only comprised cold-formed plates
which imperfection wise often differs from corresponding welded plates.

A number of researchers world-wide have since then performed investigations to investigate

if the ultimate resistance of welded plates is the same as the cold formed plates of Winter, i.e.
the Winter reduction function. However, many of these tests presented in e.g. Nishino et. al
(1967), Dwight et. al (1968), Fukumoto and Itoh (1984) showed that the Winter function tends
to overestimate the ultimate resistance of more slender welded plates. Furthermore, other
researchers, e.g. Veljkovic and Johansson (2001) has by numerical simulations shown that the
Winter function is more suitable to use for plates without residual stresses, i.e. not in as-welded
condition. Though the Winter function is still used in the EN 1993-1-5 to estimate the buckling
resistance of both cold-formed and welded plates under in-plane compression.

2
 Introduction

1.2. Patch loading

Another type of plate buckling frequently encountered in practice, is buckling of a girder web
subjected to a locally applied in-plane compressive load. Local in the sense of not being
distributed over the whole width of the plate, in this case the girder web. Examples of when this
load case may occur may be found in numerous structural applications, e.g. wheel loads on
gantry girders, purlins on main frame structures, crane girders and as in line with the focus of
the ComBri project, during launching of bridge girders.

Emphasizing that steel structures usually are made slender on economical basis, the reader
understands in which manner a modern steel bridge, composite or pure steel, is designed. The
common way to ensure that the buckling resistance of a slender bridge girder web are sufficient,
may be either to increase the web thickness of web or by using stiffeners. The choice is in most
cases based on total economy, e.g. labour costs for the extra welding needed to reinforce the
web with a stiffener versus the cost for increasing the web thickness. However, vertical
stiffeners are commonly used to resist the static support reactions (patch loading) from dead
weight of the bridge and external loads in the final state. Though, when constructing a large
bridge, the common erection procedure is to incrementally launch the bridge in place. The
bridge girders are assembled at one end and pushed out over the intermediate supports along the
span of the bridge.

When a bridge girder is launched, the support reactions is not statically applied as in the final
state, but is moving along the span of the bridge. Thus, the support reactions is not possible to
manage using vertical stiffeners. Furthermore, since the bridge girder will be supported as a
console beam during most of the launching, large bending moments are added to the patch
loading. For girders with a depth up to approximately 3 m, the buckling resistance is commonly
ensured increasing the web thickness. However, regarding deeper cross-sections and larger
spans, the bending moments may increase in such an extent that the most efficient way to
guarantee the buckling resistance of the web is to reinforce the web by one or several
longitudinal stiffeners. Reinforcing a girder web with longitudinal stiffeners not only increases
the bending resistance but has also as shown by many researchers a beneficial effect on the
patch loading resistance, e.g. Rockey et. al (1978), Bergfelt (1979) and Janus et. al (1988).

The ultimate patch loading resistance of an unstiffened steel girder web has over the years
been quite thoroughly investigated. One of the more recent and acknowledged publications was
Lagerqvist (1994) which also was implemented as the patch loading rules of EN 1993-1-5.
However, parts of the existing rules in EN 1993-1-5 has been questioned, and with Gozzi (2007)
a refined proposal for the patch loading resistance was presented and validated.

Regarding the ultimate patch loading resistance for longitudinally stiffened girder webs
publications as Graciano (2002), Seitz (2005) and Davaine (2005) are examples of work
focused on improving the prediction models regarding the failure mode. In EN 1993-1-5 the
patch loading resistance for a longitudinally stiffened web is predicted using a model presented

3
in Graciano (2002). However, the prediction model in EN 1993-1-5 treats open and closed
section stiffeners in the same way, furthermore the model was based on the theory for
unstiffened webs. Hence, inherited the criticized part of the resistance model of Lagerqvist
(1994).

1.3. Purpose and Aim

As previously mentioned, the work presented within this thesis has been divided into two
parts, one considering patch loading of a girder web reinforced with a longitudinal stiffener and
one focusing on plate buckling under uniformly distributed compression. Therefore, this section
was also sub-divided into parts comprising the purposes and aims for the two research areas
respectively.

The purpose of the work presented within this thesis regarding the ultimate patch loading
resistance was to

• Investigate if an ultimate patch loading resistance method for girder webs

reinforced with one longitudinal stiffener, consistent with the proposal of Gozzi
(2007) regarding unstiffened webs, could be stated.

• Examine if webs stiffened with closed section stiffeners could safely be designed
in the same manners as open section stiffeners with respect to the patch loading
resistance.

The work focusing on buckling resistance of plates with welds subjected to uniformly
distributed compressive in-plane stresses was conducted with the purpose of

• Produce experimental results using specimens made of steel with a higher strength
than commonly used in civil engineering today.

• Examine if steels with higher strength may be considered in the same manners as
more commonly used structural steels with respect to the ultimate plate buckling
resistance.

• Examine if the Winter function used in EN 1993-1-5 is applicable regarding plates

joined by welding.

The aim of this thesis was, regarding both the patch loading resistance and local buckling
resistance, to if possible

• Propose and validate an efficient and safe design procedure, improving the
prediction of the ultimate resistance in comparison to EN 1993-1-5 and previously
presented research work.

4
 Introduction

1.4. Limitations
Regarding the patch loading investigation the following limitations were imposed:

• The experimental results gathered from the literature comprises only plate I-girders
subjected to patch loading (one local load). Opposite or end patch loading was not
considered.

• The investigation presented herein only considers the patch loading resistance of a
web reinforced with one longitudinal stiffener of open or closed type.

• Possible interaction phenomena was only investigated with respect to bending

moment, e.g. shear / patch loading interaction was not considered.

Regarding the local buckling investigation the following limitations were introduced:

• The gathered data from the literature only comprises plate specimens with a square
cross-section under uniaxial compression, i.e. the individual plates were all treated
as simply supported internal compression elements.

Further, the following limitations was common for both the patch loading and the local
buckling investigation:

• The gathered data, as well as the experimental work conducted, only comprised
structural steel, i.e. no tests or specimens made of stainless steel were considered
herein.

• All experimental results gathered from the literature and presented tests herein,
comprises only welded girders or box specimens in as-welded condition, i.e. none
of the specimens were stress relieved.

1.5. Basic concepts

Within this section some basic concepts and notations used within this thesis are explained.
The notations used for describing the layout for a girder web longitudinally stiffened with an
open or closed section stiffener is described in Figure 1.1.

1.5.1. Effective cross-section of longitudinal stiffeners

The moment of inertia for a longitudinal stiffener, Ist, is used herein to determine e.g. the
relative flexural rigidity of the longitudinal stiffener. Generally the moment of inertia is
determined for the stiffener itself and a contributing part of the girder web. However, there
exists different definitions of how to estimate the Ist, e.g. Rockey et. al (1979) and Graves Smith
and Gierlinski (1982), however regarding this thesis the definition of EN 1993-1-5 according to
Figure 1.2 is adopted.

5
 Figure 1.1: Schematic description of cross-sectional notations for a girder
stiffened with an open sectioned stiffener (left) and a closed section
stiffener (right).

The section of the stiffener used as the gross area comprising the stiffener with an addition
of the web, 15Htw wide on each side of the stiffener. Though this must be compatible with the
actual dimensions of the cross-section, e.g. distance to flanges or overlapping areas.

Figure 1.2: The definition of EN 1993-1-5 regarding the effective cross-section of

longitudinal stiffeners. Left open section stiffener and to the right a
closed section stiffener.

6
 Introduction

1.5.2. Bending resistance

The bending resistance of the longitudinally stiffened girders was herein calculated with
respect to EN 1993-1-5. This with respect to cross-section classes and possible reductions to
effective sections. This was conducted for all outstand and internal elements under compressive
stresses, i.e. flanges, the part of web under compression and the stiffeners. The cross-section
was subdivided into simply supported parts, e.g. the stiffener in Figure 1.2 above would be
divided into three parts, and the rest of the web also into three parts (one above the stiffener, one
“inside” the stiffener and one from the stiffener and down to the neutral axis) all treated
individually.

Furthermore, the bending resistance was also modified with respect to the girder being of
hybrid type or not. Regarding common hybrid girders, i.e. with a flange having a higher yield
strength than the web, the approximation according to eq. (1.1) and eq. (1.2) was used to
determine the bending resistance.

M R = f yf ˜ W eff – 'W (1.1)

hw ˜ Aw f yw 2 f yw
'W = ------------------ ˜ § 1 – -------· ˜ § 2 + -------· (1.2)
12 © f yf ¹ © f yf ¹

However, some of the test data from the literature was based on girders with an “opposite”
hybrid girder, i.e. with the web having a higher yield strength than the flange. In these cases the
bending resistance was approximated assuming that the web was to reach the yield limit even
though the flange having a lower yield strength, i.e. first assuming the whole cross-section
having the yield stress of fyw. The bending resistance was then modified subtracting the
overestimation of the flange resistance, all according to eq. (1.3).

M R = f yw ˜ W eff – f yw – f yf ˜ A fl ˜ h w + t f (1.3)

1.6. Disposition of the thesis

In chapter 2 the basic plate buckling theory is briefly described. An introduction into
structural stability initiates the chapter, followed by the concepts of critical loads, effective
width (by e.g. von Kármán and Winter) with respect to local buckling. Furthermore some
models describing the ultimate patch loading resistance regarding unstiffened girder webs,
followed by models regarding webs with longitudinal stiffeners, are introduced. Formulations
regarding patch loading and bending moment interaction are also briefly presented.

Chapter 3 comprises a survey of published work regarding patch loading resistance of

longitudinally stiffened I-girder webs. Specimens with webs reinforced with open stiffeners, as
well as closed stiffeners are presented. Moreover, results from 366 numerical simulations from

7
the literature are introduced. All the gathered tests results were also re-evaluated with respect to
EN 1993-1-5 and the results are shown in this chapter.

A proposal of a modified design approach, based on the findings in the literature is presented
in chapter 4. The design model is validated by re-evaluating the test results and numerical
simulations with respect to the proposal. Furthermore, the proposed design approach is
compared with some directly comparable proposals by other authors as well as the design rules
of EN 1993-1-5. In a last step a partial safety factor in accordance to the guidelines in Annex D
of EN 1990 (2002) for the tests as well as for the numerical simulations, is introduced.

Experimental work regarding local buckling published by other authors is presented in

chapter 5. The test results gathered from the literature comprises specimens made of plates
joined by welds along their edges to a box shaped cross-section. All of the tests introduced
within this chapter is in as-welded condition and re-evaluated with respect to EN 1993-1-5.

Chapter 6 presents the experimental work regarding local buckling of box-sectioned welded
specimens preformed at LTU. The test set-up, layout of the specimens, measured quantities and
more are described. Furthermore, the results from the local buckling tests are compared to the
EN 1993-1-5 and presented in this chapter.

Chapter 7 proposes a modified reduction function for calculating the effective width
regarding plates with welds. Furthermore the proposal is validated by comparison to the
available tests results from both literature and experimental work conducted at LTU. In a last
step, the proposed reduction function is provided a partial safety factor on the same manners as
for the patch loading part of this thesis.

All the work presented in this thesis is discussed and concluded in chapter 8. Furthermore,
some proposals for future work is also introduced.

Tables containing data of the specimens used for patch loading experiments and numerical
simulations presented in the literature are displayed in Appendix A.

In Appendix B additional figures describing the test and numerical simulation data are
shown. This with respect to the herein proposed design approach, as well as the proposals by
other researchers which have been used for the comparison. The statistical evaluation of the
proposed resistance approach is also provided in this appendix.

Appendix C is detailing the local buckling experiments. This with respect to specimen data,
stress / strain figures from tensile tests, axial load / mean axial deformation figures from the
local buckling tests etc. Furthermore the used measuring equipment are briefly described and
the statistical evaluation of the partial safety factor with respect to the tests results from the
literature and LTU conducted experimental work is presented.

8
 Plate Buckling - Theory

Chapter 2:
Plate Buckling - Theory

The words “stable” or “instable” are used by people in various contexts. Almost everyone
have a relation or thought concerning the two words describing the state of something. The
terms are used in the wide range from psychology and politics to nuclear and chemical
applications. The term “stable” is often connected to something positive and rigid when
“instable” is closely linked to the possibility of an abrupt loss of something. One of the most
known and used context of the two words, which almost all people have a relation to, is when
used in medical surroundings; a stable or instable health state.

The interest in stability / instability is also a central concern regarding mechanical systems,
e.g. structural or civil engineering, see Figure 2.1. In this field the stability or instability of a
structure is often confined to regard the elastic part of the phenomena. However, as will be
shown later herein, a structural engineer may also have to consider the inelastic state. As an
example of structural instability one can consider the columns in a building made with a steel
frame. These columns have not only to withstand the vertical loads of the dead weight and e.g.
snow, but also lateral loads caused by the wind. This well known instability phenomenon is
usually referred to as column or flexural buckling.

Figure 2.1: Maybe an up-coming example of structural instability?

9
 The buckling may be of global nature, as described above, but may also be of localized
(local) type. Buckling of local sort are regional located buckling, e.g. a flange of a beam or at a
certain level of a silo, see Figure 2.2. Local buckling occur due to compressive stresses and may
in a further perspective cause global buckling because of the loss of resistance of the cross
section in question.

Figure 2.2: Different examples of buckling. Shell buckling in a silo (left), Farshad
(1994), and box shaped profile (right).

A structure or a member in an equilibrium state under e.g. compressive load may become
unstable and the structure acquires a new equilibrium state or a new trend of behaviour. When
considering classical buckling theory the critical stress level is defined as the stress at which the
perfect structure becomes unstable. This point is called the bifurcation point or bifurcation load.
Usually two more types of elastic instabilities are distinguished. These are limit equilibrium
instability (snap-through buckling) and dynamic or flutter instability.

Considering the load - displacement behaviour of a plate subjected to compressive stresses,

a load level lower than the bifurcation point corresponds to a state where buckles are of elastic
type. Hence, the secondary path in Figure 2.3 represents the post buckling stadium.

Primary path Secondary path

Load

Critical load

Bifurcation point

Deformation
Figure 2.3: Schematic description of the bifurcation of equilibrium.

10
 Plate Buckling - Theory

The bifurcation load or critical load has under the years been thoroughly investigated. As
mentioned above, the critical load is determined with elastic analysis and have been examined
theoretically by many different researchers, e.g. Timoshenko and Gere (1963).

2.1. Plate buckling theory

A thin plate is, by definition, a two-dimensional flexural element of which the thickness is
much smaller than its other two dimensions. A plane passing through the middle of the plate is
called the middle plane.

Thin plate elements are used in various structures; they may be elements in a complex
structure or may themselves constitute the major part of a structure. Examples of plate elements
are walls of containers, silos, and reservoirs, flat roofs, flat elements of vehicles and aircrafts,
and sheet piles. Examples of plates in civil engineering applications are the flanges and the web
of a beam. Plate elements may be homogeneous and isotropic or they may be stiffened and / or
have a composite construction.

Depending on the mode of application, a plate can be subjected to various lateral as well as
in-plane forces. Under certain circumstances, applied in-plane loading may cause buckling
which can be global or in some cases, have a localized nature; delamination buckling of
composite plates or buckling of a web in a steel beam are examples of local buckling. Regarding
thin plates, buckling is a phenomenon which may influence the load-bearing capacity of plate
elements. Hence, this must be taken into consideration in the design of plate elements.

2.1.1. Elastic analysis / Calculation of critical load

The theory behind the behaviour of a thin plate under compressive forces is usually divided
into two parts; firstly the calculation of the critical load and secondly the determination of the
ultimate load level. The critical load level is by definition the point were the perfect structure,
or member, in question loose its stability.

Analytical calculation of the bifurcation or critical load on the basis of the classical theory of
elasticity may be done either through solving the differential plate equation or via the energy
method. The differential equation describing the equilibrium under small deformations of a
plate loaded in its plane was established by Saint-Venant in 1870, Dubas and Gehri (1986), and
states

4 4 4 2 2 2
w w w w w w 1 w w w w w w
--------- + 2 ˜ ----------------- + --------- = ---- N x ˜ --------- + N y ˜ --------- + 2 ˜ N xy ˜ ------------ (2.1)
wx
4
wx wy
2 2
wy
4 D wx
2
wy
2 wxwy

where w is the lateral displacement and the flexural rigidity of the plate is given by

11
 3
E˜t
D = -----------------------------
2
(2.2)
12 ˜ 1 – Q

This plate equation was derived under the assumptions that the material is behaving in a
ideally elastic way, the plate is without initial imperfections such as initial curvature or residual
stresses. Furthermore, the plate deformations are assumed to be small. Under these assumptions
the plate shows no lateral deformations until the critical stress level is reached. At this point, the
deflection can either be negative or positive regarding the coordinate system of the plate, Figure
2.4.

Figure 2.4: System bifurcation at point A. The plate buckles in either a positive or
negative lateral direction, w.

The plate equation may be convenient to use when a rigorous solution of eq. (2.1) is possible.
When the plate in question is for example reinforced with stiffeners, the problem gets more
advanced. These more advanced applications led to the development of other models, better
describing the actual behaviour of plates.

In 1891 Bryan developed an strain energy expression for a plate under bending. The
approach of this method is to study the plate energy in the bifurcation point, where the plate
cease to be in its assumed perfectly flat state and instead follow its secondary equilibrium path
(see Figure 2.3) in a laterally deformed state. The energy based solution is built on the classical
correlation between the internal energy of bending and the external work done by the forces
acting in the middle plane of the plate. The expression for describing the strain energy stored in
the deformed plate is

2
1 § w 2 w w 2 w· § w2 w w2 w 2
w w ·
2
U = --- ˜ D
2 ³³ ¨ ---------2 + ---------2 ¸ – 2 ˜ 1 – Q ˜ ¨ ---------2 ˜ --------2- – §© ------------·¹ ¸ dx dy (2.3)
© wx wy ¹ © wx wy wxwy ¹

Furthermore the equation describing the work conducted by the externally applied forces is

12
 Plate Buckling - Theory

2 2 2
1 w w w w w w
T = – ---
2 ³³ N x ˜ ---------2 + N y ˜ ---------2 + 2 ˜ N xy ˜ ------------ dx dy
wx wy wxwy
(2.4)

The equations eq. (2.3) and eq. (2.4) are only valid for small deformations, which is assumed
to be the case at the bifurcation point. With Figure 2.3 in mind, the comparison between the
internal energy and external work gives, according to Timoshenko and Gere (1963), the
following information concerning the stability of the plate in question at the bifurcation point:

• If U > T, the flat form of equilibrium of the plate is stable (primary path)

• If U < T, the plate is unstable and buckling occurs (secondary path)

However, the critical load amplitude may be found by setting

T = UœU–T = 0 (2.5)

which can be solved under the condition that the change in energy potential must have a
minimum value for a stable equilibrium. This may be used for the derivation of the differential
equation form of the equilibrium, eq. (2.1). Another way to solve the problem is to apply an
expression for the lateral deformation of the plate.

2.1.2. Simply supported plates under uniform compression

Figure 2.5: Simply supported plate under uniform compressive load. Dubas and
Gehri (1986).

If considering a plate subjected to uniformly distributed forces along two of the edges,
according to Figure 2.5, the determination of the critical load level of the plate in question is
dramatically simplified comparing to the general case with loads applied in all the in-plane

13
directions. Since the only load applied on the plate, in the form of a uniform distributed
compressive force, acting along the edges x = a/2 and x = -a/2, the rest of the external applied
loads according to equation eq. (2.1) equals zero:

N y = N xy = 0 (2.6)

The assumed edge constraints of the plate leads to the following boundary conditions:

Along the edges x = a/2 and x = -a/2

2
w w
w = ---------2 = 0 (2.7)
wx

and along the edges y = 0 and y = b

2
w w
w = ---------2 = 0 (2.8)
wy

The boundary conditions implies that the deformed shape of the simply supported plate may
be described by a double trigonometric Fourier series on the form

f f
mSx nSy
w =
¦ ¦a mn sin ----------
a
- sin ---------
b
m ,n = 1 ,2 ,3} (2.9)
m = 1n = 1

By substituting the expression of the lateral deflection according to equation eq. (2.9) into
eq. (2.3) and eq. (2.4) under the above described conditions in eq. (2.6), eq. (2.7) and eq. (2.8),
and by using the relation between the external work done by the applied load and the strain
energy according to equation eq. (2.5), the following relation may after some mathematical
work be stated

mS 2 nS 2 2 mS 2 mSx nSy
D ˜ §© §© -------·¹ + §© ------·¹ ·¹ + N x ˜ §© -------·¹ ˜ a mn sin ----------- sin --------- = 0 (2.10)
a b a a b

To satisfy the equation eq. (2.10) for all positions on the plate, i.e. all values of x and y, the
following relation has to be true:

mS 2 nS 2 2 mS 2
D ˜ § § -------· + § ------· · + N x ˜ § -------· = 0 (2.11)
©© a ¹ © b¹ ¹ © a ¹

or in another form

14
 Plate Buckling - Theory

mS 2 nS 2 2
D ˜ § § -------· + § ------· ·
©© a ¹ © b¹ ¹
N x = -----------------------------------------------------
2
- (2.12)
§ mS
-------·
© a¹

The combination of the two integer parameters now have to be chosen in such a way that the
applied load, Nx, reach a minimum value, i.e. the sought critical load value, Ncr. It can be shown
that the lowest critical load is reached when the plate buckles in a shape such that one half sinus
wave is formed over the width of the plate (y-direction), hence the integer parameter n = 1,
Timoshenko and Gere (1963). With this, the equation eq. (2.12) may be evaluated to

2 2 2 2
a ˜ S ˜ D §m 1·
N cr = -----------------------
2
- ˜ ¨ ------ + -----¸
2 2
m = 1 ,2 ,3} (2.13)
m ©a b¹

in which the integer parameter m describes the number of half sinus waves over the length of
the plate (x-direction). The equation eq. (2.13) are more often formed as

2
S ˜ D-
N cr = k cr ˜ -------------
2
(2.14)
b

where the dimensionless parameter kcr is the buckling load coefficient and is given by

m˜b a 2
k cr = § ----------- + -----------· m = 1 ,2 ,3} (2.15)
© a m ˜ b¹

Furthermore, with the expression for the flexural rigidity of the plate given in eq. (2.2),
inserted in eq. (2.14) the well known expression for the critical, or bifurcation, stress may be
expressed as

2 2
S ˜E t
V cr = k cr ˜ -----------------------------
2
˜ §© ---·¹ (2.16)
12 ˜ 1 – Q b

with the insight of that

cr N
V cr = ------- (2.17)
t

The buckling load coefficient, kcr, is, as can be seen in eq. (2.15), a function of the plate width
b, the length a and the number of sinus half waves over the length, m. For different values of the
plate width and length ratio a / b, the lowest critical stress level will be found for different
numbers of half waves according to Figure 2.6.

15
 Figure 2.6: The buckling load coefficient for a simply supported thin plate.
Timoshenko and Gere (1963).

2.1.3. Initial plate imperfections

In section 2.1.1, a quite straight forward method for calculating the critical stress level is
presented. However, as always concerning theoretical models describing nature, it is important
to remember the assumptions made for the theory in question. Emphasizing the assumptions
made of a initially perfect flat plate and a perfectly isotropic behaviour in a homogenous
material the understanding of the limitations in the presented theory are obvious. All materials
have different levels inherent imperfections, also steel. A plate delivered from the steel
fabricator has an initial curvature and probably also residual stresses from uneven cooling of the
material. These facts makes the assumptions made above somewhat unrealistic, which also has
been proven experimentally and may be found in chapter 3.

Now when the assumptions are found to be a quite utopical description of the real behaviour
of the considered plates, the question arises how these initial imperfections affect the plate
behaviour before, as well as after, the bifurcation point. Figure 2.7 shows the difference in the
plate behaviour when plate imperfections are considered.

Considering Figure 2.7 two conclusions concerning how the imperfection influence the plate
behaviour may be drawn. Firstly, buckling of a plate with inherent imperfections is gradual and
the exact critical load may be difficult to determine. Hence, difficulties arises when a
comparison between theoretically and experimentally determined critical loads are to be
conducted. Secondly, as mentioned before, the plate may accept continued loading after the
bifurcation load is reached. Thus the critical load is shown to be a non-representative measure
on the ultimate resistance of the plate in question.

16
 Plate Buckling - Theory

Figure 2.7: The influence of initial plate imperfections in relation to a perfect

plate. Lateral displacement, G, and applied in-plane stress, V, in
relation to the elastic critical stress, Vcr.

2.1.4. Geometric imperfections

When considering the initial out-of-plane imperfections, i.e. initial buckles, the influence of
these on the maximal out-of-plane deformation / load correlation are shown in Figure 2.8.

The graph and the calculations behind was made by H. Nylander in 1951 and shows how an
applied initial deformed shape with the amplitude wo (in the same shape as the deformed plate)
affects the magnitude of lateral deformations under applied load. Furthermore, when the
material is assumed to be ideal elastic, the model gives no information concerning the ultimate
load. Concluded, the initial geometric imperfections primarily influences the plate stiffness and
becomes more obvious with an increased plate slenderness.

Figure 2.8: The effect of initial geometric imperfections. Relation between the
lateral deformation, w, plate thickness, d, and load, N, concerning
different amplitudes of initial imperfections wo. Nylander (1951).

17
 2.1.5. Residual stresses
Knowing that residual stresses are present in all materials, it is evident that this must affect
also the plate buckling theory. Geometrical imperfections and residual stresses in a plate under
compression mainly affects the initial phase of the loading of the plate. This since the initial
imperfections acts as an existing applied load before applying external loads. In Figure 2.9
below, a schematical distribution of residual stresses caused by edge welding a plate is shown.

Figure 2.9: Schematic distribution of residual stresses in an edge welded plate.

Considering Figure 2.9 above, the influence of the initial load due to the present residual
stresses is clear. Since the middle region of the plate before external loads are applied, already
is under compressive stresses, it is obvious that yielding of the plate in question will occur at a
lower external load level compared to a residual stress free plate, see Figure 2.10.

The effect of inherent residual stresses is more marked for stockier or intermediate slender
plates, for which yielding is the governing cause of failure. Concerning more slender plates, the
initial geometric imperfection tend to surpass the influence of residual stresses, Dubas and
Gehri (1986). Hence, the influence of residual stresses decreases with increasing plate
slenderness.

Vcr

'L / L

Figure 2.10: Schematic influence on the behaviour of a plate with (S) and without
(A) residual stresses.

18
 Plate Buckling - Theory

2.2. The effective width concept

As shown above, the estimation of the critical load may be done by a straight forward
method. However, the elastic analysis assumes, as described in previous sections, that the plate
in question is perfectly flat and that no initial stresses are present. Because of the presence of
these imperfections non-linear models were evolved. Furthermore, the initial plate
imperfections were not solely the reason to why non-linear theories had to be evolved. The
assumption concerning the constitutive relations, in this case ideal elastic material, is not
suitable to use when the ultimate resistance is sought for.

Another reason why non-linear models were established was that many researchers showed
that the ultimate load of a plate under compression may significantly surpass the critical load
level. This was especially evident concerning more slender plates. Regarding stockier plates the
resistance is often limited by yielding in the material and the ultimate load may be lower than
the critical.

In linear elastic analysis, the distribution of the load is assumed to remain uniform until the
plate buckles. However, when the plate starts to buckle, the stresses are re-distributed in the
plate. The plate behaviour under these large deformations, or post critical behaviour, is a
complicated area to describe. Some differential equations describing the phenomenon were
derived by von Kármán in 1910 but the methods for solving these are complex, Dubas and Gehri
(1986). The finite difference method, fourier series or different perturbation methods are
possible tools for this work.

Other methods may also be used for studying the post critical plate behaviour. One example
is the numerical methods, e.g. the finite element method, FEM, which probably is the most
powerful tool available today. However, other methods have been used during the years of
research. Analytical methods such as the Ritz energy method or a method based on a theory by
Skaloud and Kristek called the “Folded plate theory method” are both excellent examples.

As described above, the theory behind plate buckling is rather complicated due to the
combination between the membrane stresses from the applied load and bending stresses in the
deformed plate, as well as shear stresses due to rotation at the corners of the plate. For design
purposes the above described methods may be too advanced to use. This is why the “Effective
width approach” by von Kármán et al. (1932), is widely spread as the model for determining the
ultimate resistance of plates under compression.

2.2.1. The von Kármán effective-width formula

The starting point for the effective width approach is that the ultimate resistance is reached
when the largest edge stress reaches the yield stress level. Since the formed buckle in the middle
of the plate reduces the plates ability to carry the load, the stresses are re-distributed as shown

19
in Figure 2.11 below. The real stress distribution in the plate is approximated, or substituted,
with two strips which describes the load carrying effective width of the plate.

Figure 2.11: Stress distribution in a plate before (a) and after buckling (b).The von
Kármán assumption concerning the effective width is presented in (c).
Brush and Almroth (1975).

von Kármán’s hypothesis was that the fictitious plate with the width of beff would have the
critical stress equal to the yield stress, i.e.

V cr = fy (2.18)

Furthermore, the critical stress according to eq. (2.16) under the condition that the plate is
under uniform compression and simply supported (kcr = 4) the following expression may
describe the relation between effective width and yield stress level:

2
S ˜E t 2
4 ˜ ----------------------------- ˜ § --------· = f y (2.19)
2 © b eff¹
12 ˜ 1 – Q

or with the original plate width equal to b

20
 Plate Buckling - Theory

cr V
b eff = b ˜ ------
- (2.20)
fy

which is usually referred to as the von Kármán effective-width formula. Furthermore, the
relation

fy b fy
Op = - = 1 ,05 ˜ --- ˜ --------------
------ (2.21)
V cr t k cr ˜ E

was made as a generalization of the corresponding well known parameter for column buckling
and was called the reference slenderness of the plate. In modern design rules, when design is
done with respect to the ultimate load, this expression is the only one in which the elastic critical
load is considered, and as expressed in von Kármán et al. (1932) the following may be stated

E-
b eff = 1 ,9 ˜ t ˜ --- (2.22)
fy

or

b eff 1
-------- = ----- for Op t 1 (2.23)
b Op

under the circumstances that the plate is simply supported and under uniform compressive load.

Although, von Kármán’s theories gained reputation as a good method to use for the
determination of the ultimate load of the plate in question, the method was a method based on
plates without initial imperfections and when compared to test results it was found to be true
only for large b / t ratios. However, von Kármán still stands as the first researcher proposing a
reduction factor function.

2.2.2. The Winter function

Theodor von Kármáns work was a milestone concerning the simplified design methods
concerning plate buckling. Many researchers followed his work, aiming for an expression
describing a real plate with inherent initial imperfections. One of the more known and widely
spread in design codes, is the one proposed by Winter in 1947. Winter conducted numerous
experimental tests on cold formed specimens and suggested

b eff 1 0 ,22
-------- = ----- § 1 – ----------· for O p t 0 ,673 (2.24)
b Op © Op ¹

as a suitable function regarding the effective width, Winter (1947). Winters first suggestion

21
 was with the coefficient 0,25 but was later changed to the 0,22 used nowadays. However, it is
interesting to notice the small difference between the “original” equation eq. (2.23) and the
experimentally based eq. (2.24).

Other researchers proposed different solutions, or modifications, of the initial von Kármán
formula. Two reported in Dubas and Gehri (1986) are

b eff
-------- = 1----------
,05 § 0 ,26
1 – ----------· for Op t 0 ,55 (2.25)
b Op © Op ¹

by Faulkner in 1965 and

b eff 0 ,82-
- = ----------
------- (2.26)
b Op
0 ,85

suggested by Gerard in 1957.

Even though a lot of effort has been put into this research field, the Winter function, based
on the cold formed members survived and is nowadays set as the function used in the present
design rules in EN 1993-1-5.

In EN 1993-1-5 the plate slenderness, Op in eq. (2.21) is rewritten according to

bet
Op = -------------------------------- (2.27)
28 ,4 ˜ H ˜ k V

and His defined as

H = 235
--------- (2.28)
fy

The above stated parameter was introduced as a precaution to eventual differences in the
material characteristics considering steels with fy > 235 MPa. However, this parameter may be
debated in some senses, e.g. when used in physical interpretations of the behaviour of a cross-
section as the moment of inertia, see the discussion in chapter 8. Regarding the buckling load
coefficient, kV, for a simply supported plate under uniform compressive load, this is set to be
equal to 4.

As mentioned above, design with respect to local buckling of flat compression elements is
made through a reduction of the cross sectional area of the plate in question. Concerning internal
compression elements this is, according to EN 1993-1-5, done through the use of the expression

22
 Plate Buckling - Theory

Op – 0 ,055 ˜ 3 + \
U = ----------------------------------------------------
2
- d 1 ,0 (2.29)
Op

in which the factor <= Vmin / Vmax, represents the actual stress distribution over the plate.
Concerning uniform distribution of compressive stress this factor equals 1. Thus, the eq. (2.29)
reflects the original Winter function eq. (2.24) used for these kind of plate elements in EN 1993-
1-5.

2.3. Patch loading

Another form of buckling is the patch loading, commonly used for a load with a shorter
distribution length along a girder, applied perpendiculary to the flange in the plane of the web.
The phenomena is similar to the previously described plate buckling, however patch loading in
the elastic region does not distribute the stresses at an even magnitude as for the previously
described buckling. Early patch loading investigations date from the end of the 1930’s when the
influence of the flange stiffness on the web resistance to patch loading often was estimated using
the analogy of a beam on an elastic foundation following the old formulas of Zimmermann
(1888), Bergfelt (1979). The previously used slopewise load distribution (on a slope 1:1 from
the applied load) was replaced by the beam on an elastic foundation in order to estimate the load
acting on the edge of the web. Although the load distribution problem had two solutions, i.e. the
45 degree slope and the elastic foundation, the buckling problem was still to solve.

The aforementioned research on plate buckling regarding the gained knowledge in how to
predict the ultimate resistance, i.e. moving from the idea that the critical buckling load was a
good approximation of the ultimate resistance to actual models describing the maximum load a
plate could carry, applies also in the field of patch loading resistance. The elastic critical load is
nowadays “only” used to classify the slenderness of the girder web in order to calculate a
reduction factor. Other models does not use a reduction formulation, e.g. in the 1960’s tests in
Granholm (1960) gave a very simple and preliminary formula for the prediction of the ultimate
load with the thickness given in mm and the ultimate load in tonnes according to

2
F u = 8 ,5 ˜ t w (2.30)

This was probably one of the first ultimate patch loading resistance models to be derived
based on an empirical consideration. However, more refined models were to follow and the
resistance for a longitudinally stiffened web has often been closely linked to the unstiffened
ditto. Herein, a presentation of how the research regarding the patch loading resistance for an
unstiffened web has progressed will come first. Following the theory for an unstiffened web will
be the corresponding theory for the longitudinally stiffened webs.

23
 2.3.1. Resistance for girders without longitudinal stiffeners
3-hinge models

Bergfelt
One of the first models based on a fully mechanical approach based on the failure
mechanisms observed under experimental work was presented in Bergfelt (1979). Bergfelt
referred to the model as the “three-hinge-flange” and stated his earlier work presented in
Bergfelt (1971) as its origin. The three hinge mechanism model was derived from tests results
from his own and other researchers work.

Bergfelt describes his model as follows: “At a small load the flange behaves as a beam on
elastic foundation (consisting of the web). At increasing load a plastic hinge forms in the flange
just under the load. The web stresses start yielding below the hinge, whereafter the yielding
region extends. The negative bending moments in the flange increase, and the failure starts
when a (negative) plastic hinge forms on each side of the load.” However, Bergfelt also states
that the model in Bergfelt (1971) not seemed to be valid for tf / tw > 2 (i.e. more common girder
ratios). The authors’ idea of the reason for this problem was that for girders with more slender
webs compared to the flanges, the crippling of the web starts as buckling of the region of the
web under the applied patch load and not because of a reached yield limit of the web.

This contradiction (compared to the basic idea of the three-hinge-flange mechanism) led
Bergfelt to refine his model further, and was so done with Bergfelt (1979). Furthermore,
Bergfelt mentions that if the load is distributed through a very stiff bar, or is distributed over a
longer distance, there are possibilities that the centre plastic hinge in the flange may be replaced
by two hinges at each end of the load introducing bar/plate.

To make the model more applicable concerning “normal” girders, Bergfelt aimed towards
finding a satisfactory estimation of V w according to the model description in Figure 2.12 below.

Figure 2.12: The three-hinge-flange model according to Bergfelt (1979).

Bergfelt used the von Kármán approach, with the approximative description of the failure
stress according to eq. (2.31)

Vw = V cr ˜ f yw (2.31)

24
 Plate Buckling - Theory

Bergfelt evolved the system and end up in the expression

2
F R = 0 ,8 ˜ t w ˜ E ˜ f yw ˜ t i e t w ˜ f s s ,h w ,etc (2.32)

in which ti = tf for a web of “normal” slenderness and with a flange satisfying bf = 25tf.

In other cases with bf / tf -ratios not equal to 25 the eq. (2.33) is valid (under the restriction
that the flange has a rectangular cross-section).

bf
t i = t f ˜ 4 ------------- (2.33)
25 ˜ t f

Concerning eq. (2.34) the expression contains a number of correction terms and also terms
for including eventual influence of vertical, f(sv), and longitudinal, f(sl), stiffeners. Bergfelt also
states that the other correction factors generally lies close to 1.

f s s ,h w ,etc | f s s ˜ f h w ˜ f f yw ˜ f M E ˜ f G ˜ f s v ˜ f s l (2.34)

Regarding longitudinally stiffened girders Bergfelt proposes a resistance function of

amplification factor type which is described in section 2.3.2.

Roberts and Chong

The attentive reader may suspect that the three-hinge model probably is more applicable to
patch loading cases with a shorter loading length. Nevertheless, in Roberts and Chong (1981) a
three-hinge mechanism was proposed to be used under “distributed” patch loading. With a
distributed patch load the authors referred to a load distributed over the whole panel length. The
applicable tests of Bossert and Ostapenko (1967) were used as comparison to the proposed
model. However, different from the proposed model of Bergfelt but in line with other work
made by Roberts and e.g. Shimizu et al. (1989a,b), Roberts and Chong derived the model with
yield lines in the web.

Ungermann
A more recent publication using the three-hinge mechanism is the dissertation by
Ungermann (1990). Ungermann used a more contemporary approach to establish a patch
loading resistance model for design, i.e. by using the von Kármán approach of plate slenderness,
with the slenderness parameter according to eq. (2.35). The resistance proposal in Ungermann
(1990) according to eq. (2.36) was also verified through a comparison to tests. Furthermore, the
resistance proposal presented by Ungermann comprises two equations which are valid for two
different web slenderness values with 0,8 as the divider. This is due to the yielding of the web
regarding more stocky webs and the same idea may be found in the work by Roberts presented
later within this chapter.

25
 The two equations of Ungermann reads

F
OF = ------y- (2.35)
F cr

­ 2 2
° F R = 22 ˜ 7 H ˜ t w ˜ f yw if O F d 0 ,8
°
® § 0 ,525 0 ,375· (2.36)
° F R = 2 ˜ c u ˜ t w ˜ f yw ˜ ¨ ------------- + -------------¸ if O F ! 0 ,8
° © OF OF ¹
2
¯

in which the distance between the outermost plastic hinges is estimated as

ss s s 2 4 ˜ b f ˜ t f ˜ f yf
2 ˜ c u = ---- + § ----· + ------------------------------ (2.37)
2 © 2¹ t w ˜ f yw

and the yield resistance, Fy, of the web is calculated over this length of the web according to

F y = 2 ˜ c u ˜ t w ˜ f yw (2.38)

As may be noticed in eq. (2.36) the yield resistance of the web over the length 2cu is reduced
with a function, f(OF), i.e. the resistance is given with the reduced yield load on the form

FR = F F O F ˜ Fy (2.39)

and was the first patch loading design model based on a reduction factor dependent on the web
plate slenderness parameter, OF.

4-hinge models

Roberts and Rockey

In the end of the 70’ies, at the same time as Bergfelt developed his model, Roberts and co-
authors presented an alternative plastic mechanism solution. The background of the work was
that none of the up to then published models and / or design recommendations was entirely
satisfactory when compared to the experimental data base available hitherto. In Roberts and
Rockey (1978) and (1979) a solution for the ultimate resistance for plate girders under patch
loading was proposed. The model was based on four plastic hinges in the flange accompanied
by yield lines in the web. The idea of the ultimate failure model is described in Figure 2.13.
Furthermore, the model presented in the publications of Roberts and Rockey was in the articles
compared with available test data and, according to the authors, suitable to use for prediction of
the ultimate patch loading resistance.

26
 Plate Buckling - Theory

Figure 2.13: The definition of the failure mechanism and the position of the four
plastic hinges in the loaded flange and yield lines in the web
according to Roberts and Rockey (1979).

In 1981 Roberts presented an article himself with a revised version of the aforementioned
model. The modifications of the model presented in Roberts (1981) was mainly due to new tests
focusing on how changes of the web depth and the thickness of the flanges and the web affected
the patch loading resistance. The latter was a revised form of the four hinge model from Roberts
and Rockey (1979), and the procedure to estimate the patch loading resistance according to the
version of Roberts (1981) is presented in short manners below.

The greek symbols D, E and T in Figure 2.13 above represents the assumed position of the
yield lines in the web, the position of the outermost plastic hinges in the flange and the
deformation of the web precisely prior to failure respectively. The next step to take in order to
formulate the resistance equation is to assume that the external load deforms the girder a small
vertical distance, Gw, which implies a rotation in the plastic hinges of Gw / E and of the yield lines
in the web of magnitude Gw / 2D cosT. By summing and equating the external and internal work
done under this incremental deformation, the following equation is given.

4 ˜ M pf 4 ˜ E ˜ M pw 2 ˜ s s ˜ M pw 2 ˜ K ˜ M pw
F R = ---------------- + -------------------------- + --------------------------- – -------------------------- (2.40)
E D ˜ cos T D ˜ cos T D ˜ cos T

in which K is a definition of a length of the web under the external load which is assumed to
have yielded because of compressive membrane stresses, hence this part of the web offers no
bending resistance and is subtracted from the directly loaded length ss according to eq. (2.40).
By minimizing FR in eq. (2.40) with respect to E, the spread of the plastic hinges in the flange
may be expressed as

2 M pf ˜ D ˜ cos T
E = ---------------------------------
- (2.41)
M pw

and under the assumption that the flange deformation just before collapse may be estimated
using the theory of elasticity and that the moment distribution in the flange varies linearly

27
between +Mpf at the outer plastic hinge to - Mpf at the closest hinge at the edge of the patch load,
the maximum vertical displacement of the flange may be derived by integration to

2
M pf ˜ E
v E = ------------------- (2.42)
6 ˜ E ˜ If

Through some geometrical compatibilities, mathematical work and also on the assumptions
that fyf = fyw and the distance between the flange and the yield line, D, was set to 25tw for slender
girders, Roberts ends up in a solution for estimating the patch loading resistance according to

3
---
E ˜ f yw ˜ t f 3 ˜ ss tw 2
- ˜ 1 + ----------- ˜ § -----·
2
F R = 0 ,5 ˜ tw ˜ ---------------------- (2.43)
tw hw © tf ¹

However, Roberts recommended that the ratio ss / hw would be limited to 0,2 due to the
somewhat unrealistic assumption of a straight flange between the two inner plastic hinges when
the loaded length grows larger. Furthermore, Roberts indicates that eq. (2.43) seems to
underestimate the ultimate resistance concerning girders with very thin flanges and webs. Based
on a comparison to test data Roberts suggested that the ratio tf / tw would be limited to three to
avoid the aforementioned issues but also states that this limitation is not recommended for
practical situations.

In the same publication (Roberts (1981)) an alternative failure model is presented. This
model addresses the possibility of a failure by direct yielding of the web underneath the patch
load. With an increased web thickness, (i.e. stockier web) the ratio out of plane bending stiffness
to the compressive membrane stiffness will be raised. According to Roberts this implies that an
alternative formulation for more stocky webs would be needed and is founded on the model
described in Figure 2.14.

Figure 2.14: The failure model for stocky webs according to Roberts (1981).

Analogous to the method on which eq. (2.43) was derived, Roberts uses the external and
internal work, equating these and minimizing the expression with respect to E and ends up in
eq. (2.44) below.

F R = f yw ˜ t w ˜ s s + 2 4 ˜ M pf ˜ f yw ˜ t w (2.44)

28
 Plate Buckling - Theory

Knowing that the plastic moment resistance of the flange is given by

2
M pf = f yf ˜ b f ˜ t f e 4 (2.45)

and inserting eq. (2.45) in eq. (2.44) the following equation for prediction of the ultimate patch
loading resistance concerning a web failing due to yielding may be derived

§ f yf ˜ b f ·
F R = f yw ˜ t w ˜ ¨ s s + 2 ˜ t f ˜ ----------------
-¸ (2.46)
© f yw ˜ t w¹

All in all, Roberts used the lowest of the two described resistances (i.e. either the direct
yielding resistance or the resistance of the buckled web) as the resistance of regarded girder, that
is the smallest of eq. (2.43) and eq. (2.46) gave the actual resistance of the girder.

Lagerqvist
In 1994 Lagerqvist presented his doctoral thesis focused on the resistance of steel girders
subjected to concentrated forces. In Lagerqvist (1994) a thorough literature review was
accompanied with the presentation of experimental work, numerical simulations and in the end
the proposal of a design model. Lagerqvist addressed patch loading of three types; patch load,
opposite patch load and end patch load. Herein only the work dealing with the first of these will
be considered. The proposal of Lagerqvist was based on a von Kármán approach and consisted
of three parts, an expression for the yield resistance, the elastic critical buckling load and the
resistance function itself.

Concerning the expression for the yield resistance this was derived on the basis of 48 tests
made on welded girders made of high strength steel and moreover 12 tests on rolled beams were
included. All three load applications were tested, however the majority of tests were focused on
end patch loading. Numerical simulations by means of FEM were used to derive appropriate
elastic buckling loads, i.e buckling coefficients for the three load cases. Furthermore, the
resistance function proposed was empirically determined with the use of about 250 tests from
the literature and in a last step the whole design proposal was compared to some 540 tests and
was found to predict the patch loading resistance with a better accuracy than other models from
the literature. The work presented in Lagerqvist (1994), with respect to the patch load case
according to the mechanical model described in Figure 2.15, will be summarized herein.

In Lagerqvist (1994) the first of the three included parts in the resistance model to be
addressed was the expression of the yield load. A model according to Figure 2.15 was used,
which is similar to what earlier was proposed by Roberts concerning webs with lower
slenderness, i.e. webs suspected to fail by direct yielding. However, based on experimental
observations Lagerqvist stated that the deformed part of the loaded flange increased with an
increasing web slenderness, i.e the responding part of the web increased with the increasing web

29
slenderness. Nevertheless, this behaviour was not captured in the proposal of Roberts in which
the loaded length of the web remains the same over variations of web slenderness. This could
be shown using a formulation on the form of eq. (2.39) and the resistance expressed by Roberts,
i.e. eq. (2.46) put equal to the yield load, Fy. The loaded length in this expression (i.e. the
bracketed terms in eq. (2.46)) remains the same over variations of the web slenderness, OF.
Lagerqvist suggested a possible alternative formulation to capture this behaviour and this was
to enhance the plastic bending resistance of the outermost hinges by letting a part of the web
contribute (see Figure 2.15). By using this fictitious T-section of the outer hinges, Lagerqvist
secured the dependent relation between the slenderness of the web and the loaded length, i.e. if
hw increased the contributing part in Mo would increase and so also the loaded length.

Figure 2.15: The mechanical model for the yield resistance as suggested in
Lagerqvist (1994).

To establish an expression for the yield load Lagerqvist made the evaluation in the same
manners as Roberts did, i.e. equating external and internal work and ended up in the equation

§ f yf ˜ b f h w 2·
F y = f yw ˜ t w ˜ ¨ s s + 2 ˜ t f + 2 ˜ t f ˜ ----------------- + k ˜ § ------· ¸
2
(2.47)
© f yw ˜ t w © tf ¹ ¹

and by comparison to test results, Lagerqvist proposes k2 = 0,02 for the contributing part of the
web.

After having established an expression to determine the yield load of the web during patch
loading, Lagerqvist focuses on the second part in his design model, the elastic critical load,
eq. (2.48). The method used in Lagerqvist (1994) to determine the buckling coefficient was
based on FE analyses of a web with flanges. The model was verified through comparison to
other researchers’ published work and found proper to use for further simulations. Lagerqvist
combined his derived expressions for the buckling coefficients of both the first and second
buckling mode of the deformed web and ended up in eq. (2.49) as the best combination.

2 3
S ˜E tw
F cr = k F ˜ -----------------------------
2
˜ ------ (2.48)
12 ˜ 1 – Q h w

30
 Plate Buckling - Theory

3
hw 2 bf ˜ t f
k F = 5 ,82 + 2 ,1 ˜ § ------· + 0 ,46 ˜ 4 ---------------
- (2.49)
© a¹ 3
hw ˜ tw

Furthermore, Lagerqvist (1994) also proposed a simplified version of eq. (2.49), in which
Lagerqvist included the contribution from the flanges (i.e. the last term in eq. (2.49)) in the first
term of the expression. This simplified coefficient would be more suitable to use in design
applications. The simplified buckling coefficient was proposed as

hw 2
k F = 6 + 2 ˜ § ------· (2.50)
© a¹

The third and last part of the design model presented in Lagerqvist (1994) was the reduction
function itself. The function is dependent of the web plate slenderness, OF (see eq. (2.35)) and
was calibrated with the use of some 190 tests with M E e M R d 0 ,4 . The reduction factor function
proposed reads

0 ,47
F F OF = 0 ,06 + ---------- d 1 (2.51)
OF

and gives the patch loading resistance with use of eq. (2.39). This equation was however
simplified in Johansson et. al (2001) which presented the new design rules for plated structures
to be implemented in EN 1993-1-5. The simplified version (design version) of the reduction
function eq. (2.51) reads according to

0 ,5
F F OF = ------- d 1 (2.52)
OF

and was furthermore introduced in EN 1993-1-5. Furthermore, the term describing the
contribution of the web to the outer plastic hinges in eq. (2.47), (in EN 1993-1-5 called m2) was
restricted to only influence the resistance concerning webs more slender than 0,5. That is

hw 2
m 2 = 0 ,02 ˜ § ------· if O F ! 0 ,5
© tf ¹ (2.53)
m2 = 0 if O F d 0 ,5

Müller
Another proposal addressing the reduction function for predicting the patch loading
resistance may be found in Müller (2003). In his doctoral thesis Müller proposed a reduction
factor founded on the general plate buckling curve proposed in Maquoi and Rondal (1986). The
proposal of the latter authors was based on the consideration that any plate buckling curve

31
captures the yielding (lower slenderness) and the actual reduction curve (more slender plates).
Based on this consideration Maquoi and Rondal presented a general format for which the
buckling curves could be written

J
1–F ˜ 1–F˜O = K˜F (2.54)

in which the factor J is depending mainly on the boundary conditions for the plate under
consideration and the K is an imperfection factor dependent of the plate slenderness. The
imperfection parameter was expressed as

K = D ˜ O – O0 (2.55)

In Müller J = 1 is used to interpolate between the yielding and the von Kármán proposed
curve for reduction with increasing plate slenderness. With J =1 the solution to eq. (2.54) reads

1
F = ----------------------------- (2.56)
2
M+ M –O

and with M according to

1
M = --- ˜ 1 + D ˜ O – O 0 + O (2.57)
2

Müller proposed a reduction curve with D = 0,34 and O0 = 0,8 to be used for girders subjected
to patch loading. The curve was derived in comparison to tests results and furthermore proposed
to be used with the plate slenderness determined according to the reduced stress method of EN
1993-1-5. Müller used data from tests and numerical simulations to determine the required load
amplifiers according to the reduced stress method. Another example of derivation of a
resistance function based on the eq. (2.56) and eq. (2.57) may be found in Grotmann (1993).
However this work was proposing a geometrical imperfection factor including the properties of
high strength steel by the use of the parameter H according to eq. (2.28).

Gozzi
In Gozzi (2007) the work on patch loading related issues were taken another step further in
the refinement undertaking. The doctoral thesis by Gozzi put the patch loading resistance of
plated girders in ultimate as well as serviceability limit state in focus. Herein the work
concerning the serviceability limit state will be overlooked and only the ultimate limit state will
be regarded. When dealing with the ultimate patch loading resistance Gozzi continued and
modified the work presented in Lagerqvist (1994). This with special attention paid to the
expression for the yield resistance, in particular the assumption about the addition to plastic
moment resistance which origin is found in the added part of the web concerning the outermost
hinges in the 4-hinge model of Lagerqvist. In Gozzi (2007) a thorough numerical simulation

32
 Plate Buckling - Theory

investigation was presented without any proof of that the web contributed in the aforementioned
plastic hinges. The FE investigation comprised 19 models over which the flange thickness and
width, web thickness, aspect ratio and loaded length was varied. According to Gozzi the
numerical study could not prove any contribution from the web regarding the plastic moment
resistance in the outermost hinges. Hence Gozzi proposed that the influence of the m2 parameter
regarding the loaded length would be neglected, i.e. m2 = 0 irrespective of web slenderness. This
conclusion was also supported in Davaine (2004) which is presented in section 2.3.2.

However, the above presented conclusion by Gozzi made the yield resistance decrease hence
a new calibration of the reduction function was needed. This undertaking was presented in
Gozzi (2007) using a reduction factor function of the same type as in Müller (2003) (see
previous section). As mentioned, the yield resistance was changed which furthermore gives an
overestimation of the patch loading resistance if the factors proposed by Müller is applied. Thus,
the factors D and O0 was calibrated using a data base consisting of 184 individual patch loading
experiments with low applied bending moments compared to the design resistance. The
calibration handed a best fit curve with the factors set to DF = 0,5 and O0F = 0,6. Moreover, the
results showed that the stockier specimens still had a higher resistance, and the plateau level was
proposed to be set to 1,2. The proposition of Gozzi (2007) regarding the reduction factor
function may then be concluded as

1
F F = ------------------------------------- d 1 ,2 (2.58)
2
MF + MF – OF

and

1
M F = --- ˜ 1 + 0 ,5 ˜ O F – 0 ,6 + O F (2.59)
2

Further, the proposal was in Gozzi (2007) proved to give a prediction of the ultimate patch
loading resistance with less scatter compared to the design model implemented in EN 1993-1-
5. The proposed model was furthermore verified through a statistical evaluation according to
Annex D of EN 1990 (2002) and the derived partial safety factor, JM1, was proposed to be set
to 1,0.

2.3.2. Resistance for girders with longitudinal stiffeners

Calculating the patch loading resistance for a longitudinally stiffened plated girder has often
been estimated using the corresponding resistance for an unstiffened girder. When the patch
loading resistance for such an unstiffened web has been calculated it is multiplied with an
amplification factor to estimate the actual resistance regarding the web equipped with a
longitudinally stiffener (e.g. on the form of eq. (2.63)). However, as been presented in previous
section 2.3.1 the reduction factor approach has become the leading prediction model used in the

33
design regulations of today (e.g EN 1993-1-5). With the aim to have stringent patch loading
resistance prediction models both regarding longitudinally stiffened and the corresponding
unstiffened type, this has brought the amplification factor models somewhat out of date.
Nevertheless, a lot of effort has been put into the topic of estimating the ultimate patch loading
resistance of longitudinally stiffened webs with amplification factors (e.g. Bergfelt (1979),
Janus et. al (1988), Kutmanová and Skaloud (1992), Graciano and Edlund (2001)), and will be
presented shortly herein.

Although these are two historically predominant methods of calculating the patch loading
resistance of a longitudinally stiffened girder web there are other proposals available. One quite
unique example is a model developed by genetic programming (GP) presented in Cevik (2007).
The programming method is a self adaptable program which uses the predefined variables to fit
an expression to predict the actual test result. The GP based formulation of patch loading
resistance of longitudinally stiffened webs was calibrated towards 138 tests with 11 geometrical
and material parameters used as variables. According to Cevik the final GP-equation shows a
perfect agreement when comparing to the experimental data base used, showing a mean value
of 1,021 and a coefficient of variation of 0,156. Although the correlation to the experiments are
good the equation is somewhat complicated and lacks a physical foundation. This may make the
expression inadequate when dealing with parameters outside the interval used for the GP.
Moreover, the equation is calibrated with only open stiffeners which makes it questionable to
use for closed stiffener types. Nevertheless, the equation, with geometries in mm and material
properties in MPa, to predict the resistance in kN according to Cevik reads

3 § ·
§ cos a ˜ f yw · ¨ tf ¸
F Rl = ¨ t w + --------------------------------------------------------------¸ ˜ ¨ ---------------------------------------------------¸ (2.60)
© – 59 ,57 ˜ t st + s s – 83 ,08 ¹ ¨ 184 ,22
t + 17 ,97 + ----------------¸
© f ss ¹

f yf
˜ § ---------------------------------------------------------------------------------------------·
© h w ˜ b 1 + 65 ,81 + b f – 98 ,77 ˜ s s – f yf ¹

§ hw ·
˜ ¨ t w + -----------------------------------------------------------------------------
2
-¸
© t f – 15 ,42 ˜ t st + f yf – 34 ,76 – b st¹

Another type of direct prediction model may be found in Graciano (2002), and later also
Graciano and Edlund (2003), in which the 4-hinge model (Roberts and Rockey (1979), see
Figure 2.13), with the addition of a longitudinal stiffener, acts as foundation for the work. Under
the assumptions that fyw = fyf (the shortcoming of the yield line model mentioned in section )
and that the position of the yield line would be D = b 1 e 2 d 20 ˜ t w the resistance expression,
a model in Graciano (2002) named “Model II: Failure mechanism model”, was stated as

34
 Plate Buckling - Theory

­ 2
° 2 E ˜ f yw ˜ t f 24 ˜ E ˜ I f ˜ s s + 2 ˜ t f – K ˜ M pw b1
° F Rl = 4 ˜ t w ----------------------- + -----------------------------------------------------------------------------
2
- if ----- d 40
tw
° b1 b 1 ˜ M pf
® (2.61)
° 2
2 ˜ E ˜ t 12 ˜ E ˜ I ˜ s + 2 ˜ t – K ˜ M b
° F = 2 ˜ f ˜ t 2 -------------------f + -----------------------------------------------------------------------------
f s f pw
- if ----1- ! 40
° Rl yw w
D ˜ f yf D ˜ Mpf
2 tw
¯

the latter also the resistance for the unstiffened web plate in which the yield lines are positioned
at D = 20 ˜ t w ˜ f yw e f yf and for both cases the parameter K, according to eq. (2.62), assuming
that the collapse load is transmitted over this length of the web which is yielding due to
membrane compressive stresses.

M pw ˜ 4 ˜ E + 2 ˜ s s + 2 ˜ t f
K = -----------------------------------------------------------------------------------------
2
- (2.62)
M pf
2 ˜ M pw + f yw ˜ t w ˜ D ˜ ----------------------------------
6 ˜ E ˜ I f ˜ M pw

However, in many cases f yw z f yf and aiming for a harmonized resistance formula for both
stiffened an unstiffened girders, along with being user friendly, there might be better options for
the designer than the above presented approach.

Amplification factor methods

As stated earlier, the amplification factor method used to predict the resistance of
longitudinally stiffened girders subjected to patch loading uses the resistance for the unstiffened
girder multiplied with an amplification factor, i.e. eq. (2.63).

F Rl = F R ˜ f s l (2.63)

The ordinary way of deriving such an amplification factor was with reference to
experimental work and tests of girders with the same dimensions and only the presence of a
longitudinal stiffener as difference. Further, the difference in ultimate resistance was expressed
as with an empirically determined function. Some examples of such amplification factor
proposals presented by some authors are presented in this section.

One of the more straight forward recommendations for an amplification factor was given in
Markovic and Hajdin (1992) who suggested a linear equation according to

b1 b1
f s l = 1 ,28 – 0 ,7 ˜ ------ for 0 ,1  ------  0 ,4 (2.64)
hw hw

which was derived based on test data from the literature comprising 133 longitudinally stiffened
and 318 unstiffened girders. Using eq. (2.64), the authors compared different equations for

35
predicting the resistance regarding unstiffened girders, FR, and searched for the best model to
be used to predict the resistance for a longitudinally stiffened girder according to eq. (2.63).
Testing several equations the best one, according to the authors was the one presented by
Roberts (1981), eq. (2.43). This either in the “basic” form according to eq. (2.43) or a form
including bending moment. Furthermore, Markovic and Hajdin concluded that eq. (2.43) was
the hitherto best formula to predict the resistance for the unstiffened girders if the influence of
the loading length was diminished.

The same year another amplification factor was proposed in Kutmanová and Skaloud (1992).
The research work was founded on earlier performed experimental and theoretical work (Janus
et. al (1988) described in section 3.1.6) regarding single- and double-sided longitudinally
stiffened as well as unstiffened girders. The results of the tests were analysed with a non-linear
regression approach and the following equation was established as amplification factor

b1
f s l = 0 ,958 – 0 ,09 ˜ ln § ------· (2.65)
© h w¹

and the ultimate resistance of an unstiffened girder according to

0 ,153
ss § If f yf ·
F R = 12 ,6 ˜ t w ˜ f yw ˜ § 1 + 0 ,004 ˜ -----· ˜ ¨ ----
2
- ˜ --------
-¸ (2.66)
© t w¹ © t 4 240¹
w

These equations are similar to the ones presented in Janus et. al (1988), however slightly
modified to have a better prediction level. However, one drawback of these equations may be
that they are established using tests from only one place, e.g. the steel delivered from the same
mill, same equipment used and so on. Though, the population used for the regression analysis
comprises many individual tests which is favourable.

Bergfelt
Based on his test results (see chapter 3) Bergfelt determined an amplification factor to take
the influence of the longitudinal stiffener into account. This with respect to the ultimate patch
loading resistance. In Bergfelt (1979) the three-hinge-flange model was presented together with
a resistance function for unstiffened girders according to eq. (2.32). However the investigation
in Bergfelt (1979) was started as an attempt to determine the factor f(sl) (see eq. (2.34)) and
through the comparison with the (relatively few and scattered) test results, Bergfelt proposed
eq. (2.67) as the amplification factor for a longitudinally stiffened girder.

1 b1 sK b1
f s l = 1 + §© --- + ------·¹ ˜ -------- ; 0 ,1  ------  0 ,33 (2.67)
3 hw 3b 1 hw

36
 Plate Buckling - Theory

in which the modified distance between the outermost plastic hinges in the upper flange is
proposed to lie in the interval

2
ss
s y + s s d s K d s y + s s + ---- (2.68)
sy

and with a correction factor for the flange bending moment, K, Bergfelt proposes eq. (2.69) to
determine the distance between the outermost plastic hinges.

bf t f 2 tw f yf
s y = 5 ,2 ˜ ---- ˜ § -----· ˜ ----
- ˜ -------------------- (2.69)
K tw © ¹ ti E ˜ f yw

However, Bergfelt also proposes a more simple way to determine the amplification factor for
the presence of the longitudinal stiffener. The idea behind this formula is according to Bergfelt
that the increase in ultimate resistance due to the longitudinal stiffener partly depends on the
ratio sy / b1and partly on the effect of the welding. The alternative formulation was stated as
follows

sy sy
1 + 0 ,02 ˜ -----  f s l  1 ,1 + 0 ,02 ----- (2.70)
b1 b1

Graciano
In the aforementioned doctoral thesis Graciano (2002), two additional models were
investigated besides the previously described Model II. The “Model I: Regression analysis of
test results” was based on the customary approach of amplification factors. As base for the
ultimate resistance Graciano used the findings of Lagerqvist (1994), see section , eq. (2.47)-eq.
(2.51), as the resistance for the unstiffened case. Based on a large number of test results found
in the literature, Graciano performed a regression analysis with the ratios b1 / hw, tf / tw and fyf
/ fyw as parameters in the amplification function. The results from this regression analysis was
that the best fit would be found if using an amplification factor function according to

b 1 f yf e f yw
f s l = 0 ,556 – 0 ,277 ˜ ln § ------ ˜ § -----------------· · (2.71)
© hw © tf e t w ¹ ¹

However, Graciano also states, with reference to statistics literature, that the main
shortcoming of his empirical approach is that the actual accuracy in prediction is strongly
dependent of the population size used for the analysis.

Reduction factor methods and the elastic critical load

As shown earlier in this chapter, the elastic critical load as an ultimate load has been proved
to be inadequate to use for design. However, the elastic critical load is usually essential to

37
determine the plate slenderness for von Kármán approach reduction factor models, i.e. most
reduction factor models regarding plate buckling of today. Hence, how the elastic critical load
is determined is of great importance to achieve a good correlation between the predicted
resistance and the actual resistance of e.g. a test or a real girder. The elastic critical load has been
subjected to extensive research work, e.g. Rockey et. al (1979), Graves-Smith and Gierlinski
(1982), Kutzelnigg (1982) and Janus et. al (1988). As mentioned earlier approximate solutions
were given to estimate the elastic critical load under various support and loading conditions.
Nowadays, the elastic critical load may be estimated in complicated cases by means of different
computer aided approaches, e.g. FEM. However, for “everyday” design purposes there has to
be analytical approximations of how to estimate the critical load, or usually the buckling
coefficient regarding the considered plate, i.e. eq. (2.48). More recent research work presenting
solutions for estimating the elastic critical load for a longitudinally stiffened web subjected to
patch loading may be found in Graciano (2002) and Davaine (2005).

The work regarding the elastic critical load presented in Graciano (2002) was founded on
numerical simulations using the FE-package ABAQUS. Graciano first studied simply
supported plates with and without longitudinal stiffeners and compared to previously presented
work. Furthermore the model got more refined adding flanges to the web, and a parameter study
was conducted in order to investigate the relevance of some parameters, e.g. the relative position
and flexural rigidity of the stiffener and the contribution from the flanges. Moreover the
influence of the torsional rigidity of the longitudinal stiffener was investigated which led to that
also closed longitudinal stiffeners were included in the study. The results from the numerical
investigation were then used to modify the buckling coefficient regarding unstiffened webs,
proposed in Lagerqvist (1994), see eq. (2.49), by adding a term, ksl, which took the contribution
from the longitudinal stiffener into account (eq. (2.72)). This term was proposed as a function
of the cross section of the stiffener (i.e. open or closed) and the panel aspect ratio of the upper
(directly loaded) panel b1 / a.

3
hw 2 b f ˜ tf
k F = 5 ,82 + 2 ,1 ˜ § ------· + 0 ,46 ˜ 4 ---------------
- + k sl (2.72)
© a¹ 3
hw ˜ tw

The term ksl added the contribution from the longitudinal stiffener, taking the relative
flexural rigidity of the stiffener into account and a factor, Co, which is through regression
analysis dependent of the ratio b1 / a and the ratio torsional / flexural rigidity of the stiffener
according to

k sl = C o ˜ J st (2.73)

Summing up the results from the regression analysis Graciano ended up in two expressions
for the Co parameter, according to eq. (2.74) below. The first will normally be suited for open
stiffeners and the second for closed section stiffeners.

38
 Plate Buckling - Theory

­
° 5 ,44 ˜ b----1- – 0 ,21 §I-----st-  0 ,15·
° a © J st ¹
Co = ® (2.74)
° b1
° 6 ,51 ˜ ----- §I-----st- t 0 ,15·
¯ a © J st ¹

However, the eq. (2.74) was concluded to only be valid if

0 ,05 d b 1 e a d 0 ,3 (2.75)

and

b 1 d 0 ,3 ˜ h w (2.76)

Further, the relative flexural rigidity of the stiffener should, according to Graciano, not be
taken larger than the transition rigidity, i.e. the rigidity for which the buckling mode of the web
switches from lateral displacement stiffener to a stiffener acting as a nodal line regarding the out
of plane web buckling. Thus, ksl is limited according to

k sl d C o ˜ J st t (2.77)

in which Jst,t is the transition rigidity for open stiffeners (or Ist / Jst < 0,15) according to

a 2 ,9 b1
J st t = 14 ˜ §© ------·¹ + 211 ˜ § 0 ,3 – -----·
©
(2.78)
hw a¹

or regarding closed stiffeners (or stiffeners with Ist / Jst > 0,15) according to

a 1 ,3
J st t = 45 ˜ § ------· (2.79)
© h w¹

Graciano also states that these sets of equations were obtained with geometric interaction
between the web plate and the longitudinal stiffener taken into account. Moreover, the equations
also account for the transition from global to local buckling modes. The above described
approach to determine the buckling coefficient for longitudinally stiffened webs under patch
loading was, combined with the design proposal by Lagerqvist (1994), by Graciano named
“Model III: Post-critical Resistance Approach”. This Model III was in Graciano (2002)
proposed to be used for design purpose since it was found to be the most complete model
available and with a good agreement with experimental comparison. Moreover, it was later
somewhat modified and in EN 1993-1-5 the recommended design method to use for predicting
the patch loading resistance of a longitudinally stiffened girder (see the section EN 1993-1-5
below).

39
 Further work concerning the elastic critical load for a longitudinally stiffened web was
presented in Davaine et. al (2004), Davaine and Aribert (2005) and later the doctoral thesis of
Davaine (2005). The aforementioned work comprised not only efforts focusing on the elastic
critical load alone, but also a complete reduction factor approach which is presented later herein.

Solely focusing of the elastic critical load within this section, the work presented by Davaine
and co-authors aimed for adding buckling of the upper panel to the expression used to estimate
the elastic critical load of the web. The proposal was based on considering the upper panel,
according to Figure 2.16, as simply supported and loaded on both longitudinal edges with an
un-symmetric in-plane load.

Figure 2.16: The simply supported upper panel as proposed in Davaine (2005).

Based on an extensive FE investigation comprising 366 numerical simulations, see section

3.2.1, the authors by regression analysis with the parameters (a / b1) and ((ss + 2tf) / a) derived
an expression for the buckling coefficient regarding the upper panel according to

s + 2 ˜ tf
§ 0 ,6 ˜ s--------------------
- + 0 ,5·
ss + 2 ˜ t f a © a ¹
k F2 = § 0 ,8 ˜ § ---------------------· + 0 ,6· ˜ § -----· (2.80)
© © a ¹ ¹ © b 1¹

The transfer of the applied load through the upper panel (slope 1:1) gives that the eq. (2.80)
is only valid when

ss + 2 ˜ t f + 2 ˜ b 1 d a (2.81)

When the buckling coefficient of Graciano describes the panel as a whole, the buckling
coefficient according to eq. (2.80) together with eq. (2.82) predicts the elastic critical load for
the upper panel alone. This buckling mode / failure mode has been commonly observed in the
numerical investigations of Davaine (2005) as well as experimental work by others.

2 3
S ˜E tw
F cr2 = k F2 ˜ -----------------------------
2
˜ ----- (2.82)
12 ˜ 1 – Q b 1

40
 Plate Buckling - Theory

Finally, Davaine and co-authors proposes an interaction formula for the two buckling modes
according to

1-
------ 1 1
= ---------- + ---------- (2.83)
F cr F cr1 F cr2

in which Fcr1 denotes the buckling load according to EN 1993-1-5 presented in following
section. The interaction formulation was based on observations during incremental loading in
the numerical simulations. Davaine and co-authors noticed that the response of the stiffened
web was divided in two steps; the first corresponding to local buckling in the lower panel and
the second local buckling of the upper panel until failure.

The reduction factor approach, used to predict the ultimate resistance of a longitudinally
stiffened girder subjected to patch loading, all depends on the web slenderness as shown
previously herein. Furthermore, within this section the two most recent publications on the topic
is presented along with the recommendations of the EN 1933-1-5. The reduction factor
approach by Graciano (2002) was modified to be implemented in the EN 1993-1-5 and hence
the original proposal (aforementioned Model III) will not be regarded within this section.

Davaine
Along with the proposal of the improved estimation of the critical load, see eq. (2.82) and eq.
(2.83), a proposal for an improved reduction factor function was proposed in Davaine (2005).
The reduction factor function was calibrated with the use of the extensive numerical simulations
by Davaine, see section 3.2.1, and furthermore also justified through a comparison with
experimental data gathered in the literature. Davaine proposed to use a function on the form of
eq. (2.56) and eq. (2.57), hence the plateau length and the imperfection factor were calibrated
to fit the numerical results. The parameters were determined to be set to DF = 0,21 and O0F =
0,8. Furthermore, Davaine proposed to set the term m2 = 0 regarding the expression for the yield
resistance. Emphasizing the origin of this parameter as the contribution from the web to the
outermost plastic hinges in the 4-hinge model of Lagerqvist, Davaine observed a better
correlation with the numerical results if the contribution from the web was omitted. Even though
this was not the scope of the doctoral thesis, the questioned term could be disregarded without
the whole concept failing. Recalling the previously described findings in Gozzi (2007) which
proved that the m2-term should be neglected, further indicates that the assumption of Davaine
was correct.

Seitz
Another approach for determining the ultimate patch loading resistance regarding a
longitudinally stiffened web was presented in Kuhlmann and Seitz (2002), (2004) and later
refined and presented in the doctoral thesis Seitz (2005). The scope of this approach was to
consider local buckling of each individual panel as well as global buckling of the whole
stiffened web. The approach was motivated via the different load cases the two (considering a

41
panel with only one stiffener) sub-panels are subjected to, i.e. “opposite patch loading” for the
upper and “regular patch loading” for the lower. Further, the upper panel was expected to fail
in a column buckling mode and the lower panel in a plate buckling mode with larger post-
critical reserves, so Kuhlmann and Seitz drew the conclusion that the resistance of the stiffened
girder would be possible to define by interpolation between the plate-like and the column-like
behaviour.

By an experimental investigation and following numerical experiments Seitz presented

interpolation functions used to determine the ultimate patch loading resistance as interpolated
between the plate buckling resistance and the column buckling resistance.

EN 1993-1-5
As previously mentioned, the design recommendations of EN 1993-1-5 is a modified version
of the Model III proposal of Graciano (2002). The procedure is rather straight forward and
presented in short terms herein.

The patch loading resistance of the longitudinally stiffened web is predicted according to

FR = F F ˜ F y (2.84)

with the yield resistance of the web determined as

F y = f yw ˜ t w ˜ s s + 2 ˜ t f ˜ 1 + m 1 + m 2 < f yw ˜ t w ˜ a (2.85)

The parameters m1 and m2 are calculated according to

f yf ˜ b f
m 1 = ----------------- (2.86)
f yw ˜ t w

and according to eq. (2.53)

­ h 2
° m 2 = 0 ,02 ˜ § -----w-· if O F ! 0 ,5
® © tf ¹
°
¯ m2 = 0 if O F d 0 ,5

The slenderness ratio, OF is as usual determined according to eq. (2.35), i.e.

F
OF = ------y-
F cr

with the critical load according to

42
 Plate Buckling - Theory

3
tw
F cr = 0 ,9 ˜ k F ˜ E ˜ ------ (2.87)
hw

So forth the proposal of Lagerqvist is followed, however to take the influence of the
longitudinal stiffener into account, the modified version proposed by Graciano (2002) is used
in EN 1993-1-5 according to

hw 2 b1
k F = 6 + 2 ˜ § ------· + § 5 ,44 ˜ ----- – 0 ,21· ˜ J st (2.88)
© a¹ © a ¹

in which the relative flexural rigidity of the longitudinal stiffener is calculated with

I a 3 b
st
- d 13 ˜ § ------· + 210 ˜ § 0 ,3 – ----1-·
J st = 10 ,9 ˜ --------------- (2.89)
hw ˜ tw 3 © ¹
hw © a ¹

where the second moment of area of the stiffener, Ist, is including contributing parts of the web
according to Figure 1.2. According to EN 1993-1-5 eq. (2.88) is valid for 0 ,05 d b 1 e a d 0 ,3
and b 1 e h w d 0 ,3 . Last but not least, the reduction factor is obtained by using the eq. (2.52), i.e.

0 ,5
F F OF = ------- d 1
OF
The attentive reader may here notice the differences in the original proposals of Lagerqvist
(1994) and Graciano (2002).

2.3.3. Interaction with bending

When loading a girder, e.g. simply supported, with some load between the supports it
inevitably also subjects the girder to bending moment. The case of patch loading is not an
exception to this. Hence, researchers has over the years proposed different models to take this
into account. Usually with an interaction model based on individual resistance models, i.e. the
patch loading resistance and the bending moment resistance, treated as individual phenomena
but interacting on basis of the interaction equation. This also implies that the interaction model
used would be the same disregarding of the web is stiffened or not.

One of the earlier contemporary interaction models was presented in Bergfelt (1971) who
proposed

F E· 8 ME 2
§ ------ + § --------· = 1 (2.90)
© F R¹ © M R¹

However in a later publication, Bergfelt (1976), the author states that no interaction between
patch loading and bending moment seems to be present when ME / MR < 0,6.

43
 Moving on to the into the 90’s, the publication Lagerqvist (1994) proposed two interaction
equations; one for welded girders, eq. (2.91) and one regarding rolled beams, eq. (2.92)
according to

F M
------E + 0 ,8 ˜ -------E- = 1 ,4 (2.91)
FR MR

F E· 2 ME 2
§ ------ + § --------· = 1 (2.92)
© F R¹ © M R¹

and in EN 1993-1-5, the first equation of these two are recommended for design purposes.

2.4. Summary of the theoretical review

As seen, a lot of effort has been put into the issues concerning plate buckling related ultimate
resistance. This both regarding plates under uniformly distributed compressive stresses as well
as unevenly distributed, herein focused on patch loading. Many different proposals have been
made on how the ultimate resistance should be predicted in the most precise way, both with
respect to theory as well as experimental observations and numerical simulations. As mentioned
earlier, the presence of initial imperfections such as residual stresses from welding and
geometric imperfections may reduce the theoretically determined resistance greatly. This has
been handled by using experiments as reference and develop semi-empirical resistance models.

The today recommended design models in the EN 1993-1-5 has been presented herein, and
the author will use these as a reference to the work presented in the following chapters.
Furthermore, the work presented by Gozzi (2007), Graciano (2002) and Davaine (2005) will be
used. The latter two as reference to the herein proposed resistance approach since both of them
are focused on the ultimate patch loading resistance for a longitudinally stiffened web. The
author of this thesis also find the work presented by Seitz (2005) interesting, however
shortcoming in the german language of the author herein makes the interpretation of the model
difficult and moreover very uncertain. Hence, no comparisons with this proposed approach will
be conducted herein.

44
 Patch Loading - Test Results

Chapter 3:
Patch Loading - Test Results

All since the 1950’ies experimental work focused on which and how the parameters of a
welded I-girder influences the patch loading resistance has been investigated. One of the more
internationally known early publications, considering longitudinally stiffened girder webs,
would by many researchers said to be the work of Allan Bergfelt in the end of the 70’ies and the
beginning of the 80’ies. The publication Rockey et. al (1978), with Bergfelt as co-author was
the start of an extensive investigation with many experimental tests. Bergfelt (1979) and
Bergfelt (1983) were continuing the previously conducted work. One of the larger, if not the
largest, test series was presented in Janus et. al (1988) which presented a test programme
comprising over 150 individual specimens, both stiffened and unstiffened.

In 1990 Dubas and Tschamper presented an investigation comprising also closed

longitudinal stiffeners. These stiffeners has a higher torsional stiffness and so forth also in many
cases a more favourable type of stiffener to use for web stiffening. The work in Dubas and
Tschamper (1990) used webs with open stiffeners as a reference to the V-shaped closed stiffener
specimens.

From open stiffeners and via V-shaped stiffeners, Carretero and Lebet in 1998 presented a
series of tests with girders stiffened with closed stiffeners of trapezoidal (TRP) type. The TRP-
type stiffeners are probably the most used today when a longitudinal stiffener is applied to e.g.
a bridge girder. More tests on specimens reinforced with TRP-stiffeners was presented in
Walbridge and Lebet (2001) and Kuhlmann and Seitz (2004). The latter also included FE
simulations verified via the experimental work and in Seitz (2005) proposals were made for an
improved design procedure (see section 2.3.2).

An extensive FE investigation into longitudinally stiffened girder webs was presented in

Davaine (2005). The work was focused on improving design codes and the proposals for
improved design is presented in section 2.3.2.

Within this thesis, some of the experimental work presented by some of the aforementioned
authors and more has been used to evaluate the EN 1993-1-5 patch loading resistance

45
recommendations. In Table 3.1 below the data used and its origin is presented as well as some
of the characteristics of the specimens used in the experiments. A total of 140 specimens with
open longitudinal stiffeners, 24 with closed stiffeners and 366 FE simulations were gathered
surveying the literature. The authors’ publications in which the test data and/or simulations were
presented is briefly described in this chapter.

The gathered data was in a first step evaluated with respect to EN 1993-1-5. However, the
current limitations in EN 1993-1-5 was not followed in here, i.e. the validation statements
regarding eq. (2.88) was not taken into account.

Table: 3.1: Characteristic data from the experiments and simulations gathered
from the published material introduced within this chapter.

Author No. of tests Open / Closed a / hw h1 / hw ss / hw

stiffener
Rockey et. al (1978) 4 Open 1,00 0,20-0,21 0,05
Bergfelt (1979) 9 Open 0,75-3,24 0,20 0,05-0,06
Bergfelt (1983) 6 Open 1,50-4,08 0,20-0,34 0,05-0,16
Galea et. al (1987) 2 Open 1,40 0,21-0,26 0,54
Shimizu et. al (1987) 1 Open 1,00 0,20 0,30
Janus et. al (1988) 101 Open 1,00-2,00 0,10-0,50 0,10-0,20
12 Open
Dubas and Tschamper (1990) 24 1,76-2,48 0,15-0,20 0,04-0,24
12 V-shape
Dogaki et. al (1990) 2 Open 1,00 0,20 0,10
Carretero and Lebet (1998) 6 6 TRP-shape 1,31-2,25 0,20-0,38 0,25-0,38
3 Open
Walbridge and Lebet (2001) 5 1,43 0,11-0,23 0,29
2 TRP-shape
Kuhlmann and Seitz (2004) 4 4 TRP-shape 2,00 0,25-0,30 0,58
Davaine (2005) (FEA) 366 Open 1,33-4,00 0,10-0,40 0,20-1,00

3.1. Patch loading experiments on longitudinally stiffened girders

3.1.1. Rockey et. al (1978)

During the first half of 1977 Rockey and Bergfelt made the first test in, what was to be, an
extensive investigation regarding longitudinally stiffened webs behaviour under patch loading.
The test series “R” comprised 8 ultimate load tests on a total of four specimens and was
presented in Rockey et. al (1978). Half of the test series was conducted on unstiffened girders
used as reference to the other half which was fitted with open longitudinal stiffeners welded on
to the webs. The four specimens were in other means identical, except differences in the flange
dimension. The purpose of the experimental work was to investigate the influence of the
longitudinal stiffener upon the ultimate patch load resistance. The tests were carried out in such

46
 Patch Loading - Test Results

a way that every girder was tested twice; the girder was after the first test turned so the patch
load could be applied on the undamaged flange.

Test setup
The two girders R2 and R4 was fitted with longitudinal stiffeners during the first set of tests
and hence used in the evaluation in this thesis. The other two girders (R1 and R3) tested in
Rockey et. al (1978) were also equipped with longitudinal stiffeners, but this only after the first
set of tests in which these were tested as unstiffened. In the following tests the girders R1 and
R3 were fitted with longitudinal stiffeners but in the tension zone. This in an attempt to reduce
the influence of the fact that the girders already had been loaded to failure due to the patch load.
The same procedure was followed concerning the stiffened girders R2 and R4. These were also
stiffened with an additional longitudinal stiffener in the new compression zone and the old
damaged part was now on the tension side of the girder. The exact dimensions of these four tests
on two girders may be found in Appendix A.

The tests were made with the girders simply supported and the patch load applied in the
centre of the flange. Strains were measured with rosette gauges, lateral deformations, initial as
well as during the tests, were also measured. Vertical deformation on both flanges of the girders
was measured in the centre of the specimens with transducers.

Test results and conclusions

The test results from Rockey et. al (1978) has here been evaluated with respect to EN 1993-
1-5 (Figure 3.1 below).

1.2
Rockey et. al (1978)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.1: The four tests on two girders with open stiffeners from Rockey et. al
(1978) evaluated with respect to EN 1993-1-5. Fexp/Fy as a function
of the slenderness, OF.

47
 The main conclusion in Rockey et. al (1978) was that it was shown that by using a
longitudinal stiffener positioned at one-fifth of the web depth, the patch loading resistance could
be significantly increased.

3.1.2. Bergfelt (1979)

The work initiated in Rockey et. al (1978), described in section 3.1.1, was with Bergfelt
(1979) taken a step further. In this publication two new test series (series “A” and “B”) were
accompanying the previously tested series “R” presented earlier.

Test setup
Test series “A” was intended to investigate the influence of a distance variation between
vertical stiffeners for both unstiffened and longitudinally stiffened girders. The series comprised
a total of 3 girders which was first tested unstiffened. After the first test a longitudinal stiffener
of open type was welded to the web and the same girder was tested once again, though rotated
so the patch load was applied on what was the tension flange when unstiffened. The girder was
then sectioned into two girders with a length between 510 and 1200 mm, cutting away
approximately 700 mm of the mid part (i.e. what was defined as the damaged part from the
previous patch loading). As a last step these smaller girders were equipped with vertical
stiffeners at the ends and tested two times (i.e. one test with the patch load applied on each
flange).

Initial out-of-plane deformations were measured with transducers as well as the propagating
buckling during the tests.

Test results and conclusions

The conclusions drawn in Bergfelt (1979) were summarized in three sections. First, the load
bearing capacity regarding patch loading on a girder with a slender web was increased through
the usage of a longitudinal stiffener. Second, the author concluded that the prediction of the
failure load is strongly dependent of the distance between the outermost formed plastic hinges
in the loaded flange. At last, Bergfelt concluded that the three-hinge-flange theory that was used
for the evaluation of the test results was very approximative, though giving a fair picture of the
beam behaviour immediately before failure. However, Bergfelt opinion was that this model had
to undergo some corrections before being used for quantitative calculations.

A total of nine tests on longitudinally stiffened girders were conducted and these tests were
evaluated with respect to the EN 1993-1-5 and may be found below in Figure 3.2. Furthermore,
the publication presented the test series “B” which comprised a total of 9 tests on unstiffened
girders. Though, these tests were not regarded herein because of the unstiffened webs.

48
 Patch Loading - Test Results

1.2
Bergfelt (1979)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.2: The nine tests on girders with open stiffeners from Bergfelt (1979)
evaluated with respect to EN 1993-1-5. Fexp/Fy as a function of the
slenderness, OF.

3.1.3. Bergfelt (1983)

As an extension to the previously described tests in Bergfelt (1979) the author in Bergfelt
(1983) presented six additional tests aiming to clarify some results and give a better basis for
calculations.

Test setup
A single sided open stiffener with varying upper panel depth were used on all the six tests
presented in this publication. The specimen layout and test set-up used were the same as
presented in section 3.1.2 also the same procedure in specimen fabrication was used (i.e.
dividing of one main girder into two smaller). The two main specimens had a panel length, a,
of 3000 mm and the four other smaller specimens panel lengths of 1100 mm. The longitudinal
stiffeners were placed at either h1 / hw = 0,2 or 0,34. The rest of the specimen dimensions may
be seen in Appendix A.

Test results and conclusions

The main conclusion in the publications concerning these new tests was that placing the
stiffener closer to the loaded flange had a larger beneficial influence on the patch loading
resistance.

The six specimens used herein were evaluated with respect to EN 1993-1-5 and the results
may be found in Figure 3.3.

49
 1.2
Bergfelt (1983)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.3: The six additional specimens from Bergfelt (1983) evaluated with
respect to EN 1993-1-5. Fexp/Fy as a function of the slenderness,OF.

3.1.4. Galea et. al (1987)

The work presented in Galea et. al (1987) was an investigation into how the presence of
bending moments and a longitudinal stiffener would influence the patch loading resistance.
Moreover the position of the stiffener and possible changes in the resistance was investigated.
The authors presented a test series consisting of four test girders; two stiffened and two without
any longitudinal stiffeners. Three of the experiments were performed with extra long span to
achieve the required amount of bending moment to investigate the interaction patch loading -
bending moment.

Test setup
The specimens used in the experimental work all had the same dimensions and the only
difference was the presence of an open longitudinal stiffener on two of the girders (P2 and P3).
These two girders are the ones used herein for further evaluation. The yield stress measured to
be between 244 - 286 MPa. Further information about dimensions and material properties may
be found in Appendix A.

The two specimens R2 and R3 were tested with an extra applied moment, or if put in another
way, with extra span; a total length of 15,4 m. The setup was of three-point bending type; the
beam simply supported with an external concentrated force applied at the centre of the beam.
The patch load was applied with a loading device consisting of four rollers spread over a load
length of 690 mm. As for the tests described in Shimizu et. al (1987) these tests were made with
specimens with extension girders to reach the required span regarding bending moment. The

50
 Patch Loading - Test Results

girders were prevented from lateral rotation, i.e. no risk for lateral-torsional buckling of the
beam.

The instrumentation in the test series comprised out-of-plane deflection measurement on

points situated in the vincinity of the centre-line of the beam. Measurements were made on both
surfaces of the web. Moreover, vertical deflection was measured with transducers placed on the
top and bottom flange. Strains were measured with rosette gauges, as well as uni-axial gauges,
on the web, the flanges and the transversal and longitudinal stiffeners.

Test results and conclusions

The two herein regarded girders R2 and R3 failed at loads of 720 and 730 kN respectively.
These loads accompanied with the dimensions of the specimens were used to evaluate the tests
with respect to EN 1993-1-5 and is shown in adjacent Figure 3.4. EN 1993-1-5 underestimates
the patch load resistance with approximately 50%.

1.2
Galea et. al (1987)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.4: Girders R2 and R3 from Galea et. al (1987) evaluated with respect to
EN 1993-1-5. Fexp/Fy as a function of the slenderness, OF.

The main conclusions by the authors were that the longitudinal stiffener increased the
ultimate load with approximately 37%. Furthermore, the authors concluded that the position of
the stiffener (i.e. at 1/4th or 1/5th of the depth) did not have any significant influence on the load
carrying capacity.

3.1.5. Shimizu et. al (1987)

In Shimizu et. al (1987) the authors presented a study comprising tests of 10 specimens
reinforced with one to three longitudinal stiffeners of open type.The aim of the study was to
clarify the buckling or collapse behaviour of stiffened web panels during a simulated launching

51
procedure. All the specimens were fabricated of SS41 steel (fy = 235,2 MPa). The total length
of the girders was either 6 or 9 meters. This in order to study the interaction moment-patch
loading behaviour. Herein the specimen with a single stiffener was taken into account, denoted
EL1.

Test setup
The specimens were attached to extension girders with a bolted connection to get a longer
girder and also a larger applied moment. These end beams were used in all the tests and the
small mid part (the actual specimen) was replaced to form a new test setup. All specimens had
a depth of 1 m and the stiffener size was 80 x 6 mm. Additional supports were used to prevent
lateral-torsional buckling of the girders. Further details of the geometry may be found in
Appendix A.

Strains in the girders web were measured with rosette gauges on both surfaces of the web.
Furthermore, uni-axial strain gauges were placed on the flanges to measure the axial strains in
these. Out-of-plane deflections of the webs were measured with a transducer. This was also
conducted prior to the test to determine the initial curvature of the web plates. Vertical
deformations were measured on both the top and the bottom flange.

Test results and conclusions

The test of the girder EL1 was compared to an unstiffened sibling and was shown to have a
resistance 31% higher than the unstiffened version of the same beam. The test result from the
girder EL1 were used for an evaluation with respect to the EN 1993-1-5 and this is shown in
Figure 3.5 below.

1.2
Shimizu et. al (1987)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.5: EL1 with open stiffener from Shimizu et. al (1987) evaluated with
respect to EN 1993-1-5. Fexp/Fy as a function of the slenderness, OF.

52
 Patch Loading - Test Results

From the experimental work presented the authors drew the conclusions that a smaller span
length increases the maximum loads. Furthermore a wider launching shoe (load length) was also
concluded to be beneficial concerning the patch load resistance.

3.1.6. Janus et. al (1988)

One of the most comprehensive studies based on experimental work was presented during
the late 80-ies by Janus et. al (1988). A total of 152 tests on steel girders under patch loading
were made. The specimens were of unstiffened type as well as longitudinally stiffened with
single- and double-sided open stiffeners. A total of 101 specimens were of stiffened model,
hence also used for the evaluation herein.

The authors’ aims for the study was to achieve a general understanding regarding the
behaviour of a longitudinally stiffened steel girder subjected to patch load and possible
differences between the stiffened and unstiffened types. Furthermore the influence of the
stiffener rigidity was examined with respect to any correlation to the patch loading resistance.
Moreover the position of the stiffener and the influence of changes in this parameter was
investigated.

Test setup
The test series was divided into four sets in which different parameters and their influence
on the ultimate patch loading capacity was examined. The parameters/quantities varied
throughout the test series were the position of the longitudinal stiffener (i.e. b1), the size of the
stiffener, the height-to-thickness ratio of the web (i.e. hw / tw), the aspect ratio of the web
(a / hw) and the size of the loaded flange. All of the dimension and material characteristics of
the 101 stiffened specimens used in this evaluation is listed in Appendix A. During all the tests
the ratio load length / panel length (ss / a) was held equal to 0,1.

The setup of the tests were of three-point type with the girder simply supported and the patch
load applied in the beam centre. Strains were measured on a number of positions on both the
web and the longitudinal stiffener. Displacement out-of-plane concerning the web buckling as
well as vertical deformation of both flanges and the longitudinal stiffener were measured with
electrical transducers. Furthermore the initial curvature of the web was measured prior to the
test on all specimens.

Test results and conclusions

Janus et. al (1988) concluded that all of the test girders failed with a segmental plastic hinge
line under the patch load in the web and three point plastic hinges were also developed in the
loaded flange. The authors’ also concluded that the presence of a longitudinal stiffener
substantially increased the patch loading resistance only if the stiffener was located in the
vincinity of the loaded flange with b1 / hw < 0,25. Furthermore the tests presented in Janus et.
al (1988) led to the establishment of a patch loading resistance model of the amplification factor
type, i.e. eq. (2.63).

53
 The results from the tests presented by Janus et. al (1988) was used to be evaluated with
respect to the EN 1993-1-5 and is shown in adjacent Figure 3.6. A total of 101 specimens
equipped with open stiffness were used in the comparison. Noticable in Figure 3.6 is not only
the large scatter within the tests, it is also evident that the EN 1993-1-5 overestimates the patch
loading resistance in some cases. This seems to be evident for more stocky webs, i.e. web
slenderness, OF < 0,7.

1.2
Janus et. al (1988)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.6: The 101 tests on specimens with open longitudinal stiffeners from
Janus et. al (1988) evaluated with respect to EN 1993-1-5. Fexp/Fy as
a function of the slenderness, OF.

3.1.7. Dubas and Tschamper (1990)

An extensive test programme comprising 48 unstiffened webs and 24 panels with
longitudinally stiffened webs subjected to patch loading were presented in Dubas and
Tschamper (1990). Herein the 24 panels equipped with both open and closed (V-shaped)
stiffeners were evaluated and each specimens data may be found in Appendix A.

Test setup
The study of Dubas and Tschamper was investigating how the torsional rigidity of the
stiffeners influenced the ultimate patch loading resistance. Moreover, the interaction bending
moment / patch loading was studied through the application of additional load pairs analogous
with the setup used by Kuhlmann and Seitz (2004) described in section 3.1.11.

Test results and conclusions

The longitudinally stiffened test girders showed failure modes with buckling of the upper
(loaded) panel. Also the patch loading resistance was shown to be improved by the use of the
stiffeners, especially concerning the more torsional stiff closed V-shape stiffener. Common for

54
 Patch Loading - Test Results

the panels was that the relative position of the stiffeners, h1 / hw, concerning the closed stiffeners
was kept constant at 0,2. Regarding the open stiffeners the ratio b1 / hw was 0,15 or 0,2.

The 24 tests used in the evaluation herein were evaluated with respect to EN 1993-1-5 and
the results may be found in Figure 3.7. The results in plotted Figure 3.7 all points out an
underestimation of the patch loading resistance when calculating according to the EN 1933-1-
5. Generally speaking, it seems to be more evident for the closed section stiffened panels than
for the ones equipped with open stiffeners.

1.2
Dubas and Tschamper (1990) (OS)
Dubas and Tschamper (1990) (CS)
EN 1993-1-5
0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.7: The 24 experiments from Dubas and Tschamper (1990) evaluated
with respect to EN 1993-1-5. Fexp/Fy as a function of the slenderness,
OF.

3.1.8. Dogaki et. al (1990)

Plate girders reinforced with longitudinal stiffeners and subjected to patch loading was the
focus in Dogaki et. al (1990). The work presented in the publication was a part of a wider patch
loading investigation regarding stiffened and unstiffened girders. In Dogaki et. al (1990)
experimental work comprising three girders, two stiffened and one without stiffener, was
presented. The paper presented the ultimate load tests from the three girders as well as
comparison with theoretical approaches to predict the failure loads presented by other authors.

Test setup
The test specimens which were equipped with a longitudinal stiffener (model 4 and model 5)
had an open stiffener of thickness 4,5 mm. The width of the stiffener was 30 and 38 mm
respectively. The other dimensions was nominally the same, but the measured values may be
viewed in Appendix A.

55
 The test setup was of simply supported type with the patch load applied in the centre of the
girder. Lateral rotations were prevented by supports at the sides of the beam, as well as with help
of the hydraulic jack used for application of the load.

Numerous strain gauges of uniaxial and rosette type were used to monitor and measure the
developing strains during the tests. The strains were measured on both flanges and in the web
(with rosette gauges). Furthermore, the vertical displacement was measured at both upper and
lower flange at the centre of the span. The out-of-plane deformations of the web was also
measured with transducers.

Test results and conclusions

The ultimate patch load of the model 4 and model 5 was reached at approximately 105 and
110 kN respectively. These experimental loads was with the specimen dimensions evaluated
with respect to EN 1993-1-5 and presented in Figure 3.8 below. As may be seen, the EN 1993-
1-5 somewhat underestimates the resistance, however not by more than approximately 30%.

Dogaki et. al (1990) concluded that the post-critical strength of longitudinally stiffened
girder under patch loading was remarkable. Moreover, the buckling of the web was all localized
to the upper panel and the longitudinal stiffeners seemed to be stiff enough to form a nodal line
for the out-of-plane deformation of the web. As for the comparison of the theoretical models
three predictions out of four were underestimating the ultimate load. These when comparing
with the models of Janus et. al (1988) and a model previously developed by the authors of
Dogaki et. al (1990).

1.2
Dogaki et. al (1990)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.8: The two specimens from Dogaki et. al (1990) evaluated with respect
to EN 1993-1-5. Fexp/Fy as a function of the slenderness, OF.

56
 Patch Loading - Test Results

3.1.9. Carretero and Lebet (1998)

In Carretero and Lebet (1998) an investigation of the behaviour of slender webs subjected to
patch loads was presented. The results of an experimental investigation were compared to four
different resistance models; a model presented in Dubas and Tschamper (1990), a swiss design
norm named SIA 161, the patch load resistance model by Lagerqvist (1994) and the modified
model in EN 1993-1-5 described in section 2.3.2.

Test setup
A total of 6 composite beams were tested under a concentrated load. Concerning the study
presented herein, 6 of the panels in the beams were longitudinally stiffened with a TRP stiffener
and furthermore used in the evaluation. The dimensions of the specimens may be found in
Appendix A.

Test results and conclusions

A general conclusion of the comparison was that all four models seemed to give conservative
predictions of the patch load resistance, however the model by Lagerqvist (1994) was concluded
to be the best prediction model and the modified resistance model in EN 1993-1-5 gave
predictions approximately 25% more conservative. Furthermore, Carretero and Lebet
concluded that longitudinal stiffening of slender girder webs increased the patch load resistance
with approximately 25 - 60% dependent of the placing of the stiffener (i.e. the depth of the upper
panel, b1). The test results regarded in this thesis were evaluated according to EN 1993-1-5 and
presented in Figure 3.9.

1.2
Carretero and Lebet (1998)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.9: The 6 panels with a closed stiffener from Carretero and Lebet (1998)
evaluated with respect to EN 1993-1-5. Fexp/Fy as a function of the
slenderness, OF.

57
 As shown in Figure 3.9 the actual patch loading resistance, Fexp, compared to EN 1993-1-5,
FR, was up to 2,2 times the predicted load. However, a large scatter between the 6 tests may be
seen.

3.1.10.Walbridge and Lebet (2001)

The experimental work presented in Walbridge and Lebet (2001) consists of tests of a total
of 6 specimens; two fitted with a closed TRP-shaped longitudinal stiffener, three with an open
stiffener and one girder was unstiffened as reference to the stiffened ones. The girders were
made as composite beams with a lower flange reinforced with concrete according to Figure
3.10. Herein the results from the five stiffened girders will be taken into account for further
evaluations.

Figure 3.10: The cross-section layout of the specimens tested in the experimental
investigation presented in Walbridge and Lebet (2001).

Test setup
The tests setup was made with the intention to simulate the launching of a bridge girder with
the concrete already cast on the flange. The loaded length of the upper flange was held constant
for all tests, i.e. 200 mm and the web depth was in all cases 700 mm over a panel length of 1000
mm. The longitudinal stiffeners were placed that the depth of the upper panel was 75, 100 or
125 mm. The complete list of dimensions and some material properties of the specimens may
be found in Appendix A.

Test results and conclusions

Walbridge and Lebet (2001) concluded that the torsional stiffness of the closed stiffeners was
capable to restrict the out-of-plane deformations of the web in a more efficient way then
compared to the open type with less torsional stiffness. Using the TRP-stiffener at 75 mm
distance from the upper flange increased the ultimate patch loading resistance with 64%
compared to the unstiffened reference girder meanwhile an open stiffener at the same position
increased the resistance by 31% only.

58
 Patch Loading - Test Results

The results from Walbridge and Lebet (2001) was evaluated with respect to EN 1993-1-5 and
are shown in Figure 3.11. Noticeable is the high resistance when compared to the one predicted
by EN 1993-1-5.

1.2
Walbridge and Lebet (2001) (OS)
Walbridge and Lebet (2001) (CS)
EN 1993-1-5
0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.11: The five specimens from Walbridge and Lebet (2001) evaluated with
respect to EN 1993-1-5. Two specimens with closed stiffener and 3
with an open stiffener. Fexp/Fy as a function of the slenderness, OF.

3.1.11.Kuhlmann and Seitz (2004)

Girders reinforced with closed section stiffeners subjected to patch loading was the topic in
Kuhlmann and Seitz (2004). Experimental work as well as FE simulations were conducted with
the aim of improving design methods for predicting the ultimate patch load capacity. Larger
loading lengths than most of the researchers investigated previously were used to fill this gap in
knowledge. Pure patch loading was examined as well as the interaction of patch loading and a
larger bending moment. The results from the experimental work comprising a total of 7 tests on
5 girders were evaluated and used to verify the FE simulations.

Test setup
The test series consisted of 3 smaller girders with a span of 2,4 m and two larger specimens
with a length of 9,6 m. The latter was used for introducing bending moment high enough to
examine the patch loading - bending moment interaction. All of the girders except one had
longitudinally stiffened webs. All stiffeners were of closed TRP type. Furthermore, two of the
patch loading tests were made on panels with two longitudinal stiffeners. The test setup and the
loading frame is schematically described in Figure 3.12. The girders with a single longitudinal
stiffener were used in the evaluation herein and the dimensions and other characteristics of these
girders used in the evaluation herein may be found in Appendix A.

59
 Figure 3.12: The test setup used for the patch loading experiments with “extra”
applied moment. The load pair denoted “Q” used to introduce a
higher bending moment. Kuhlmann and Seitz (2004).

Kuhlmann and Seitz measured out-of-plane web deformations, initial as well as growth
during loading, on a grid consisting of 23 x 15 measuring points. The vertical displacement was
also measured, in this case with a pair of transducers in each of the four loading points. Hence
vertical deflection was measured in 4 points on both side of the loading rig. Strains were
measured with uniaxial, as well as rosette gauges. The rosette gauges were applied on both sides
on the web and the uniaxial gauges were used to measure the axial strains in the flanges.

Test results and conclusions

In Kuhlmann and Seitz (2004) it was concluded that by the use of a closed longitudinal
stiffener the patch loading resistance could be substantially increased. The authors showed that
with a TRP stiffener positioned at h 1 = 0 ,25 ˜ h w the patch load resistance increased with
about 56% when compared to an unstiffened girder. At h 1 = 0 ,3 ˜ h w the increase was
somewhat lower, about 44%. However, using two stiffeners increased the patch loading
resistance even more; about 86% higher resistance when comparing to the unstiffened case the
authors concluded.

Concerning the patch loading - bending moment interaction, the authors could not observe
any significant differences in the patch load resistance when applied under a bending moment.
Only when the combination bending, shear and patch load was examined, a reduction in the
patch load resistance could be noticed.

Regarding the evaluation of tests results herein, the four tests on girders with a single closed
longitudinal stiffener were used. Two of these were tested with an extra applied bending
moment. The patch loading resistance evaluated with respect to EN 1993-1-5 is shown in Figure
3.13.

60
 Patch Loading - Test Results

1.2
Kuhlmann and Seitz (2004)
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.13: Four specimens with a closed longitudinal stiffener from Kuhlmann
and Seitz (2004) evaluated with respect to EN 1993-1-5. Fexp/Fy as a
function of the slenderness, OF.

3.2. Numerical simulations

3.2.1. Davaine (2005)

In 2005 an extensive study based on numerical FE simulations was presented in the doctoral
thesis of Davaine (2005). The study was focused on steel girders with deep webs, e.g. up to 5
m. The aim of the study was to justify the approach in EN 1993-1-5 also for longitudinally
stiffened girders with webs this deep. The author found that the experimental data presented by
other researchers mainly comprised specimens with girder depths up to 1,2 m. The work
presented in Davaine (2005) was based on the findings from simulations on 366 specimens with
different geometries, listed in Appendix A, however all of the girder webs were reinforced with
a stiffener with open cross-section. The FE-model was validated with respect to results from
previously tested girders. Not only patch loading was investigated, but also the interaction of
patch loading - bending moment.

The FE-simulations was first used to re-formulate the critical load Fcr as stated in EN 1993-
1-5. In a second step, the resistance function was calibrated, some of these findings are
presented in section 2.3.2. Furthermore, the proposals were calibrated with the statistical
procedure described in Annex D of EN 1990 (2002).

Within this thesis the 366 simulations were used to be evaluated and the evaluation with
respect to EN 1993-1-5 is presented in Figure 3.14.

61
 1.2
Numerical simulations
EN 1993-1-5

0.8
Fexp / Fy

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 3.14: The 366 numerical simulations of Davaine (2005) evaluated with
respect to EN 1993-1-5. Fexp/Fy as a function of the slenderness, OF.

3.3. Summary of the experimental review

Summarizing the conclusions from the previously conducted research work, the authors to
the above presented publications all seems to agree that adding a longitudinal stiffener to the
web increases the patch loading resistance. If the stiffener if of closed type, this effect seems to
be even larger due to the increased stiffness both with respect to out-of-plane bending and
torsion. However, the actual gain of resistance using longitudinal stiffeners all depends on the
girder cross-section, placement of the stiffener and the stiffener cross-section.

A total of 140 individual tests made on specimens with open longitudinal stiffeners were
used in the evaluation herein. Comparing the test results with respect to the by EN 1993-1-5
predicted resistance in Figure 3.15, it seems like the majority of the tests are on the safe side.
Nevertheless, some of the stockier tests ( O F a 0 ,6 ) by Janus et. al (1988) seems to be
overestimated with respect to their resistance. Furthermore, the scatter amongst the individual
tests by all the authors are noticable. Also, specimens with web slenderness, OF > 2 would
benefit from a raised reduction curve when looking into Figure 3.15.

Considering Figure 3.16 containing the 24 specimens with a closed stiffener type, all tests
seems to be on the safe side of the reduction curve of EN 1993-1-5. Though, one test by
Carretero and Lebet (1998) exactly coincides with the curve. However, it seems that the margin
of safety regarding the prediction of EN 1993-1-5 seems to be rather high and moreover the tests
also in this case seems to show a large scatter.

62
 Patch Loading - Test Results

Tests from literature (OS)

3 EN 1993-1-5
Fexp / FR

0
0 1 2 3 4 5
Slenderness, OF

Figure 3.15: The 140 specimens with open section longitudinal stiffeners. Ultimate
experimental load, Fexp, compared to the EN 1993-1-5 recommended
design procedure.

Tests from literature (CS)

3 EN 1993-1-5
Fexp / FR

0
0 1 2 3 4 5
Slenderness, OF

Figure 3.16: The 24 specimens with closed section longitudinal stiffeners. Ultimate
experimental load, Fexp, compared to the EN 1993-1-5 recommended
design procedure.

Regarding the numerical simulations by Davaine (2005) most of the 366 numerical
simulations seems to keep together in a cluster, see Figure 3.17, though some simulations seems

63
to have much larger resistance than predicted by the EN 1993-1-5. Furthermore, regarding the
simulations with a web slenderness, OF > 1,5 the EN 1993-1-5 curve could have been raised to
better coincide with the simulations.

3
Fexp / FR

1
Numerical simulations
EN 1993-1-5

0
0 1 2 3 4 5
Slenderness, OF

Figure 3.17: The 366 numerical simulations with open section longitudinal
stiffeners. Ultimate experimental load, Fexp, compared to the EN
1993-1-5 recommended design procedure.

All in all, the tests regarding both open and closed stiffeners together with the numerical
simulations, could benefit from a modification of the resistance function, better shaped to fit the
more slender specimens, i.e. the prediction of the ultimate patch loading resistance could be
improved for more slender girders. Further, the scatter amongst the tests could possibly be
reduced by a better estimation of the buckling load of the stiffened webs.

64
 Patch Loading - Design Proposal

Chapter 4:
Patch Loading - Design Proposal

The aim herein is to find a design approach for longitudinally stiffened girders subjected to
patch loading consistent with the one proposed in Gozzi (2007) concerning unstiffened girders.
This involves the yield resistance of the web, the elastic critical load to determine the
slenderness of the web and as third component a reduction function to determine the reduction
factor as a function of the slenderness. The basics of the reduction factor approach used to
determine the ultimate patch loading resistance have been presented in chapter 2 along with
proposals from other authors as well as the today recommended approach given in EN 1993-1-5.

The previously presented tests reported by other authors (see chapter 3) are used to validate
the proposal. Hence, the herein proposed ultimate patch loading resistance model was validated
using data from tests on girders with both open and closed sections stiffeners, as well as
numerical simulations. Furthermore, this chapter also contains a comparison of the proposed
model with the most recent published directly comparable models of Graciano (2002) and
Davaine (2005). Also a comparison to the EN 1993-1-5 proposed approach to predict the
ultimate patch loading resistance is conducted.

4.1. Yield resistance

As stated in chapter 2, some of the work presented in Gozzi (2007) was focused on making
an improvement concerning the expression for the yield resistance. Emphasizing previously
mentioned risen questions regarding the web contribution to the plastic resistance in the
outermost hinges in the model by Lagerqvist (1994), Gozzi investigated if this criticism was
justified.

The findings of the conducted research work of Gozzi concluded that the questioned part of
the yield resistance expression should be neglected, i.e. the contribution from the web to the
bending moment resistance of the outer plastic hinges should be omitted in the mechanism
model. Furthermore, since the mechanism regarding an unstiffened web subjected to patch
loading is profoundly the same as for a web longitudinally stiffened, the work by Gozzi should
be applicable also regarding longitudinally stiffened girders and would not induce any direct

65
sources of resistance prediction issues. Moreover, in a historical perspective, some design codes
e.g. EN 1995-1-5 have recommended use of the same equations for the yield resistance both for
stiffened and unstiffened girders. Since this way of designing has been used it would be
preferable, from a designers point of view, to maintain this correlated design recommendations,
i.e. basically the same equations to use regardless to if the web is longitudinally stiffened or not.
Hence, the expression in eq. (4.1) will herein be applied as the yield resistance of the
longitudinally reinforced web subjected to patch loading.

§ § f yf ˜ b f · ·
F y = f yw ˜ t w ˜ ¨ s s + 2 ˜ t f ˜ ¨ 1 + ----------------
-¸ ¸ (4.1)
© © f yw ˜ t w¹ ¹

However since the model is considering one panel subjected to patch loading, the effective
loaded length ly, expressed in the brackets of eq. (4.1), is limited to the panel width a. Hence,
the yield resistance will inhere be determined as

§ § f yf ˜ b f · ·
F y = f yw ˜ t w ˜ ¨ s s + 2 ˜ t f ˜ ¨ 1 + ----------------
-¸ ¸ d f ˜ t ˜ a (4.2)
© © f yw ˜ t w¹ ¹ yw w

4.2. Elastic critical load

One of the more debated and difficult parts to determine of the three in the resistance function
approach is probably the elastic critical load for the stiffened web. Nowadays there are a
multitude of ways to determine the critical load, e.g. numerical methods, software solely
developed to predict the critical load or, maybe most commonly used, by hand calculations
using different models. As presented in section 2.3.2 the most recent work regarded herein on
this topic was presented in Graciano (2002) and Davaine (2005). The work by Graciano was
after some modifications implemented in EN 1993-1-5. However, the elastic critical load or
buckling coefficient of Graciano only regarded the whole web panel. This was in Davaine
(2005) further improved to also include a consideration of the critical load for the upper panel
alone, eq. (2.82), using a buckling coefficient according to eq. (2.80). This critical load uses a
theory based on a non-uniform opposite patch loading of the upper panel. Furthermore, the
actual load distribution in the upper panel is assumed to be in on a 1:1 slope which leads to that
s s + 2 ˜ t f + 2 ˜ b 1 d a for the buckling coefficient to be valid, otherwise the usually present
vertical stiffeners will carry a larger portion of the load and the results will be conservative.
Davaine (2005) also stated that during the numerical simulations performed, an interaction
behaviour between the two buckling modes was observed, i.e. the buckling mode described by
the equations of Graciano / EN 1993-1-5 and the one regarding the upper panel of Davaine.
However, the author of this thesis finds it hard to justify an elastic critical load, used for
characterizing the slenderness of the web in the elastic region, determined as an interaction of
two different buckling modes. The most intuitive way must, according to the author, be using
the lowest of the two critical loads, i.e.

66
 Patch Loading - Design Proposal

­ F cr1
F cr = min ® (4.3)
¯ F cr2

in which the first is proposed to be calculated according to eq. (2.87) and the second according
to the proposal of Davaine, i.e. eq. (2.82) and eq. (2.80).

4.3. Reduction function

As previously stated, many different forms of reduction functions have over the years been
proposed and furthermore proven to be well suitable to use under certain circumstances.
Though, since the reduction factor usually is determined as a function of the slenderness the
calibration of these together are of great importance. Since a modified expression of the yield
resistance (section 4.1), and furthermore accompanied with changes of the way to calculate
elastic critical load according to section 4.2, was adopted, the reduction function has to be tuned
to fit these new proposals. To achieve the aim of using the same reduction factor for both
longitudinally stiffened and unstiffened girders as historically done, the work presented in
Gozzi (2007) was investigated as a solution. This since the proposed reduction function on the
form originally proposed by Müller (2003), already has been proven to be able to predict the
ultimate resistance of unstiffened girders using the proposed modifications regarding the yield
load. A proposal on the form of Müller has, due to the use of two modifiable parameters, a
superior flexibility inherent compared to a function on the form of EN 1993-1-5, i.e. eq. (2.52).

The proposed function of Gozzi (2007) was calibrated using 184 patch loading experiments
with unstiffened webs with induced bending moments according to ME / MR < 0,4. Furthermore,
the limit of the reduction factor concerning stocky webs, i.e. with a slenderness lower than what
is needed to reduce the resistance with respect to buckling, was chosen to be set to 1,2 instead
of the usually used 1,0 of e.g. eq. (2.52). This was based on the fact that the more stocky
specimens showed a resistance higher than the yield resistance and a better fit of the curve
(prediction of the actual real behaviour) could be achieved in this way. However, the limitation
of the reduction factor in the lower region of web slenderness would actually not be needed
when compared to the experimental work presented in Gozzi (2007), though a bit
unconventional. Nevertheless, the proposed reduction function for unstiffened girders, as
presented in section 2.3.1, reads

1
F F = ------------------------------------- d 1 ,2
2
MF + MF – OF

with

1
M F = --- ˜ 1 + 0 ,5 ˜ O F – 0 ,6 + O F
2

67
 A comparison of the above presented resistance function with respect to some other usually
used or historically famous may be studied in Figure 4.1.

1.6
0,5 / O
1/O
1.2
Reduction factor, F

1 / O2
Winter function
Proposal
0.8

0.4

0
0 1 2 3 4 5
Slenderness, O
Figure 4.1: Comparison of reduction curves, e.g. the Winter function, the herein
proposed reduction function.

4.4. Proposal of design approach

Collecting the three parts regarding the yield load, the critical load and the reduction factor
function a complete ultimate patch loading resistance approach may be stated. As mentioned
earlier, the different parts have been proposed based on numerical simulations and/or
comparisons with experimental work. Common for both approaches are that they are limited to
comprise specimens, numerical as well as experimental, with properties spread over an interval.
This interval is usually used as the interval for which one states the proposal to be valid over.
However, inhere the limitations of previous research was disregarded if not directly violating
statements in proposed equations or physical limitations concerning the specimen. Hence, based
on the research of Gozzi (2007) the yield load will be calculated according to eq. (4.2).

Regarding the critical load, the lowest of the critical load for the whole panel and the upper
panel is suggested to be used, i.e.

­ F cr1
F cr = min ®
¯ F cr2

with Fcr1 according to EN 1993-1-5 and Graciano (2002) which states

68
 Patch Loading - Design Proposal

3
tw
F cr1 = 0 ,9 ˜ k F1 ˜ E ˜ ------ (4.4)
hw

with the buckling coefficient for the whole stiffened web according to

hw 2
k F1 = 6 + 2 ˜ § ------· + k st (4.5)
© a¹

with the influence from an open stiffener is calculated as

b1
k st = § 5 ,44 ˜ ----- – 0 ,21· ˜ J st (4.6)
© a ¹

If the above stated addition to the buckling coefficient for the whole unstiffened web not
should be given the possibility to be negative, i.e. the buckling coefficient for a stiffened web
would be lower than the corresponding unstiffened, the following relation is needed

b 1 0 ,21
----- t ---------- | 0 ,039 (4.7)
a 5 ,44

In EN 1993-1-5 this possibility is given, i.e. if the panel is wide enough it would be possible
that the buckling coefficient for a stiffened panel would be lower than for an unstiffened with
the same dimensions. To avoid this oddity, the buckling coefficient for the unstiffened panel is
herein proposed as a lower bound for kF1. Hence, eq. (4.7) can be disregarded if the influence
of the longitudinal stiffener is limited to

b1
k st = § 5 ,44 ˜ ----- – 0 ,21· ˜ J st t 0 (4.8)
© a ¹

As for closed section stiffeners, the proposal of Graciano (2002) is proposed to be used
herein.

b1
k st = 6 ,51 ˜ ----- ˜ J st (4.9)
a

Regarding the relative flexural rigidity of the stiffener, Jst, this is calculated as

E˜I st
J st = --------------
- (4.10)
D ˜ hw

69
with the moment of inertia of the stiffener, Ist, including the contributing parts of the web
according to Figure 1.2. However, when regarding a longitudinal stiffener of open type the
relative flexural rigidity is limited by the transition rigidity according to

a 3 b
J st d J st ,t = 13 ˜ §© ------·¹ + 210 ˜ §© 0 ,3 – ----1-·¹ (4.11)
hw a

At this point EN 1993-1-5 uses the limitation of the ratio b1 / a < 0,3, otherwise the last
bracketed product in eq. (4.11) would be negative. However, inhere this aspect is disregarded
and if b1 / a > 0,3 the transition rigidity is set to the first term, i.e.

­ a-· 3 b1 b1
° 13 ˜ § ----- + 210 ˜ § 0 ,3 – -----· if ----- d 0 ,3
° © h w¹ © a¹ a
J st d J st ,t = ® (4.12)
° a 3 b1
° 13 ˜ §© ------·¹ if ----- ! 0 ,3
¯ hw a

Regarding closed section stiffeners, the statement for the transition rigidity of Graciano
(2002) is kept to be calculated as

a 1 ,3
J st d J st ,t = 45 ˜ §© ------·¹ (4.13)
hw

Regarding the elastic critical load for the upper panel, the only limit regarding dimensions of
this proposal is governed here. The derived expression of Davaine (2005) is proposed to be used
under the restriction that the responding length (loaded length) of the lower part of the upper
panel has to be smaller than the actual panel width, that is

s s + 2 ˜ t f + 2 ˜ b1 d a

The buckling coefficient for the upper panel regarding panels with both types of stiffeners is
proposed to be calculated in line with the work of Davaine, i.e. eq. (2.80) and eq. (2.82).

At last the yield load and the elastic critical load determines the slenderness of the stiffened
web according to the von Kármán approach stated in eq. (2.35) and further determines the
reduction factor according to eq. (2.58) and eq. (2.59).

4.5. Validation of the design proposal

To validate the proposed design procedure the data from the in chapter 3 presented
experimental work was used. However, four tests had to be excluded from the data base due to
the restrictions of the load distribution in the upper panel. I.e. s s + 2 ˜ t f + 2 ˜ b 1 d a was not
satisfied which may lead to that the vertical stiffeners would probably carry much of the applied

70
 Patch Loading - Design Proposal

load. Furthermore, the statement would probably be satisfied for most common dimensions of
girders, hence excluding the tests TG1-1 to TG1-3 by Janus et. al (1988) and the EL1 of Shimizu
et. al (1987) (see Appendix A) would not be detrimental to the design applicability. After
removing these four tests, 136 specimens with open stiffeners, 24 test with closed stiffeners and
366 numerical simulations remains to be used in the proposal validation.

Regarding the relative flexural rigidity of the stiffeners, these were calculated according to
eq. (4.10) with a Young’s modulus set to 210 GPa and a Poisson’s ratio of 0,3. The results from
the experiments in relation to the proposed ultimate resistance approach are shown in Figure 4.2
(open stiffeners), Figure 4.3 (closed stiffeners) and Figure 4.4 (numerical simulations).

Tests from literature (OS)

3 Proposal
Fexp / FR

0
0 1 2 3 4 5
Slenderness, OF
Figure 4.2: Fexp/FR for the respective specimen slenderness according to
the proposal. 136 tests with open stiffeners.

As seen the scatter amongst the tests are somewhat large, however this may be the case when
using a larger amount of test data produced at various test institutes. Experience shows that
when using test data from one or a few laboratories, the scatter is often decreased. This may
have its origin in different measurement equipment, test setups etc. However, an other
possibility may be that the scatter may be native of parameters not included in the model.
Though, this negative side of the scatter may also be turned to something that may be counted
as a strength of a prediction model, i.e. if the model may predict the inhomogeneous test
population safely the reliability in design work would be higher than if using a model only based
on for example numerical simulations or tests made at one laboratory.

When concluding the two graphs over the tests regarding the open and closed stiffeners, the
conclusion that all of the 136 + 24 specimens are safely predicted by the proposal. However,
only the characteristic ultimate resistance are yet regarded. Further, the scatter amongst the

71
relatively few closed stiffener experiments seems to be comparable to what was shown
considering the much larger data base of open stiffener experiments.

Tests from literature (CS)

3 Proposal
Fexp / FR

0
0 1 2 3 4 5
Slenderness, OF
Figure 4.3: Fexp/FR for the respective specimen slenderness according to
the proposal. 24 tests with closed stiffeners.

Regarding the substantial amount of numerical simulations results used from Davaine (2005)
the Figure 4.4 below shows that the most of the simulations are predicted conservatively, i.e.
safe, by the herein proposed prediction model.

Numerical simulations
3 Proposal
Fexp / FR

0
0 1 2 3 4 5
Slenderness, OF
Figure 4.4: Fexp/FR for the respective specimen slenderness according to
the proposal. 366 tests with open stiffeners.

72
 Patch Loading - Design Proposal

When the results presented Figure 4.2 - Figure 4.4 above is put in statistical figures, the mean
value, the coefficient of variation and the standard deviation of the three sub-groups are
presented in Table 4.1.

Table: 4.1: Statistical interpretation of the results shown in Figure 4.2 - Figure
4.4. Experimental results, Fexp with reference to the predicted
ultimate resistance, FR.

Open Closed Numerical

stiffeners stiffeners simulations
Number of tests 136 24 366
Mean 1,496 1,499 1,410
Standard deviation 0,251 0,271 0,235
Coefficient of variation 0,168 0,180 0,167
Lower 5-percent fractile 1,162 1,060 1,125
Upper 5-percent fractile 1,975 1,879 1,793

As seen in Table 4.1 the two groups of test data (open and closed stiffeners) seems to be
comparable with each other. When regarding the numerical simulations the statistical
parameters seems to be somewhat better which would be explained if pointing out that more
than two times as many specimens have been used.

Regarding the neglection of the upper limit of the ratio b1 / a the Figure 4.5 containing both
open and closed section stiffeners, shows that this assumption would not jeopardize the safety
of the model. The prediction model seems to underestimate the actual resistance somewhat
more for these, nevertheless, the model may be used without restrictions regarding b1 / a and
still be safe. However, there are only four individuals tests with a upper panel depth / width ratio
above 0,3.

73
 Tests from literature (OS)
3 Tests from literature (CS)
Fexp / FR Proposal

0
0 0.1 0.2 0.3 0.4
b 1/ a
Figure 4.5: Fexp/FR as a function of the ratio b1 / a. 160 tests with
longitudinal stiffeners.

As a last step in the validation process a statistical evaluation of the proposed design model
was conducted according to the recommendations of Annex D in EN 1990 (2002). This
evaluation ends up in a partial safety factor, JM1, to be used for calculation of the design
resistance. This evaluation may be studied in its full extents in Appendix B.2. Within this
chapter only the final result is given, i.e. the partial safety factor for the proposed design model,
evaluated using the 136 tests of girders with an open stiffener and the 24 tests with a closed
longitudinal stiffener. The partial safety factor was determined to JM1 = 1,0 which herein will
be given as the recommendation for design purposes. Moreover, the same partial safety factor
was determined on basis of the numerical simulations comprising the 366 simulations. The
result from this evaluation is enclosed in Appendix B.2.3 with the evaluation of the partial factor
which was determined to JM1 = 1,0.

4.6. Comparison with other models

The herein proposed ultimate patch loading resistance approach was furthermore compared
to the proposal of Davaine (2005), the un-modified proposal of Graciano (2002) and the
recommended approach of EN 1993-1-5. The comparison was conducted using the same
specimens and the aforementioned equation restrictions concerning the ratio b1 / a. The
respective statistical parameters of the other three models may be studied in Table 4.2.

74
 Patch Loading - Design Proposal

Table: 4.2: Comparison of the proposals by Graciano (2002), Davaine (2005),

the recommendations in EN 1993-1-5 and the herein proposed
approach. The 136 tests with open stiffeners were regarded.

Graciano Davaine EN 1993-1-5 Proposal

(2002) (2005)
Mean 1,249 1,476 1,456 1,496
Standard deviation 0,245 0,264 0,337 0,251
Coefficient of variation 0,196 0,179 0,271 0,168
Lower 5-percent fractile 0,861 1,152 0,892 1,162
Upper 5-percent fractile 1,626 1,980 2,078 1,975

As seen, the model by Graciano (2002) is the only one showing a lower standard deviation
combined with a lower mean value than the proposed approach. However, when studying the
results more closely, it can be concluded that the model by Graciano seems to overestimate the
ultimate resistance regarding some of the more stocky tests. Regarding the corresponding
statistical parameters the closed section stiffener specimens and the numerical tests, these may
be studied in Table B.1 and Table B.2. Though, a conclusion from these tables and their
respective graphs Figure B.14 - Figure B.21 is that the proposed model of Davaine (2005) seems
to be the best one to predict the ultimate resistance of the numerical experiments used herein.
However, this may not be regarded as a complete surprise since the model of Davaine was
calibrated with respect to these numerical experiments.

4.7. Interaction with bending moment

Regarding the possible interaction with bending, the bending moment resistance was
calculated according to the specifications of EN 1993-1-5 with respect to the specimens cross-
section class respectively. Possible reductions due to local buckling was also regarded for the
stiffeners; open stiffeners was treated as an “outstand flange” and the parts of a closed section
stiffener was handled as an “internal compression part”.

Furthermore, some, or rather most, of the specimens in the data base was of hybrid type, i.e.
with different yield resistances for the web and flanges respectively. However, for those
experiments where the stiffener material properties were given, these were in all cases the same
as for the web. The differences in the strength of the girder parts was also taken into account,
both regarding the “common” hybrid girder with stronger flanges than web and also for the
cases when the web was stronger than the flanges. These procedure was also presented in
chapter 1.

The model originally proposed by Lagerqvist (1994) and furthermore implemented in EN

1993-1-5 to be recommended to use for interaction between patch loading and bending moment
was presented in section 2.3.3 and eq. (2.91). The eq. (2.91) was proposed to be used regarding
welded girders, hence the one used in the evaluation herein.

75
 2.5
Tests from literature (OS)
Tests from literature (CS)
2 EN 1993-1-5

1.5
Fexp / FR

0.5

0
0 0.25 0.5 0.75 1 1.25
M E / MR
Figure 4.6: Fexp/FR as a function of ME/MR for the 160 tests with open and
closed longitudinal stiffeners.

3
Numerical simulations
2.5 EN 1993-1-5

2
Fexp / FR

1.5

0.5

0
0 0.25 0.5 0.75 1 1.25
ME / MR
Figure 4.7: Fexp/FR as a function of ME/MR for the 366 numerical
simulations with open longitudinal stiffeners.

As seen in Figure 4.6 no obvious interaction between bending moment and patch loading
may be observed for the experimental results. However, the experiments under a high bending

76
 Patch Loading - Design Proposal

moment utilization are few, hence no conclusions regarding this matter will be drawn herein.
However, even though the aforementioned tests are few, they are still safely predicted using the
interaction equation recommended by EN 1993-1-5 according to eq. (2.91). Considering the
numerical simulations no further conclusions regarding a potential moment - patch loading
interaction may be drawn. Figure 4.7 shows no real interaction behaviour, though the
simulations subjected to larger bending moment utilization is, as for the experiments, too few
to make any statement.

4.8. Summary of the proposed design procedure

In this section the proposed design procedure is summarized with its different steps. The
following calculation procedure is proposed to be used for the prediction of the patch loading
resistance of a longitudinally stiffened I-girder. The longitudinal stiffener may be of open or
closed type, however, the proposal is only validated for panels stiffened with one stiffener. Also,
the girders used in the evaluation was all of welded type. However, the proposed resistance
model is only valid for girders satisfying eq. (4.14).

s s + 2 ˜ tf + 2 ˜ b1 d a (4.14)

The yield resistance of the girder web is calculated using

§ § f yf ˜ b f · ·
F y = f yw ˜ t w ˜ ¨ s s + 2 ˜ t f ˜ ¨ 1 + ----------------
-¸ ¸ d f ˜ t ˜ a (4.15)
© © f yw ˜ t w¹ ¹ yw w

which is multiplied with the reduction factor according to

F R = FF ˜ Fy (4.16)

The reduction function is proposed to be on the form

1
F F = ------------------------------------- d 1 ,2 (4.17)
2
MF + MF – OF

in which

1
M F = --- ˜ 1 + 0 ,5 ˜ O F – 0 ,6 + O F (4.18)
2

The slenderness of the girder web is calculated using the von Kármán approach according to

F
OF = ------y- (4.19)
F cr

77
 The elastic critical load is proposed to be calculated as the lowest of the one regarding the
upper panel and the whole panel respectively, i.e. as

­ F cr1
F cr = min ® (4.20)
¯ F cr2

in where the critical load for the whole panel should be calculated according to

3
tw
F cr1 = 0 ,9 ˜ k F1 ˜ E ˜ ------ (4.21)
hw

with the buckling coefficient kF1

hw 2
k F1 = 6 + 2 ˜ § ------· + k st (4.22)
© a¹

Regarding the upper panel, the elastic critical load is proposed to be calculated using

2 3
S ˜E tw
F cr2 = k F2 ˜ -----------------------------
2
˜ ----- (4.23)
12 ˜ 1 – Q b 1

with a buckling coefficient of

s + 2 ˜ tf
§ 0 ,6 ˜ s--------------------
- + 0 ,5·
ss + 2 ˜ t f a © a ¹
k F2 = § 0 ,8 ˜ § ---------------------· + 0 ,6· ˜ § -----· (4.24)
© © a ¹ ¹ © b 1¹

The improved stiffness of the panel due to the presence of a longitudinal stiffener, kst, is
calculated differently regarding open and closed section stiffeners. Regarding open stiffeners
this term is proposed to be calculated according to

b1
k st = § 5 ,44 ˜ ----- – 0 ,21· ˜ J st t 0 (4.25)
© a ¹

with the relative flexural rigidity of the stiffener according to

E˜I st
J st = --------------
- (4.26)
D ˜ hw

in which the moment of inertia of the stiffener, Ist, includes the contributing parts of the web
according to Figure 1.2.

78
 Patch Loading - Design Proposal

However the transition rigidity of the stiffener, Jst,t, is set as an upper limit of the relative
flexural rigidity, according to

­ a-· 3 b1 b1
° 13 ˜ § ----- + 210 ˜ § 0 ,3 – -----· if ----- d 0 ,3
° © h w¹ © a¹ a
J st d J st ,t = ® (4.27)
° a 3 b1
° 13 ˜ §© ------·¹ if ----- ! 0 ,3
¯ hw a

Regarding closed section stiffeners, the added stiffness from the longitudinal stiffener is
proposed to be calculated as

b1
k st = 6 ,51 ˜ ----- ˜ J st (4.28)
a

with the relative flexural rigidity, Jst, according to eq. (4.26) however with a limiting transition
rigidity according to

a 1 ,3
J st d J st ,t = 45 ˜ §© ------·¹ (4.29)
hw

The design resistance is the predicted using

F Rd = F F ˜ F y e J M1 (4.30)

with the partial safety factor according to Appendix B, i.e. JM1 = 1,0

4.9. Concluding remarks

Summing up the results presented in this chapter, an ultimate patch loading resistance model
regarding longitudinally stiffened webs was presented which better predicted the ultimate patch
loading resistance than herein compared models. The recommendations of EN 1993-1-5, the
proposal of Graciano (2002) and the proposal of Davaine (2005) were used for the comparisons.
Furthermore, the herein proposed model uses the same reduction factor function as for
unstiffened webs proposed in Gozzi (2007), which makes it more suitable for design purposes,
i.e. one comprehensive resistance function to be used regardless of the presence of longitudinal
stiffeners or not.

The proposed model was validated through a comparison with tests results and numerical
simulations, though only regarding girders reinforced with one longitudinal stiffener. However,
the model was found to be applicable for webs with open as well as closed section stiffeners,
see Figure 4.8 and Figure 4.9 below. The model was found to be relevant also disregarding the
validation ratio limits b1 / a and b1 / hw used in the compared models, see Appendix B.1.

79
 1.6
Tests from literature (OS)
Tests from literature (CS)
1.2 Proposal
Fexp / Fy

0.8

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 4.8: The 136 experiments with open stiffeners and 24 with closed stiffeners
compared to the proposed ultimate patch loading resistance model.

When comparing the proposed design model with the model of Graciano (2002) the latter
model seems to overestimate the results for b1 / hw > 0,3 (see Figure B.14 and Figure B.16).
This is also outside the validation limit of the model by Graciano, however comparing the
aforementioned two figures with Figure B.1 and Figure B.12 it seems that using the herein
proposed model will hand predictions on the safe side when b1 / hw > 0,3. Studying Figure B.15
and Figure B.17 showing how the ratio b1 / a influences the level of prediction for the model of
Graciano, it may be observed that outside the valid interval 0,05 < b1 / a < 0,3 there are some
tests and numerical simulations on the unsafe side. Figure 4.5 and Figure B.12 showing the
corresponding results by the herein proposed model, the latter seems to predict the experimental
results safely disregarding the ratio 0,05 < b1 / a < 0,3.

Comparing the proposed approach with the model proposed in Davaine (2005) on the same
basis as above, the model by Davaine seems to predict all of the experimental results safely (see
Figure B.18 and Figure B.19) however with reference to Table 4.2 and Table B.2 the proposal
of this thesis seems to be a model with less scatter and a better mean value for the closed section
stiffened panels. However, regarding the open stiffeners the mean value of the model by
Davaine seems to be slightly better, though with a larger scatter amongst the individual tests.

Regarding the numerical simulations, a few results from the simulations seems to be
overestimated with respect to the ultimate patch loading resistance. The reason for this are
difficult to point out, since there seems to be no special differences in geometry, material,
loading conditions etc. when compared to the ones predicted safely.

80
 Patch Loading - Design Proposal

1.6
Numerical simulations
Proposal
1.2
Fexp / Fy

0.8

0.4

0
0 1 2 3 4 5
Slenderness, OF
Figure 4.9: The 366 numerical simulations with open stiffeners compared to the
proposed ultimate patch loading resistance model.

All in all, the herein proposed model seems to perform better than the models used in the
comparison regardless of the ratio b1 / a and b1 / hw. Furthermore the proposed model was found
to be applicable both to webs stiffened with one open or one closed section longitudinal
stiffener. The partial safety factor was evaluated with respect to both the test results and the
numerical simulations and found to be in both cases 1,0.

81
82
 Local Buckling - Test Results

Chapter 5:
Local Buckling - Test Results

The plate buckling phenomena has, as mentioned in previous chapters, been quite thoroughly
investigated. This also on a strictly experimental basis. The research work is in a continuous
state since new steel grades and design rules enter the field of constructional applications.

The articles and papers presented in this chapter have been chosen to be comparable to the
tests in chapter 6. This with respect to specimen layout, welding conditions, support conditions,
steel grades and other comparable similarities. Furthermore, all the test results presented in this
chapter are evaluated with respect to the Winter function discussed in chapter 2 and according
to the EN 1993-1-5 specifications concerning plate slenderness values.

5.1. Nishino et. al (1967)

An investigation aiming to clarify how residual stresses influence the resistance against local
buckling was presented by Nishino et al. (1967). Specimens used in this research work were
fabricated of plates welded together to form a square cross section, see Figure 5.1, and tested in
as-welded condition.

Figure 5.1: Specimen layout and weld detailing. Nishino et al. (1967).

Two different steel grades were used for the specimens, ASTM A7 (sheared specimen plates)
and ASTM A514 (flame-cut specimen plates) with properties according to adjacent Table 5.1.
In addition to the buckling tests the residual stress condition in the specimens were measured
with the sectioning method.

83
 Table 5.1: Results from tension coupon tests.The average compressive residual
stresses was estimated regarding each plate (side) individually.
Nishino et al. (1967).

Specimen Material Yield Strength, Average compressive Ratio

No. fy [MPa] residual stress, Vrc [MPa] Vrc / fy
1 A7 273,0 83 - 97 0,23
2 A7 266,1 69 - 76 0,16
3 A514 799,8 76 - 83 0,10
4 A514 710,8 97 - 103 0,15

5.1.1. Test Setup

The tests of the specimens were divided into four sets, each comprising two specimens with
the same geometrical properties and made of the same steel. The width - thickness ratios were,
according Nishino et al., selected such that the critical loads were reached in either the elastic
range or the elastic-plastic range. Furthermore, the length to width ratio of the plates were
between 4,35 to 7,2. This would according to the authors guarantee that

• the buckling mode corresponding to the lowest critical load would be developed
and

• short enough to prevent column buckling to be the governing failure mode.

The buckling tests were performed with the specimens under uniformly distributed
compressive force as the specimens were equipped with rigid end plates, milled flat to simplify
the alignment in the test rig. Simply supported conditions were assumed to be valid constraints
for the plates in the welded specimen.

5.1.2. Test results and conclusions

Regarding the conclusions drawn by the authors in their article, the most interesting
concerning this theses were:

• The effect of residual stresses on the buckling strength of a plate is less pronounced
for A 514 steel than it is for A 7 steel.

• Considerable post-buckling strength exists in a plate buckled in the elastic range,

while a plate buckled in the elastic-plastic range has a relatively small reserve of
post-buckling strength.

• The plate elements of square columns of A 514 steel are stronger than those of A 7
steel when compared on a non dimensional basis (compared to the yield strength of
each grade).

84
 Local Buckling - Test Results

Furthermore the test results presented by Nishino et al. were re-evaluated herein with respect
to the Winter function and EN 1993-1-5 and presented in Figure 5.2.

1.5
Specimens of grade A7
Specimens of grade A514
EN 1993-1-5
1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.2: Test results from all the 8 specimens from Nishino et al.(1967). The
results are re-evaluated with respect to the Winter function, eq.
(2.24). Plate slenderness according to EN 1993-1-5.

5.2. Dwight et. al (1968)

The local buckling tests presented in Dwight et al. (1968) comprised tests of square box,
rectangular box and cruciform sections. The square box section tests were conducted with the
aim of filling gaps in previously presented tests reported by J.D. Harrison and also presented by
Dwight and Moxham (1969), see section 5.3.1. A total of 49 columns were tested in as-welded
and stress relived condition. However only four of the tests were in as-welded condition and of
square-box model, hence the ones used in the evaluation in this thesis.

The mechanical properties of the steel used for fabrication of the specimens were determined
through compression tests. The length of the specimens were set to 3,5 to 4 times the plate width
and tested under uniform compressive stress. In this evaluation the result from four of these tests
were used and the yield stress in compression was measured to 354 and 403 MPa respectively.
The test results from these tests were re-evaluated herein with respect to the Winter function and
EN 1993-1-5 and presented in Figure 5.3.

85
 1.5
Dwight et. al (1968)
EN 1993-1-5

1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.3: Test results from 4 as-welded specimens reported in Dwight et al.
(1968). The results were re-evaluated with respect to the Winter
function. Plate slenderness according to EN 1993-1-5.

5.2.1. Conclusions
Dwight et al. concluded that the difference between the resistance of an as-welded specimen
compared to a stress relieved specimen could be in the order of 10 to 15%. This considering a
considerable range of width to thickness ratios and with the higher resistance concerning the
stress relieved specimens.

5.3. Dwight and Moxham (1969)

Another survey of work by different researchers in the field of plate buckling were presented
by Dwight and Moxham (1969). The paper focused on investigating how well the British
standards BS 153 and BS 449 of 1969 were describing the actual behaviour of plate buckling
and was somewhat a continuation of the work described in section 5.2. Special efforts were put
into investigating how the weld induced residual stresses affected the ultimate resistance with
respect to local buckling. Dwight and Moxham gathered test results from over 40 welded
column specimens of square box sections with yield strengths in the range of 232 to 402 MPa.
The tests applicable to this theses are listed below.

5.3.1. Tests made by J.D. Harrison

Dwight and Moxham reported results from 20 experiments made by J.D. Harrison and J.B.
Dwight. These specimens were in as-welded as well as in annealed condition. The length of the
specimens were about 4 times the plate width and the specimens were loaded with uniformly
distributed compressive stress. In this thesis the only regarded tests are the as-welded ones (10

86
 Local Buckling - Test Results

specimens). Furthermore, the test results from these tests were re-evaluated with respect to the
Winter function and EN 1993-1-5 to be comparable within this thesis. These re-evaluated
results are presented in Figure 5.4.

5.3.2. Tests made by K.E. Moxham

In Dwight and Moxham (1969) three additional tests were collected for the evaluation. These
tests were made by K.E. Moxham and conducted in a similar way to the one described above.
However, these tests were made in a somewhat larger scale with bigger specimens (plate
thickness of 12,7 mm) but still in as-welded condition and thereby possible to compare with the
other tests reported herein. The re-evaluation of the three specimens, with a yield strength of
312 MPa, are in Figure 5.4 presented along with the others from the same publication.

1.5
Dwight and Moxham (1969)
EN 1993-1-5

1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.4: Test results from the 13 as-welded specimens reported in Dwight and
Moxham (1969). The results were re-evaluated with respect to the
Winter function. Plate slenderness according to EN 1993-1-5.

5.3.3. Conclusions
Several conclusions were drawn concerning the work presented by Dwight and Moxham
(1969). Concerning this thesis relevant conclusions are:

• Residual stresses caused by welding may reduce the strength of fabricated

members in relation to the size of the welds.

• The load - deformation curve for a web containing residual stresses is less peaky
than that for a stress free web.

87
5.4. Fukumoto and Itoh (1984)
A comprehensive investigation regarding uniformly compressed steel plates was presented
in a paper by Fukumoto and Itoh (1984). The purpose of the work was to review and store data
of experimental investigations under clearly defined and described conditions. The authors
collected data from 793 individual tests of a variety of cross sections such as single plates,
welded square boxes, square and rectangular tubes, welded rectangular sections and cruciform
specimens. Data concerning specimens in as-welded as well as annealed condition were
regarded. 13% of the data collected was regarding specimens made of steel with higher yield
strength than 430 MPa (definition of high strength steel in the paper).

Data concerning initial geometrical imperfections, residual stress levels and ultimate loads
were presented in form of histogram plots. Concerning the residual stresses, Fukumoto and Itoh
states that the magnitude of the residual compressive stress may not be influenced of the yield
stress of the base material. This statement was founded on 32 residual stress measurements on
specimens made of high strength steel which showed that the Vrc / fy ratio was lower for the high
strength steel specimens compared to the rest of the data set.

Fukumoto and Itoh collected results from 383 plates with inherent residual stresses. The
plates were of the type with welds along the unloaded edges (in tubes or as single plates) or as-
welded box sections. The authors made a non-linear regression analysis with an assumed
uniform variance on the data and the mean function presented with a standard deviation of
0,0871 was

Vu 0 ,968 0 ,286 0 ,0338

------ = ------------- – ------------- + ---------------- for Op t 0 ,571 (5.1)
fy Op O
2
O
3
p p

Furthermore, the authors made the same analysis for 172 plates without residual stresses.
These plates were as-cut, annealed or annealed box sections. The result from this analysis was

Vu 1 ,133 0 ,384 0 ,0468

------ = ------------- – ------------- + ---------------- for Op t 0 ,658 (5.2)
fy Op O
2
O
3
p p

with a standard deviation of 0,104. Herein both the equations eq. (5.1) and eq. (5.2) are
compared to the Winter function in Figure 5.5.

88
 Local Buckling - Test Results

1.5
Plates with residual stresses (eq. 5.1)
Plates without residual stresses (eq. 5.2)
EN 1993-1-5
1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.5: Mean functions of plates with eq. (5.1) and without eq. (5.2) residual
stresses from Fukumoto and Itoh (1984) compared to the Winter
function.

Several interesting conclusions were drawn by Fukumoto and Itoh concerning their
experimental data-base approach. Conclusions among others were:

• No clear difference between the plate strengths determined through single plate
tests and square boxes could be pointed out.

• Annealed plates showed larger variations in strength than as-welded plates.

• Further experimental investigations were needed concerning plates of high

strength steel.

5.5. Rasmussen and Hancock (1992)

An investigation with the aim of determining if high strength steel with yield stress in the
range 450 - 700 MPa could be designed according to existing Australian design rules was
presented in Rasmussen and Hancock (1992). A test programme comprising box welded
sections and cruciform shaped specimens as well as I-shaped sections were used to examine if
the design codes had to be modified or if they were usable also for the grades with higher
strength (a similar aim as for this thesis, except the difference in regarded codes). The
investigation focused on whether the yield slenderness limits for welded uniformly compressed
plates supported along one or both longitudinal edges were applicable to the high strength steels.
However, since this thesis solely focus on plates supported along both sides, these test results
are the only ones regarded herein. Furthermore, the intention of the investigation by Rasmussen

89
and Hancock (1992) may not be completely in line with the aim of this thesis, still the test results
from the paper in question are valuable and re-evaluated with respect to the Winter function.

The test programme was divided into three parts; measurement of the material properties
with tension and compression coupons, residual stress measurement through specimen
sectioning and compression tests of the specimens. The specimens were all made of
BISALLOY 80 steel which, according to Rasmussen and Hancock, is equivalent to the ASTM
A514 grade. The through coupon tests measured mechanical properties of the BISALLOY 80
grade are presented in Table 5.1.

Table 5.1: Nominal and measured mechanical properties of BISALLOY 80.

Rasmussen and Hancock (1992).

Nominal plate Type of test Youngs modulus, Measured values,

thickness [mm] E [GPa] fy / fu [MPa]
5 Tension 211 670 / 775
5 Compression 211 750 / -

5.5.1. Test setup

The box specimens used in the test programme were all fabricated by weld joining four plates
(Figure 5.6) with nominal thickness of 5 mm and with 3 different nominal widths (plate
slenderness values in Figure 5.7). Gas metal arc welding with a Lincoln L50 wire were used for
all the welds.

Figure 5.6: Specimen layout and weld detailing. Rasmussen and Hancock (1992).

The specimens were milled flat at the ends to allow a proper seating in the test rig. The
bottom plate was fixed to prevent rotation and the top plate was mounted on a spherical seat.
Furthermore the length of the specimens were chosen to allow unrestrained development of
local buckles and short enough to prevent overall instability phenomena (column buckling).

5.5.2. Residual stress measurement

The longitudinal residual stresses were measured with the sectioning method and readings
were made with use of strain gauges. Gauges were applied near the centreline of each plate of

90
 Local Buckling - Test Results

the box specimen and the mean values of the measured compressive stresses on the four plates
are presented for each specimen in Table 5.2.

Table 5.2: Measured residual stresses of box columns. The average compressive
residual stresses was estimated regarding each specimen
individually. Rasmussen and Hancock (1992).

Specimen Width, Thickness, Average compressive Yield Strength, Ratio

b [mm] t [mm] residual stress, Vrc [MPa] fy [MPa] Vrc / fy
B1RS 80 5 169 670 0,25
B2RS 110 5 114 670 0,17
B3RS 140 5 73 670 0,11

5.5.3. Test results and conclusions

The test results from Rasmussen and Hancock were re-evaluated herein with respect to the
Winter function and EN 1993-1-5 and presented in Figure 5.7.

1.5
Rassmusen and Hancock (1992)
EN 1993-1-5

1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.7: Test results from all the 6 specimens from Rasmussen and Hancock
(1992). The results were re-evaluated with respect to the Winter
function, eq. (2.24). Plate slenderness according to EN 1993-1-5.

The investigation of the high strength steel sections presented by Rasmussen and Hancock
rendered in the following conclusions regarding the box sectioned specimens:

• The strength of slender welded high strength steel plates exceeds that of welded
ordinary steel plates when compared on a non dimensional basis (compared to the
yield strength of each grade). The test results suggest that the difference in the non

91
 dimensional strength may be greater for plates supported along one longitudinal
edge than for plates supported along both.

• More slender plates are more affected of the presence of the residual stresses than
stockier ones. This is due to the fact that the more stocky plates may be almost
completely plastified at the ultimate load level.

5.6. Möller and Johansson (1995)

Buckling tests on six specimens made of high strength steel were presented by Möller and
Johansson (1995). The aim of the investigation was to investigate the buckling behaviour of the
newly developed steel “WELDOX 1100” from SSAB Oxelösund. The yield stress of this grade
was measured to 1349 MPa.

The specimens were of stub column type with a box shaped cross section, Figure 5.8, and the
height of the specimens was chosen to 3,5 times the specimen width. This to prevent column
buckling, avoid clampening effects from the end supports and to allow the specimen to buckle
in such a way that the lowest buckling load would be acquired. Furthermore, the specimens were
tested in as-welded condition.

Figure 5.8: Specimen layout and weld detailing. Möller and Johansson (1995).

5.6.1. Test setup

The tests were performed under uniform compression of the specimens between two rigid
end plates. The deformation speed was chosen such that the nominal stress would reach the
yield strength within 30 seconds. Furthermore the deformation of the specimens were carried
on until a 50% load drop from the ultimate load was acquired. Deformation and load data was
sampled during the tests.

92
 Local Buckling - Test Results

5.6.2. Test results and conclusions

The test results from Möller and Johansson (1995) were re-evaluated with respect to the
Winter function and EN 1993-1-5 and presented in Figure 5.9.

1.5
Möller and Johansson (1995)
EN 1993-1-5

1
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 5.9: Test results from all the 6 specimens from Möller and Johansson
(1995). The results were re-evaluated with respect to the Winter
function. Plate slenderness according to EN 1993-1-5.

5.7. Concluding remarks

Regarding the presented results collected through the literature survey the predominantly
chosen steel grades seems to be of a type with lower strength (i.e. yield strength below 460
MPa). However, with respect to the tests presented later in this thesis, these gathered test results
are of great importance to be used as a reference.

Furthermore, the re-evaluation (or use of test data) was made with respect to EN 1993-1-5
and the Winter function. This was done in order to be able to do a comparison between the
different experimental results. Even though this procedure was conducted, some differences
considering the results are still present. One obvious difference is that in some cases the yield
strength of the steel was measured in compression. Usually the compressive strength is slightly
higher compared to steel in tension. This influences not only the evaluation considering the
reduction factor, but also the plate slenderness. Emphasizing the definition of plate slenderness
according to EN 1993-1-5, described in eq. (2.27) and eq. (2.28), the yield strength of the
material in question is regarded. An increased yield strength implies a higher plate slenderness,
i.e. the plate will be considered more slender than it would be if the yield strength in tension
would be used. Furthermore a higher yield strength also implies a lower relative resistance.

93
 Regarding the presented results some conclusions may be drawn when contemplating the
conducted work shown in the sections above.

• The effect of the presence of residual stresses is evidently decreasing the local
buckling resistance.

• Most of the stockier specimens seems to have a resistance surpassing the resistance
predicted by the Winter function, i.e. EN 1993-1-5.

• More slender specimens tend to have a lower resistance than predicted by the
Winter function, i.e. EN 1993-1-5.

94
 Local Buckling - Experimental Work

Chapter 6:
Local Buckling - Experimental Work

The local buckling phenomenon has over the years been quite thoroughly investigated by
numerous of different researchers, e.g. see chapter 5. However, the field of local buckling
research concerning members made of steels with higher strength has yet not been fully
evaluated. This topic has been the focus of the experimental work presented herein and, in some
way, a step towards filling these gaps in knowledge and further enhance the possibility of using
high strength steel in constructional work of today.

6.1. Background
During the winter and spring of 2004 a local buckling test programme, comprising stub
column tests and uniaxial tests, were performed at the division of structural engineering, Luleå
university of technology, LTU. The tests were a part of the RFSC funded project “LiftHigh -
Efficient Lifting Equipment with Extra High Strength Steel” and with focus on the second
project work package: “Global and local buckling of hollow sections and welded boxes”.

With focus on this work package, 48 specimens with box cross section have been tested at
LTU, solely with respect to the local buckling phenomena. This was complemented with
uniaxial tension tests for the determination of the mechanical properties of the steel in question.
Furthermore, measurements of the residual stress state in the specimens (as-welded condition)
was conducted and presented in Clarin (2004).

The specimens were fabricated by SSAB Oxelösund and made of extra high strength steel,
as well as of a more commonly used steel grade. The specimens were designed to simulate four
individual plates under uniform compression and simply supported along their boundaries.

6.2. Experimental investigation

The aim for the test programme was to investigate if steels with yield strength > 460 MPa
behaves different than ordinary steel grades with respect to local buckling. This is something
that has not been examined to such a great extent before. The specimens for evaluating the local
buckling resistance were fabricated out of three different steel grades; 3 mm thick Domex 420

95
(hot rolled) and the two Weldox grades 700 (quenched and tempered) and 1100 (quenched),
both of plates with a nominal thickness of 4 mm.

In addition to the buckling tests, 18 coupon tension tests were conducted with the purpose of
determine the properties of the three different grades needed for further evaluation of the
buckling test data.

6.3. Uniaxial tests

The mechanical properties of the steel used for the fabrication of the local buckling
specimens (see section 6.4) were determined through tensile coupon tests. The tests were made
according to the test standardization in EN 10002-1 (2001). A total of 18 coupons were laser-
cut from the same virgin plates as used for the fabrication of the buckling test specimens.
Furthermore, because the rolling direction of the steel was altered between being along and
perpendicular to the loading direction in the buckling tests, the mechanical properties were also
determined in these directions, Figure 6.1.

6.3.1. Specimens
The thicknesses of the plates used for the fabrication of the box specimens, hence also
concerning the coupons, were nominally 3 mm for the Domex grade and 4 mm concerning the
Weldox.

Figure 6.1: Plate with laser-cut coupons along and transverse the rolling
direction.

Prior to each tension test the coupon was measured to determine the geometry of the
specimen. The length of the coupons was 379 mm and the width 39 mm for the gripping part of
the coupon (the ends) and 24,9 mm (mean value for all 18 coupons) for the notched area in the
middle of the coupon specimen. Furthermore the plate thickness was determined to 3,05 mm for
the Domex plates, 4,09 mm for the Weldox 700 and 3,98 mm for the Weldox 1100 plates.

96
 Local Buckling - Experimental Work

6.3.2. Test setup

The tension tests were made in a 600 kN servo - hydraulic DARTEC rig and the test data was
acquired through software enclosed with the rig. The load and axial elongation was measured
until failure of the coupon specimens.

Figure 6.2: The coupon equipped with an extensometer in the test rig.

6.3.3. Test results

In Figure 6.3 the typical stress-stain relations are shown for the three different grades and in
Table 6.1 the results from the 18 tested coupons are presented in numbers. The Domex 420
grade shows a classic steel stress - strain relation behaviour, with a distinct yield plateau.
Therefore the yield strength is stated for these specimens. However, the Weldox grades show a
strongly non-linear behaviour and has no well identifiable yield plateau. In this case the 0,2%
proof stress are used as the yield criterion. All of the uniaxial stress - strain curves from the 18
coupons may be found in Appendix C.

1600
S420, Coupon D1
W700, Coupon W1
V,Tensile stress [MPa]

1200 W1100, Coupon W7

800

400

0
0 10 20 30 40
H, Strain [%]
Figure 6.3: Typical stress - strain relation for Domex 420, Weldox 700 and 1100.
All specimens oriented along the rolling direction.

97
 Concerning the material behaviour of the Domex 420 and Weldox 700 it is evident that the
yield or 0,2% proof stress and ultimate resistance is higher when tested transverse to the rolling
direction. The Weldox 1100 seems to behave contradictive to the other two grades, with an
almost equal 0,2% proof stress and ultimate strength in the both directions, maybe with a
slightly higher strength along the rolling direction. This was also concluded by Gozzi (2004).

Table: 6.1: Results from the uniaxial tensile coupon tests.0o indicates rolling
direction along the loading direction and 90o transverse.

Specimen Yield Strength, Proof Stress, Ultimate Strength, A5 [%]

fy [MPa] Rp0.2 [MPa] Rm [MPa]

D1 442 - 529 30,1

D2 439 - 526 29,8
Domex 420
0o

D3 443 - 530 30,1

D4 469 - 533 30,5
90o

D5 473 - 533 28,8

D6 471 - 532 28,7
W1 - 769 821 15,2
Weldox 700

W2 - 774 828 15,6

o
0

W3 - 775 826 14,4

W4 - 791 824 14,6
o

W5 - 800 834 14,7

90

W6 - 791 826 15,0

W7 - 1345 1477 9,5
Weldox 1100

W8 - 1350 1480 8,6

o
0

W9 - 1357 1489 *
W10 - 1326 1457 *
o

W11 - 1359 1512 8,7

90

W12 - 1320 1485 8,6

* Indicates failure outside of the range of the extensometer.

6.4. Buckling tests

6.4.1. Specimens
The specimens were made of four identically designed plates, welded together along their
edges, see Figure 6.4 and Figure 6.5. The design of the specimens were conducted with the
purpose to allow the plates to act as simply supported along the longitudinal edges (edges in the
loading direction). Furthermore, the aim was to have these simply supported plates subjected to
an uniformly distributed compressive stresses. This was achieved through welding flat milled
rigid end plates to the top and bottom of the box section. These end plates were assumed to be
thick enough (thickness > 15 mm) to distribute the applied load evenly to the four plates of the
welded box specimen.

98
 Local Buckling - Experimental Work

To prevent column buckling, the height of the specimens were limited to 3 times the plate
width. This would also minimize the influence of possible clamping effects (moment restraints)
from the end plates. Furthermore the rolling direction of the steel was varied between being
along and perpendicular to the loading axis of the specimen.

A A

A-A

Figure 6.4: Specimen layout and weld positions.

All specimen fabrication work, along with the production of the Weldox plates, were made
by SSAB Oxelösund. The Domex plates were fabricated by SSAB Tunnplåt in Borlänge. The
test ready box specimens were delivered to LTU along with plates of the three grades for
fabrication of the coupons needed for the uniaxial tests.

Figure 6.5: Specimens S30-0a (left) and W73-0a (right) after test.

The 48 specimens were divided into three sets, each comprising one of the steel grades
Domex 420, Weldox 700 or Weldox 1100. The nominal plate slenderness values, Op, were

99
chosen to 0,7, 0,85, 1,0 and 1,5 and the nominal thickness was 3 mm (Domex) and 4 mm
(Weldox). The width of the plates was calculated with respect to EN 1993-1-5 on basis of the
pre-defined plate slenderness.

The different slenderness “groups” comprised four specimens for each steel grade. Two of
these had the rolling direction oriented in the axial, or loading, direction of the specimen,
denoted 0o. The other two were designed with the rolling direction perpendicular to the loading
direction, marked 90o. The different specimens setup and geometries are shown in Appendix C.

Welds
All welds were of fillet type and had a nominal throat thickness (a) equal to the plate
thickness, i.e. 3 and 4 mm respectively. Gas metal arc welding was used for the welds and two
different electrodes were used with respect to the different steel grades, see Table 6.2 below for
electrode properties.

Table: 6.2: Nominal electrode properties provided by SSAB Oxelösund.

Electrode Steel Nominal Yield Nominal Ultimate Elongation [%]

Type Grade Strength, fye [MPa] Strength, fue [MPa]
AWS A5.18-93
Domex 470 560 26
(D=1 mm)
AWS A5.28-79
Weldox 690 770 20
(D=1 mm)

However, the heat input of 0,33 kJ / mm, welding speed 340 mm / min., current 155 A and
the voltage 15,3 V were all the same for all specimens. Mison 25 (77% Ar and 23% CO2) was
used as protective gas for all the welds.

6.4.2. Test setup

All the box specimens were tested in an INSTRON I450, 4,5 MN rig, see adjacent Figure
6.7. The specimens were uniaxially loaded with a deformation speed of 0,072 mm / min. until
the ultimate load had been reached.

The deformation speed was kept until the load response had decreased with 10% of the
ultimate load. At this point the deformation speed was doubled and the test was run until the
load had decreased to approximately 70% of the ultimate load.

100
 Local Buckling - Experimental Work

Figure 6.6: A box specimen placed in the INSTRON I450, 4,5 MN test rig.

6.4.3. Measurements
During testing data was sampled over six channels. The load was measured with a load cell
from DARTEC with a measuring range up to 2 MN. The deformation in the loading direction
was measured with four 11 mm LVDT's in four points located at the corners of one of the end
plates of the specimens, e.g. see Figure 6.7. Four LVDT’s were used to be able to calculate the
mean axial deformation of the end plate which in further evaluations was used as the mean axial
plate deformation. The out of plane plate deflection, or buckle growth, was also measured. This
was done with a 25 mm LVDT at the mid point of one side of the specimen, see Figure 6.7.

During all the tests the sample rate of data was 2 Hz and a 600 Hz Spider 8 from HBM was
used for interpreting the signals from the gauges to PC environment. For information
concerning the specifications of the equipment used for acquiring data, see Appendix C.5.

Figure 6.7: The test setup with all the LVDT’s and the load cell. The specimen
was deformed from the lower side and the load measured by the load
cell on the upper side (left). To the right the out of plane deflection
LDVT is pictured as well as some of the axial deformation LVDT’s.

101
 Prior to test start the specimens position in the rig was measured to ensure that the loading
axis was in the centre of the specimens, hence the risk of introducing forces due to eccentricity
of the specimen was minimized. As an extra precaution to unwanted influences, a small hole
was drilled through one of the end plates of the specimens. This was to ensure that the air
pressure inside the closed specimen was equal to the surrounding air at all times during the
deformation of the specimen. Furthermore, possible pressure differences due to the welding
(heated air) was also avoided through this procedure.

Additional measurements concerning the geometry of the specimens were also conducted.
The plate dimensions were measured prior to the buckling tests and are enclosed in Appendix
C.2. The plate width was measured on three positions on all four plates in every specimen. In
addition to this, the plate height was measured on one position on all four sides. All dimensions
were measured between the weld edges, i.e the effective width and height of the simply
supported plates.

6.4.4. Results
The test data essential to the aim of this investigation was the ultimate load registered
concerning respective specimen. The typical load - mean deformation behaviour for the
specimens made of the three different grades is presented in Figure 6.8. All of the load - mean
axial deformation curves are shown in Appendix C.3.

1600

1200
F, Axial Load [kN]

800

400 S40-0a
W74-0b
W114-0a

0
0 2 4 6 8 10
G, Mean axial deformation [mm]
Figure 6.8: Typical load - mean axial deformation behaviour for the box
specimens made of the three different steel grades.

The cross section area for the stress comparison was calculated from the data presented in
Table C.1. The weld area was added to the plate section area. The weld areas were set to 19 mm2
for the Domex specimens, 34 mm2 for the Weldox 700 specimens and 32 mm2 for the Weldox

102
 Local Buckling - Experimental Work

1100 specimens. All weld areas were theoretically determined with respect to their individual
measured plate thicknesses. The mean 0,2% proof stress, Rp0.2, was calculated from the tension
coupon test results presented in Table 6.1.

Three specimens, one from each grade, were removed from the buckling test programme.
These specimens, S20-0b, W72-0b and W112-0b, were put aside to be used for the
measurement of longitudinal residual stresses, presented in Clarin (2004). Furthermore SSAB
Oxelösund delivered some extra specimens of the stubbier type with a nominal plate
slenderness of 0,7. These specimens were made of the two Weldox grades and the results are
presented with the other results from the ordinary specimens.

Figure 6.9: Specimen W74-0a with deformed end plate (left) and specimen W111-
0b with ruptured weld in upper left corner (right).

Unfortunately, the results from the specimen W74-0a had to be removed from the evaluation
because of some problems regarding end plate deformation, Figure 6.9 (left). The specimen
never reached its ultimate load due to the plastic deformation of one of the end plates. In
addition to this, problems concerning specimen W111-0b occurred. This specimen reached its
ultimate load, but shortly thereafter one of the corner welds failed (see Figure 6.9 (right)) and
the load dropped very fast. However, the load - deformation curve shows a somewhat different
behaviour and are shown Appendix C, but since the ultimate load were reached without
problems, the results from this specimen was evaluated and presented among the other results.

6.5. Test evaluation

The test results were evaluated with respect to the EN 1993-1-5, i.e. eq. (2.27) and eq. (2.29) in
chapter 2. The calculations are based on the mean values for each specimen, i.e. the mean width
for all four plates and the mean values concerning the mechanical properties. This is also the
case concerning the plate thickness, which is determined through measurement of the coupons
used in the material tests.

In Figure 6.10 and Figure 6.11 the results are plotted in comparison to the Winter function, i.e.
eq. (2.24).

103
 1.5

Domex 420
Weldox 700
Weldox 1100
1 EN 1993-1-5
Fexp / Fy

0.5

0
0 0.5 1 1.5 2 2.5 3
Slenderness, OP
Figure 6.10: The evaluated results from the 48 specimens along with the Winter
function. Plate slenderness calculated according to EN 1993-1-5.

2
Domex 420
Weldox 700
1.5 Weldox 1100
EN 1993-1-5
Fexp / FR

0.5

0
0 0.4 0.8 1.2 1.6 2
Slenderness, OP
Figure 6.11: The experimental results, Fexp,with respect to the predicted
resistance, FR, according to EN 1993-1-5, i.e. the Winter function.
The plate slenderness calculated according to EN 1993-1-5.

Regarding the evaluated test data some things are important to be pointed out. Firstly, the
scatter between the results for each group of tests is small for the more slender specimens. Some
differences can be noticed for the slender specimens, especially concerning the slenderness
value of the plates. The origin of these differences is mostly dependent of the different strength

104
 Local Buckling - Experimental Work

of the steel, due to the differences in the rolling direction. Though, these differences seem to be
less pronounced with increasing steel strength. Furthermore, the scatter between each test tend
to be larger for the specimens with Op< 0,9.

6.6. Concluding remarks

Considering Figure 6.10 the evaluated test results seems to be consistent within each test
group, i.e. plate slenderness value concerning each grade. This is even more obvious regarding
the most slender specimens where the four tests of each grade are nearly four repetitions of each
test (the same ultimate load reached for each specimen). Moreover, another conclusion may be
drawn based on this fact and this is simply that the test procedure seems to have been consistent
with small differences between each specimen test.

When comparing the test results with the Winter function (i.e. EN 1993-1-5) the more stocky
plates, Op < 0,9, seems to be spread around with the reduction function, see Figure 6.11. The
resistance may even be somewhat higher than predicted through the Winter function.
Considering these more stocky specimens, the ones of the “lower” strength steel seems to have
a higher resistance than the high strength steel specimens which is positioned closer or on the
Winter function in Figure 6.11. However, this may have its origin in the difference in
mechanical behaviour and how the material properties are regarded as discussed above.

Regarding the other range of specimens, Op > 0,9, the opposite has to be concluded. The
Winter function seems to overestimate the resistance concerning more slender plates. This
completely independent of steel grade. However, if Figure 6.11 is considered along with the
Figure 6.10, the specimens of high strength steel seems to coincide better with the Winter
function.

Considering all of the evaluated and presented test results the following may be concluded:

• The Winter function tends to describe the mean value of the resistance of stockier
specimens. In this case plates with Op < 0,9.

• The Winter function overestimates the resistance of more slender plates. In this
case plates with Op > 0,9.

• Plates made of high strength steel may be treated in the same way as “ordinary”
grades with respect to the local buckling resistance.

• With respect to the Winter function, no difference between the specimens with
different rolling direction could be concluded.

Considering the evaluation of the test results, one obvious difference, regarding the
mechanical properties of the steel, in the evaluation procedure has to be mentioned. The slight
difference between the specimens of the Domex and Weldox specimens concerns the used

105
material properties, i.e. yield strength for Domex and 0,2% proof stress concerning Weldox
specimens. This different approach is dependent of the lack of well defined yield plateau
considering the Weldox grade, still the hardening properties of these grades influences the
evaluation. This in the way that the difference between the ultimate strength and the stress
defined as fy is larger for the steels with lower strength, i.e. a well defined yield limit. This leads
to that the calculated critical stress level considering the Weldox specimens will be in an
unfavourable position since the stress level defined as yield stress is closer to ultimate strength
of the steel. In the evaluation of the experimental work, this leads to a lower reduction factor,
hence a lower position if plotted with the Winter function as a reference.

When considering the actual experimental work some things are important to state. First, the
measurement of the buckling growth has not been implemented nor evaluated in this thesis. This
data was herein excluded due to the fact that this test data was considered to give no further
valuable information or possibilities to conclude with respect to the aim of this thesis.
Furthermore, the measured initial geometric plate imperfections were neither implemented
herein.

106
 Local Buckling - Design Proposal

Chapter 7:
Local Buckling - Design Proposal

7.1. Background
The results from the experimental investigation regarding local buckling of the box-
sectioned specimens presented in previous chapter 6, the Winter function seems to be a
somewhat inappropriate choice when calculating the ultimate resistance to uniformly
distributed compressive stresses for plates heavy welded or with large welds compared to the
dimensions of the considered plate. Emphasizing that the Winter function was based on tests of
cold-formed specimens, the actual residual stress magnitude and distribution of a plate with
large welds may not be a plate represented by the Winter function. When cold-forming profiles,
one do not only produce the wanted profile without welding, but also changes the material
properties as well as inducing residual stresses. Introducing these changes the material
behaviour, may lead to an incompatibility issue when compared directly with welded plates.
Even though the boundary conditions of the respective plates are the same, i.e. simply supported
around all edges in this case, the differences on a deeper level may leave the researcher
astonished when comparing their test results with the Winter function. The cause of this may be
the different residual stress state between welded plates and cold-formed. Furthermore, cold-
forming induces plastic strains into the material. Experimental work has shown that cold-
formed profiles have significantly higher proof stresses and ultimate strength levels in the area
of forming, i.e. corners of a box section, Gardner (2002) and Talja (2002). These increases in
the material resistance of course affects the over-all behaviour of such a specimen.

Furthermore, if considering the residual stress state in a cold-formed profile compared to a

welded section of the same dimensions, the magnitudes of compressive stresses in the flat sides
of the former seems to be lower than the corresponding ones in the welded profile, Ingvarsson
(1977). Lower levels of compressive residual stresses increases the resistance. Moreover, the
definition of the plate width regarding a cold-formed profile also influences the outcome, e.g. if
the plate width is the flat side alone (inner or outer side) or as the distance between the centres
of the corners. Thus, the Winter function may not on a phenomenological basis be comparable
with welded plates. However, the influence of the residual stresses in the plates should have less
influence on the resistance when the plate slenderness is lower (more plastic buckling).

107
 Veljkovic and Johansson (2001) comprises FE studies of plates with and without residual
stresses and concluded that the Winter function is more suitable to use for plates without
significant residual stresses or stress relieved. This is not the case concerning plates in as-
welded condition. Similar conclusions were also drawn by Rusch and Lindner (2001).

When considering Figure 7.1, comprising the collected data from the literature (see chapter
5) and the experimental results presented in chapter 6, the outcome seem to coincide, or at least
describe the mean value of the results discussed above. Regarding plates of different steel
grades, it seems like the Winter function may be a more suitable function to use when the plate
slenderness is lower. Regarding more slender plates, Op> 0,9, the Winter function overestimate
the resistance concerning all the plates in the herein used data base.

Domex 420
1.5
Weldox 700
Weldox 1100
Tests from literature
Mean function, (eq. 5.1)
1 EN 1993-1-5
Fexp / Fy

0.5

0
0 0.4 0.8 1.2 1.6 2
Slenderness, OP
Figure 7.1: The evaluated test data from 48 box specimens along with data
acquired from relevant literature resulting in a total of 85 specimens
tested with respect to local buckling.

Nevertheless, the results presented in this thesis shows that steel with higher strength may be
treated in the same way as “ordinary” steel grades. The high strength steel may even coincide a
bit better with the Winter function than steels with lower strength (fy < 460 MPa), see chapter 6.

7.2. Proposal and validation of new reduction function

Studying the work conducted regarding the ultimate patch loading resistance, or more
specifically the reduction function proposed regarding the patch loading (see chapter 4 and
Figure 4.1) this reduction function seems suitable to use also concerning local buckling,
however with the plateau set to 1,0. Hence, the proposed reduction function will be

108
 Local Buckling - Design Proposal

1
F P = ------------------------------------- d 1 ,0 (7.1)
2
MP + MP – OP

and

1
M P = --- ˜ 1 + 0 ,5 ˜ O P – 0 ,6 + O P (7.2)
2

Compared to the Winter function (Figure 7.1) and the experiments from literature and
presented herein the proposed reduction function according to eq. (7.1) and eq. (7.2) seems to
better predict the ultimate resistance of plates with larger welds, Figure 7.2. However, the more
stocky specimens of the 13 tests by Dwight and Moxham (1969) seems to present ultimate loads
surprisingly low regarding the specimens used, see Figure 5.4. The reason for this is difficult to
herein explain or debate.

Domex 420
1.5
Weldox 700
Weldox 1100
Tests from literature
Proposal
1
Fexp / Fy

0.5

0
0 0.4 0.8 1.2 1.6 2
Slenderness, OP
Figure 7.2: The experimental ultimate load Fexp in relation to the yield load, Fy,
as a function of the plate slenderness Op put in comparison to the
proposed reduction function, eq. (7.1).

However, comparing the proposed reduction function in Figure 7.2 with the Winter function
used in EN 1993-1-5 shown in Figure 7.1 it is clear that predicting the ultimate plate buckling
load according to the proposal is more accurate. This only regarding plates with larger welds
and not stress relieved in any way, i.e. as-welded condition. Further, the proposed reduction
function is evidently proper to use regardless of steel strength or at least within the interval of
the data from literature and experiments, i.e. from a yield stress, fy, of approximately 258 MPa
to Rp0.2 of around to 1350 MPa.

109
 Regarding the design resistance, FR, this is determined in the same manners as in EN 1993-
1-5, i.e. according to eq. (7.3).

F Rd = F P ˜ F y e J M1 (7.3)

As the last step in the validation work of the proposed model, a partial safety factor was
calculated in accordance to the recommendations of Annex D in EN 1990 (2002). The method
of the statistical evaluation is shown in the patch loading oriented Appendix B, and the results
of the local buckling results of how the partial safety factor, JM1 = 1,07 is shown in Appendix
C. Even though, this safety factor is higher than the 1,0 recommended in EN 1993-1-5, the
herein presented material shows that the proposed reduction function is much more appropriate
to use regarding plates with large welds.

7.3. Concluding remarks

Within this chapter a reduction function suitable to use for plates with large welds was
presented. The Winter function used in EN 1993-1-5 was based on tests of cold-formed
specimens which usually have lower compressive residual stresses inherent, than compared to
plates with large welds.

Furthermore, it was shown that steel with a higher strength have the same resistance with
respect to local buckling when compared to more commonly used structural steel. Moreover,
the high strength steels may even be better to use regarding local buckling related issues as
shown in Figure 7.3 when compared to steels with a lower strength.

The experiments used for this evaluation was partly found in the literature (37 tests) and
partly (48 tests) made within the scope of this thesis. Different authors have presented a large
quantity of tests results, from the 1960’ies until today. As discussed previously herein, using test
data from many different laboratories may induce some uncertainties with respect to what is
presented in the actual published report, how quantities are measured, deformation rates etc.
The evaluation of the local buckling resistance was made on measured values, e.g. the yield
strength measured by the authors. Though, was the yield strength measured in the same manners
in 1968 than it was in 2004 with respect to the recommendations in EN 10002-1 (2001)? Was
the welding techniques comparable with the fashions of which plates are welded today, e.g.
filler material, energy input etc.? Since the yield strength is a dominant factor when predicting
the ultimate plate buckling resistance of a plate, this also makes the evaluation of the tests results
inherit the same uncertainties. With this in mind an additional statistical evaluation was
conducted, comprising only contemporary results which in this case meaning results from the
1990’ies and forward. With this manoeuvre, the data base comprised 60 individual specimens
for which the partial safety factor was determined with respect to the proposed reduction
function. The key values of this evaluation are displayed in Appendix C, though the corrected
partial safety factor was calculated to 1,03.

110
 Local Buckling - Design Proposal

2.5
Tests made at LTU
Tests from the literature
2
Proposal

1.5
Fexp / FR

0.5

0
0 200 400 600 800 1000 1200 1400
fy [MPa]
Figure 7.3: The experimental results from the tests made at LTU and gathered
from the literature, Fexp, in relation to the predicted resistance of the
proposal, FR, as a function of the yield strength of the steel in the
specimens. A total of 85 specimens are shown.

Regarding the definition of how large the weld have to be in relation to the plate dimensions
to be stated as “large” have not been studied herein. Though, the swedish design code
Handboken Bygg (1994) recommends two different reduction functions for welded plates. For
plates with welds > 0,5t a larger reduction is made than for plates with welds smaller than that.
However, the other dimensions of the plate are probably not unimportant, i.e. if the weld is
positioned along the edge of a wide plate the negative influence of the buckling resistance in the
middle of the plate would be small. Nevertheless, the herein proposed reduction function was
within this chapter shown to be more suitable to use regarding plates with welds when compared
to the Winter function.

111
112
 Discussion and Conclusions

Chapter 8:
Discussion and Conclusions

8.1. Patch loading - Discussion

Previously within this thesis, a modified reduction function was proposed for use to predict
the ultimate patch loading resistance of longitudinally stiffened webs. Moreover, a proposal
regarding the prediction of the elastic critical load of the stiffened web was introduced. The
design approach was based on the renowned method for unstiffened webs of Lagerqvist (1994)
which later was slightly simplified and introduced as the design rule in EN 1993-1-5. Though,
parts of the EN 1993-1-5 rules for prediction of the patch loading resistance have been
questioned and with Gozzi (2007) a modified design approach regarding unstiffened webs was
presented. The proposal of Gozzi eliminated the questioned part and was shown to predict the
patch loading resistance for unstiffened webs well.

The mechanical model of Lagerqvist was also used and evaluated in Graciano (2002) to be
compatible concerning stiffened webs. The work presented in Graciano was also introduced in
the EN 1993-1-5 and design rules for webs stiffened with one longitudinal stiffener. However,
the EN 1993-1-5 only regards the elastic critical buckling load for the stiffened web as a whole
which is contradictive to the observations of some experimental work presented. Furthermore,
numerical work presented in Davaine and Aribert (2005) showed that the rules of EN 1993-1-5
predicted the patch loading resistance heavily conservative regarding simulations with a higher
b1 / hw ratio. Thus the authors introduced a critical buckling load estimation concerning the
upper (directly loaded) panel alone. The buckling load for the web was then proposed to be
estimated using an interaction formulation of the EN 1993-1-5 critical load for the whole web
and the critical load for the upper panel.

The herein proposed resistance model is consistent with the proposal for unstiffened girder
webs of Gozzi (2007) with respect to reduction function and mechanical model describing the
failure behaviour of a girder subjected to patch loading. Furthermore, the elastic critical load
was herein proposed to be estimated as the lowest buckling load comparing the EN 1993-1-5
(the whole panel) and the upper panel. This was a subjective choice by the author of this thesis
since the idea of two buckling modes interacting seems somewhat contradictive or inappropriate
to use with the von Kármán concept of determining the slenderness of a plate. The concept

113
proposed herein which instead uses the minimum value of two possible buckling modes is
probably more uncomplicated and obvious to a user (e.g. structural designer), moreover
consistent with respect to design rules concerning other phenomena.

Another not too obvious ingredient in the EN 1993-1-5 is, according to the author, the use of
the H- parameter, i.e. 235 e f y > MPa @ . This parameter was, to the best of the authors
knowledge, introduced as a modifier to uncertainties regarding the behaviour of steel with
fy > 235 MPa. However, the herein presented experimental work regarding local buckling
shows that the parameter may not be adequate. Moreover, regarding patch loading resistance
the use of the parameter to determine the contributory part of the web to the longitudinal
stiffener is debatable. The rule implies that if two geometrically identical girders, though one
being made of a steel with higher yield strength of the web compared to the other, will have
different second moment of area for the stiffener, Ist, when designed according to the EN 1993-
1-5.

Considering the experimental data gathered from the literature, some of the specimens were
tested two times. First patch loading on one flange until failure, then turning the girder and
performing a second test applying patch loading on the un-damaged flange. Inspecting the
results from these tests, it is evident that the second test is influenced by the first since in most
cases the ultimate patch loading resistance is lower the second time. Nevertheless, considering
these sequel experiments on the same basis as other tests only loaded one time, the choice of
doing this should be safe since the resistance should be higher than the experimental results
shows.

A possible interaction between patch loading and bending moment of longitudinally

stiffened girders could not be proved herein. However, patch loading experiments with
longitudinally stiffened girders with a high utilization of the bending moment resistance are not
common in the literature. Hence any conclusions regarding the matter will not be stated within
this thesis. Though, Graciano and Casanova (2004) presented numerical simulations used to
study this matter, and according to the authors the patch loading resistance of a longitudinally
stiffened girder could be reduced with more than 60% if the bending moment resistance was
utilized in larger extent.

8.2. Local buckling - Discussion

As presented previously within this thesis, in Europe the common procedure to predict the
ultimate local buckling resistance of simply supported plates subjected to compression stresses
is based on the effective width concept. Calculating the effective width according to the
Eurocode EN 1993-1-5 will include the use of the Winter function, established by testing cold-
formed specimens, Winter (1947).

114
 Discussion and Conclusions

Previous conducted and presented research work have shown that by cold-forming a
structural profile, not only the residual stress distributions and magnitudes differs when
compared to a welded specimen, but the material properties may also be significantly different.
Gardner (2002) and Talja (2002) both shows that parts of the cold-formed profiles have a higher
proof strength if compared to welded specimens. Regarding the residual stresses present in a
cold-formed profile compared to a similar welded profile, the compressive stresses in the webs
of the cold-formed profile are commonly lower than in the welded ditto, Ingvarsson (1977).
Since the compressive residual stresses decreases the ultimate load resistance, a plate with
welds along its edges will in all likelihood have a lower local buckling resistance than compared
to the same plate in a cold-formed profile.

As shown in chapter 5 and chapter 6 the Winter function is evidently not appropriate to use
considering plates with larger welds. In this case large welds in the meaning that the welding
procedure governs compressive residual stresses of such an extent that these surpass the actual
initial stress state in the cold-formed profiles used deriving the Winter function. The
applicability of the Winter function has been questioned, not only herein, but supporting FE
studies may be found in Veljkovic and Johansson (2001). This article presents numerical
simulations of plates with and without residual stresses and concludes that the Winter function
is more suitable to use regarding plates without any significant residual stresses, e.g. cold-
formed or stress relieved plates. Furthermore, Rasmussen and Hancock (1992) also concludes
that the Winter function overestimates the resistance of welded plates and the mean curve for
as-welded plates of Fukumoto and Itoh (1984) further proves the statement. The swedish design
code BSK 99 (2003) uses a distinctive limit of what should be treated as a plate with “large”
welds or not. The throat thickness of the weld, a, is compared to the plate thickness and if
a > 0,5t the effective width of the plate is estimated to be smaller in comparison to the contrary.
Similar approaches may be found in other design regulations, e.g. the Australian Standard AS
4100 (1990) which uses two curves (LW - Light welded and HW - Heavily welded). Though,
since also the other plate dimensions influences the magnitude and distribution of weld induced
residual stresses, a curve decision solely based on the weld thickness in comparison to the plate
thickness may be a bit subjective.

When it comes to the data base of the square hollow sections, both from the literature and
conducted tests, it was shown that disregarding tests made before the 1990’ies (25 specimens)
improved the performance of the proposed reduction function in comparison to the test results.
Gathering tests results from the literature often brings uncertainties regarding the specimen data
and results of the experiments, e.g. “how was it measured, what techniques were used?”.
Furthermore, the results from the tests of some of the more stocky tests were surprisingly low
which may be a result of differences between how the experiments were performed at the
different laboratories and maybe also differences in followed experimental guides, norms and /
or regulations used. An example may be how the yield strength is measured. It is well known
that the out-come of tension tests, usually yield and ultimate strength, of steel coupons are

115
strongly dependent of the deformation/strain rate. Hence, questions rise whether determining
the mechanical properties was made in the same manners regarding tests made in the 60’ies as
tests during the 90’ies? Further, since the prediction of the ultimate buckling load is strongly
dependent of the yield strength of the considered plate, an answer to the low resistance may
possibly be found here.

Regarding the experiments conducted on specimens made of high strength steel, it was
shown that square hollow sections made of steels with higher strength may have a higher
resistance to local buckling when compared to specimens fabricated of more commonly used
structural steels. However, the fact that the magnitude and distribution of the compressive
residual stresses strongly influences the ultimate buckling resistance is well known. Though,
this higher resistance of the high strength steel specimens may be a result of lower residual
stresses in relation to the strength of the steel. In Clarin (2004) the residual stresses was
measured and compared regarding the specimens used for the local buckling tests presented
herein. When the three different steel grades were compared, it seemed like the magnitude of
the residual stresses was the same for all the specimens. However in relation to the yield strength
or proof stress the ratio was lower for the high strength steels. This could be an explanation of
why the high strength steel specimens have a better resistance to local buckling. Though, it is
important to remember that the yield strength of the filler material in the weld is influencing the
magnitude of the residual stresses. According to the knowledge of the author, there are no
matching filler material regarding high strength steels as of today. Hence, the high strength steel
specimens were welded with an undermatching weld which possibly led to unconventionally
low residual stresses when compared to the common steel grade specimens.

8.3. Conclusions
Within this thesis, the following conclusions are drawn from the presented work regarding
patch loading resistance of longitudinally stiffened webs and local buckling resistance of simply
supported plates.

• The experimental work comprising 48 stub columns with a square hollow section
made of three different steel grades was concluded to have been conducted with
success. The scatter within the different test groups were small and the tests results
overall comparable with results found in the literature.

• Based on experimental results, plates made of high strength steel was concluded to
perform slightly better than common structural steel plates, compared with sole
respect to the local buckling resistance. Thus, it was concluded that no further
special attention has to be paid when using high strength steel in buckling related
design.

116
 Discussion and Conclusions

• The proposed reduction function regarding the effective plate width was, by
comparison to 85 individual tests, concluded to be more suitable to use for heavily
welded plates than the Winter function used in EN 1993-1-5.

• When compared to EN 1993-1-5 the proposed patch loading resistance model was
shown to reduce the scatter and improve the prediction accuracy. Furthermore, the
proposal was concluded to safely predict the resistance of webs stiffened with a
stiffener with open as well as a closed cross-section.

• The proposed patch loading resistance approach, consistent with the model for
unstiffened girders of Gozzi (2007), was statistically evaluated with respect to
Annex D of EN 1990 (2002). In comparison to 160 experiments and 366 numerical
simulations the partial safety factor JM1 = 1,0 is proposed.

8.4. Proposals for future work

Usually when a topic is investigated with effort more questions appears along the way.
Working on this thesis was not an exception to this and within this section some proposals of
further studies are posted.

The performance of proposed reduction function simply supported welded plates still needs
to be investigated with respect to its other applications, e.g. effective cross-section calculations
of girders subjected to bending moments. Furthermore, the ultimate resistance with respect to
local buckling is strongly influenced of the present residual stress state in the plate. Thus, it
would be of interest to investigate if plates of high strength steel, heavily welded with matching
filler material, would be in line with the conclusions herein; that plates made of high strength
steel seems to resist buckling better than common structural steel grades.

Regarding the ultimate patch loading resistance for webs stiffened with a closed stiffener, the
data found in the literature was limited. Furthermore, most of the evaluated tests seemed to be
predicted conservatively with respect to the proposed model. Questions rose wether the
estimation of the elastic critical load for such a girder could be predicted too low. A higher
critical load would lower the slenderness value of the web, hence also a lower reduction due to
buckling. An improved estimation of the critical buckling load regarding such stiffened webs
would be interesting to perform, by means of numerical simulations and / or experiments.

Further, the herein proposed design procedure was only validated with respect to one
longitudinal stiffener. To widen the applicability further, this could also be verified through
comparison to webs with additional stiffeners. The most straightforward approach of this would
according to the author be by further investigate the critical load for such girders.

As no conclusions could be drawn regarding the possible patch loading / bending moment
interaction, this would be possible to study further. The limited amount of experimental work

117
studying this creates a void to fill. Furthermore, numerical simulations regarding the issue have
been conducted, e.g. Graciano and Casanova (2004), nevertheless not, to the best knowledge of
the author, presented in such a way that these could be further evaluated by other researchers.
Hence, more experiments regarding the possible interaction would be beneficial for the steel
researchers around the world.

This thesis was focused only on patch loading and not opposite or end patch loading. Hence
the herein proposed resistance approach still needs to be validated with respect to these two load
cases.

118
 References

Chapter 9:
References

AS4100-1990 - Steel Structures. (1990). Standards Association of Australia, Sydney,

Australia. (ISBN: 0-7262-6493-8).

Bergfelt, A. (1971). “Studies and tests on slender plate girders without intermediate
stiffeners”, Proceedings of IABSE Colloquium, London, 1971. pp. 67-83.

Bergfelt, A. (1979). “Patch Loading on a Slender Web - Influence of Horizontal and Vertical
Web Stiffeners on the Load Carrying Capacity”, Publication S 79:1, Department of
Structural Engineering, Division of Steel and Timber Structures, Chalmers University
of Technology, Göteborg.

Bergfelt, A. (1983). “Girder Web Stiffening for Patch Loading”, Publication S 83:1,
Department of Structural Engineering, Division of Steel and Timber Structures,
Chalmers University of Technology, Göteborg. (ISSN: 0534-0411).

Bossert, T.W. and Ostapenko, A. (1967). “Buckling and ultimate loads for plate girder web
plates under edge loading”, Fritz Engineering Laboratory Report No. 319.1,
Department of Civil Engineering, Lehigh University, Bethlehem-PA, USA.

Brush, D.O. and Almroth, B.O. (1975). “Buckling of Bars, Plates and Shells”, McGraw-Hill
Inc., USA. (ISBN: 0-07-008593-5).

BSK 99 - Swedish Regulations for Steel Structures. (2003). Boverket, Sweden.

(ISBN: 91-7147-732-2).

Carretero, A. and Lebet, J.-P. (1998). “Introduction de Forces Concentrées dans les Poutres
Élancées”, Construction Métallique, No. 1, 1998. pp. 5-18. (In French).

Cevik, A. (2007). “A new formulation for longitudinally stiffened webs subjected to patch
loading”, Journal of Constructional Steel Research, Vol. 63. pp. 1328-1340.

Clarin, M. (2004). “High Strength Steel - Local Buckling and Residual Stresses”, Lic.Thesis
2004:54, Division of Structural Engineering - Steel Structures, Luleå University of
Technology, Luleå. (ISRN: LTU-LIC--04/54--SE).

119
 Davaine, L., Raoul, J. and Aribert, J.-M. (2004). “Patch load resistance of longitudinally
stiffened bridge girders”, Proceedings of the International symposium on steel bridges
“Steelbridge 2004”, Millau, France, June 23-25, 2004.

Davaine, L. and Aribert, J.-M. (2005). “Launching of steel girder bridge - Patch load
resistance of longitudinally stiffened webs”, Proceedings of 4thEuropean Conference
on Steel and Composite Structures, Maastricht, The Netherlands, June 8-10, 2005.

Davaine, L. (2005). “Formulations de la réststance au lancement d’une âme métallique de

pont raide longitudinalement”, Doctoral Thesis D05-05, INSA de Rennes, France. (In
French).

Davids, A.J. and Hancock, G.J. (1986). “Compression Tests of Short Welded I-Sections”,
Journal of Structural Engineering, Volume 112, No. 5, May 1986.

Dogaki, M., Murata, M., Nishijima, Y., Okumura, T. and Yonezawa, H. (1990). “Ultimate
strength of plat girders with longitudinal stiffeners under patch loading”, Technology
Reports of Kansai University, Vol. 33, 1990. pp. 121-132.

Dubas, P. and Gehri, E. (1986). “Behaviour and Design of Steel Plated Structures”, ECCS -
Technical committee 8 - Structural stability, Technical working group 8.3 - Plated
Structures, ECCS printing No. 44, 1st edition. Swiss Federal Institute of Technology,
Zürich, Switzerland.

Dubas, P. and Tschamper, H. (1990). “Stabilité des Âmes Soumises à une Charge
Concentrée et à une Flexion Globale”, Construction Métallique, No. 2, 1990. pp. 25-
39. (In French).

Dwight, J.B., Chin, T.K. and Ractliffe, A.T. (1968). “Local Buckling of Thin-walled
Columns, Effect of Locked-in Welding Stresses”, CIRIA, Res. Rep. No. 12, May 1968.

Dwight, J.B. and Moxham, K.E. (1969). “Welded Steel Plates in Compression”, The
Structural Engineer, No. 2, February 1969.

EN 10002-1:2001 (2001) - “Metallic materials - Part 1: Method of test at ambient

temperature”, CEN, European Committee for Standardization, Brussels, Belgium.

EN 1990. (2002). “Eurocode - Basis of structural design”, CEN, European Commitee for
Standardization, Brussels, Belgium.

EN 1993-1-5. (2006). “Eurocode 3 - Design of steel structures - Part 1.5: Plated structural
elements”, CEN, European Committee for Standardization, Brussels, Belgium.

Farshad, M. (1994). “Stability of Structures”, Developments in civil engineering Vol. 43,

Elsevier Science B.V, Amsterdam, The Netherlands. (ISBN: 0-444-81698-4).

120
 References

Fukumoto, Y. and Itoh, Y. (1984). “Basic Compressive Strength of Steel Plates from Test
Data”, Proceedings Japanese Soc. Civ. Eng., No. 344, April 1984. pp. 129-139.

Galea, Y., Godart, B., Radouant, I. and Raoul, J. (1987). “Tests of buckling of panels
subjected to in-plane patch loading”, ECCS Colloquium on Stability of Plates and Shell
Structures, Ghent University, Belgium. 6-8 April. pp. 65-71.

Gardner, L. (2002). “A New Approach to Structural Stainless Steel Design”, Ph.D. Thesis,
Department of Civil and Environmental Engineering, Imperial Collede of Science,
Technology and Medicine, London, U.K.

Gozzi, J. (2004). “Plastic Behaviour of Steel - Experimental Investigation and Modelling”,

Lic.Thesis 2004:51, Division of Structural Engineering - Steel Structures, Luleå
University of Technology, Luleå. (ISRN: LTU-LIC--04/51--SE).

Gozzi, J. (2007). “Patch loading resistance of plated girders - Ultimate and serviceability
limit state”, Doctoral Thesis 2007:30, Division of Structural Engineering - Steel
Structures, Luleå University of Technology, Luleå. (ISRN: LTU-DT--07/30--SE).

Graciano, C. (2002). ”Patch Loading - Resistance of Longitudinally Stiffened Steel Girder

Webs”, Doctoral Thesis 2002:18, Division of Steel Structures, Luleå University of
Technology, Luleå. (ISSN: 1402-1544).

Graciano, C.A. and Edlund, B. (2003). “Failure mechanism of slender girder webs with a
longitudinal stiffener under patch loading”, Journal of Constructional Steel Research,
Vol. 59. pp. 27-45.

Graciano, C. and Casanova, E. (2005). “Ultimate strength pf longitudinally stiffened I-girder

webs subjected to combined patch loading and bending”, Journal of Constructional
Steel Research, Vol. 61. pp. 93-111.

Graves Smith, T.R. and Gierlinski, J.T. (1982). “Buckling of stiffened webs by local edge
loads”, ASCE - Journal of the Structural Division, No. 108. pp. 1357-1366.

Grotmann, D. (1993). “Proposal for ‘European plate buckling curves’ including high
strength steel property parameters”, Lehrstuhl für Stahlbau, RWTH Aachen.

Ingvarsson, L. (1977). “Cold-Forming Residual Stresses and Box Columns Built Up by two
Cold-Formed Channel Sections Welded Together”, Bulletin No. 121 of The
Department of Building Statics and Structural Engineering, The Royal Institute of
Technology, Stockholm, Sweden, July 1977.

Janus, K., Kárníková, I. and Skaloud, M. (1988). “Experimental investigation into the
ultimate load behaviour of longitudinally stiffened steel webs under partial edge
loading”, ACTA Technica CSAV, No. 2. pp. 158-195.

121
 Johansson, B. and Lagerqvist, O. (1995). ”Resistance of plate edges to concentrated forces”,
Journal of Constructional Steel Research, Vol. 32. pp. 69-105.

Johansson, B., Maquoi, R. and Sedlacek, G. (2001). “New design rules for plated structures
in Eurocode 3”, Journal of Constructional Steel Research, Vol. 57. pp. 279-311.

von Kármán, T., Sechler, E.E. and Donnell, L.H. (1932). “The Strength of Thin Plates in
Compression”, Transactions ASME 1932, Applied Mechanics, APM-54-5.

Kutzelnigg, E. (1982). “Beulwerte nach der linearen Theorie für längsversteifte Platten unter
Längsrandbelastung”, Stahlbau 51, Heft 3, 1982. (In German).

Kutmanová, I. and Skaloud, M. (1992). “Ultimate Limit State of Slender Steel Webs Subject
to (i) Constant and (ii) Repeated Partial Edge Loading”, Journal of Constructional
Steel Research, Vol. 21. pp. 147-162.

Kuhlmann, U. and Seitz, M. (2002). “Behaviour of longitudinally stiffened girder webs

subjected to patch loading”, Proceedings of the 3rd European conference on steel
structures, Coimbra, Portugal. 19-20 September. pp. 581-590.

Kuhlmann, U. and Seitz, M. (2004). “Longitudinally Stiffened Girder Webs Subjected to

Patch Loading”, Proceedings of the International symposium on steel bridges
“Steelbridge 2004”, Millau, France, 23-25 June.

Lagerqvist, O. (1994). “Patch Loading - Resistance of steel girders subjected to

concentrated forces”, Doctoral Thesis 1994:159 D, Division of Steel Structures, Luleå
University of Technology, Luleå. (ISSN: 0348-8373).

Maquoi, R. and Rondal, J. (1986). “From Thick to Thin or from Thin to Thick?”,
Proceedings of IABSE Colloquium, Stockholm, 1986. pp. 59-65.

Markovic, N. and Hajdin, N. (1992). “A Contribution to the Analysis of the Behaviour of

Plate Girders Subjected to Patch Loading”, Journal of Constructional Steel Research,
Vol. 21. pp. 163-173.

Müller, C. (2003). “Zum Nachweis ebener Tragwerke aus Stahl gegen seitliches
Ausweichen”, Heft 47, Lehrstuhl für Stahlbau, RWTH Aachen. (ISBN: 3-8322-1574-
3). (In German).

Möller, M. and Johansson, B. (1995). “Buckling Tests on Rectangular Plates made of two
Different Types of Weldox 1100 Steel”, Division of Steel Structures, Luleå University
of Technology, Internal Printing 1995:05.

Nishino, F., Ueda, Y. and Tall, L. (1967). “Experimental Investigation of the Buckling of
Plates with Residual Stresses”, ASTM Special Technical Publication, No. 419.
pp. 12-30.

122
 References

Nylander, H. (1951). “Initially deflected thin plate with initial deflections affine to additional
deflection”, IABSE Publications 1951.

Rasmussen, K.J.R. and Hancock, G.J. (1992). “Plate Slenderness Limits for High Strength
Steel Sections”, Journal of Constructional Steel Research, Vol. 23. pp. 73-96.

Roberts, T.M. and Rockey, K.C. (1978). “Méthode pour prédire la charge de ruine d’une
poutre a âme mince soumise a une charge semi-répartie dans le plan de l’âme”,
Construction Métallique, No. 3. pp. 3-13. (In French).

Roberts, T.M. and Rockey, K.C. (1979). “A mechanism solution for predicting the collapse
loads of slender plate girders when subjected to in-plane patch loading”, Proceedings
of the Institution of Civil Engineers, Part 2, 67. pp. 155-175.

Roberts, T.M. (1981).”Slender plate girders subjected to edge loading”, Proceedings of the
Institution of Civil Engineers, Part 2, 71. pp. 805-819.

Roberts, T.M. and Chong, C.K. (1981). “Collapse of plate girders under edge loading”,
ASCE - Journal of the Structural Division, No. 107. pp. 1503-1509.

Rockey, K.C., Bergfelt, A. and Larsson, L. (1978). “Behaviour of longitudinally reinforced

plate girders when subjected to inplane patch loading”, Publication S 78:19,
Department of Structural Engineering, Division of Steel and Timber Structures,
Chalmers University of Technology, Göteborg.

Rockey, K.C., Samuelsson, A. and Wennerström, H. (1979). “The Buckling of

Longitudinally Reinforces Web Plates Loaded by a Central In-plane Patch Load”,
Stability problems in engineering structures and componets, Richards, T.H. and
Stanley, P. editors, Applied Science Publishers LTD., London, 1979. pp. 75-88.

Rusch, A. and Lindner, J. (2001). “Tragfähigkeit von beulgefährdeten

Querschnittselementen unter Berücksichtigung von Imperfektionen”, Stahlbau 70,
Heft 10, 2001. (In German).

Seitz, M. (2005). “Tragverhalten längsversteifter Blechträger unter quergerichteter

Krafteinleitung”, Doctoral Thesis Nr. 2005-2, Institute of Structural Design,
Universität Stuttgart, Germany. (ISSN: 1439-3751). (In German).

Shimizu, S., Yoshida, S. and Okuhara, H. (1987). “An experimental study on patch-loaded
web plates”, ECCS Colloquium on Stability of Plates and Shell Structures, Ghent
University, Belgium. 6-8 April. pp. 85-94.

Shimizu, S., Yabana, H. and Yoshida, S. (1989a). “A New Collapse Model for Patch-Loaded
Web Plates”, Journal of Constructional Steel Research, Vol. 13. pp. 61-73.

Shimizu, S., Horii, S. and Yoshida, S. (1989b). “The Collapse Mechanisms of Patch Loaded
Web Plates”, Journal of Constructional Steel Research, Vol. 14. pp. 321-337.

123
 Talja, A. (2002). “Test Results of RHS, Tophat and Sheeting Profiles”, Technical Research
Centre of Finland, VTT Building and Transport, Internal Report: RTE50-IR-28/2002.

Timoshenko, S.P. and Gere, J.M. (1963). “Theory of Elastic Stability”, 2nd edition, McGraw-
Hill International Editions. (ISBN: 0-07-Y85821-7).

Ungermann, D. (1990). “Bemessungsverfahren für volland- undkastenträger unter

besonderer berücksichtigung des stegverhaltens”, Heft 17, Lehrstuhl für Stahlbau,
RWTH Aachen. (ISSN: 0722-1037). (In German).

Veljkovic, M. and Johansson, B. (2001). “Design for Buckling of Plates due to Direct
Stress”, Proceedings of Nordic Steel Construction Conference, Helsinki, Finland, June
2001. pp. 729-736.

Walbridge, S. and Lebet, J.-P. (2001). “Patch loading tests of bridge girders with
longitudinal web stiffeners”, Rapport d’essais ècole Polyechnique Fédérale de
Lausanne, ICOM 447.

Winter, G. (1947). “Strength of Thin Steel Compression Flanges”. Transactions, American

Society of Civil Engineers, Vol. 112. pp. 527-554.

124
 APPENDIX A: Patch Loading - Specimen Data

APPENDIX A:
Patch Loading - Specimen Data

In this appendix some specimen data of the herein used tested girders is presented.
Geometries, some mechanical properties and some of the evaluated results is showed in
tables comprising the specimens by author / authors.

In Appendix A.1 some of the data for the specimens used herein is presented. This
Appendix comprises all specimens reinforced with an open stiffener which counts 140
individual plated girders of which 136 has been used in the evaluation.

Appendix A.2 contains the specimen data for the plated girders equipped with a
closed section stiffener of V- or TRP-shape. The tables in this Appendix contains 24
individual plated girders and a schematic figure presenting the stiffener cross-section.

Some of the important data of the numerical simulations used herein is presented in
Appendix A.3. A total of 366 individual simulated plate girders with an open sectioned
longitudinal stiffener are presented in the tables.

125
126
 APPENDIX A: Patch Loading - Specimen Data

A.1: Data for specimens with open stiffeners

Table A.1: 56 of the specimens presented in Janus et. al (1988). TG 1-1, TG 1-2
and TG 1-3 were excluded from the evaluation as presented in
section 4.5.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
TG 1-1 2 505 505 236 5 50 439 50,5 250 5 12 30 0,003 -
TG 1-2 2 505 505 239 5 50 439 50,5 250 5 20 35 0,004 -
TG 1-3 2 505 505 231 5 50 453 50,5 250 5 30 33,5 0,004 -
TG 2-1 2 505 505 234 5 50 453 50,5 100 5 12 36,5 0,055 1,39
TG 2-2 2 505 505 232 5 50 446 50,5 100 5 20 35,6 0,054 1,36
TG 2-3 2 505 505 233 5 50 458 50,5 100 5 30 41 0,061 1,56
TG 3-11 2 505 505 236 5 50 485 50,5 50 5 12 35 0,049 1,31
TG 3-12 2 505 505 234 5 50 466 50,5 50 5 12 42 0,061 1,59
TG 3-21 2 505 505 239 5 50 467 50,5 50 5 20 39 0,056 1,40
TG 3-22 2 505 505 232 5 50 471 50,5 50 5 20 42 0,060 1,53
TG 3-31 2 505 505 231 5 50 461 50,5 50 5 30 47,5 0,068 1,68
TG 3-32 2 505 505 233 5 50 481 50,5 50 5 30 42,5 0,059 1,49
TG 11-1 2 1005 502,5 191 5 50 293 100,5 250 5 12 32 0,015 1,98
TG 11-2 2 1005 502,5 210 5 50 472 100,5 250 5 20 34 0,012 1,93
TG 11-3 2 1005 502,5 215 5 50 476 100,5 250 5 30 37,5 0,014 2,10
TG 12-1 2 1005 502,5 204 5 50 295 100,5 100 5 12 32,5 0,140 1,26
TG 12-2 2 1005 502,5 218 5 50 461 100,5 100 5 20 38 0,115 1,30
TG 12-3 2 1005 502,5 218 5 50 470 100,5 100 5 30 38,2 0,114 1,22
TG 13-11 2 1005 502,5 191 5 50 303 100,5 50 5 12 29 0,122 1,21
TG 13-12 2 1005 502,5 204 5 50 293 100,5 50 5 12 33 0,140 1,35
TG 13-21 2 1005 502,5 210 5 50 475 100,5 50 5 20 44 0,128 1,67
TG 13-22 2 1005 502,5 218 5 50 469 100,5 50 5 20 34 0,099 1,28
TG 13-31 2 1005 502,5 215 5 50 478 100,5 50 5 30 43 0,123 1,59
TG 13-32 2 1005 502,5 218 5 50 473 100,5 50 5 30 40 0,115 1,48
TG 31-1 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 315 0,243 1,71
TG 31-1' 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 300 0,232 1,63
TG 31-2 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 342 0,227 1,33
TG 31-2' 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 327 0,217 1,27
TG 31-3 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 370 0,264 1,23
TG 31-3' 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 395 0,282 1,31
TG 32-1 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 285 0,220 1,54
TG 32-1' 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 295 0,228 1,60
TG 32-2 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 290 0,193 1,12
TG 32-2' 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 299 0,199 1,16
TG 32-3 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 351 0,250 1,16
TG 32-3' 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 338 0,241 1,12
TG 33-1 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 296 0,229 1,60
TG 33-1' 6 622,5 500 256,4 12 120 241,7 62,25 200 5 80 276 0,213 1,50
TG 33-2 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 300 0,199 1,16
TG 33-2' 6 622,5 500 256,4 12 120 241,7 62,25 125 5 80 282 0,187 1,09
TG 33-3 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 372 0,265 1,23
TG 33-3' 6 622,5 500 256,4 12 120 241,7 62,25 75 5 80 399 0,285 1,32
TG 021-0 2,4 499,2 499,2 223,7 5,1 100 292 49,92 100 5,2 13,3 40 0,050 1,07
TG 021-1 2,2 499,4 499,4 238,2 6,1 119,9 309,4 49,94 100 5,5 31,5 55 0,050 1,52
TG 021-2 2,2 499,4 499,4 238,2 6 119,8 309,4 49,94 100 5 40,5 57,5 0,053 1,60
TG 021-3 2,2 499,4 499,4 238,2 6 120,2 309,4 49,94 100 5,3 50,2 62 0,057 1,72
TG 022-1 2,2 499,4 499,4 238,2 11,7 119,9 238,7 49,94 100 5,7 30,9 65 0,043 1,43
TG 022-2 2,2 499,4 499,4 238,2 11,9 119,3 238,7 49,94 100 5 40,5 66,5 0,044 1,45
TG 022-3 2,2 499,4 499,4 238,2 11,9 119,7 238,7 49,94 100 5,1 50,1 59 0,039 1,29
TG 041-0 4,4 501,6 501,6 361,8 8,5 118,6 262,2 50,16 100 5,1 17,2 192 0,123 1,18
TG 041-1 4 500 500 360 8,4 119,3 262,2 50 100 8,1 40,7 190 0,127 1,39
TG 041-2 4 500 500 360 7,8 119,1 262,2 50 100 8,1 50 202 0,142 1,52
TG 041-3 4 500 500 360 8,5 119,2 262,2 50 100 8,5 60,5 193,5 0,128 1,41
TG 042-1 4 500 500 360 20 102,9 285,4 50 100 7,8 39,3 315 0,114 1,59
TG 042-2 4 498,4 498,4 360 20 120,6 285,4 49,84 100 8,4 50,7 290 0,092 1,42
TG 042-3 4 498,4 498,4 360 20 120,4 285,4 49,84 100 8,4 60,4 276 0,088 1,35

127
 Table A.1: 45 of the specimens with an open stiffener presented in Janus et. al
(1988).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
TG 061-0 5,6 498,4 498,4 425,7 11,9 120,2 238,7 49,84 100 7,9 22,8 339 0,16 1,13
TG 061-1 5,6 498,4 498,4 425,7 12,3 89,7 276,5 49,84 100 10 32,2 387 0,20 1,30
TG 061-2 5,5 500,5 500,5 454,5 12,3 89,7 276,5 50,05 100 11 51,1 408 0,20 1,37
TG 061-3 5,5 500,5 500,5 454,5 12,1 89,4 276,5 50,05 100 9,9 60,1 420 0,21 1,42
TG 062-1 5,6 498,4 498,4 425,7 30,4 99 253,5 49,84 100 10 32,4 564 0,16 1,21
TG 062-2 5,6 498,4 498,4 425,7 30,5 100 253,5 49,84 100 10 50,7 592 0,16 1,27
TG 062-3 5,6 498,4 498,4 425,7 30 100,1 253,5 49,84 100 10 60,5 610 0,17 1,32
TG 121-1 2 500 500 243,8 6 119,6 274 50 100 5,2 16,6 55 0,06 1,84
TG 121-2 2 500 500 243,8 6 119,9 274 50 100 5,2 20,2 50 0,05 1,67
TG 121-3 2 500 500 243,8 6,1 119,9 274 50 100 5,2 24,9 57 0,06 1,89
TG 122-1 2 500 500 243,8 12,1 120,7 254,1 50 100 5,1 15,8 84 0,05 2,12
TG 122-2 2 500 500 243,8 12,1 120,7 254,1 50 100 5,2 20,2 72 0,04 1,81
TG 122-3 2 500 500 243,8 12,1 120,8 254,1 50 100 5,1 25,2 76 0,05 1,91
TG 141-1 4 500 500 283,3 8,4 120,1 294,3 50 100 8,1 20,2 171 0,11 1,35
TG 141-2 4 500 500 283,3 8,5 120,2 294,3 50 100 8,3 30,6 156 0,10 1,22
TG 141-3 4 500 500 283,3 8,3 120,4 294,3 50 100 8,1 34,4 185 0,12 1,46
TG 142-1 4 500 500 283,3 20,3 120,9 270,3 50 100 8,2 19,8 256,5 0,09 1,37
TG 142-2 4 500 500 283,3 20,4 120,9 270,3 50 100 8,4 30,2 248 0,08 1,32
TG-142-3 4 500 500 283,3 20,2 120,7 270,3 50 100 8 35,9 257 0,09 1,37
TG-161-1 5,4 502,2 502,2 395,6 12,4 90,7 272,1 50,22 100 10 25,7 336 0,18 1,23
TG-161-2 5,4 502,2 502,2 395,6 12,3 90,8 272,1 50,22 100 11 30,2 387,5 0,21 1,42
TG 161-3 5,4 502,2 502,2 395,6 12,4 90,7 272,1 50,22 100 11 34,7 399 0,21 1,46
TG 162-1 5,4 502,2 502,2 395,6 30,4 99,6 269,3 50,22 100 10 25,6 610 0,16 1,44
TG 162-2 5,4 502,2 502,2 395,6 30,6 99,3 269,3 50,22 100 11 30,3 600 0,16 1,39
TG 162-3 5,4 502,2 502,2 395,6 30,6 100,2 269,3 50,22 100 11 33,9 605 0,16 1,40
TG 241-1 4,1 500,2 500,2 303,9 8,3 120,3 277,5 50,02 50 5,2 16 201 0,14 1,53
TG 241-1' 4,1 500,2 500,2 303,9 8,3 120,4 277,5 50,02 50 5,1 16 196 0,13 1,49
TG 241-2 4,1 500,2 500,2 303,9 8,2 120,8 277,5 50,02 50 8,1 20,8 186 0,12 1,39
TG 241-2' 4,1 500,2 500,2 303,9 8,1 120,6 277,5 50,02 50 8,1 21,5 199 0,13 1,49
TG 241-3 4,1 500,2 500,2 303,9 8,2 120,3 277,5 50,02 50 8,3 25,5 199 0,13 1,47
TG 241-3' 4,1 500,2 500,2 303,9 8,2 120,8 277,5 50,02 50 8,2 25,5 186 0,12 1,37
TG 241-4 4,1 500,2 500,2 303,9 8,1 120,7 277,5 50,02 50 8,2 30,3 187 0,13 1,37
TG 241-4' 4,1 500,2 500,2 303,9 8,4 120,6 277,5 50,02 50 8,2 30,4 210 0,14 1,51
TG 241-5 4,1 500,2 500,2 303,9 8,1 120,7 277,5 50,02 50 7,9 35,1 192 0,13 1,38
TG 241-6 4,1 502,2 502,2 303,9 8,2 120,7 277,5 50,22 50 8,1 40,5 208 0,14 1,47
TG 242-1 4,1 502,2 502,2 303,9 19,7 118,2 244,4 50,22 50 5,2 15,4 243 0,09 1,35
TG 242-1' 4,1 502,2 502,2 303,9 19,7 118,6 244,4 50,22 50 5,2 15,8 237 0,09 1,31
TG 242-2 4,1 502,2 502,2 303,9 19,8 118,5 244,4 50,22 50 8,1 20,6 267 0,10 1,45
TG 242-2' 4,1 502,2 502,2 303,9 19,9 118,4 244,4 50,22 50 8,1 20,4 259 0,10 1,40
TG 242-3 4,1 502,2 502,2 303,9 19,8 118,6 244,4 50,22 50 8,3 24,6 255 0,10 1,36
TG 242-3' 4,1 500,2 500,2 303,9 19,9 118,7 244,4 50,22 50 8,3 24,5 261 0,10 1,39
TG 242-4 4,1 500,2 500,2 303,9 19,7 118,6 244,4 50,22 50 8,3 30,2 264 0,10 1,39
TG 242-4' 4,1 500,2 500,2 303,9 19,6 118,3 244,4 50,22 50 8,2 30,9 266 0,10 1,40
TG 242-5 4,1 500,2 500,2 303,9 19,6 118,3 244,4 50,22 50 8 35,1 270 0,10 1,40
TG 242-6 4,1 500,2 500,2 303,9 19,6 118,4 244,4 50,02 50 8,1 40,4 285 0,11 1,45

Table A.2: Two tests on girders with an open stiffener from Rockey et. al (1978).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
R2 2,1 802 798 266 15,55 300,5 286 40 168 6,1 60 71 0,01 1,65
R4 2 800 798 266 5,07 120,4 285 40 162 4 40 45 0,05 2,16

128
 APPENDIX A: Patch Loading - Specimen Data

Table A.3: 11 specimens with open stiffeners from Bergfelt (1979).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
R22 ss 2,1 802 798 266 15,55 300,5 286 40 168 6,1 60 68,5 0,01 1,59
R42 ss 2 800 798 266 5,07 120,4 285 40 162 4 40 42,5 0,04 2,04
A12 s 2 2500 800 300 15 300 295 40 160 6 60 80 0,04 1,84
A14 s 2 1200 800 300 15 300 295 40 160 6 60 78 0,02 1,72
A16 s 2 600 800 300 15 300 295 40 160 6 60 92 0,01 2,30
A22 s 3 2500 800 245 12 250 265 40 160 6 60 132,6 0,12 1,95
A24 s 3 1200 800 245 12 250 265 40 160 6 60 97,5 0,04 1,31
A26 s 3 600 800 245 12 250 265 40 160 6 60 121,4 0,03 1,86
A32 s 2 2200 680 354 5 120 290 40 136 4 40 45,8 0,15 1,85
A34 s 2 1020 680 354 5 120 290 40 136 4 40 54,4 0,08 1,97
A36 s 2 510 680 354 5 120 290 40 136 4 40 54,7 0,04 2,27

Table A.4: The two specimens with an open stiffeners from Dogaki et. al (1990).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
Model 4 3,2 897,1 899,1 270 8,005 181,2 266 90 180 4,7 29,7 105,42 0,16 1,49
Model 5 3,2 892,2 901,5 270 7,957 180,4 266 90 180 4,7 38,2 110,36 0,16 1,56

Table A.5: Two specimens with open stiffeners from Galea et. al (1987).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
P2 6 1780 1274 279 40 230 244 690 333 12 110 720 0,85 1,18
P3 6 1780 1274 2786 40 230 267 690 270 12 110 730 0,79 1,22

Table A.6: One specimen from Shimizu et. al (1987). The specimen was
excluded in the evaluation as mentioned in section 4.5.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
EL1 6 600 1000 325,2 9 300 235,2 300 200 6 80 438,2 1,02 -

Table A.7: Six specimens with open stiffeners from Bergfelt (1983).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
731 3 3000 735 252 12 250 277 40 250 6 60 93,3 0,10 1,42
732 3 1100 735 252 12 250 277 40 250 6 60 92,4 0,04 1,74
733 3 1100 735 252 12 250 277 120 250 6 60 101 0,04 1,56
734 3 3000 735 252 12 250 277 40 150 6 60 104,7 0,12 1,55
735 3 1100 735 252 12 250 277 40 150 6 60 101,8 0,04 1,30
736 3 1100 735 252 12 250 277 120 150 6 60 106,3 0,04 1,10

129
 Table A.8: 12 of the specimens with open stiffeners from Dubas and Tschamper
(1990).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
VT07-4 3,8 2480 1000 375 8,35 150 281 40 200 6 90 135 0,86 1,45
VT07-5 3,8 1760 1000 375 8,35 150 281 40 200 6 90 165 0,12 1,74
VT07-6 3,8 1760 1000 375 8,35 150 281 40 200 6 90 170 0,13 1,79
VT08-4 3,8 2480 1000 358 8,3 150 328 240 200 6 90 199 0,88 1,40
VT08-5 3,8 1760 1000 358 8,3 150 328 240 200 6 90 229 0,16 1,49
VT08-6 3,8 1760 1000 358 8,3 150 328 240 200 6 90 235 0,16 1,52
VT09-4 3,8 2480 1000 371 12 150 283 40 150 6 90 145 0,74 1,47
VT09-5 3,8 1760 1000 371 12 150 283 40 150 6 90 184 0,11 1,72
VT09-6 3,8 1760 1000 371 12 150 283 40 150 6 90 180 0,11 1,68
VT10-4 3,8 2480 1000 380 12 150 275 240 150 6 90 240 0,90 1,69
VT10-5 3,8 1760 1000 380 12 150 275 240 150 6 90 275 0,16 1,78
VT10-6 3,8 1760 1000 380 12 150 275 240 150 6 90 288 0,17 1,86

Table A.9: The three specimens with open stiffeners from Walbridge and Lebet
(2001).

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
Panel4-C2 5 1000 700 392 20 225 355 200 125 10 80 520,6 0,10 1,32
Panel5-C3 5 1000 700 392 20 225 355 200 75 10 80 559,9 0,11 1,59
Panel6-C3 5 1000 700 392 20 225 355 200 100 10 80 582,1 0,12 1,55

130
 APPENDIX A: Patch Loading - Specimen Data

A.2: Data for specimens with closed stiffeners

Table A.10: 12 specimens presented in Dubas and Tschamper (1990). The

layout of the closed stiffeners according to adjacent figure.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
VT07-1 3,8 2480 1000 375 8,35 150 296 40 150 2 130 0,96 1,34
VT07-2 3,8 1760 1000 375 8,35 150 296 40 150 2 176 0,13 1,67
VT07-3 3,8 1760 1000 375 8,35 150 296 40 150 2 172 0,13 1,64
VT08-1 3,8 2480 1000 358 8,3 150 292 240 150 2 160 0,96 1,09
VT08-2 3,8 1760 1000 358 8,3 150 292 240 150 2 280 0,21 1,76
VT08-3 3,8 1760 1000 358 8,3 150 292 240 150 2 300 0,22 1,89
-
VT09-1 3,8 2480 1000 371 12 150 286 40 150 2 130 0,85 1,17
VT09-2 3,8 1760 1000 371 12 150 286 40 150 2 198 0,12 1,65
VT09-3 3,8 1760 1000 371 12 150 286 40 150 2 210 0,12 1,75
VT10-1 3,8 2480 1000 380 12 150 282 240 150 2 247 0,85 1,55
VT10-2 3,8 1760 1000 380 12 150 282 240 150 2 330 0,20 1,90
VT10-3 3,8 1760 1000 380 12 150 282 240 150 2 315 0,19 1,82

Table A.11: Six plated girders from Carretero and Lebet (1998). The
layout of the stiffeners are according to the adjacent figure.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
G1-P2 4 1200 800 405 10 160 371 300 160 5 436,5 0,22 1,68
G2-P2 6 1200 800 447 15 200 364 300 300 5 632,1 0,19 1,59
G4-P4 6 1200 800 483 20 300 399 300 230 5 590,3 0,08 1,03
-
G4-P6 6 1800 800 483 20 300 399 300 160 5 698 0,15 1,05
G5-P1 6 1050 800 483 20 300 399 200 230 5 645,1 0,08 1,35
G6-P2 6 1050 800 483 20 300 399 200 160 5 777,9 0,09 1,22

131
 Table A.12: Four specimens presented in Kuhlmann and Seitz (2002).
The stiffener layout are according to the adjacent figure.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
II 6 2400 1200 367 20 260 396 700 250 4 1034 0,21 1,54
III 6 2400 1200 367 20 260 378 700 310 4 949 0,20 1,44
-
IV 6 2400 1200 373 20 260 378 700 250 4 991 0,96 1,47
V 6 2400 1200 373 20 260 378 700 310 4 958 0,94 1,44

Table A.13: Two specimens from Walbridge and Lebet (2001). The
stiffener layout are according to the adjacent figure.

tw hw fyw tf bf fyf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [MPa] [mm] [mm] [MPa] [mm] [mm] [mm] [mm] [kN] MR proposal
Panel2-C1 5 1000 700 392 20 225 355 200 75 5 699,1 0,14 1,77
-
Panel3-C2 5 1000 700 392 20 225 355 200 125 5 507,4 0,10 1,15

132
 APPENDIX A: Patch Loading - Specimen Data

A.3: Data for numerical simulations

The numerical simulations used for the evaluations herein is more thoroughly
presented in Davaine (2005). All the 366 simulations used herein was defined with an
open stiffener. In all simulations the yield limit for all steel plates was set to 355 MPa.

133
134
 APPENDIX A: Patch Loading - Specimen Data

Table A.14: 63 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P101 20 8000 3000 60 900 1000 285 30 300 6 948 0,20 1,61
P102 20 8000 3000 60 900 1000 435 30 300 6 987 0,21 1,52
P103 20 8000 3000 60 900 1000 585 30 300 6 558 0,19 1,34
P104 20 8000 3000 60 900 1000 735 30 300 6 348 0,19 1,23
P105 20 8000 3000 60 900 1000 885 30 300 6 009 0,18 1,18
P106 20 8000 3000 60 900 1000 1035 30 300 5 770 0,17 1,28
P107 20 8000 3000 60 900 1000 1185 30 300 5 714 0,17 1,41
P111 20 4000 3000 60 900 1000 285 30 300 7 906 0,12 1,56
P112 20 4000 3000 60 900 1000 435 30 300 7 646 0,11 1,38
P113 20 4000 3000 60 900 1000 585 30 300 7 045 0,10 1,20
P114 20 4000 3000 60 900 1000 735 30 300 6 705 0,10 1,21
P115 20 4000 3000 60 900 1000 885 30 300 6 290 0,09 1,32
P116 20 4000 3000 60 900 1000 1035 30 300 6 100 0,09 1,45
P117 20 4000 3000 60 900 1000 1185 30 300 6 061 0,09 1,62
P121 20 12000 3000 60 900 1000 285 30 300 6 385 0,28 1,49
P122 20 12000 3000 60 900 1000 435 30 300 6 652 0,30 1,56
P123 20 12000 3000 60 900 1000 585 30 300 6 559 0,29 1,48
P124 20 12000 3000 60 900 1000 735 30 300 6 499 0,29 1,40
P125 20 12000 3000 60 900 1000 885 30 300 6 212 0,28 1,28
P126 20 12000 3000 60 900 1000 1035 30 300 5 945 0,27 1,25
P127 20 12000 3000 60 900 1000 1185 30 300 5 568 0,25 1,30
P131 20 8000 3000 60 900 1000 287,5 25 250 6 691 0,20 1,55
P132 20 8000 3000 60 900 1000 437,5 25 250 6 838 0,20 1,51
P133 20 8000 3000 60 900 1000 587,5 25 250 6 479 0,19 1,37
P134 20 8000 3000 60 900 1000 737,5 25 250 6 326 0,19 1,28
P135 20 8000 3000 60 900 1000 887,5 25 250 6 068 0,18 1,20
P136 20 8000 3000 60 900 1000 1037,5 25 250 5 869 0,17 1,31
P137 20 8000 3000 60 900 1000 1187,5 25 250 5 715 0,17 1,41
P141 20 4000 3000 60 900 1000 287,5 25 250 7 649 0,11 1,54
P142 20 4000 3000 60 900 1000 437,5 25 250 7 456 0,11 1,39
P143 20 4000 3000 60 900 1000 587,5 25 250 6 888 0,10 1,20
P144 20 4000 3000 60 900 1000 737,5 25 250 6 584 0,10 1,19
P145 20 4000 3000 60 900 1000 887,5 25 250 6 228 0,09 1,31
P146 20 4000 3000 60 900 1000 1037,5 25 250 6 097 0,09 1,46
P147 20 4000 3000 60 900 1000 1187,5 25 250 6 071 0,09 1,62
P151 20 12000 3000 60 900 1000 287,5 25 250 6 222 0,27 1,46
P152 20 12000 3000 60 900 1000 437,5 25 250 6 466 0,29 1,51
P153 20 12000 3000 60 900 1000 587,5 25 250 6 472 0,29 1,47
P154 20 12000 3000 60 900 1000 737,5 25 250 6 478 0,29 1,42
P155 20 12000 3000 60 900 1000 887,5 25 250 6 334 0,28 1,35
P156 20 12000 3000 60 900 1000 1037,5 25 250 6 332 0,28 1,33
P157 20 12000 3000 60 900 1000 1187,5 25 250 5 664 0,26 1,32
P161 20 8000 3000 60 900 1000 290 20 200 6 469 0,19 1,50
P162 20 8000 3000 60 900 1000 440 20 200 6 688 0,20 1,50
P163 20 8000 3000 60 900 1000 590 20 200 6 414 0,19 1,39
P164 20 8000 3000 60 900 1000 740 20 200 6 334 0,19 1,33
P165 20 8000 3000 60 900 1000 890 20 200 6 134 0,18 1,25
P166 20 8000 3000 60 900 1000 1040 20 200 6 030 0,18 1,34
P167 20 8000 3000 60 900 1000 1190 20 200 5 674 0,17 1,40
P171 20 4000 3000 60 900 1000 290 20 200 7 441 0,11 1,53
P172 20 4000 3000 60 900 1000 440 20 200 7 304 0,11 1,42
P173 20 4000 3000 60 900 1000 590 20 200 6 779 0,10 1,25
P174 20 4000 3000 60 900 1000 740 20 200 6 510 0,10 1,18
P175 20 4000 3000 60 900 1000 890 20 200 6 192 0,09 1,31
P176 20 4000 3000 60 900 1000 1040 20 200 6 092 0,09 1,46
P177 20 4000 3000 60 900 1000 1190 20 200 6 077 0,09 1,63
P181 20 12000 3000 60 900 1000 290 20 200 6 088 0,27 1,42
P182 20 12000 3000 60 900 1000 440 20 200 6 299 0,28 1,47
P183 20 12000 3000 60 900 1000 590 20 200 6 387 0,28 1,46
P184 20 12000 3000 60 900 1000 740 20 200 6 458 0,29 1,44
P185 20 12000 3000 60 900 1000 890 20 200 6 474 0,29 1,42
P186 20 12000 3000 60 900 1000 1040 20 200 6 144 0,27 1,31
P187 20 12000 3000 60 900 1000 1190 20 200 5 835 0,26 1,36

135
 Table A.15: 63 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P1101 20 8000 3000 60 900 1000 297,5 5 50 6 043 0,18 1,40
P1102 20 8000 3000 60 900 1000 447,5 5 50 6 113 0,18 1,41
P1103 20 8000 3000 60 900 1000 597,5 5 50 6 064 0,18 1,39
P1104 20 8000 3000 60 900 1000 747,5 5 50 6 126 0,18 1,40
P1105 20 8000 3000 60 900 1000 897,5 5 50 6 118 0,18 1,39
P1106 20 8000 3000 60 900 1000 1047,5 5 50 6 139 0,18 1,39
P1107 20 8000 3000 60 900 1000 1197,5 5 50 6 155 0,19 1,53
P1201 20 8000 3000 60 900 1000 270 60 300 7 840 0,23 1,81
P1202 20 8000 3000 60 900 1000 420 60 300 7 640 0,23 1,65
P1203 20 8000 3000 60 900 1000 570 60 300 7 109 0,21 1,43
P1204 20 8000 3000 60 900 1000 720 60 300 6 767 0,20 1,27
P1205 20 8000 3000 60 900 1000 870 60 300 6 148 0,18 1,19
P1206 20 8000 3000 60 900 1000 1020 60 300 5 672 0,17 1,25
P1207 20 8000 3000 60 900 1000 1170 60 300 5 700 0,17 1,39
P1301 20 8000 3000 60 900 1000 275 50 400 7 760 0,23 1,79
P1302 20 8000 3000 60 900 1000 425 50 400 7 568 0,22 1,60
P1303 20 8000 3000 60 900 1000 575 50 400 7 022 0,21 1,35
P1304 20 8000 3000 60 900 1000 725 50 400 6 662 0,20 1,18
P1305 20 8000 3000 60 900 1000 875 50 400 6 076 0,18 1,18
P1306 20 8000 3000 60 900 1000 1025 50 400 5 673 0,17 1,25
P1307 20 8000 3000 60 900 1000 1175 50 400 5 698 0,17 1,40
P1401 20 4000 3000 60 900 1000 297,5 5 50 6 596 0,10 1,42
P1402 20 4000 3000 60 900 1000 447,5 5 50 6 690 0,10 1,43
P1403 20 4000 3000 60 900 1000 597,5 5 50 6 615 0,10 1,41
P1404 20 4000 3000 60 900 1000 747,5 5 50 6 702 0,10 1,41
P1405 20 4000 3000 60 900 1000 897,5 5 50 6 639 0,10 1,41
P1406 20 4000 3000 60 900 1000 1047,5 5 50 6 723 0,10 1,62
P1407 20 4000 3000 60 900 1000 1197,5 5 50 6 677 0,10 1,80
P1501 20 4000 3000 60 900 1000 270 60 300 8 685 0,13 1,73
P1502 20 4000 3000 60 900 1000 420 60 300 8 385 0,12 1,53
P1503 20 4000 3000 60 900 1000 570 60 300 7 855 0,12 1,34
P1504 20 4000 3000 60 900 1000 720 60 300 7 533 0,11 1,34
P1505 20 4000 3000 60 900 1000 870 60 300 6 758 0,10 1,40
P1506 20 4000 3000 60 900 1000 1020 60 300 6 005 0,09 1,42
P1507 20 4000 3000 60 900 1000 1170 60 300 6 001 0,09 1,58
P1601 20 4000 3000 60 900 1000 275 50 400 8 633 0,13 1,71
P1602 20 4000 3000 60 900 1000 425 50 400 8 311 0,12 1,51
P1603 20 4000 3000 60 900 1000 575 50 400 7 757 0,12 1,32
P1604 20 4000 3000 60 900 1000 725 50 400 7 402 0,11 1,32
P1605 20 4000 3000 60 900 1000 875 50 400 6 669 0,10 1,39
P1606 20 4000 3000 60 900 1000 1025 50 400 6 011 0,09 1,42
P1607 20 4000 3000 60 900 1000 1175 50 400 6 006 0,09 1,59
P1701 20 12000 3000 60 900 1000 297,5 5 50 5 766 0,26 1,35
P1702 20 12000 3000 60 900 1000 447,5 5 50 5 828 0,26 1,36
P1703 20 12000 3000 60 900 1000 597,5 5 50 5 796 0,26 1,35
P1704 20 12000 3000 60 900 1000 747,5 5 50 5 824 0,26 1,36
P1705 20 12000 3000 60 900 1000 897,5 5 50 5 812 0,26 1,35
P1706 20 12000 3000 60 900 1000 1047,5 5 50 5 835 0,26 1,35
P1707 20 12000 3000 60 900 1000 1197,5 5 50 5 844 0,27 1,37
P1801 20 12000 3000 60 900 1000 270 60 300 7 255 0,32 1,70
P1802 20 12000 3000 60 900 1000 420 60 300 7 377 0,33 1,73
P1803 20 12000 3000 60 900 1000 570 60 300 7 021 0,31 1,57
P1804 20 12000 3000 60 900 1000 720 60 300 6 871 0,31 1,46
P1805 20 12000 3000 60 900 1000 870 60 300 6 443 0,29 1,30
P1806 20 12000 3000 60 900 1000 1020 60 300 5 642 0,25 1,17
P1807 20 12000 3000 60 900 1000 1170 60 300 5 569 0,25 1,28
P1901 20 12000 3000 60 900 1000 275 50 400 7 169 0,32 1,68
P1902 20 12000 3000 60 900 1000 425 50 400 7 304 0,32 1,71
P1903 20 12000 3000 60 900 1000 575 50 400 6 898 0,31 1,52
P1904 20 12000 3000 60 900 1000 725 50 400 6 642 0,30 1,36
P1905 20 12000 3000 60 900 1000 875 50 400 6 119 0,28 1,18
P1906 20 12000 3000 60 900 1000 1025 50 400 5 519 0,25 1,15
P1907 20 12000 3000 60 900 1000 1175 50 400 5 548 0,25 1,28

136
 APPENDIX A: Patch Loading - Specimen Data

Table A.16: 59 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P2001 14 4000 2000 40 900 500 485 30 300 2 923 0,10 1,12
P2002 16 4000 2000 40 900 500 485 30 300 3 494 0,12 1,05
P2003 20 4000 2000 40 900 500 485 30 300 4 709 0,16 0,94
P2004 14 8000 2000 40 900 500 485 30 300 2 830 0,20 1,20
P2005 16 8000 2000 40 900 500 485 30 300 3 452 0,24 1,17
P2006 20 8000 2000 40 900 500 485 30 300 4 896 0,33 1,13
P2007 14 4000 2000 120 900 500 485 30 300 4 487 0,06 1,10
P2008 16 4000 2000 120 900 500 485 30 300 5 457 0,07 1,05
P2009 20 4000 2000 120 900 500 485 30 300 7 618 0,09 0,98
P2010 14 8000 2000 120 900 500 485 30 300 4 377 0,11 1,25
P2011 16 8000 2000 120 900 500 485 30 300 5 359 0,13 1,23
P2012 20 8000 2000 120 900 500 485 30 300 7 482 0,19 1,17
P2013 14 4000 2000 40 900 1000 485 30 300 4 031 0,14 1,20
P2014 16 4000 2000 40 900 1000 485 30 300 4 910 0,17 1,14
P2015 20 4000 2000 40 900 1000 485 30 300 6 779 0,23 1,04
P2016 14 8000 2000 40 900 1000 485 30 300 3 817 0,27 1,36
P2017 16 8000 2000 40 900 1000 485 30 300 5 152 0,36 1,46
P2018 20 8000 2000 40 900 1000 485 30 300 7 294 0,49 1,39
P2019 14 4000 2000 120 900 1000 485 30 300 4 939 0,06 1,08
P2020 16 4000 2000 120 900 1000 485 30 300 6 035 0,08 1,03
P2021 20 4000 2000 120 900 1000 485 30 300 8 414 0,10 0,97
P2022 14 8000 2000 120 900 1000 485 30 300 4 684 0,12 1,23
P2023 16 8000 2000 120 900 1000 485 30 300 5 765 0,14 1,20
P2024 20 8000 2000 120 900 1000 485 30 300 8 119 0,20 1,15
P2025 14 4000 2000 40 900 2000 485 30 300 8 048 0,28 1,90
P2026 16 4000 2000 40 900 2000 485 30 300 9 812 0,34 1,79
P2027 20 4000 2000 40 900 2000 485 30 300 13 558 0,46 1,63
P2028 14 8000 2000 40 900 2000 485 30 300 5 195 0,37 1,47
P2029 16 8000 2000 40 900 2000 485 30 300 6 635 0,46 1,48
P2030 20 8000 2000 40 900 2000 485 30 300 8 570 0,58 1,27
P2031 14 4000 2000 120 900 2000 485 30 300 9 837 0,12 1,91
P2032 16 4000 2000 120 900 2000 485 30 300 12 039 0,15 1,79
P2033 20 4000 2000 120 900 2000 485 30 300 16 788 0,21 1,65
P2034 14 8000 2000 120 900 2000 485 30 300 5 810 0,15 1,33
P2035 16 8000 2000 120 900 2000 485 30 300 7 164 0,18 1,30
P2036 20 8000 2000 120 900 2000 485 30 300 10 006 0,25 1,22
P2037 14 4000 2000 40 900 500 185 30 300 3 611 0,13 1,66
P2038 16 4000 2000 40 900 500 185 30 300 4 303 0,15 1,54
P2039 20 4000 2000 40 900 500 185 30 300 5 728 0,19 1,37
P2040 14 8000 2000 40 900 500 185 30 300 3 427 0,24 1,68
P2041 16 8000 2000 40 900 500 185 30 300 4 121 0,28 1,58
P2042 20 8000 2000 40 900 500 185 30 300 5 483 0,37 1,39
P2043 14 4000 2000 120 900 500 185 30 300 4 906 0,06 1,51
P2044 16 4000 2000 120 900 500 185 30 300 5 922 0,07 1,44
P2045 20 4000 2000 120 900 500 185 30 300 8 047 0,10 1,31
P2046 14 8000 2000 120 900 500 185 30 300 4 406 0,11 1,45
P2047 16 8000 2000 120 900 500 185 30 300 5 324 0,13 1,38
P2048 20 8000 2000 120 900 500 185 30 300 7 418 0,18 1,29
P2049 14 4000 2000 40 900 1000 185 30 300 4 394 0,15 1,69
P2050 16 4000 2000 40 900 1000 185 30 300 5 323 0,18 1,59
P2051 20 4000 2000 40 900 1000 185 30 300 7 253 0,24 1,43
P2052 14 8000 2000 40 900 1000 185 30 300 3 905 0,27 1,61
P2053 16 8000 2000 40 900 1000 185 30 300 4 692 0,32 1,50
P2054 20 8000 2000 40 900 1000 185 30 300 6 252 0,42 1,31
P2055 14 4000 2000 120 900 1000 185 30 300 5 142 0,06 1,45
P2056 16 4000 2000 120 900 1000 185 30 300 6 230 0,08 1,38
P2057 20 4000 2000 120 900 1000 185 30 300 8 517 0,11 1,26
P2058 14 8000 2000 120 900 1000 185 30 300 4 508 0,11 1,36
P2059 16 8000 2000 120 900 1000 185 30 300 5 482 0,14 1,30

137
 Table A.17: 13 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P2060 20 8000 2000 120 900 1000 185 30 300 7 748 0,19 1,22
P2061 14 4000 2000 40 900 2000 185 30 300 8 668 0,30 2,65
P2062 16 4000 2000 40 900 2000 185 30 300 10 552 0,36 2,49
P2063 20 4000 2000 40 900 2000 185 30 300 14 441 0,48 2,22
P2064 14 8000 2000 40 900 2000 185 30 300 4 328 0,30 1,41
P2065 16 8000 2000 40 900 2000 185 30 300 5 604 0,38 1,41
P2066 20 8000 2000 40 900 2000 185 30 300 6 926 0,46 1,14
P2067 14 4000 2000 120 900 2000 185 30 300 10 158 0,13 2,55
P2068 16 4000 2000 120 900 2000 185 30 300 12 321 0,15 2,37
P2069 20 4000 2000 120 900 2000 185 30 300 16 910 0,21 2,14
P2070 14 8000 2000 120 900 2000 185 30 300 4 842 0,12 1,28
P2071 16 8000 2000 120 900 2000 185 30 300 5 912 0,15 1,21
P2072 20 8000 2000 120 900 2000 185 30 300 8 574 0,21 1,16

138
 APPENDIX A: Patch Loading - Specimen Data

Table A.18: 59 of the numerical simulations presented in Davaine (2005)

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P3001 16 6000 3000 40 1000 1000 735 30 300 4 167 0,12 1,21
P3002 20 6000 3000 40 1000 1000 735 30 300 5 888 0,17 1,13
P3003 25 6000 3000 40 1000 1000 735 30 300 8 455 0,23 1,12
P3004 16 6000 3000 120 1000 1000 735 30 300 5 311 0,06 1,10
P3005 20 6000 3000 120 1000 1000 735 30 300 7 598 0,08 1,09
P3006 25 6000 3000 120 1000 1000 735 30 300 10 733 0,12 1,07
P3007 16 8000 3000 40 1000 1000 735 30 300 4 075 0,16 1,25
P3008 20 8000 3000 40 1000 1000 735 30 300 6 027 0,23 1,27
P3009 25 8000 3000 40 1000 1000 735 30 300 10 295 0,37 1,47
P3010 16 8000 3000 120 1000 1000 735 30 300 5 287 0,08 1,19
P3011 20 8000 3000 120 1000 1000 735 30 300 7 532 0,11 1,18
P3012 25 8000 3000 120 1000 1000 735 30 300 10 595 0,15 1,14
P3013 16 6000 3000 40 1000 2000 735 30 300 5 433 0,16 1,20
P3014 20 6000 3000 40 1000 2000 735 30 300 7 835 0,22 1,18
P3015 25 6000 3000 40 1000 2000 735 30 300 11 791 0,32 1,21
P3016 16 6000 3000 120 1000 2000 735 30 300 6 384 0,07 1,14
P3017 20 6000 3000 120 1000 2000 735 30 300 9 524 0,10 1,17
P3018 25 6000 3000 120 1000 2000 735 30 300 14 188 0,15 1,21
P3019 16 8000 3000 40 1000 2000 735 30 300 5 211 0,20 1,27
P3020 20 8000 3000 40 1000 2000 735 30 300 8 474 0,32 1,40
P3021 25 8000 3000 40 1000 2000 735 30 300 11 398 0,41 1,27
P3022 16 8000 3000 120 1000 2000 735 30 300 6 415 0,09 1,26
P3023 20 8000 3000 120 1000 2000 735 30 300 9 595 0,14 1,29
P3024 25 8000 3000 120 1000 2000 735 30 300 13 787 0,20 1,27
P3025 16 6000 3000 40 1000 3000 735 30 300 6 635 0,20 1,25
P3026 20 6000 3000 40 1000 3000 735 30 300 9 528 0,27 1,22
P3027 25 6000 3000 40 1000 3000 735 30 300 14 774 0,40 1,29
P3028 16 6000 3000 120 1000 3000 735 30 300 7 650 0,08 1,22
P3029 20 6000 3000 120 1000 3000 735 30 300 11 555 0,13 1,27
P3030 25 6000 3000 120 1000 3000 735 30 300 17 734 0,19 1,34
P3031 16 8000 3000 40 1000 3000 735 30 300 6 060 0,24 1,26
P3032 20 8000 3000 40 1000 3000 735 30 300 10 666 0,40 1,49
P3033 25 8000 3000 40 1000 3000 735 30 300 12 424 0,45 1,17
P3034 16 8000 3000 120 1000 3000 735 30 300 7 473 0,11 1,32
P3035 20 8000 3000 120 1000 3000 735 30 300 11 606 0,17 1,40
P3036 25 8000 3000 120 1000 3000 735 30 300 15 939 0,23 1,30
P3037 16 6000 3000 40 1000 1000 285 30 300 5 389 0,16 1,94
P3038 20 6000 3000 40 1000 1000 285 30 300 7 319 0,20 1,74
P3039 25 6000 3000 40 1000 1000 285 30 300 9 832 0,26 1,55
P3040 16 6000 3000 120 1000 1000 285 30 300 6 503 0,07 1,72
P3041 20 6000 3000 120 1000 1000 285 30 300 8 708 0,09 1,55
P3042 25 6000 3000 120 1000 1000 285 30 300 11 609 0,12 1,38
P3043 16 8000 3000 40 1000 1000 285 30 300 5 093 0,20 1,95
P3044 20 8000 3000 40 1000 1000 285 30 300 6 833 0,25 1,71
P3045 25 8000 3000 40 1000 1000 285 30 300 8 927 0,32 1,47
P3046 16 8000 3000 120 1000 1000 285 30 300 5 991 0,09 1,69
P3047 20 8000 3000 120 1000 1000 285 30 300 7 998 0,12 1,49
P3048 25 8000 3000 120 1000 1000 285 30 300 10 922 0,16 1,35
P3049 16 6000 3000 40 1000 2000 285 30 300 6 443 0,19 1,84
P3050 20 6000 3000 40 1000 2000 285 30 300 8 840 0,25 1,65
P3051 25 6000 3000 40 1000 2000 285 30 300 11 738 0,31 1,44
P3052 16 6000 3000 120 1000 2000 285 30 300 7 093 0,08 1,63
P3053 20 6000 3000 120 1000 2000 285 30 300 9 770 0,11 1,49
P3054 25 6000 3000 120 1000 2000 285 30 300 13 203 0,14 1,34
P3055 16 8000 3000 40 1000 2000 285 30 300 5 801 0,22 1,76
P3056 20 8000 3000 40 1000 2000 285 30 300 7 692 0,29 1,51
P3057 25 8000 3000 40 1000 2000 285 30 300 9 813 0,35 1,25
P3058 16 8000 3000 120 1000 2000 285 30 300 6 514 0,10 1,59
P3059 20 8000 3000 120 1000 2000 285 30 300 8 742 0,13 1,41

139
 Table A.19: 62 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P3060 25 8000 3000 120 1000 2000 285 30 300 12 071 0,17 1,28
P3061 16 6000 3000 40 1000 3000 285 30 300 7 129 0,21 1,74
P3062 20 6000 3000 40 1000 3000 285 30 300 9 860 0,27 1,57
P3063 25 6000 3000 40 1000 3000 285 30 300 13 124 0,35 1,36
P3064 16 6000 3000 120 1000 3000 285 30 300 7 822 0,09 1,62
P3065 20 6000 3000 120 1000 3000 285 30 300 10 821 0,12 1,48
P3066 25 6000 3000 120 1000 3000 285 30 300 14 761 0,16 1,33
P3067 16 8000 3000 40 1000 3000 285 30 300 6 290 0,24 1,63
P3068 20 8000 3000 40 1000 3000 285 30 300 8 272 0,31 1,38
P3069 25 8000 3000 40 1000 3000 285 30 300 10 531 0,38 1,14
P3070 16 8000 3000 120 1000 3000 285 30 300 7 027 0,10 1,54
P3071 20 8000 3000 120 1000 3000 285 30 300 9 536 0,14 1,37
P3072 25 8000 3000 120 1000 3000 285 30 300 13 149 0,19 1,23
P4001 20 6000 4000 60 1200 1000 985 30 300 6 161 0,08 1,32
P4002 25 6000 4000 60 1200 1000 985 30 300 8 607 0,10 1,21
P4003 30 6000 4000 60 1200 1000 985 30 300 11 236 0,13 1,13
P4004 20 6000 4000 150 1200 1000 985 30 300 7 731 0,04 1,20
P4005 25 6000 4000 150 1200 1000 985 30 300 10 952 0,06 1,13
P4006 30 6000 4000 150 1200 1000 985 30 300 14 490 0,08 1,07
P4007 20 8000 4000 60 1200 1000 985 30 300 5 970 0,10 1,24
P4008 25 8000 4000 60 1200 1000 985 30 300 8 465 0,14 1,18
P4009 30 8000 4000 60 1200 1000 985 30 300 11 366 0,18 1,17
P4010 20 8000 4000 150 1200 1000 985 30 300 7 594 0,06 1,15
P4011 25 8000 4000 150 1200 1000 985 30 300 10 874 0,08 1,15
P4012 30 8000 4000 150 1200 1000 985 30 300 14 393 0,10 1,12
P4013 20 6000 4000 60 1200 2000 985 30 300 7 911 0,10 1,19
P4014 25 6000 4000 60 1200 2000 985 30 300 11 469 0,14 1,19
P4015 30 6000 4000 60 1200 2000 985 30 300 15 487 0,18 1,18
P4016 20 6000 4000 150 1200 2000 985 30 300 8 890 0,05 1,09
P4017 25 6000 4000 150 1200 2000 985 30 300 13 009 0,07 1,11
P4018 30 6000 4000 150 1200 2000 985 30 300 17 766 0,09 1,12
P4019 20 8000 4000 60 1200 2000 985 30 300 7 546 0,13 1,25
P4020 25 8000 4000 60 1200 2000 985 30 300 11 194 0,18 1,27
P4021 30 8000 4000 60 1200 2000 985 30 300 16 122 0,25 1,33
P4022 20 8000 4000 150 1200 2000 985 30 300 8 693 0,06 1,17
P4023 25 8000 4000 150 1200 2000 985 30 300 13 046 0,09 1,21
P4024 30 8000 4000 150 1200 2000 985 30 300 17 868 0,13 1,21
P4025 20 6000 4000 60 1200 3000 985 30 300 9 630 0,12 1,26
P4026 25 6000 4000 60 1200 3000 985 30 300 14 119 0,17 1,26
P4027 30 6000 4000 60 1200 3000 985 30 300 19 244 0,22 1,26
P4028 20 6000 4000 150 1200 3000 985 30 300 10 439 0,06 1,15
P4029 25 6000 4000 150 1200 3000 985 30 300 15 528 0,08 1,19
P4030 30 6000 4000 150 1200 3000 985 30 300 21 678 0,12 1,22
P4031 20 8000 4000 60 1200 3000 985 30 300 8 851 0,15 1,27
P4032 25 8000 4000 60 1200 3000 985 30 300 13 409 0,22 1,31
P4033 30 8000 4000 60 1200 3000 985 30 300 20 504 0,32 1,45
P4034 20 8000 4000 150 1200 3000 985 30 300 9 887 0,07 1,20
P4035 25 8000 4000 150 1200 3000 985 30 300 15 514 0,11 1,29
P4036 30 8000 4000 150 1200 3000 985 30 300 21 913 0,16 1,34
P4037 20 6000 4000 60 1200 1000 385 30 300 7 481 0,09 1,71
P4038 25 6000 4000 60 1200 1000 385 30 300 10 028 0,12 1,54
P4039 30 6000 4000 60 1200 1000 385 30 300 12 884 0,15 1,42
P4040 20 6000 4000 150 1200 1000 385 30 300 9 676 0,05 1,66
P4041 25 6000 4000 150 1200 1000 385 30 300 13 278 0,07 1,54
P4042 30 6000 4000 150 1200 1000 385 30 300 17 009 0,09 1,43
P4043 20 8000 4000 60 1200 1000 385 30 300 7 119 0,12 1,75
P4044 25 8000 4000 60 1200 1000 385 30 300 9 826 0,16 1,60
P4045 30 8000 4000 60 1200 1000 385 30 300 12 777 0,20 1,49
P4046 20 8000 4000 150 1200 1000 385 30 300 9 252 0,07 1,71
P4047 25 8000 4000 150 1200 1000 385 30 300 12 528 0,09 1,55
P4048 30 8000 4000 150 1200 1000 385 30 300 15 973 0,11 1,42
P4049 20 6000 4000 60 1200 2000 385 30 300 10 017 0,12 1,87

140
 APPENDIX A: Patch Loading - Specimen Data

Table A.20: 47 of the numerical simulations presented in Davaine (2005).

tw hw tf bf ss b1 tst bst Fexp ME/ F /F

a exp R
Specimen [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [mm] [kN] MR proposal
P4050 25 6000 4000 60 1200 2000 385 30 300 13 626 0,16 1,69
P4051 30 6000 4000 60 1200 2000 385 30 300 17 522 0,20 1,55
P4052 20 6000 4000 150 1200 2000 385 30 300 11 017 0,06 1,67
P4053 25 6000 4000 150 1200 2000 385 30 300 15 204 0,08 1,55
P4054 30 6000 4000 150 1200 2000 385 30 300 19 421 0,10 1,42
P4055 20 8000 4000 60 1200 2000 385 30 300 9 085 0,15 1,83
P4056 25 8000 4000 60 1200 2000 385 30 300 12 415 0,20 1,64
P4057 30 8000 4000 60 1200 2000 385 30 300 15 914 0,24 1,49
P4058 20 8000 4000 150 1200 2000 385 30 300 10 239 0,07 1,67
P4059 25 8000 4000 150 1200 2000 385 30 300 13 844 0,10 1,50
P4060 30 8000 4000 150 1200 2000 385 30 300 17 690 0,13 1,37
P4061 20 6000 4000 60 1200 3000 385 30 300 11 175 0,14 1,81
P4062 25 6000 4000 60 1200 3000 385 30 300 15 478 0,18 1,66
P4063 30 6000 4000 60 1200 3000 385 30 300 20 088 0,23 1,53
P4064 20 6000 4000 150 1200 3000 385 30 300 12 079 0,07 1,66
P4065 25 6000 4000 150 1200 3000 385 30 300 16 882 0,09 1,55
P4066 30 6000 4000 150 1200 3000 385 30 300 21 818 0,12 1,43
P4067 20 8000 4000 60 1200 3000 385 30 300 10 119 0,17 1,76
P4068 25 8000 4000 60 1200 3000 385 30 300 13 831 0,22 1,58
P4069 30 8000 4000 60 1200 3000 385 30 300 17 578 0,27 1,41
P4070 20 8000 4000 150 1200 3000 385 30 300 11 156 0,08 1,65
P4071 25 8000 4000 150 1200 3000 385 30 300 15 202 0,11 1,48
P4072 30 8000 4000 150 1200 3000 385 30 300 19 480 0,14 1,35
P5001 25 8000 5000 60 1200 1000 1235 30 300 8 283 0,10 1,35
P5002 30 8000 5000 60 1200 1000 1235 30 300 10 860 0,13 1,26
P5003 25 8000 5000 150 1200 1000 1235 30 300 10 551 0,06 1,27
P5004 30 8000 5000 150 1200 1000 1235 30 300 13 976 0,08 1,20
P5005 25 8000 5000 60 1200 2000 1235 30 300 10 304 0,13 1,24
P5006 30 8000 5000 60 1200 2000 1235 30 300 13 924 0,16 1,23
P5007 25 8000 5000 150 1200 2000 1235 30 300 12 135 0,07 1,20
P5008 30 8000 5000 150 1200 2000 1235 30 300 16 767 0,09 1,22
P5009 25 8000 5000 60 1200 3000 1235 30 300 11 851 0,15 1,23
P5010 30 8000 5000 60 1200 3000 1235 30 300 16 158 0,19 1,22
P5011 25 8000 5000 150 1200 3000 1235 30 300 13 784 0,08 1,23
P5012 30 8000 5000 150 1200 3000 1235 30 300 19 750 0,11 1,29
P5013 25 8000 5000 60 1200 1000 485 30 300 9 643 0,12 1,68
P5014 30 8000 5000 60 1200 1000 485 30 300 12 528 0,15 1,57
P5015 25 8000 5000 150 1200 1000 485 30 300 12 698 0,07 1,68
P5016 30 8000 5000 150 1200 1000 485 30 300 16 353 0,09 1,56
P5017 25 8000 5000 60 1200 2000 485 30 300 12 946 0,16 1,83
P5018 30 8000 5000 60 1200 2000 485 30 300 16 808 0,19 1,69
P5019 25 8000 5000 150 1200 2000 485 30 300 14 560 0,08 1,69
P5020 30 8000 5000 150 1200 2000 485 30 300 18 644 0,10 1,55
P5021 25 8000 5000 60 1200 3000 485 30 300 14 710 0,18 1,79
P5022 30 8000 5000 60 1200 3000 485 30 300 18 751 0,22 1,62
P5023 25 8000 5000 150 1200 3000 485 30 300 16 171 0,09 1,69
P5024 30 8000 5000 150 1200 3000 485 30 300 20 684 0,11 1,54

141
142
 APPENDIX B: Patch Loading - Further Evaluation

APPENDIX B:
Patch Loading - Further Evaluation

In Appendix B.1 additional statistical information and supplementary graphs is

provided. Three sections focused on the herein proposed design model, the design model
by Graciano and the design model by Davaine are enclosed in this appendix. This
additional material is presented as complement to the presented tables and graphs in
chapter 4.

Appendix B.2 contains a statistical evaluation of the herein proposed design model.
The method for calculating the partial safety factor JM1 by the recommendations of EN
1990 (2002) is presented along with the results of the corresponding evaluation with
respect to the experimental data base of 160 tests (136 + 24) and the 366 numerical
simulations.

143
 B.1 Evaluation of design models
Within this section the influence of various parameters on the ultimate patch loading
resistance are shown. The section is divided into three sub-sections containing
supplementary information regarding the design proposals of this thesis, the thesis
Graciano (2002) and Davaine (2005) respectively. Some additional statistical
evaluations are also enclosed in table format. The notation FR in each section refers to
the predicted ultimate patch loading resistance for the model respectively. Further,
special findings made with aid of these additional graphs have been presented in chapter
4 and / or chapter 8. The tests used are the same 136 with open stiffeners, the 24 with
closed stiffeners and the 366 numerical simulations as used in the evaluation presented
in chapter 4.

B.1.1 The proposed design model, section 4.4

The design model labelled as “proposal” are according to section 4.4. Within this
section additional information regarding how some key parameters influences the
prediction level of the model, i.e. the ratio Fexp / FR.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 0.2 0.4 0.6
b 1/ h w

Figure B.1: Fexp / FR as function of the ratio b1 / hw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

144
 APPENDIX B: Patch Loading - Further Evaluation

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 2 4 6 8
tw [mm]
Figure B.2: Fexp / FR as function of the web thickness, tw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 0.5 1 1.5 2 2.5
fyf / fyw

Figure B.3: Fexp / FR as function of the yield stress ratio fyf / fyw. The 136 tests
with open stiffeners and the 24 with closed section stiffener.

145
 Tests from literature (OS)
3 Tests from literature (CS)
Fexp / FR Proposal

0
0 1 2 3
ss / b1

Figure B.4: Fexp / FR as function of the ratio ss / b1. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 0.1 0.2 0.3 0.4
ss / a

Figure B.5: Fexp / FR as function of the ratio ss / a. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

146
 APPENDIX B: Patch Loading - Further Evaluation

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 0.2 0.4 0.6
ss / hw
Figure B.6: Fexp / FR as function of the ratio ss / hw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 0.4 0.8 1.2 1.6
hw / a

Figure B.7: Fexp / FR as function of the ratio ss / a. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

147
 Tests from literature (OS)
3 Tests from literature (CS)
Fexp / FR Proposal

0
0 5 10 15 20 25
bf / tf

Figure B.8: Fexp / FR as function of the ratio bf / tf. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 2 4 6 8 10
tf / tw
Figure B.9: Fexp / FR as function of the ratio tf / tw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

148
 APPENDIX B: Patch Loading - Further Evaluation

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 100 200 300 400 500
hw / tw
Figure B.10: Fexp / FR as function of the ratio hw / tw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Proposal
Fexp / FR

0
0 40 80 120 160
b1 / tw

Figure B.11: Fexp / FR as function of the ratio b1 / tw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

149
 Numerical simulations
3 Proposal
Fexp / FR

0
0 0.2 0.4 0.6
b1/ hw

Figure B.12: Fexp / FR as function of the ratio b1 / hw. The 366 numerical
simulations with open stiffeners.

Numerical simulations
3
Proposal
Fexp / FR

0
0 0.1 0.2 0.3 0.4
b 1/ a
Figure B.13: Fexp / FR as function of the ratio b1 / a. The 366 numerical
simulations with open stiffeners.

150
 APPENDIX B: Patch Loading - Further Evaluation

B.1.2 The proposed design model of Graciano (2002)

In this section some additional information regarding the performance of the model
presented in Graciano (2002) is provided. In section 4.6 a statistical comparison
regarding the 136 tests with open stiffeners was presented. In this section the same
comparative study is presented with respect to the 24 tests on panels with closed section
stiffeners and the 366 numerical simulations. The proposed approach of Graciano (2002)
is compared to the approach presented in section 4.4. Further some additional graphs
presenting the experimental results in relation to the prediction of the model by Graciano
is provided within this section.

Table B.1: Statistical comparison between the proposal of Graciano (2002) and
the herein proposed design approach with respect to the ratio Fexp /
FR. Tests with closed section stiffeners (CS) and the numerical
simulations (FEA).

CS with CS with FEA with FEA with

respect to respect to respect to respect to
proposal Graciano proposal Graciano
Number of tests 24 24 366 366
Mean 1,499 1,265 1,410 1,344
Standard deviation 0,271 0,286 0,235 0,234
Coefficient of variation 0,180 0,226 0,167 0,174
Lower 5-percent fractile 1,060 0,874 1,125 1,029
Upper 5-percent fractile 1,879 1,623 1,793 1,701

151
 Tests from literature (OS)
3 Tests from literature (CS)
Fexp / FR Graciano (2002)

0
0 0.2 0.4 0.6
b1/ hw

Figure B.14: Fexp / FR as function of the ratio b1 / hw. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Tests from literature (OS)

3 Tests from literature (CS)
Graciano (2002)
Fexp / FR

0
0 0.1 0.2 0.3 0.4
b 1/ a

Figure B.15: Fexp / FR as function of the ratio b1 / a. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

152
 APPENDIX B: Patch Loading - Further Evaluation

Numerical simulations
3 Graciano (2002)
Fexp / FR

0
0 0.2 0.4 0.6
b1/ hw

Figure B.16: Fexp / FR as function of the ratio b1 / hw. The 366 numerical
simulations with open stiffeners.

Numerical simulations
3 Graciano (2002)
Fexp / FR

0
0 0.1 0.2 0.3 0.4
b 1/ a
Figure B.17: Fexp / FR as function of the ratio b1 / a. The 366 numerical
simulations with open stiffeners.

B.1.3 The proposed design model of Davaine (2005)

The corresponding additional information presented in the previous section is in this
section presented with the focus on the model of Davaine (2005). In this section the same
comparative study is presented with respect to the 24 tests on panels with closed section

153
stiffeners and the 366 numerical simulations. The proposed approach of Davaine (2005)
is compared to the approach presented in section 4.4.

Table B.2: Statistical comparison between the proposal of Davaine (2005) and
the herein proposed design approach with respect to the ratio Fexp /
FR. Tests with closed section stiffeners (CS) and the numerical
simulations (FEA).

CS with CS with FEA with FEA with

respect to respect to respect to respect to
proposal Davaine proposal Davaine
Number of tests 24 24 366 366
Mean 1,499 1,525 1,410 1,330
Standard deviation 0,271 0,285 0,235 0,162
Coefficient of variation 0,180 0,187 0,167 0,122
Lower 5-percent fractile 1,060 1,059 1,125 1,113
Upper 5-percent fractile 1,879 1,890 1,793 1,576

Tests from literature (OS)

3 Tests from literature (CS)
Davaine (2005)
Fexp / FR

0
0 0.2 0.4 0.6
b1/ hw
Figure B.18: Fexp / FR as function of the ratio b1 / hw. The 136 tests with open
stiffeners and the 24 with closed section stiffener

154
 APPENDIX B: Patch Loading - Further Evaluation

Tests from literature (OS)

3 Tests from literature (CS)
Davaine (2005)
Fexp / FR

0
0 0.1 0.2 0.3 0.4
b 1/ a

Figure B.19: Fexp / FR as function of the ratio b1 / a. The 136 tests with open
stiffeners and the 24 with closed section stiffener.

Numerical simulations
3 Davaine (2005)
Fexp / FR

0
0 0.2 0.4 0.6
b1/ hw

Figure B.20: Fexp / FR as function of the ratio b1 / hw. The 366 numerical
simulations with open stiffeners.

155
 Numerical simulations
3 Davaine (2005)
Fexp / FR

0
0 0.1 0.2 0.3 0.4
b 1/ a
Figure B.21: Fexp / FR as function of the ratio b1 / a. The 366 numerical
simulations with open stiffeners.

156
 APPENDIX B: Patch Loading - Further Evaluation

B.2 Statistical evaluation of the proposed design model

To ensure that the proposed patch loading resistance model for longitudinally
stiffened webs (according to section 4.8) is safe to use in design, an evaluation according
to the recommendations in EN 1990 (2002), Annex D was carried out. The target of this
evaluation is to establish the partial safety factor JM1 to be used with eq. (4.30) to
determine the design resistance according to the proposed approach. Variations in the
geometry and the yield resistance, fy, were assumed to be the parameters influencing the
ultimate resistance the most, hence treated as stochastic variables in this evaluation.

B.2.1 Evaluation procedure according to EN 1990 (2002)

The evaluation procedure used herein is the standard procedure for statistical
determination of resistance models according to Annex D in EN 1990 (2002). The
evaluation is initiated putting the established resistance function, i.e. according to
section 4.8, on the form

r t = g rt X (B.1)

in which grt symbolizes the equations in the proposed approach for predicting the
ultimate patch loading resistance and the stochastic variables are denoted with X.
Further, the probabilistic model of the resistance is put according to

r = b ˜ rt ˜ G (B.2)

in which G is an error term for each individual experimental value (divergence between
the experimental and predicted values), and b is a correction factor estimated by the
“Least Square”-best fit to the experimental values, i.e.

b =
¦ r ˜re
---------------------
-
t
(B.3)
2
¦r t

A mean value of the proposed resistance function is calculated using the mean values
of the basic variables, Xm, according to

r m = b ˜ r t ˜ X m ˜ G = b ˜ g rt X m ˜ G (B.4)

Determine the coefficient of variation of the error term is the following step to take.
This is done using the error term for each individual experimental value, Gi, calculated
according to

157
 rei
G i = ------------ (B.5)
b ˜ r ti

An estimated value for the coefficient of variation of the error should then be
determined according to the following four equations, eq. (B.6) - eq. (B.9).

'i = ln G i (B.6)

n
1
' = ---
n ¦' i (B.7)
i=1

n
2 1 2
s ' = ------------ ˜
n–1 ¦ ' –' i (B.8)
i=1

and finally the coefficient of the variation of the error terms according to

2
VG = e s' – 1 (B.9)

The herein proposed resistance function is on the form, by EN 1990 (2002) called,
“more complex” according to

r = b ˜ r t ˜ G = b ˜ g rt ˜ X 1 } X j ˜ G (B.10)

and the coefficient of variation for the resistance function is obtained by

j
2 VAR > g rt X m @ 1 § w--------
g rt · 2
V rt = ----------------------------------
2
g rt X m
- # ------------------
2
g rt X m
u
¦ © wX i i¹
-V (B.11)
i=1

The attentive reader here grasps the complexity of the partial derivative according to
eq. (B.11) and the resistance model. However, in Müller (2003) a conservative value of
Vrt = 0,08 has been used concerning similar issues. The coefficient of variation denoted Vrt
regards the variations in the geometry and the yield resistance. By using this the coefficient of
variation for the probabilistic model may be calculated according to

2 2
Vr = V G + V rt (B.12)

158
 APPENDIX B: Patch Loading - Further Evaluation

Furthermore, since the population of tests comprises more than 100 individual
experiments the characteristic resistance is calculated according to

2
– k f ˜ Q – 0 ,5 ˜ Q
r k = b ˜ g rt X m ˜ e (B.13)

in which

2
Q = V ln r = ln V r + 1 (B.14)

The characteristic fractile factor, kn, for a population comprising a large number of
experimental values, i.e. n o f , the fractile factor k f in eq. (B.13) is set to 1,64. To
acquire the design resistance basically the same equation as eq. (B.13) is used, however
with a different fractile factor.

2
– k d f ˜ Q – 0 ,5 ˜ Q
r d = b ˜ g rt X m ˜ e (B.15)

The design fractile factor, k d n , for a population comprising a large number of tests,
n o f is denoted k d f and set to 3,04. Then the partial factor for the resistance, JM, may
be determined according to

2
– k f ˜ Q – 0 ,5 ˜ Q
rk e
JM = ---- = ----------------------------------------2- (B.16)
rd – k d f ˜ Q – 0 ,5 ˜ Q
e

and further, the corrected partial factor, J M , is calculated after

r
J M = ----n = k c ˜ J M (B.17)
rd

in where the resistance using the nominal values of the basic variables is denoted rn and
the error term for this case, kc, taking into account that fy is a minimum value and not an
average, is determined as

2
– 2 ˜ V f – 0 ,5 ˜ V f
y y
1 e
k c = --- ˜ --------------------------------------
2
- (B.18)
b – k ˜ Q – 0 ,5 ˜ Q
e f

B.2.2 Calculation of the partial safety factor - Experiments

The herein enclosed statistical evaluation focused on the experimental values only.
The tests on the welded girders wit both open and closed section stiffeners have been

159
chosen to be evaluated as one population tests. Hence the whole population evaluated
comprises a total of 160 individual tests, i.e. all of the tests evaluated in section 4.5. The
coefficient of variation regarding the yield resistance Vfy was set to 0,07 and the lumped
coefficient of variation of the geometry and the yield resistance, Vrt was as previously
mentioned set to 0,08. Furthermore the variables was assumed to be log-normal
distributed. The 160 test results as a function of the proposed prediction model is showed
in Figure B.22.

According to eq. (B.3) the mean value of the correction factor, b, was determined and
the coefficient of variation regarding the error term, VG, was determined using eq. (B.5)
- eq. (B.9).

b = 1,498
VG = 0,171

With use of the in section B.2.1 discussed coefficient of variation considering the
resistance function, Vrt, the coefficient of variation of the probabilistic model is calculated
according to eq. (B.12), i.e.

2 2 2 2
Vr = V G + V rt = 0 ,171 + 0 ,08 = 0 ,189

1200

800
re [kN]

400

Tests from literature (OS)

Tests from literature (CS)

0
0 400 800 1200
rt [kN]

Figure B.22: The 160 tests results denoted re as a function of the predicted
resistance, rt, according to the proposed design model.

160
 APPENDIX B: Patch Loading - Further Evaluation

The parameter Q is the calculated according to eq. (B.14)

2 2
Q = V ln r = ln V r + 1 = ln 0 ,189 + 1 = 0 ,187

which is used for the next step, calculating the partial factor for the resistance according
to eq. (B.16) with the values of k f and k d f used.

2 2
– k f ˜ Q – 0 ,5 ˜ Q – 1 ,64 ˜ 0 ,187 – 0 ,5 ˜ 0 ,187
rk e e
JM = ---- = ----------------------------------------2- = -------------------------------------------------------2- = 1 ,299
rd – k d f ˜ Q – 0 ,5 ˜ Q – 3 ,04 ˜ 0 ,187 – 0 ,5 ˜ 0 ,187
e e

In order to calculate the corrected partial factor, kc needs to be determined using eq.
(B.18) according to

2
– 2 ˜ V f – 0 ,5 ˜ V f – 2 ˜ 0 ,07 – 0 ,5 ˜ 0 ,07
2
y y
1 e 1 e
k c = --- ˜ --------------------------------------
2
- = ------------- ˜ -------------------------------------------------------2- = 0 ,801
b – k f ˜ Q – 0 ,5 ˜ Q 1 , 498 – 1 ,64 ˜ 0 ,187 – 0 ,5 ˜ 0 ,187
e e

When this is determined the corrected partial factor of the resistance may be
determined according to eq. (B.17)

r
J M = ----n = k c ˜ J M = 1 ,299 ˜ 0 ,801 = 1 ,040
rd

Hence, based on the evaluated tests, comprising 160 specimens with open and closed
longitudinal stiffeners, the partial safety factor to be used for determining the design
resistance according to the resistance model presented in section 4.4 is proposed to be
approximated to 1,0.

B.2.3 Calculation of the partial safety factor - Numerical simulations

The same procedure of evaluation as for the experiments presented in the previous
section was performed concerning the 366 simulations of Davaine (2005) used herein.
Since the procedure of evaluation should be know to the reader at this point, only the
calculated key values and the graph showing the simulated loads versus the predicted
(see Figure B.23) are presented within this section.

b = 1,395
VG = 0,142
Vr = 0,163

161
 The parameter Q is the calculated to

Q = 0 ,162

25000

20000

15000
re [kN]

10000

5000
Numerical simulations

0
0 5000 10000 15000 20000 25000
rt [kN]
Figure B.23: The 366 simulation tests results denoted re as a function of the
predicted resistance, rt, according to the proposed design model.

The partial factor JM was determined to

J M = 1 ,254

further

k c = 0 ,821

was used to determine the corrected partial factor to

J M = 1 ,254 ˜ 0 ,821 = 1 ,030

The partial safety factor based on the numerical simulations was determined to 1,030.
The population comprises a larger number of individual tests and is a more
heterogeneously composed group why the safety factor for these simulations are lower
than compared to the experiments. However, this partial safety factor is approximated to
1,0, i.e. in line with the proposed factor regarding the experiments.

162
 APPENDIX C: Local Buckling - Further Evaluation

APPENDIX C:
Local Buckling - Further Evaluation

In Appendix C.1 the stress - strain curves from the tensile coupon tests described in
chapter 6 are enclosed. This in the form of 6 figures containing the results from three
coupon tests each.

Appendix C.2 contains the measured dimensions of the 48 box specimens used for
the local buckling tests. Furthermore, for each specimen, the calculated plate slenderness
according to EN 1993-1-5 is provided.

All of the load - mean axial deformation graphs are enclosed in Appendix C.3. This
in the form of 14 figures describing the behaviour of all the 48 specimens tested.

In Appendix C.4 the evaluated test results from the local buckling tests are enclosed.
Furthermore, the cross section areas with included weld areas are shown along with
measured ultimate loads and evaluated ultimate stress levels.

The Appendix C.5 displays the measurement equipment used in the experimental
work. All of the gauges and other equipment are described individually.

The statistical evaluation of the partial safety factor for the proposed reduction
function with respect to local buckling is presented in Appendix C.6.

163
164
 APPENDIX C: Local Buckling - Further Evaluation

C.1 Stress - strain curves from uniaxial tests

600
V, Tensile stress [MPa]

400

200

D1
D2
D3
0

0 10 20 30 40

H, Strain [%]
Figure C.1: Stress - strain curves from tension tests along the rolling direction on
Domex 420.

600
V, Tensile stress [MPa]

400

200

D4
D5
D6
0

0 10 20 30 40

H, Strain [%]
Figure C.2: Stress - strain curves from tension tests transverse the rolling
direction on Domex 420.

165
 1000

V, Tensile stress [MPa]

800

600

400

200 W1
W2
W3

0 4 8 12 16

H, Strain [%]
Figure C.3: Stress - strain curves from tension tests along the rolling direction on
Weldox 700.

1000
V, Tensile stress [MPa]

800

600

400

200 W4
W5
W6
0

0 4 8 12 16

H, Strain [%]
Figure C.4: Stress - strain curves from tension tests transverse the rolling
direction on Weldox 700.

166
 APPENDIX C: Local Buckling - Further Evaluation

1600
V, Tensile stress [MPa]

1200

800

400
W7
W8
W9
0

0 4 8 12

H, Strain [%]
Figure C.5: Stress - strain curves from tension tests along the rolling direction on
Weldox 1100.

1600
V, Tensile stress [MPa]

1200

800

400
W10
W11
W12
0

0 4 8 12

H, Strain [%]

Figure C.6: Stress - strain curves from tension tests transverse the rolling
direction on Weldox 1100.

167
 C.2 Measured dimensions - Box specimens
Table C.1: Specimen dimensions, measured mechanical properties and
according to EN 1993-1-5, the calculated plate slenderness values.

Specimen Mean Mean Mean Plate Yield Proof Stress, Plate

Width, Height, Thickness, Strength, Rp0.2 [MPa] Slenderness,
bw [mm] h [mm] t [mm] fy [MPa] Op
S10-0a 82,4 268,1 441,3 0,65
S10-0b 82,5 268,0 441,3 0,65
S10-90a 82,5 268,1 471,0 0,67
S10-90b 82,1 268,3 471,0 0,67
S20-0a 101,3 325,5 441,3 0,80
S20-90a 101,3 327,6 471,0 0,83
Domex 420

S20-90b 101,2 323,5 471,0 0,83

S30-0a 119,4 379,7 3,05 441,3 - 0,94
S30-0b 119,5 380,8 441,3 0,95
S30-90a 119,5 379,8 471,0 0,98
S30-90b 119,4 380,7 471,0 0,98
S40-0a 181,3 571,1 441,3 1,43
S40-0b 180,8 571,1 441,3 1,43
S40-90a 181,5 570,1 471,0 1,48
S40-90b 181,1 571,2 471,0 1,48
W71-0a 89,4 276,7 772,7 0,70
W71-0b 90,0 276,1 772,7 0,70
W71-0c 89,5 276,6 772,7 0,70
W71-90a 90,0 276,7 794,0 0,71
W71-90b 89,5 277,3 794,0 0,71
W71-90c 89,5 277,1 794,0 0,71
W72-0a 109,3 336,6 772,7 0,85
Weldox 700

W72-90a 109,5 335,9 794,0 0,87

W72-90b 109,7 336,7 4.09 - 794,0 0,87
W73-0a 129,6 395,8 772,7 1,01
W73-0b 129,6 396,8 772,7 1,01
W73-90a 129,2 396,0 794,0 1,02
W73-90b 129,4 396,6 794,0 1,02
W74-0a 196,2 593,8 772,7 1,53
W74-0b 196,0 594,8 772,7 1,53
W74-90a 195,4 593,0 794,0 1,55
W74-90b 195,2 594,0 794,0 1,54
W111-0a 70,3 220,2 1350,7 0,75
W111-0b 70,4 221,3 1350,7 0,75
W111-0c 69,8 220,6 1350,7 0,74
W111-90a 70,1 218,8 1335,0 0,74
W111-90b 69,4 220,3 1335,0 0,73
W111-90c 69,5 220,2 1335,0 0,73
Weldox 1100

W112-0a 85,5 266,8 1350,7 0,91

W112-90a 85,5 265,2 1335,0 0,90
W112-90 85,3 267,6 3,98 - 1335,0 0,90
W113-0a 101,3 312,4 1350,7 1,07
W113-0b 101,3 312,2 1350,7 1,07
W113-90a 101,3 312,1 1335,0 1,07
W113-90b 101,2 312,1 1335,0 1,07
W114-0a 154,9 471,9 1350,7 1,64
W114-0b 154,9 469,6 1350,7 1,64
W114-90a 154,8 472,2 1335,0 1,63
W114-90b 155,2 472,2 1335,0 1,64

168
 APPENDIX C: Local Buckling - Further Evaluation

C.3 Load - mean axial deformation curves

600

400
Load [kN]

200 S10-0a
S10-0b
S10-90a
S10-90b
0
0 2 4 6

G, Mean deformation [mm]

Figure C.7: Load - mean deformation curves for Domex 420 specimens with
nominal plate slenderness of 0,7.

600

400
Load [kN]

200

S20-0a
S20-90a
S20-90b
0
0 2 4 6

G, Mean deformation [mm]

Figure C.8: Load - mean deformation curves for Domex 420 specimens with
nominal plate slenderness of 0,85. S20-0b saved for residual stress
measurements.

169
 600

400
Load [kN]

200 S30-0a
S30-0b
S30-90a
S30-90b
0
0 2 4 6

G, Mean deformation [mm]

Figure C.9: Load - mean deformation curves for Domex 420 specimens with
nominal plate slenderness of 1,0.

600

400
Load [kN]

200 S40-0a
S40-0b
S40-90a
S40-90b
0
0 2 4 6

G, Mean deformation [mm]

Figure C.10: Load - mean deformation curves for Domex 420 specimens with
nominal plate slenderness of 1,5.

170
 APPENDIX C: Local Buckling - Further Evaluation

1600

1200
Load [kN]

800

W71-0a
400 W71-0b
W71-90a
W71-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.11: Load - mean deformation curves for Weldox 700 specimens with
nominal plate slenderness of 0,7.

1600

1200
Load [kN]

800

400 W72-0a
W72-90a
W72-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.12: Load - mean deformation curves for Weldox 700 specimens with
nominal plate slenderness of 0,85. W72-0b saved for residual stress
measurements.

171
 1600

1200
Load [kN]

800

W73-0a
400 W73-0b
W73-90a
W73-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.13: Load - mean deformation curves for Weldox 700 specimens with
nominal plate slenderness of 1,0.

1600

1200
Load [kN]

800

400 W74-0b
W74-90a
W74-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.14: Load - mean deformation curves for Weldox 700 specimens with
nominal plate slenderness of 1,5. Specimen W72-0a removed due to
testing problems.

172
 APPENDIX C: Local Buckling - Further Evaluation

1600

1200
Load [kN]

800

400

W71-0c
W71-90c
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.15: Load - mean deformation curves for Weldox 700 specimens with
nominal plate slenderness of 0,7. Extra tests.

2000

1600
Load [kN]

1200

800
W111-0a
W111-0b
400
W111-90a
W111-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.16: Load - mean deformation curves for Weldox 1100 specimens with
nominal plate slenderness of 0,7. W111-0b failed in weld after
ultimate load was reached.

173
 2000

1600
Load [kN]

1200

800

W112-0a
400
W112-90a
W112-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.17: Load - mean deformation curves for Weldox 1100 specimens with
nominal plate slenderness of 0,85. W112-0b saved for residual stress
measurements.

2000

1600
Load [kN]

1200

800
W113-0a
W113-0b
400
W113-90a
W113-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.18: Load - mean deformation curves for Weldox 1100 specimens with
nominal plate slenderness of 1,0.

174
 APPENDIX C: Local Buckling - Further Evaluation

1600

1200
Load [kN]

800

W114-0a
400 W114-0b
W114-90a
W114-90b
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.19: Load - mean deformation curves for Weldox 1100 specimens with
nominal plate slenderness of 1,5.

2000

1600
Load [kN]

1200

800

400
W111-0c
W111-90c
0
0 2 4 6 8 10

G, Mean deformation [mm]

Figure C.20: Load - mean deformation curves for Weldox 1100 specimens with
nominal plate slenderness of 0,7. Extra tests.

175
 C.4 Test Results - Buckling Tests
Table C.2: Evaluated test results. Cross section areas with included weld areas.
Yield strength used for Domex 420 and 0,2% proof stress for Weldox
specimens.
Specimen Ultimate Load, Area of cross Ultimate stress, Ratio Vu/fy or
Fexp [kN] section [mm2] Vu [MPa] Vu/Rp0.2
S10-0a 502,3 1023,9 490,5 1,11
S10-0b 502,2 1025,1 489,9 1,11
S10-90a 514,9 1024,5 502,6 1,07
S10-90b 530,6 1020,5 520,0 1,10
S20-0a 505,9 1254,5 403,3 0,91
S20-90a 517,6 1254,9 412,4 0,88
Domex 420

S20-90b 492,6 1253,2 393,1 0,83

S30-0a 468,4 1475,7 317,4 0,72
S30-0b 484,0 1476,1 327,9 0,74
S30-90a 496,1 1476,6 336,0 0,71
S30-90b 487,1 1474,9 330,3 0,70
S40-0a 502,4 2230,1 225,3 0,51
S40-0b 484,2 2224,5 217,7 0,49
S40-90a 492,1 2232,9 220,4 0,47
S40-90b 493,9 2228,6 221,6 0,47
W71-0a 1186,5 1496,2 793,0 1,03
W71-0b 1193,5 1505,3 792,9 1,03
W71-0c 1191,6 1497,7 795,6 1,03
W71-90a 1254,4 1505,2 833,4 1,05
W71-90b 1246,3 1497,2 832,4 1,05
W71-90c 1216,8 1497,6 812,5 1,02
Weldox 700

W72-0a 1269,8 1820,8 697,4 0,90

W72-90a 1289,2 1824,5 706,6 0,89
W72-90b 1310,8 1828,5 716,8 0,90
W73-0a 1182,6 2153,0 549,3 0,71
W73-0b 1192,7 2154,0 553,7 0,72
W73-90a 1228,1 2146,7 572,1 0,72
W73-90b 1222,6 2150,9 568,4 0,72
W74-0b 1241,1 3239,6 383,1 0,50
W74-90a 1253,4 3231,0 387,9 0,49
W74-90b 1260,4 3227,4 390,5 0,49
W111-0a 1433,5 1151,0 1245,4 0,92
W111-0b 1490,8 1151,9 1294,2 0,96
W111-0c 1428,7 1142,2 1250,8 0,93
W111-90a 1378,5 1147,0 1201,9 0,90
W111-90b 1413,4 1136,0 1244,1 0,93
W111-90c 1523,5 1138,1 1338,6 1,00
Weldox 1100

W112-0a 1650,6 1393,6 1184,4 0,88

W112-90a 1607,1 1392,6 1154,1 0,86
W112-90 1667,7 1389,0 1200,7 0,90
W113-0a 1529,7 1645,1 929,9 0,69
W113-0b 1543,2 1644,8 938,2 0,69
W113-90a 1522,1 1643,8 925,9 0,69
W113-90b 1551,0 1643,2 943,9 0,71
W114-0a 1591,6 2497,6 637,3 0,47
W114-0b 1560,9 2497,8 624,9 0,46
W114-90a 1538,6 2496,8 616,2 0,46
W114-90b 1557,4 2502,5 622,4 0,47

176
 APPENDIX C: Local Buckling - Further Evaluation

C.5 Gauges used in Tests

All 5 position gauges (LVDT) were from Measurements Group, U.K. LTD., Vishay.

The specifications for the 4 Welwyn HS10B LVDT’s used for

measurement of axial displacement are:

Gauge No. 9554: L = 11,0 mm, Non.linearity 0,1%,

Sensibility 4,9 mV/V.
Gauge No. 9556: L = 11,0 mm, Non.linearity 0,1%,
Sensibility 5,1 mV/V.

Gauge No. 9952: L = 10,9 mm, Non.linearity 0,1%,

Sensibility 4,8 mV/V.

Gauge No. 10544: L = 11,1 mm, Non.linearity 0,1%,

Sensibility 4,9 mV/V.

The specifications for the Welwyn HS25B LVDT used for the
measurement of buckle growth is:

Gauge No. 10168: L = 25,8 mm, Non.linearity 0,1%,

Sensibility 6,4 mV/V.

A SPIDER 8, 600 Hz from HBM were used to

sample and translate measurements to PC-
environment. Serial No. F02439.

The load cell from DARTEC used for load measurement was
calibrated in 2004 with a measurement error of < 0,6% in the
whole measurement range up to 2 MN. Serial No. 89086/A.

177
 C.6 Calculation of the partial safety factor - Tests
The same procedure of evaluation as for the experiments regarding the ultimate patch
loading resistance (presented in Appendix B) was used to evaluate the proposal of reduction
function regarding local buckling resistance. Since the procedure of evaluation should be know
to the reader at this point, only the calculated key values and the graph showing the
experimentally determined ultimate loads versus the predicted loads (see Figure C.21) are
presented within this section. A total of 85 specimens were used in this evaluation.

b = 1,143
Vd = 0,078
Vr = 0,111

The parameter Q is the calculated to

Q = 0 ,111

8000

6000
re [kN]

4000

2000
Tests made at LTU
Tests from the literature

0
0 2000 4000 6000 8000
rt [kN]
Figure C.21: The 85 experimental results, re, as a function of the predicted
resistance, rt, according to the proposed reduction function.

The partial factor JM was determined to

J M = 1 ,168

further

178
 APPENDIX C: Local Buckling - Further Evaluation

k c = 0 ,916

was used to determine the corrected partial factor to

J M = 1 ,168 ˜ 0 ,916 = 1 ,07

The partial safety factor based on the 85 individual tests results was determined to 1,07.

However, according to the discussion in chapter 7 this evaluation was also conducted using
only tests made from the 1990’ies and forward. Disregarding the earlier conducted tests, the
statistical evaluation of the partial safety factor will be determined to 1,03 according to
following calculations. A total of 60 individual specimens are left when the earlier tests are
excluded.

b = 1,165
Vd = 0,071
Vr = 0,107

The parameter Q is the calculated to

Q = 0 ,107

The partial factor JM was determined to

J M = 1 ,161

further

k c = 0 ,891

was used to determine the corrected partial factor to

J M = 1 ,161 ˜ 0 ,891 = 1 ,03

179