Truss

Evolutionary optimization of steel truss
footbridges
A study of parametric design and optimization in Karamba 3D,
Galapagos and Octopus for different truss geometries
Master’s thesis in Structural Engineering and Building Technology
VICTOR ANDERSSON
CECILIA HILLBERG
Department of Architecture and Civil Engineering
Division of Structural Engineering
CHALMERS UNIVERSITY OF TECHNOLOGY
Gothenburg, Sweden 2018
Master’s thesis ACEX30-18-35
MASTER’S THESIS ACEX30-18-35
Evolutionary optimization of steel truss footbridges

VICTOR ANDERSSON
CECILIA HILLBERG

CHALMERS UNIVERSITY OF TECHNOLOGY
VICTOR ANDERSSON
CECILIA HILLBERG
© VICTOR ANDERSSON , CECILIA HILLBERG, 2018
Master’s thesis ACEX30-18-35

Chalmers University of Technology
SE-412 96 Gothenburg
Sweden
Telephone: +46 (0)31-772 1000
Cover: :
Warren sub model in Karamba 3D representing the case study over Genevadsån.

VICTOR ANDERSSON
CECILIA HILLBERG
Chalmers University of Technology
A BSTRACT
In the early design phase of complex structures, there is often an urge of structural engineers to
get detailed and informative preliminary designs, without compromising the constraint of time.
According to the Swedish parliament and government, the environmental goals to have a sustainable
energy provision and zero net emissions to the atmosphere in 2050, also affects the infrastructural
sector (Ahston et al., 2017) and it is of importance to already in an early stage formulate viable
propositions of designs with as little material as possible.
In this thesis a computational script and model of steel truss footbridges is created in Grasshopper
and Karamba 3D respectively, which is based on the Swedish norms and regulations. The purpose is
to optimize the models with the single- and multi-objective optimization plug-ins, Galapagos and
Octopus, to minimize the mass and maximize the transparency of the structures.
Investigations, regarding steel truss bridges and parametric design, initiates the thesis constructing a
knowledge which the next steps build upon. The model in Karamba 3D is first based on a real case
from Ramböll, which it is verified with in FEM-Design and with hand calculations. Two additional
geometries are added into Karamba 3D and verified the same way as before. All three sub models,
which compose the three most common truss geometries Warren, Pratt and Howe, set the basis for
an investigation of the optimization of the structures. The objectives of this thesis are to investigate
if the results from Karamba 3D are reliable and what the difference in results from the single- and
multi-objective optimization plug-ins Galapagos and Octopus are. In a second part, the most optimal
bridge geometry is sought regarding given span lengths of 10 m, 20 m and 30 m, while having the
width and a height constant for all sub models.
The results show that the model in Karamba 3D is reliable if the bridge deck is weakened with a
factor 10 000. However, it needs to be stated that actions needed to be taken compensating for the
differences in the results of the verification considering the utilization ratio of the elements and
the global LT-buckling of the top chord. Additionally, the results proved that Octopus is the most
beneficial plug-in and that the Howe sub model is the least suitable for all span lengths. The Pratt sub
model proves to be most suitable for span length of 10 m and the Warren sub model for span length
20 m and 30 m, both considering mass and transparency. To conclude, the model, consisting of
three sub models, can be used for a preliminary design to give an initial estimation of cross-sections
and topology and also gives the opportunity to be further developed, adding more constraints if desired.
Keywords: Parametric design, Steel truss footbridge, Optimization, Karamba 3D, Galapagos, Octopus
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 i

Evolutionerande optimering av fackverks-GC-broar i stål
En studie om parametrisk design och optimering i Karamba 3D, Galapagos och Octopus for olika
fackverksgeometrier
Examensarbete inom Building and Civil Engineering
VICTOR ANDERSSON
CECILIA HILLBERG
Instutition för Arkitektur och Samhällsbyggnadsteknik
Avdelning för Konstruktionsteknik
Chalmers Tekniska Högskola
SAMMANFATTNING
Behovet av att kunna formulera tidiga designförslag som är pålitliga och informativa är något som
eftersträvas i tidiga skeden. Det är Trafikverkets vision att det år 2050 inte ska produceras några
utsläpp överhuvud taget, vilket förstås också påverkar byggandet av infrastrukturen i Sverige. För att
stödja dessa mål är det viktigt att kunna optimera och föreslå desginförslag som använder sig av så
lite material som möjligt.
Målet med det här arbetet är att göra en parameterstyrd model av fackverks-GC-broar av stål i
Karamba 3D som klarar de svenska normerna och kraven. Modellen ska därefter optimeras med två
tilläggsmoduler, Galapagos och Octopus, för att kunna reducera massan och maximera genomsiktlig-
heten av broarna.
En litteraturstudie om stålbroar och parametrisk design lägger grunden för att initiera modellering
i Karamba 3D. Den första modellen baseras på en verklig bro som byggts över Genevadsån och
har beräknats av Ramböll. Modellen i Karamba 3D verifieras med det verkliga fallet i FEM-Design
och med handberäkningar. Därefter introduceras ytterligare två geometrier och verifieras på samma
sätt. I nästa del ska dessa tre delmodeller - Warren, Pratt och Howe - optimeras med geometrin och
profilerna av elementen som variabla parametrar. Delmålen i det här arbetet består av att avgöra
graden på pålitligheten gällande kraven från Karamba 3D, undersöka vad som skiljer i resultaten
mellan optimmeringen i Galapagos och Octopus och slutligen hitta den mest optimala geometrin och
de bästa profilerna för en spannlängd av 10 m, 20 m och 30 m med en konstant bredd och höjd på bron.
Resultaten visar att modellen i Karamba 3D går att lita på om brobaneplattan försvagas med en faktor
10 000. Värt att nämna är skillnaderna som påvisades i verifieringen, huvudsakligen gällande
utnyttjandegraden av vissa element samt den globala böjvridknäckningen av överramen på GC-
broarna, som behövdes kompenseras för. Resultaten påvisar även att Octopus är den mest gynnsamma
och tidseffektiva tilläggsmodulen vid optimering. Den mest optimala GC-bron för 10 m visar sig
vara Pratt GC-bron och för 20 m och 30 m är det Warren GC-bron; detta med hänsyn till både
massa och genomsiktlighet. Slutsatsen är att modellen kan användas för en preliminär design av tre
fackverks-GC-broar som kan utgöra ett ungångsläge för vidare beräkningar och detaljer.
Nyckelord: GC-bro, Fackverk, Parametrisk design, Optimering, Karamba 3D, Galapagos, Octopus
ii , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

CONTENTS
Abstract i
Sammanfattning ii
Contents iii
Preface ix
Abbreviations xi
Nomenclature xi
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Demarcations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.5 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Steel truss footbridges 4

2.1 Steel as a construction material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Steel in relation to its emissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Standard dimensions of steel load-bearing profiles . . . . . . . . . . . . . . . . . . . . 8
3 Optimization and parametric design 10

3.1 Optimization in structural engineering . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.1 Optimization techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.2 Metaheuristic algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Parametric design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.1 Rhinoceros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.2 Grasshopper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.3 Karamba 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.4 Galapagos and Octopus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4 Development and verification of the model in

Karamba 3D 19
4.1 Description of the case which the model is to be verified with . . . . . . . . . . . . . . 19
4.2 Development of the Warren sub model in Karamba 3D . . . . . . . . . . . . . . . . . 20
4.2.1 Development of the geometry in Grasshopper . . . . . . . . . . . . . . . . . . . . . 20
4.2.2 Structural analysis in Karamba 3D . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.3 Definition of loads and load combinations . . . . . . . . . . . . . . . . . . . . . . . 22
4.3 Verification of the Warren sub model in Karamba 3D with the case study in FEM-Design 24
4.4 Development and verification of the Pratt and Howe
sub models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.5 Discussion of the verification results for the Warren, Pratt and Howe truss sub models . 33
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 iii

5 Development of the optimization features 36
5.1 Constraints when optimizing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.2 Settings for the optimization features . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.3 Optimization procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6 Results of optimization of Warren, Pratt and

Howe trusses 41
6.1 Part 1 - Galapagos versus Octopus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.2 Part 2 - Optimization of the Warren, Pratt and Howe sub models for given span lengths 44
7 Discussion 50
7.1 Part 1 - Galapagos versus Octopus . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
7.2 Part 2 - The Warren, Pratt and Howe sub models for given span lengths . . . . . . . . 52
7.3 The model in Karamba 3D and its verification results . . . . . . . . . . . . . . . . . . 53
7.4 Overall optimization results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8 Conclusion 57
8.1 Further studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
References 59
Appendix A Gradient-free and gradient-based methods 65

A.1 Gradient-free method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
A.2 Gradient-based method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.2.1 First-order methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.2.2 Second-order methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Appendix B Plans of the case study over Genevadsån 69
Appendix C Cross-sections inserted in Karamba 3D 70
Appendix D Verification of the Warren sub model in Karamba 3D with FEM-Design 75

D.1 Verification of the sectional forces and the maximum displacement . . . . . . . . . . . 75
D.2 Verification of the utilization ratios of the elements . . . . . . . . . . . . . . . . . . . 82
Appendix E Verification of the natural frequency (mode 1) of the Warren sub model 89
Appendix F Verification of the Pratt sub model in Karamba 3D with FEM-Design 91
Appendix G Verification of the Howe sub model in Karamba 3D with FEM-Design 94
Appendix H Calculation of the global lateral torsional buckling of the top chord 97
Appendix I All optimization results for part 1 and part 2 101
iv , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

List of Figures
2.1 Explanation of the different parts in a truss bridge. . . . . . . . . . . . . . . . . . . 4
2.2 Three main categories (start from top): Through, deck and half-through truss bridge. 5
2.3 A visualization of different truss bridge concepts. To the right, only considering
the self-weight, the load distribution of the members of the truss bridge concepts is
presented. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.1 Classification of optimization methods, modified after an original by (Sahab, Toropov,
& Gandomi, 2013). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 A graph explaining the Pareto-optimal set, which is represented by the black points,
A, B and D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Schematic overview of the software that form the parametric design script. . . . . . 16
3.4 (a) Example of how a truss structure can be created in Grasshopper. (b) Truss
geometry output, visualized in Rhino. . . . . . . . . . . . . . . . . . . . . . . . . 16
3.5 (a) Galapagos solver (b) Settings in the Galapagos editor. . . . . . . . . . . . . . 18
3.6 Octopus solver with solutions marked as cubes and the settings listed to the right. . 18
4.1 Reference project over the stream Genevadsån, picture from Ramböll (2018). . . . 19
4.2 The reference bridge modeled in FEM-Design. The distances are shown in meters. 20
4.3 The final geometry of the Warren truss footbridge in Grasshopper. The support
conditions are shown as arrows (locked direction) and the number of n𝑐𝑏 , within one
diagonal triangle, are shown as numbers 1 to 5. . . . . . . . . . . . . . . . . . . . 21
4.4 Final appearance of the Warren sub model in Karamba 3D with the the applied beam
and shell elements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 Normal force diagram of the bottom chord from Karamba 3D. . . . . . . . . . . . 25
4.6 Convergence study of the normal force in the bottom chord at position f for different
weakening factors in the bridge deck. . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.7 Normal force diagram of the bottom chord from Karamba 3D with a weakening
factor 10 000. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.8 Moment diagram of the bottom chord from Karamba 3D. . . . . . . . . . . . . . . 27
4.9 Shear force diagram of the bottom chord from Karamba 3D. . . . . . . . . . . . . 27
4.10 The natural frequency (mode 1) of the Warren sub model in Karamba (a) and FEM-
Design (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.11 The Pratt truss bridge with position identifications. . . . . . . . . . . . . . . . . . 32
4.12 Position identifications of the truss elements for the Howe truss bridge. . . . . . . . 33
5.1 The Warren sub model without the top chords. A unit load of 1 000 kN is applied to
calculated the displacement in y-direction. . . . . . . . . . . . . . . . . . . . . . . 37
5.2 (a) Grasshopper script that penalize the fitness functions if the solutions violates any
of the implemented constraints. (b) Optimization procedure consisting of many loops. 39
6.1 Results from the case study and the solutions for Warren 20 m optimized with
Galapagos and Octopus with specified fitness values. . . . . . . . . . . . . . . . . 42
6.2 Zoom-in of the solution with lowest mass in Table 6.3. . . . . . . . . . . . . . . . 43
6.3 All solutions from Octopus, optimizing the Warren, Pratt and Howe sub models with
a span length of 10 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.4 The most optimal topology for a span length of 10 m. . . . . . . . . . . . . . . . . 44
a span length of 20 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 v

a span length of 30 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
A.1 Gradient-free method search process, based on (Chinneck, 2001). . . . . . . . . . . 65
A.2 Example of a search process used in the steepest decent method, based on
(Rothwell, 2017). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
B.1 Used plans from the case study over Genevadsån. . . . . . . . . . . . . . . . . . . 69
D.1 Identification of the positions of the Warren sub model. . . . . . . . . . . . . . . . 75
D.2 Normal force, moment and shear force diagram of the top chord from FEM-Design
(left) and Karamba 3D (right) with a bridge deck which is weakened with a factor 1. 80
D.3 Normal force, moment and shear force diagram of the top chord from FEM-Design
(left) and Karamba 3D (right) with a bridge deck which is weakened with a factor 10
000. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
D.4 Normal force, moment and shear force diagram of the bottom chord from FEM-
Design (left) and Karamba 3D (right) with a bridge deck which is weakened with a
factor 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
D.5 Normal force, moment and shear force diagram of the bottom chord from FEM-
Design (left) and Karamba 3D (right) with a bridge deck which is weakened with a
factor 10 000. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
E.1 Natural frequency (mode 1) for the Warren sub model in Karamba 3D. . . . . . . . 89
E.2 Natural frequency (mode 1) for the Warren sub model in FEM-Design. . . . . . . . 89
F.1 The Pratt truss bridge modeled in FEM-Design. The distances are shown in meters. 91
F.2 Position identifications for the Pratt truss sub model. . . . . . . . . . . . . . . . . 91
G.1 The Howe truss bridge modeled in FEM-Design. The distances are shown in meters. 94
G.2 Position identifications for the Howe truss sub model. . . . . . . . . . . . . . . . . 94
H.1 The three sub model without the top chords. A unit load of 1 000 kN is applied to
calculated the displacement in y-direction. (a) The Warren sub model (b) The Pratt
sub model (c) The Howe sub model . . . . . . . . . . . . . . . . . . . . . . . . . 97
List of Tables
4.1 An overview of the included loads in this thesis. . . . . . . . . . . . . . . . . . . . 22
4.2 Relevant factors for the permanent and variable loads. . . . . . . . . . . . . . . . . 23
4.3 Factors for permanent and variable loads in the three load combinations considered. 23
4.4 All load combinations with the factors for the self-weight and crowd load respectively,
which are used in the created model in Karamba 3D. . . . . . . . . . . . . . . . . 24
4.5 Input data of the developed Warren sub model, which is in accordance to the input
data of the FEM-Design model of the case study. . . . . . . . . . . . . . . . . . . 24
4.6 Comparison of the reaction forces of the two models in a first attempt of verification. 25
4.7 Comparison of the absolute values of the normal force from position a to f of the
bottom chord from Karamba 3D with FEM-Design. . . . . . . . . . . . . . . . . . 25
4.8 Comparison of the absolute values of the normal force at position a to f in the bottom
chord from Karamba 3D with FEM-Design with a weakening factor 10 000(*). . . 26
4.9 Comparison of the absolute values of the moment at position a to f in the bottom
chord from Karamba 3D with FEM-Design. . . . . . . . . . . . . . . . . . . . . . 27
vi , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

4.10 Comparison of the absolute values of the shear forces at position a to f in the bottom
chord from Karamba 3D with FEM-Design. . . . . . . . . . . . . . . . . . . . . . 28
4.11 Comparison of the absolute values of the maximum displacement at position f in the
bottom chord from Karamba 3D with FEM-Design. . . . . . . . . . . . . . . . . . 28
4.12 Considered controls in Karamba 3D and FEM-Design. . . . . . . . . . . . . . . . 28
4.13 Comparison of the largest utilization ratios for each element, based on load case
6.10𝑏1 , from Karamba 3D with FEM-Design. . . . . . . . . . . . . . . . . . . . . 29
4.14 Maximum utilization ratios of each sectional force and the difference between them.
The load case applied is 6.10𝑏1 and the element considered is the mid cross beam at
position f. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.15 Results of the normal resistance (𝑁𝑅𝑑 ), critical normal force capacity (𝑁𝑐𝑟 ), moment
resistance (𝑀𝑦𝑅𝑑 ), critical moment capacity (𝑀𝑐𝑟 ), shear resistance (𝑉𝑦𝑅𝑑 , 𝑉𝑧𝑅𝑑 ), the
lateral torsional reduction factor (𝜒𝐿𝑇 ) and all interaction values (𝑘𝑦𝑦 , 𝑘𝑧𝑧 , 𝑘𝑦𝑧 , 𝑘𝑧𝑦 )
of the cross beam at position f from Karamba 3D and FEM-Design after adjustment
of the buckling lengths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.16 Comparison of maximum utilization ratios for the bottom chord, cross beams and
end cross beams in Karamba and FEM-Design with adjusted buckling lengths. . . . 30
4.17 Natural frequency for mode 1 from Karamba 3D compared with the results from
FEM-Design and hand calculations. . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.18 Reaction force of one support and the maximum displacement of the Pratt truss sub
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.19 Maximum utilization ratios for each member with the perquisites of the case study
and load case 6.10𝑏1 for the Pratt truss sub model. . . . . . . . . . . . . . . . . . . 32
4.20 Reaction force of one support and the maximum displacement of the Howe truss sub
model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.21 Maximum utilization ratios for each member with the perquisites of the case study
and load case 6.10𝑏1 for the Howe truss sub model. . . . . . . . . . . . . . . . . . 33
5.1 These are the used constraints when optimizing the Karamba 3D models. . . . . . 36
5.2 The displacement in y-direction after applying a unit load of 1 000 kN and the
maximum normal capacity 𝑁𝑏𝑅𝑑 of the top chord. . . . . . . . . . . . . . . . . . . 37
5.3 Variable parameters when optimizing. The span length, width and height of the
bridge are kept constant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.4 Settings in: Single-objective optimization plug-in Galapagos (a), multi-objective
optimization plug-in Octopus (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.5 Procedure of the optimization with Galapagos and Octopus, called part 1. For part 1
only the Warren sub model with a span length of 20 m is considered. . . . . . . . . 40
5.6 Procedure for the second part of the optimization, where the most optimal solution
for a given span length is sought, referred to as part 2. . . . . . . . . . . . . . . . . 40
6.1 Optimization results of the Warren truss bridge with the mass as fitness value. . . . 41
6.2 Optimization results of the Warren truss bridge with the transparency as fitness value. 41
6.3 The best results from the optimization of the Warren truss sub model compared with
the case study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6.4 The best solutions of the Warren, Pratt and Howe sub models considering mass for
10 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
6.5 The best solutions of the Warren, Pratt and Howe sub models considering transparency
for 10 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 vii

20 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
for 20 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
30 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
for 30 m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.1 Procedure of the optimization with Galapagos and Octopus, called part 1. For part 1
only the Warren sub model with a span length of 20 m is considered. . . . . . . . . 50
D.1 All load combinations used in the created model in Karamba 3D. . . . . . . . . . . 75
E.1 Natural frequency of mode 1 compared to FEM-Design and hand calculations. . . . 89
I.1 Procedure of the optimization of part 1. . . . . . . . . . . . . . . . . . . . . . . . 101
I.2 Procedure of the optimization of part 2. . . . . . . . . . . . . . . . . . . . . . . . 101
viii , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

PREFACE
This master thesis has been written at the Department of Architecture and Civil Engineering, Division
of Structural Engineering, at Charmers University of Technology as well as in collaboration with
Ramböll Bridge Division in Gothenburg during the spring of 2018.
First, we would like to thank the Bridge Division at Ramböll, which gave us a lot of memories and
joyful contributions during this master’s thesis. Secondly, we would especially like to thank our
supervisors at Ramböll, Georgi Nedev and Gustav Good, who showed great commitment and interest
in this project.
Finally we would like to thank Joosef Leppänen, our supervisor and examiner at Chalmers University
of Technology, who conducted and supported us during this semester.
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 ix

Abbreviations
AM Additive manufacturing.
BOF Blast oxygen furnace.
CAD Computer aided design.
DLF Direct laser fabrication.
EA Evolutionary algorithm.
EAF Electric arc furnace.
FEM Finite element modeling.
GA Genetic algorithm.
HypE Hypervolume estimation algorithm for multi-objective optimization.
LENS Laser engineered net shaping.
LSF Laser solid forming.
MO Multi-objective optimization.
MOA Multi-objective optimization algorithm.
MOEA Multi-objective evolutionary algorithms.
MOP Multi-objective optimization problem.
NURBS Non-uniform rational b-spline.
SLS Serviceability limit state.
SPEA-2 Strength Pareto evolutionary algorithm 2.
ULS Ultimate limit state.
Nomenclature
Greek letters 𝐶𝑚𝐿𝑇 Factor for equivalent constant moment
𝜒𝐿𝑇 𝑚𝑜𝑑 Modified reduction factor for consider- 𝐶𝑚𝑦 Factor for equivalent constant moment
ing lateral torsional buckling 𝐶𝑚𝑧 Factor for equivalent constant moment
𝜒𝐿𝑇 Reduction factor for considering lateral 𝑘𝑦𝑦 Interaction factor
torsional buckling 𝑘𝑦𝑧 Interaction factor
𝛾𝐺 Partial coefficient for permanent load 𝑘𝑧𝑦 Interaction factor
𝛾𝑄 Partial coefficient for variable load 𝑘𝑧𝑧 Interaction factor
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 xi

1 Introduction
In the following sections the scope of the project is explained, starting with the background to the
study and the aim and objective of the thesis, followed by a clarification of the demarcations and the
method used.
1.1 Background
In the early design of complex structures, such as bridges, structural engineers and architects, often
having conflicting goals, need to combine their special skill sets to design and analyze different
conceptual designs. This process is often time consuming and little information is known about the
project in its early stages, which makes it difficult to make informed decisions. The design choices
made in the preliminary design phase will have precautions down the whole design chain, often
resulting in more costly changes executed in a later phase than the preliminary design phase (Sharafi,
Hadi, & Teh, 2013). This is an issue which the bridge department at Ramböll Sweden has reflected
upon, resulting in a request to solve this matter.
Environmental, economical and time related profits are of importance, both considering the infras-
tructure globally and regionally, in Sweden. The Swedish parliament and government has a vision,
based on the United Nation’s framework for environmental goals (United Nations, 2018), to have a
sustainable energy provision and zero net emissions to the atmosphere in 2050 (Ahston et al., 2017);
this also affecting the infrastructural sector. These goals put an particular pressure on the ability to
optimize structures correctly and time effectively. It needs to be stated that it is the responsibility of
the consultant to already in an early stage consider the environmental and economical aspects and
optimize them if possible.
Many studies have been made on the topic of structural optimization using varying kinds of software
and methods (Rothwell, 2017; Camp, Pezeshk, & Cao, 1998; X. S. Yang, 2010; Sharafi et al., 2013;
Nanakorn & Nimtawat, 2013). Most of the studies on optimization of bridges have been carried out
on truss bridges (Balling, Briggs, & Gillman, 2006) and the parameter optimized is often the weight
of the structure. However, there are optimization algorithms that can be used to optimize multiple
parameters (Martini, 2015; Coello Coello, 2006). There have also been advances in the field of
parametric design which is the design using changeable parameters instead of statically input values.
Parametric design allows the designer to quickly analyze different design parameters without redoing
the whole design process. The contribution of the advancing technologies in the steel industries,
allowing more precise and complex parts to be produced, also pushes the optimization in production
towards the realization of the preliminary design. The combination of all parts - parametric design,
optimization tools and the ability to produce the wanted sections - finally leads to the possibility to
make informed decisions early, resulting in environmental, economical and time related benefits.
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 1

1.2 Aim
The aim of the thesis is to create a computational model of steel truss footbridges in Rhinoceros
and Grasshopper with the plug-in Karamba 3D, based on current norms and regulations in Sweden.
This model is based on a case study, which gives the possibility to verify it with a model in FEM-
Design and hand calculations. The created model in Karamba 3D is then optimized with the plug-ins
Galapagos and Octopus, which entails to minimize the steel tonnage an to maximize the bridge
transparency as much as possible presuming that all the requirements, for ultimate limit state (ULS)
and serviceability limit state (SLS), are fulfilled. The intention of the optimization is to reduce the
material use for propositions of steel truss footbridges in an early design, leading to a profit in terms
of environmental and economical aspects. Additionally, it is of interest to evaluate the reliability of
the optimized results and its potential use for further projects in the future, to get answers which type
of truss is most efficient.
1.3 Objective
This thesis starts with accomplishing a literature study, studying the steel in its relation to its emissions
during production and the standard dimensions of steel profiles and its place in the optimization world.
Further, optimization techniques and parametric design are investigated to identify possibilities and
good choices when optimizing models. Subsequently, the answers to following questions are sought,
which reassemble the objectives of this thesis:
• Are the parametric models in the FE-solver Karamba 3D and its results reliable?
• Is there a difference between the results from the single-objective optimization plug-in
Galapagos and the multi-objective optimization plug-in Octopus?
• Which bridge geometry - Warren, Pratt or Howe - is the most optimal one for the given span
lengths of 10 m, 20 m and 30 m, having the bridge width and height constant?
• Can this created script in Grasshopper in the end contribute to more precise, informative and
also time efficient results in an early stage of preliminary design?
1.4 Demarcations
In the scope of this report the designed bridge is presumed to be located in Sweden and the focus is
restricted to the superstructure of steel truss footbridges. Truss types studied in this report are the
Warren, Pratt and the Howe truss. The spans treated are 10 m, 20 m and 30 m and the loads used in
the calculations are the self-weight of the construction (permanent load) and the crowd load (variable
load). The study is focusing on the global behaviour of the truss bridges, meaning that local effects
and details are neglected as it is not of prior interest in a preliminary design.
2 , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

1.5 Method
The first part of this thesis consists of a literature study, giving fundamental knowledge to proceed
and start developing a script in Grasshopper. The first sub model, created with Karamba 3D, is a
Warren truss bridge, based on an existing case recently executed by Ramböll over Genevadsån. The
verification of this model with the Strusoft FEM-Design model of the case includes a convergence
study investigating the impact of the stiffness of the bridge deck on the normal force of all elements
of the bridge. Further, the sectional forces, the maximum displacement, the natural frequencies and
utilization ratio of all elements from Karamba 3D are compared with the results in FEM-Design.
To broaden the range of choices, two additional geometries, the Pratt and Howe trusses, are added
to the script in Grasshopper and the results from Karamba 3D verified with FEM-Design and hand
calculations.
Once the three developed sub models are verified, the structures are optimized using the evolutionary
single- and multi-objective optimization plug-ins Galapagos and Octopus. To achieve an optimal
solution, the model needs to fulfill all design criterion, when reducing the mass and increasing the
transparency of the footbridge. The following criterion were considered during optimization:
• Maximum allowable deflection of the footbridge (SLS)
• Cross-sectional capacity of the elements (ULS)
• Global lateral torsional buckling of the top chord (ULS)
Part 1 of the optimization comprises to evaluate the differences of the results from Galapagos and
Octopus, having the case study as an initial condition when optimizing. In part 2 of the optimization,
the focus lies in the achievement of the optimal solution considering given span lengths of 10 m, 20
m and 30 m of the footbridges, having the height and width constant for all three sub models.

2 Steel truss footbridges
The truss design belongs to the modern era and was rather slow developed compared to the other
four basic bridge types: The beam, the cantilever, the suspension and the arch (Troitsky, 1994). It is
said that Leonardo da Vinci was the first one to realize the strength of triangular structures which the
truss is composed of. The diagonal and/or vertical members, see Figure 2.1, also referred to as web
members, carry the shear force, either in tension or compression (Duan, 2017). The top and bottom
chord, see Figure 2.1, carry the bending moment. The truss bridge became popular in the 19th and
20th centuries since the design is material efficient and can carry large loads (Rutgersson, 2008). An
explanation of the most relevant parts of a truss bridge are described more in Figure 2.1.
Figure 2.1: Explanation of the different parts in a truss bridge.
The truss bridges can be categorized into three main categories, depending on the location of the
bridge deck and where the supports are put: Deck, half-through and through truss bridges, which
can be seen in Figure 2.2 (Duan, 2017). The through bridge indicates that the bridge deck is put on
the bottom chord, whereas the deck truss bridge has the bridge deck located on the top chord. For
larger bridges, the section is closed in case of a through truss bridge, meaning that the users of the
bridge must be able to pass under the top chord. The half-through bridge is a combination of the two
mentioned before, having the deck between the top and and bottom chords.

Figure 2.2: Three main categories (start from top): Through, deck and half-through truss bridge.
As can be seen in figure Figure 2.3 there are different types of concepts when it comes to bridge
trusses (Kassimali, 1993). The Warren truss bridge is the most common one in which its web
members take either tension or compression (Duan, 2017). The Pratt truss bridge is another concept,
which concludes in the diagonals taking tension and the vertical members carrying compression.
The Howe truss is similar to the Pratt truss, having the diagonals rotated 90 degrees, making them
take compression and the verticals tension. There are several other concepts, being versions of the
ones mentioned above, which are further subdivided in triangular formations. Two examples are the
Baltimore truss and the K truss.
Figure 2.3: A visualization of different truss bridge concepts. To the right, only considering the
self-weight, the load distribution of the members of the truss bridge concepts is presented.
The drawbacks of steel truss bridges are that they are expensive to maintain and require a lot of field
labour because of the jointed layout of the members, which also makes the design hard to access for
maintenance (Troitsky, 1994). The joints are considered as a risk zone often resulting in some kind
of damage (Mehrjoo, Khaji, Moharrami, & Bahreininejad, 2008).

For instance, fatigue damage is often triggered by micro cracks caused by for example welding at
the joint locations. These are also hard to visually recognize during inspections. Another drawback
that is common for all light weight structures is that they are susceptible to dynamic problems
(Craigh & Kurdila, 2006). Some of the advantages using a truss design is that the construction of the
bridge can be built in a factory without interfering with the traffic (Ronnenbrant & Glans, 1996). The
design of truss bridges is also very efficient in taking care of the loads and at the same time allowing
some level of transparency. Nowadays truss bridges in Sweden are a common choice when designing
footbridges.
2.1 Steel as a construction material

Metals have been used by mankind for a very long time, even giving names to historic periods like
the Bronze Age and Iron Age (Domone & Illston, 2010). The properties are strong, hard and offered
good perquisites, ideal for weapons for instance (Wagner, 1996). In the 18th century the production
of iron was possible in a larger scale, probably first produced in China. The high temperature needed,
to melt iron, was achieved using a forced air blast using coke as fuel (Domone & Illston, 2010).
The product, binding with carbon from the fuel, is called pig iron or cast iron and could easily be
shaped and used for larger structures. The manufacturing of iron was also significant in the early
19th century as the Industrial Revolution took place. The production of steel is a necessary part of
many structures in peoples everyday life and the amount increases as the magnitude of the world’s
population does (Miljönytta, 2018).
Starting at a microscopic plane of description, metals consist of an atomic structure, where a

net positively charged nucleus is surrounded by a cloud of orbiting, negatively charged electrons
(Domone & Illston, 2010). Overall, the atom is electrically neutral, presuming that the number of
protons and electrons are equal. The property of an element is very dependent on the number of
electrons in the outermost layer of an atom. These so called valence electrons are the ones most likely
to bond with other atoms, resulting in different properties of an element. Metallic atoms, compared
to other atoms, have valence electrons which are able to move freely and therefore build structures,
which are arranged in a crystal lattice (Berns & Theisen, 2008). This state gives metals the ability of a
high thermal and electric conductivity (McGuire, 2008). Due to the freely moving valence electrons,
it is possible to alloy different metals, having similar or slightly smaller sizes of the atoms, resulting
in different properties of a metal (Mandal, 2015).
There is a great range of possible compositions of steel (Bhadeshia & Honeycombe, 2006). Only
a small amount of carbon is enough to form iron to the strong and mostly tough alloy called steel,
compared to iron which is rather soft. Steel is often categorized into two groups, one being plain
carbons and the other being alloy steels (Mandal, 2015). During the production of plain carbon
steel, no alloying components are added intentionally; thus small amounts of alloying substances can
appear and be implemented during production. Alloys can be defined as a composition of iron and
other components, but alloys without carbon are usually called irons (Krauss, 2005).

2.2 Steel in relation to its emissions
The environmental sustainability has been an important issue for years and forms the everyday
life, inter alia it influences the infrastructure in Sweden (Ahston et al., 2017). The road and rail-
way infrastructure is responsible for 1-2 millions emissions of carbon dioxide equivalents, which
represents 5-10 % of the total emissions from vehicles, infrastructure and traffic in Sweden. The
Swedish parliament and government has a vision, based on the United Nation’s framework for
environmental goals, especially sub section 11 (United Nations, 2018), to have a sustainable energy
provision and zero net emissions to the atmosphere in 2050 (Ahston et al., 2017).
To reach this vision, Trafikverket (Swedish Transport Administration) has decided to demand a
decrease of the emissions of 15 % until 2020 and 30 % until 2025, which finally ends up in a zero
release of emissions in 2050 (Ahston et al., 2017). These percentages are referring to a representative
value from 2015. It is also the goal of the Swedish Transport Administration to ensure an economical
benefit throughout all processes, without compromising the long-term infrastructural goals with
respect to the environmental and social aspects. The goals are to create a society with a developing
infrastructure, entailing an improvement in efficiency concerning time, costs and performance.
Steel production is energy, resource and emission intensive (World Steel Association, 2018b), a lot
depending on for instance the way of production, the coal and iron ore used and composition of the
steel product (World Steel Association (WSA), 2015). Approximately 93 % of all greenhouse gas
emissions caused by the steel industry consist of carbon dioxide emissions, which is an issue to be
solved (World Steel Association, 2018b). The advantage is that steel components of structures can
be recycled almost infinitely without deteriorating in quality (Domone & Illston, 2010).
There are mainly two ways to produce carbon steel (World Steel Association, 2018a). One way of
manufacturing steel entails retrieving iron from iron ore in the blast oxygen furnace (BOF), which
according to World Steel Association is the most common method used globally, representing 70 % of
the production. In the blast oxygen furnace the pig iron is molten in a furnace, where oxygen is blown
into (Rennie, 2016). The temperatures reach about 1250 degrees and the final product is molten
steel, which is poured off. The BOF way of production is mainly used to produce both long and
flat steel products, where in this case high-quality scrap (pre-consumer scrap) also can be recycled
(Xylia, Silveira, Duerinck, & Meinke-Hubeny, 2017).
The second way of production is meant for the purpose to recycle the increasing amount of steel
scrap, where an electric arc furnace (EAF) is used (Xylia et al., 2017). In the electric arc furnace an
electric arc induced by electrodes is created, producing heat, which melts the metal (Rennie, 2016).
Low-quality scrap (post-consumer scrap) can be recycled and is mostly formed into long products, but
also used for the production of special kinds of steel as for instance stainless steel (Xylia et al., 2017).
In this way of production and with the increasing recycling of steel scrap, less energy is used, to be
exact 56 % less energy than with the BOF, ending up in an advantage not only environmentally, but
also considering the economic energy costs. In the year 2050 it is predicted that the scrap availability
of steel versus the production of crude steel, which means steel in its first solid state after it has been
melted, will be at the same amount, indicating the importance of an environmental friendly disposal.

Considering the crude steel production globally, it has increased quite dramatically during the last 50
years, mostly due to the industrialization in China, Brazil, India, Iran and Mexico
(World Steel Association, 2017). In 1950 approximately 189 million tonnes of crude steel were
produced, with Europe as the leading producer representing 33.5 % of the total world’s production
and North America ranked second representing 25.8 %. In 2016 the production had increased to
1630 million tonnes of crude steel, having China as the main producer (49.6 %) and Europe ranked
second representing 12.3 % of the total. Today many efforts are made to decrease the magnitude of
emission, which in fact has resulted in a decrease of 50 % in specific energy consumption in iron and
steel production, considering the last 30 years (Xylia et al., 2017).
In Sweden steel is produced in 3 ore based plants, 10 plants based on scrap and 15 plants focusing on
the processing of steel (Miljönytta, 2018). The ore based plants cover approximately two thirds of the
steel production in Sweden. The steel industry being a big part in the export business, representing a
value of ca 41 billions of Swedish crowns, corresponding to 3.4 % of the total export of goods to
approximately 140 different countries. The unsolved issue is the release of carbon dioxide, mainly
caused during the reduction from iron to pure iron using coal or coke. This has started a cooperation
project in northern Sweden in 2016 with the aim to produce iron without any carbon dioxide emissions
at all. This could be achieved using hydrogen gas instead of coal and coke, which nowadays partly is
imported from Australia (LKAB, 2018). The plan is to use hydrogen gas, produced with Swedish
fossil free electricity, in which case the rest product would be water. If this vision becomes true, it
would entail a reduction of the total carbon dioxide emissions in Sweden by 10 %.
2.3 Standard dimensions of steel load-bearing profiles

The purpose to optimize a structure at an early stage is that it can be realized in the end,
implying that it is necessary to have the possibility to manufacture the optimized shapes wanted
(Bates & Fistructe, 1991). The standard dimensions of structural elements as they are known today,
started to evolve as regulations and standards were introduced. These were introduced in the early
19th century as the quantity and size of structural sections were increasingly produced in line with the
rapid development of usage of steel as a structural material. Standard dimensions, generally cast and
forged, can represent a limitation if a complex structure is longed for (Zietala et al., 2016). Parts need
to be welded and connected to each other, resulting in weak spots exposed to corrosion for instance.
Today there are technologies available, allowing complex parts to be produced (Zietala et al., 2016).
So called laser-additive manufacturing entails producing a layer-by-layer reproduction of a computer
aided design (CAD), using powder, which is melted and added. The additive manufacturing (AM),
also called 3D printing, has been developing since the 1980s (Gu, 2015). Today the AM is most
commonly used with plastics, metals and alloys. There are several ways of laser additive manufactur-
ing today, for instance direct laser fabrication (DLF), laser solid forming (LSF) and laser engineered
net shaping (LENS) (Zietala et al., 2016).
The advantages of the laser-additive manufacturing is the time efficient production as the procedure is
executed in one step (Gu, 2015). Not only is this a way of production allowing complex shapes to be
manufactured, but also it is cost effective bypassing the need to transport products; rather producing
them closely to the site where it is needed. Additionally, considering the environmental aspect, the
laser-additive manufacturing leaves no waste behind, as the product is built, adding layer-by-layer,

with the exact amount needed. The costs of AM versus traditional manufacturing is depending
on the quantity of the production (Dumitrescu & Tanase, 2016). Regarding the traditional way
of manufacturing, costs for tools are included and implying an increased production to result in a
decrease of costs. The costs for the AM, which does not require tools, are kept constant, regardless
the extent of production. This entails that a small production with AM is less expensive compared to
the traditional manufacturing, whilst a large production could mean that AM is more expensive.
The additive manufacturing method is nowadays mainly applied in the aerospace, automobile,
biomedical and defense sector (Gupta, 2017). There are few examples where the method is used in
the construction sector to build bridges though (Eindhoven University of Technology, 2017) and
(infra, 2017). In Europe there have been some examples, where footbridges have been 3D-printed, one
example being a cycle bridge in Gemert in the Netherlands, where a 8 m long bridge was printed with
pre-stressed, reinforced concrete. Another example in the Netherlands, more accurate considering
the focus of this thesis, is a 12 m long footbridge made out of stainless steel, which is printed in a lab
and is to be installed in 2018 (Dunning, 2017). It is the first 3D-printed bridge in this scale and is to
be placed across the Oudezijds Achterburgwal in Amsterdam.
The development of 3D-printing is improving and the prediction is that it will be more adapted and
used in a large number of industries in the future (Kianian, Tavassoli, Larsson, & Diegel, 2016).
Whilst the increased utilization of AM world wide, including countries close by like Germany and
the United Kingdom, Sweden has shown a retraction regarding the use of AM, beyond its use for
prototyping. A reason could be the lack of standards for the AM (Larsson, 2016). A study from
2016 is thus stating that the majority of the AM users in Sweden are expanding, finally leading to the
possibility to participate in the global competitiveness (Kianian et al., 2016).

3 Optimization and parametric design
In the following chapters different optimization techniques as well as the concept of parametric design
are explained. Additionally some software that can be used for parametric design are introduced and
described.
3.1 Optimization in structural engineering

Any optimization problem can be defined by a maximization or minimization problem
(Hasançebi, Çarbaş, Doǧan, Erdal, & Saka, 2009). Even though there is a large variety of
optimization methods available for designers to use, the designers are often still bound to the standard
dimensions of steel produced by the steel mills. Due to the improvements and evolution of the steel
industry explained in Section 2.3, there are now potentially more reasons to optimize structures.
Optimization in the early stages of the design process can help saving material (Rothwell, 2017). The
weight reduction can help creating light weight structures to counteract the high cost of today’s high
performance materials. Reducing the weight of the cross-section not only generates less material
costs but also decreases the manufacturing and operational costs (COX, 1965). However, material
reduction have some drawbacks as well (Rothwell, 2017). Light weight structures entail an increased
slenderness of the structure having the consequence of an elevated risk of buckling. This is an
example of two conflicting goals, where a decrease of the buckling risk implies that more material
is needed, resulting in a higher cost. The conflict between these two goals can be handled using
optimization tools (Beck & Gomes, 2012).
Conventional design methods, as stated in Eurocode (EC), are used to get as close to the recommended
design values as possible. The design process is carried out and then repeated, parameters modified
and analyzed until satisfactory terms are met, for instance an acceptable utilization ratio of the
cross-sections (Nanakorn & Nimtawat, 2013). This process often demands many iterations until the
design criterion is met (Rothwell, 2017). Drawbacks with the conventional design process are time
consuming calculation processes and the selected focus on only one parameter. In the design of a
structure there are often several conflicting design parameters which make it hard to find a optimal
design using the conventional design method (Coello Coello, 2006). To get a more efficient structure,
design methods that can handle and optimize several design parameters are needed.
3.1.1 Optimization techniques

The optimization techniques can basically be divided into two categories, one being the deterministic
algorithms, the other being the stochastic algorithms (X.-S. Yang, 2010). The deterministic method
can be described as a procedure which can be repeated, resulting in the same outcome if the same
starting values are used. The stochastic algorithm on the other hand includes some random numbers,
resulting in different outcomes each time the procedure is repeated. In the end, the results of the two
methods are similar.
As can be seen in Figure 3.1, the different optimization techniques can be divided into one-dimensional
and multi-dimensional methods (Sahab, Toropov, & Gandomi, 2013). The multi-dimensional methods
are divided further into continuous and discrete methods (Black, Hashimzade, & Myles, 2017) and

(Colman, 2015). The difference between discrete and continuous methods is that variables used, can
only take a certain number of values in the discrete case, whereas an infinite number of possible
values within a given span can be taken in the continuous case.
Figure 3.1: Classification of optimization methods, modified after an original by (Sahab, Toropov, &
Gandomi, 2013).
In order to understand the different optimization techniques, the fundamentals need to be explained,
which can be done using a problem defined with Equations (3.1) to (3.4), also called the generic
form of a problem (X.-S. Yang, 2010). The functions 𝑓𝑖 (𝑥), ℎ𝑗 (𝑥) and 𝑔𝑘 (𝑥) are called objective or
cost functions, which describe the design vector. It is also possible to shift the objective functions
representing a maximization problem instead of a minimization problem, where 𝑔𝑘 (𝑥) could be larger
than 0, see Equation (3.3). The design or decision variables x in Equation (3.4) are real numbers and
the variable i describes the search or design space as is expressed in Equation (3.1).
𝐹 𝑖𝑛𝑑 𝕩𝕚 ∈ ℝ 𝑡𝑜 𝑚𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑓𝑖 (𝑥), (𝑖 = 1, 2, ..., 𝑀) (3.1)
ℎ𝑗 (𝑥) = 0, (𝑗 = 1, 2, ..., 𝐽 ) (3.2)
𝑔𝑘 (𝑥) ≤ 0, (𝑘 = 1, 2, ..., 𝐾) (3.3)
𝕩𝕚 = (𝑥1 , 𝑥2 , ..., 𝑥𝑛 )𝑇 (3.4)

A problem can either be single-objective, where 𝑀 = 1, or multi-objective, where 𝑀 > 1
(X.-S. Yang, 2010). If 𝐽 and 𝐾, also called constraints, are equal to zero, the problem is unconstrained.
Should there not be any objective at all, but only constraints, it is called a feasibility problem. Some of
the most basic unconstrained numerical methods are zero-order methods, see Figure 3.1, also called
gradient-free methods as for instance the Hooke’s and Jeeves’ method, explained in Appendix A.1
(Rothwell, 2017). Gradient-based methods, also called derivative-based, like first and second-order
methods, further explained in Appendix A.1, which can also be seen in Figure 3.1, are related to
the unconstrained optimization problems (X. S. Yang, 2010). First-order methods, like the steepest
descent method, give a linear approximation to the objective function, whereas second-order methods,
like the Quasi-Newton method, uses the second-order derivatives to improve the search direction
at each iteration (Rothwell, 2017). The gradient-based methods result in an improved convergence
regarding the optimization process compared to the gradient-free method (Martini, 2015).
Logically if there are values for 𝐽 , 𝐾 or both, the problem is constrained (Sahab et al., 2013).
In structural optimization design the building codes, for instance Eurocode, usually represent the
constraints. There are two subcategories to the constrained optimization: The indirect and direct
method, see Figure 3.1. The indirect method starts to find optimal conditions of a function and
discretizes it afterwards, whereas for the direct method it is the other way around (Potschka, 2014).
The advantage of the direct method is that the result is generic.
3.1.2 Metaheuristic algorithms
Metaheuristic algorithms are a family of stochastic algorithms and have become popular in civil
engineering to optimize truss structures, seismic analyses, traffic flow analyses, lay-out designs of
floor beams and many more applications (Gandomi, Yang, Talatahari, & Alavi, 2013b). Metaheuristic
means higher level and are often stochastic algorithms with randomization and global exploration.
Metaheuristic algorithms can be used in a variety of optimization problems, for instance for discrete,
constrained and unconstrained optimization as can be seen in Figure 3.1, which is a factor why it is
an advantageous technique commonly used.
Compared to traditional numerical optimization, solvers based on metaheuristic algorithms are

inspired by nature, mimicking optimization processes in nature, one example being Darwin’s law
of "survival of the fittest" (Gandomi, Yang, Talatahari, & Alavi, 2013a). In a metaheuristic algo-
rithm each solution is called an individual and forms a large pool of solutions called a population
(Rothwell, 2017). One benefit with metaheuristic algorithms is that due to their stochastic behaviour
they cover a larger part of the design space than traditional solvers. As individuals will be generated
all over the design space, it increases the possibility to find a local optimum, where a local optimum
is a solution that is optimal within its closest neighboring solutions. Since the first population defined
by metaheuristic algorithms is stochastically generated, there is no need for an initial guess to start
the computation (Sahab et al., 2013).
The metaheuristic algorithms use two different features to search for solutions, the intensification
and the diversification (X.-S. Yang, 2010). Intensification, also called exploitation, means directing
the focus to a local area, concentrating to select the best solution (Gandomi et al., 2013a). The
diversification on the other hand, called exploration, is the process of generating many different types

of solutions to facilitate the global search space, making it more efficient. Diversification is used to
make sure the solutions do not get trapped in a local optimum. It is important to find a balance between
the two features, avoiding difficulties due to convergence if the intensification is too dominant, or
complications finding the global optimum if the diversification is influencing more (X.-S. Yang, 2010).
One of the metaheuristic algorithms is the evolutionary algorithm (EA), which is combining the
intensification and diversification. Evolutionary algorithms are based on the inspiration on the bio-
logical evolution in nature (Sahab et al., 2013). The most commonly used evolutionary algorithm is
the genetic algorithm (GA), which is a solver trying to replicate a natural selection process where
the "best" (fittest) solutions, also called individuals, in a population is evolved (Simon, 2013). This
algorithm is, as in the natural selection process, based on the facts that the population is diverse and
that each individual has a finite life span and is able to reproduce. In a GA the first generation of
individuals is randomly generated, resulting in some of them having high and some having low fitness
(Sahab et al., 2013). Individuals with high fitness are more likely to survive and reproduce, which
is why the individuals with low fitness are removed (Simon, 2013). To create a second generation,
the individuals from the first generation (parents), having a high fitness, will cross-over (mate) with
each other and thus create a new population (children). The cross-over process gives the children a
combination of "genetic" information from the parents. This process is replicating Darwin’s theory
of natural selection. Since the individuals with high fitness will mate, their offspring will achieve a
higher fitness and thus get closer to an optimum solution.
Another way to increase the fitness of a population is to use a so called penalty function
(Smith & Coit, 1995). The idea of a penalty function is to penalize bad or unfeasible solutions
so that the fitness of the population is not decreased. The penalty function can be seen as a barrier
that hinders solutions that violates the constraints from being considered. Once a solution violates
the constraints the penalty function changes the fitness value of the solution so it gets a really low
fitness, thus hindering the solution from being considered.
If there is a small variety in a population there is a risk of inbreeding. In the terms of evolution-
ary computing, this means that the solution would not converge and the iteration hits a dead-end
(Simon, 2013). To avoid inbreeding, some sort of mechanism is needed to apply more diversity
to the population (Camp et al., 1998). Thus the concept of mutation is introduced, which in the
context of GA entails changing one of the parameters in the offspring when cross-over takes part,
creating new unique individuals (Simon, 2013). This creates a larger diversity in the population and
reduces the risk of inbreeding. A balance is sought to avoid inbreeding, having a too small rate of
mutation, and to bypass the behaviour of a random number generator when having a too large amount
of mutation. The diversification will affect the convergence of a GA, but it is also highly dependent
on the choice of population size, fitness function and other parameters (Gandomi et al., 2013a). GAs
are commonly robust but take long time to reach a global optimum, meaning a slow convergence rate
(Mitropoulou, Fourkiotis, Lagaros, & Karlaftis, 2013).
Other than GA there are several other evolutionary algorithms, for example evolutionary strategies
and evolutionary programming (Simon, 2013). All algorithms are based on the same framework but
have different foci on the population size, the recombination, the extension of mutation and fitness
evaluation for instance.

Problems in civil engineering are seldom so easy that they can be defined by one simple goal and thus
the need arises for multi-objective optimization algorithm (MOA) (Simon, 2013). Often EAs are
modified and used in multi-objective optimization (MO) (Mitropoulou et al., 2013). When optimizing
single-objective problems the goal is to find the best solution to a single objective (Tapabrata, 2018).
Even though the process might be straight forward, there are a lot of parameters that need to be
taken into account even when using single-objective optimization (Simon, 2013). Logically, for a
multi-objective optimization problem (MOP) the complexity is further increased. In MOP there
is no single solution that is optimal but rather several solutions that are compromises of different
objectives (Mitropoulou et al., 2013). To validate if a specific combination of solutions in a MOP
is good enough, a different approach to standard single-objective optimization is needed. In MOP
the best trade-off between several objectives is explained using a set of Pareto-optimal solutions
(Martini, 2015).
Looking at Figure 3.2, the Pareto-optimal solutions can be explained (Martini, 2015). The two axes
represent two objective functions, 𝑓1 and 𝑓2 , and the points (A to E) show different solutions of the
problem (Martini, 2015), as there is hardly ever just one single solution that will be an optimum when
using MOAs (Mitropoulou et al., 2013). The set of solutions are often compromises between different
objectives, which implies that a change of an objective cannot be increased without decreasing another
objective. As the minimum value is sought in the example, see Figure 3.2, B is a better solution
than C, and D is better than E, having the smaller value in either one or both objective functions, i.e
lower fitness (Martini, 2015). This implies that B is dominant compared to C, and D dominates E,
thus solutions C and E are called dominated solutions. Looking at the solutions B and D, neither
of them dominates the other, showing an example of non-dominance. The solutions A, B and D
are called non-dominated front or Pareto-optimal set, as they represent a set of solution, which are
non-dominant towards each other.
Figure 3.2: A graph explaining the Pareto-optimal set, which is represented by the black points, A, B
and D.
There are classical gradient based methods which can solve MOPs, thus not very efficient yet,
which can result in a single solution instead of a set of solutions (Tapabrata, 2018). Gradient based
methods have difficulties handling the shape of the Pareto-optimal set, if the shape is concave

(Coello Coello, 2006) or if the function is discontinuous (Tapabrata, 2018). EAs do not have this
problem and can handle the convex shape of the Pareto-front, which is one advantage of EAs in MOP,
also known as multi-objective evolutionary algorithms (MOEA).
Even though non-dominated solutions have been found in a generation it is not sure that this solution
is the best for another generation (Coello Coello, 2006). There is a risk that good solutions can be
lost in the process of mutation or cross-over, making the new offspring having lower fitness than
the parents (Gro, Oltean, & Oltean, no date). To make sure that the best non-dominated solution is
always found and kept some MOEA have a method for storing non-dominated solutions in a archive,
also know as an secondary population, from a generation and comparing these to other generations.
This concept is called elitism and is used to find the best non-dominated solutions for all generations.
If better individuals are found, the secondary population is updated with these individuals.
One of the more advanced MOEAs that uses elitism are the strength Pareto evolutionary algorithm 2
(SPEA-2) (Coello Coello, 2006). In addition to the archive with previously found non-dominated
solutions SPEA-2 uses a ranking system called strength. The strength is calculated based on how
many other individuals an individual is dominating as well as how many individuals are dominating
it. The fitness is calculated based on the strengths of all non-dominated solutions in the current and
secondary population. When calculating the fitness, SPEA-2 also considers how close the solutions
are to the Pareto-front as well as the distribution of individuals. This is used to keep the individuals
evenly distributed on the front. The algorithm also preserves solutions that are boundary solutions.
Another advanced MOEA is the hypervolume estimation algorithm for multi-objective optimization
(HypE) (Bader, Zitzler, & Rudolph, 2010). HypE optimization methods works similarly to SPEA-2,
but instead of using the strength of the solutions to determine the fitness, it uses the hypervolume,
meaning it measures the area underneath the Pareto-front, which represents the area dominated
by the Pareto-optimal solutions (Bader & Zitzler, 2011). Since HypE wants to find the largest
hypervolume it turns the multi-objective optimization problem into a single-objective problem, as
only the hypervolume is optimized. The unique thing about hypervolume calculation is that it is
non-decreasing, monotonic, with respect to Pareto-dominance, meaning that a Pareto-optimal front is
achieving the largest hypervolume. The main drawback of hyper volume based optimization methods
is that they are highly demanding with regard to computer power.
3.2 Parametric design

In parametric design the geometry of a structure is based on parameters that can be changed
(Dominik Holzer, Richard Hough, 2007). This is opposed to static analysis where the input values will
remain the same through the design process. Using parametric design the variables and algorithms
used allows the designer to quickly change the layout of the design by changing the initial parameters
without redoing the design.
The benefit of using parametric design tools to integrate the work of both architects and struc-
tural engineers is that the structural optimization can be done during the structural design phase
(Dominik Holzer, Richard Hough, 2007). In this thesis different parametric design software are used
to optimize steel truss bridges, which are explained in the following chapters.

3.2.1 Rhinoceros
Rhinoceros (Rhino) is a 3D modeling tool commonly used by architects and designers in the early
design phase (McNeel, 2018). Rhino uses non-uniform rational b-spline (NURBS) which are
mathematical representations of a 3D geometry. NURBS are very flexible and uses little information
to represent a shape (Ruiz-Geli, 2017). In this thesis Rhinoceros version 5.0 has been used. In
Figure 3.3, the different software used in this thesis are displayed. It can be seen that Rhino is the
basis software, having a first plug-in to Rhino called Grasshopper. As plug-ins into Grasshopper,
Karamba 3D, Galapagos and Octopus are used. The different plug-ins are explained in the subsections
that follow.
Figure 3.3: Schematic overview of the software that form the parametric design script.
3.2.2 Grasshopper
Grasshopper is a visual algorithm editor and is widely popular due to the ease with plug-ins which can
be used in the software; some plug-ins being Karamba 3D, Octopus and Galapagos, see Figure 3.3
(Davidson, 2018). Unlike other programming languages like Python or C++, Grasshopper is a visual
programming language. This implies that no scripting is needed, even though possible; instead
components and outputs are created on a canvas and connected together with inputs, see Figure 3.4a.
(a) (b)
Figure 3.4: (a) Example of how a truss structure can be created in Grasshopper. (b) Truss geometry
output, visualized in Rhino.
As can be seen in Figure 3.4a the inputs to the left, which are changeable - in this case starting points
for the truss - are connected via components, to achieve the wanted output in the end. The middle
part in white represents all the components that are used to create the geometry of the truss. To the

right, in green, there is a geometry component containing the final truss geometry. The generated
truss geometry can be seen in Figure 3.4b. Grasshopper version August 27 2014 build 0.9.0076 has
been used in this thesis.
3.2.3 Karamba 3D
Karamba 3D is a structural analysis plug-in to Grasshopper which makes it possible to combine
the parametric design environment of Grasshopper with structural analysis (Preisinger & Bollinger-
Grohmann-Schneider, 2018). Karamba 3D takes the geometry created in Grasshopper and translates
it into beams, shells, joints and other cross sections. Like in any structural analysis, supports as
well as loads can be specified. All the parts can thereafter be assembled and analyzed with first or
second-order theory. Karamba 3D can handle phenomenons like buckling and natural frequency. It
contains also a result viewer where displacements and reaction forces can be displayed. In this thesis
Karamba 3D version 1.3.0 WIP (180304) is used.
Some known limitations with Karamba 3D are pointed out in the Karamba 3D manual and listed
below (Preisinger, 2016):
• Karamba 3D does not take global buckling into consideration.
• 𝑘𝑧 and 𝑘𝑦 are set to 1 for lateral torsional buckling
• The interaction factors, used in EC3 to calculate interaction between bending moment and
normal force, 𝐶𝑚𝑦 , 𝐶𝑚𝑧 and 𝐶𝑚𝐿𝑇 , are restricted to values between 0.9 and 1.
• It is not possible to view global forces, like the bending moment across the whole structure.
Instead Karamba 3D displays the internal forces.
3.2.4 Galapagos and Octopus

Galapagos is also a plug-in to Grasshopper that can optimize one parameter using single-objective
EA. Since Galapagos comes together with Grasshopper it uses the same version as Grasshopper. In
Galapagos, see Figure 3.5a, the user can first specify a genome which is the parameters that can be
changed to optimize the structure. Secondly, the fitness can be chosen, for instance minimizing the
vertical displacement. Galapagos then uses GA explained in Section 3.1.2 to generate a population
and trying to optimize the fitness function.
In Galapagos there are some settings that can be changed to specify the speed or accuracy of the
optimization, see Figure 3.5b. One of these settings are the size of the population, a large population
means larger possibility to find local maximum and minimum but it increases the computation time.
Another setting is the initial boost which is a factor that sets the population of the first iteration to
a larger value to cover more of the design space. Drawbacks with a large initial boost is increased
computation time. However, it reduces the chance of Galapagos being stuck in local maximum or
minimum. Finally, the number of max stagnant means after how many iterations Galapagos stops if
it does not find a better fitness value.

(a) (b)
Figure 3.5: (a) Galapagos solver (b) Settings in the Galapagos editor.
Octopus is another optimization software, that has two different optimization algorithms, SPEA-2
and HypeE both explained in Section 3.1.2. Since both SPEA-2 and HypE are MOEA:s, they have
the possibility to optimize several fitness functions at the same time. Thus, Octopus allows the user
to find the best trade off between several goals. Worth mentioning is that since Octopus uses a MOA,
there will not be a single optimal solution but many solutions that are all optimal. Thus, it is up to
the designer to chose which of these solutions that should be chosen. Since Octopus is a plug-in to
Grasshopper the optimization can be used in combination with the structural analysis in Karamba 3D.
For the scope of this report SPEA-2 is used for the multi-objective optimization in Octopus. HypE is
disregarded since it is highly demanding regarding computational power. The version of Octopus
used in this thesis is version 0.3.6.
Octopus has several settings that can be adjusted to change the behaviour of the optimization algorithm,
see Figure 3.6. As for Galapagos, the first generation in Octopus starts with a larger population, in
this case twice the preset population. The elitism setting means how much of the non-dominated
solutions are stored into the secondary population. Mutation probability controls how likely it is
that an individual mutates. The cross-over rate determines how much of the genes from the parent
solutions is combined. A cross-over rate of 0.5 means half of the genes come from either parent.
Finally, the size of the population can be chosen.
Figure 3.6: Octopus solver with solutions marked as cubes and the settings listed to the right.

4 Development and verification of the model in
Karamba 3D
In this chapter a steel truss bridge over Genevadsån, an accomplished case from Ramböll, is presented
and the development of a parametric model with the same conditions as the case, created in Rhino,
Grasshopper and Karamba 3D, is described. In a next step the developed model is compared with
the FEM-model from the case study with regard to sectional forces, natural frequency, maximum
displacement of the bridge and utilization ratio of specified members of the bridge. Subsequently,
two additional geometries are implemented and all the verification results are discussed.
4.1 Description of the case which the model is to be verified with

The considered case is a steel truss footbridge built over the stream Genevadsån, see Figure 4.1, which
is located just south of Halmstad along side the highway E6 and represents a part of the planned cycle
path between Gothenburg and Helsingborg - the so called Kattegattsleden (Alexandersson, 2014).
The truss bridge was built in a factory and placed on site 14th of February 2018.
Figure 4.1: Reference project over the stream Genevadsån, picture from Ramböll (2018).
The bridge is a one span Warren truss bridge in steel and has a span length of 20 m, a width of 2.7 m
and a height of 1.4 m. Hot-rolled hollow profiles (VKR) are used for the structural elements, except
for the cross beams, for which IPE profiles are used. The bridge deck is a 10 mm thick steel plate
and the steel quality for all elements is S355. The end cross beams are constructed to withstand a
replacement of the bearings used for the supports. Further details are shown in Appendix B.

The reference bridge has been modeled in FEM-Design by Ramböll, see Figure 4.2, and to some
extent simplified and adjusted to the focus of this thesis. FEM-Design is an advanced Finite el-
ement modeling (FEM) tool that allows the user to do detailed structural analysis and design
(StruSoft, 2018) and is based on Eurocode with the accompanying national adjustments. All structural
elements are modeled as steel beams except the bridge deck which is modeled as a shell structure. In
the adjusted FEM-Design model all the members are fixed and the support conditions of the bridge
are simply supported, see Figure 4.2, where the directions locked are shown as arrows. In this model
the self-weight and crowd load are included as well as a line load from the railings of 0.5 kN/m
along each bottom chord. This model is used to retrieve sectional forces, maximum displacements,
natural frequencies and the utilization ratios, which are then compared with the model developed in
Karamba 3D, described in the next section.
Figure 4.2: The reference bridge modeled in FEM-Design. The distances are shown in meters.
4.2 Development of the Warren sub model in Karamba 3D

In this section the geometry and the chosen structural properties of the Warren steel truss sub model
in Karamba 3D are described more in detail. The loads and load combinations are also presented
concluding in the assembly of all parts into the final model.
4.2.1 Development of the geometry in Grasshopper

The first step when creating a parametric model is defining the geometry of the studied bridge. From
the presented bridges in Figure 2.3, it was decided to go further with the Warren truss bridge as it
is a common footbridge type and since the reference bridge presented in Section 4.1 is a Warren
truss bridge. The geometry of the bridge is created in Grasshopper using components as explained
in Section 3.2.2. Starting with a single point in the origin (0/0/0) the geometry is created with
components, duplicating and moving points into a final geometry, seen in Figure 4.3.

Figure 4.3: The final geometry of the Warren truss footbridge in Grasshopper. The support conditions
are shown as arrows (locked direction) and the number of n𝑐𝑏 , within one diagonal triangle, are shown
as numbers 1 to 5.
To adjust the appearance of the bridge, the designer can define the input parameters as the length,
width and height of the bridge. Additionally the number of diagonal triangles (n△ ) and number of
spacings between the cross beams (n𝑐𝑏 ) within a diagonal triangle can be defined by the designer. It
is arranged that the designer first chooses the n△ , which are defined as 5 in the case of Figure 4.3.
Afterwards n𝑐𝑏 can be set, which is also 5 in Figure 4.3. The same n𝑐𝑏 is then set at all the other truss
triangles. At the start and end diagonal triangle, n𝑐𝑏 is half of the amount as for the ones in the mid
sections. Should n𝑐𝑏 be an uneven number, the number is rounded up to the next number, i.e. to three
in the case of Figure 4.3.
The geometry is divided into six main components of the bridge: The four longitudinal chords, the
diagonals, the cross beams, the end cross beams, the vertical end posts and the bridge deck. The
complete geometry is used as an input to Karamba 3D, where the structural analysis is executed.
4.2.2 Structural analysis in Karamba 3D

Even though the bridge geometry is created in Grasshopper, the finished geometry still needs to be
translated into structural components and be given material properties. In Karamba 3D all the lines
are translated into beams, which includes the longitudinal chords, the diagonals, the cross beams and
the vertical end posts, which can be seen in Figure 4.4. The bridge deck, which is simulated as a
mesh between the four corner points is translated into a shell.
Karamba 3D offers the user a list of cross-sections to apply to the geometry. There is also a possibility
to add additional cross-sections, which is done in this case. VKR and circular hollow sections
(KCKR) are added into Karamba 3D by specifying cross-sectional parameters into an excel-sheet
where all cross-sections used in Karamba 3D are stored. The cross-sections used in this thesis for the
beam elements are IPE, VKR and KCKR, which are all adjusted and based on the input values from
(TIBNOR, 2011), which can be found in Appendix C. Worth mentioning is that the effective shear
areas in y- and z-direction respectively for the IPE profiles are considered so that 𝐴𝑦 only includes
the area of the flange and 𝐴𝑧 only includes the area of the web. The shear area for the VKR profiles
are calculated according to EN 1993-1-1 chapter 6.2.6. VKR and IPE cross-sections are included
in the model since they are used in the reference project. The circular cross-sections represent an
addition to broaden the range of cross-sections. The final geometry with applied cross-sections is
shown in Figure 4.4.

Figure 4.4: Final appearance of the Warren sub model in Karamba 3D with the the applied beam and
shell elements.
The material and material class for the elements is specified to steel S355 and the Shear and Young’s
modulus used are 81 GPa and 210 GPa respectively. The density of the steel is set to 7 850 kg/m3
and the gravitational constant to 9.82 m/s2 . Four supports are introduced in accordance to the case
study, which are locked in the directions of the arrows, seen in Figure 4.3.
4.2.3 Definition of loads and load combinations

In case of footbridges there are a several loads that need to be considered, which are - according to
Eurocode SS-EN 1991-2 table A2.2 - described in Table 4.1. The loads considered in this thesis are
the self-weight of the construction and the crowd load. The uniformly distributed load is considered
as a crowd load of 5 kN/m2 and is a recommended value according to SS-EN 1991-2. The vehicle
load representing a service vehicle is not included in this thesis, the reason being that it is not always
considered, which is the case for the case over Genevadsån. Additionally, the wind and temperature
load are not studied as the thesis is focusing on the global preliminary analysis. According to
Eurocode, snow loads need not to be combined with load groups gr1 and gr2 and are therefore
neglected.
Table 4.1: An overview of the included loads in this thesis.

Loads Description Symbol Considered in this thesis
Permanent load Self-weight of construction 𝑄𝑐 Yes
Load group 1 gr1 Yes
Traffic load Concentrated load 𝑄𝑓 𝑤𝑘 No
Load group 2 gr2 No
Variable load
Snow load 𝑄𝑠𝑛𝑘,𝑘 No
Wind load 𝐹𝑊 ,𝑘 No
Temperature load 𝑇𝑘 No

In Karamba 3D the self-weight of the structure can be retrieved automatically as the chosen element
material and geometry is defined. However, the crowd load is simulated as line loads on each cross
beam, which should then carry the load to the bottom chords and finally to the supports. Each cross
beam is defined to take the load of the area with half the distance to the next cross beam on each side,
times the width of the bridge.
The load combinations considered are ULS 6.10a, ULS 6.10b and SLS 6.15a-b (frequent) and the
bridge is considered to be in safety class 3. The recommended values for the 𝜓-factors and the
𝛾-values for the permanent and variable loads are shown in Table 4.2.
Table 4.2: Relevant factors for the permanent and variable loads.
𝜓-factors
Permanent loads
Max Min
Self-weight of construction,
1 1
characteristic nominal value
Comb Freq Quasi

Variable load 𝛾𝑄
𝜓0 𝜓1 𝜓2
Load gr1:
Uniformly distributed
0.4 0.4 0 1.5
load (𝑞𝑓 𝑘 ) +
horizontal forces (𝑄𝑓 𝑙𝑘 )*
*neglected in this thesis
The final factors for the permanent and the variable loads for each load combination are shown in
Table 4.3. Primary loads are the leading variable actions, i.e. the load is the main load. The secondary
loads are accompanying variable actions, meaning that another load is the main load.
Table 4.3: Factors for permanent and variable loads in the three load combinations considered.
6.10a ULS 6.10b ULS 6.10ab SLS
Permanent loads
Max Min Max Min Max Min
𝛾𝐺 - factors 1.35 1 1.2 1 1 1
Primary Secondary Primary Secondary Primary Secondary

Variable loads
𝛾𝑄 • 𝜓0 𝛾𝑄 • 𝜓0 𝜓0 𝛾𝑄 • 𝜓0 𝜓1 𝜓2
Load gr1:
Uniformly distributed
0.6 0.6 1.5 0.6 0.4 0
load (𝑞𝑓 𝑘 ) +
horizontal forces (𝑄𝑓 𝑙𝑘 )*
*neglected in this thesis

The load combinations used in Karamba 3D are implemented in Grasshopper with the factors given
in Table 4.3. The load combinations shown in Table 4.4 are all possible combinations for the loads
used. The worst load combination is 6.10𝑏1 , which is the one used for all the verification results in
the next section.
Table 4.4: All load combinations with the factors for the self-weight and crowd load respectively,
which are used in the created model in Karamba 3D.
Load combination Type Name Self-weight Crowd load
6.10𝑎1 1.35 0.6
6.10a ULS
6.10𝑎2 1 0.6
6.10𝑏1 1.2 1.5
6.10𝑏2 1.2 0.6
6.10b ULS
6.10𝑏3 1 1.5
6.10𝑏4 1 0.6
6.10𝑎𝑏1 1 0.4
6.10a-b SLS
6.10𝑎𝑏2 1 0
4.3 Verification of the Warren sub model in Karamba 3D with

the case study in FEM-Design
To see if the bridge developed in Karamba 3D is working properly it needs to be verified. First, the
similarity of the input data of the developed model in Karamba 3D and the model in FEM-Design of
the case study, see Section 4.1, is stated, which can be seen in Table 4.5, and is in accordance to each
other.
Table 4.5: Input data of the developed Warren sub model, which is in accordance to the input data of
the FEM-Design model of the case study.
Element Cross-section Dimensions [mm] Steel quality Quantity
Top chord VKR 150x100x6.3 S355J2H 2
Bottom chord VKR 150x100x6.3 S355J2H 2
Diagonals VKR 100x60x5 S355J2H 20
Vertical end post VKR 100x100x5 S355J2H 4
Cross beam IPE 140 S355J2 25
End cross beam IPE 180 S355J2 2
Bridge deck Plate (LxWxH) 20 100 x 2 500 x 10 S355J2 1
Further, the structural analysis is of interest and to be compared for the two software. Using the
components in Karamba 3D, a first order, plastic analysis is made to calculate the reaction and
sectional forces, the maximum displacement and the utilization ratios of the bridge, which are
compared with the ones from the FEM-model from the case study. Also the natural frequency is
verified with the results from FEM-Design and hand calculations. As a first attempt to verify the
results of the two models, the reaction forces of a simple equilibrium of the bridge is investigated,
which can be seen in Table 4.6. The difference between the results is rather small and sets the basis
for further verification of the results.

Table 4.6: Comparison of the reaction forces of the two models in a first attempt of verification.
Reaction forces [kN]
Load applied Karamba 3D FEM - Design Difference
Self-weight 24.8 24.4 2%
Crowd load 62.5 62.5 0%
In a next step the normal force 𝑁, the moment 𝑀𝑦 and the shear force 𝑉𝑧 of the two models are
compared. This is executed with the preconditions that the railing of 0.5 kN/m is included in the
permanent load and load case 6.10𝑏1 , see Table 4.4, is used. In this case, 6.10𝑏1 is chosen as it is the
worst case in ULS when only considering the crowd load. The choice of load combination is of no
importance in the verification, as long as it is the same in the two software. Due to symmetry, only
one half of the bridge is investigated and its results presented. First, the normal force diagram of the
bottom chord is displayed in Figure 4.5 and the associated numbers shown in Table 4.7.
Figure 4.5: Normal force diagram of the bottom chord from Karamba 3D.
Table 4.7: Comparison of the absolute values of the normal force from position a to f of the bottom
chord from Karamba 3D with FEM-Design.
Normal force N [kN] (absolute values)
Member Position Karamba 3D FEM-Design Difference
a 1.0 1.8 -42 %
b (left) 31.1 91.3 -66 %
Bottom b (right) 77.6 141.5 -45 %
chord d (left) 32.0 0.5 6 709 %
d (right) 86.7 122.7 -29 %
f 71.6 73.9 -3 %
As can be seen in Table 4.7, the normal forces in the bottom chord do not seem to match, which
indicates a difference between the two models. The difference in results originates from the behaviour
of the bridge deck. In both software, the bridge deck is included with the same supporting conditions,
namely fixed. Reasoning that the design of the bridge deck is of no interest when optimizing, focusing
on the structure and behaviour of the truss footbridge, it is decided to weaken the bridge deck. When
dividing the Young’s and Shear modulus with a weakening factor, the stiffness of the bridge deck is
weakened, but the mass of it remains. In Figure 4.6 a convergence study is presented, showing the
influence of the stiffness of the bridge deck on the normal force at position f in the bottom chord by
increasing the weakening factor.

Figure 4.6: Convergence study of the normal force in the bottom chord at position f for different
weakening factors in the bridge deck.
As can be seen in Figure 4.6, the normal force at position f of the bottom chord seems to start
converging with a weakening factor 1 000. Due to this fact, being somewhat on the safe side, a
weakening factor of 10 000 is henceforth chosen for all results, if not mentioned otherwise. A more
detailed investigation of all normal forces with different weakening factors is found in Appendix D.
The new results of the normal force in the bottom chord with a weakening factor of 10 000 is shown
in Figure 4.7 and Table 4.8. As can be seen, the results now cohere well to each other, having the
largest difference of 4 %.
Figure 4.7: Normal force diagram of the bottom chord from Karamba 3D with a weakening factor
10 000.
Table 4.8: Comparison of the absolute values of the normal force at position a to f in the bottom
chord from Karamba 3D with FEM-Design with a weakening factor 10 000(*).
Normal force N [kN] (absolute values)
Member Position Karamba 3D, 10 000* FEM - Design, 10 000* Difference
a 3.2 3.0 4%
Bottom
c 266.3 270.1 -1 %
chord
f 403.0 411.2 -2 %

Secondly, the moment distribution in the bottom chord is analyzed and presented in Figure 4.8 and
the results, compared to the ones in FEM-Design, are shown in Table 4.9. The results are quite in
accordance, having the largest deviation of 6 % at position a.
Figure 4.8: Moment diagram of the bottom chord from Karamba 3D.
Table 4.9: Comparison of the absolute values of the moment at position a to f in the bottom chord
from Karamba 3D with FEM-Design.
Moment My [kNm] (absolute values)
Member Position Karamba 3D FEM - Design Difference
a 2.5 2.3 6%
b 10.3 10.4 -1 %
Bottom c 10.2 10.5 -3 %
chord d 14.0 14.4 -3 %
e 9.1 9.3 -2 %
f 12.2 12.6 -3 %
Thirdly, the shear forces of the bottom chord are investigated, shown in Figure 4.9 and the results and
comparison with FEM-Design displayed in Table 4.10. As can be seen, these results also cohere well
to each other, having the largest deviation of 5 %.
Figure 4.9: Shear force diagram of the bottom chord from Karamba 3D.

Table 4.10: Comparison of the absolute values of the shear forces at position a to f in the bottom
chord from Karamba 3D with FEM-Design.
Shear force Vz [kN] (absolute values)
a 6.0 5.7 4%
b (left) 12.3 11.7 5%
Bottom b (right) 17.7 18.3 -3 %
chord d (left) 19.6 20.3 -4 %
d (right) 19.0 19.7 -4 %
f 18.2 18.9 -3 %
In Table 4.11, the maximum displacement downwards in the middle of the Warren sub model (position
f see Figure 4.9) can be seen. Worth mentioning is that still load case 6.10𝑏1 (ULS) is applied when
retrieving the results for the comparison. The displacements between the two models cohere well to
each other. Further results of the sectional forces of all elements and maximum displacements with
an unweakened bridge deck (weakening factor 1) are shown in Appendix D.
Table 4.11: Comparison of the absolute values of the maximum displacement at position f in the
bottom chord from Karamba 3D with FEM-Design.
Max displacement [mm] (absolute values)
All f 52.2 53.2 -2 %
Considering the utilization ratio of the bridge members, a small study is executed to verify if the
utilization component in Karamba 3D works as intended. In an attempt to verify the utilization ratio,
the results from Karamba 3D are compared with the ones from FEM-Design from the case study.
The utilization ratios are calculated according to SS-EN 1993-1-1, sections 6.2-6.3. In both Karamba
3D and FEM-Design following controls are considered, see Table 4.12.
Table 4.12: Considered controls in Karamba 3D and FEM-Design.

Control Chapter in EN 1993-1-1
Normal stress 6.2.1
Normal capacity 6.2
Flexural buckling 6.3.1
Torsional-flexural buckling 6.3.1
Lateral torsional buckling 6.3.2.2
Interaction between normal force
6.3.3
and bending, 1st and 2nd order
Shear resistance 6.2.6, 6.2.8
Torsional resistance 6.2.7
When looking at all elements, the worst maximum utilization ratio for each element is detected in
both Karamba 3D and FEM-Design and presented in Table 4.13.

Table 4.13: Comparison of the largest utilization ratios for each element, based on load case 6.10𝑏1 ,
from Karamba 3D with FEM-Design.
Maximum utilization ratio
Element Karamba 3D FEM-Design Difference
Bottom chord 62 % 67 % -6 %
Top chord 51 % 53 % -2 %
Diagonals 65 % 60 % 5%
Vertical end posts 31 % 30 % 1%
Cross beams 18 % 47 % -29 %
End cross beams 9% 15 % -7 %
Looking at Table 4.13, the largest difference of the maximum utilization ratios between the two
models is detected to be for the cross beams, which is why they are investigated further. In the case of
load combination 6.10𝑏1 , see Table 4.4, the maximum utilization ratios of the normal, moment and
shear resistances of the mid cross beam at position f are compared between the two models, which
can be seen in Table 4.14.
Table 4.14: Maximum utilization ratios of each sectional force and the difference between them. The
load case applied is 6.10𝑏1 and the element considered is the mid cross beam at position f.
Maximum utilization
Member Karamba 3D FEM-Design Difference
Normal force N 0% 0% 0%
Moment My 18 % 47 % -29 %
Moment Mz 1% 0% 1%
Shear force Vy 0% 0% 0%
Shear force Vz 7% 5% 2%
Additionally, the details retrieved from the utilization component in Karamba 3D are compared with
the corresponding ones from FEM-Design, which is shown in Table 4.15, still looking at the mid
cross beam with load case 6.10𝑏1 . During this step, the buckling length of the cross beams, including
the end cross beams, in x- and y-direction is detected to be 0.27 m in Karamba 3D. This is wrong
and therefore changed to 2.7 m in both x- and y-direction, which is the same as in the FEM-Design
model. The calculation procedure from FEM-Design of the cross beam at position f can also be
found in Appendix D.2. In Table 4.15 the normal resistance (𝑁𝑅𝑑 ), critical normal force capacity
(𝑁𝑐𝑟 ) moment resistance (𝑀𝑦𝑅𝑑 ), critical moment capacity (𝑀𝑐𝑟 ), shear resistance (𝑉𝑦𝑅𝑑 , 𝑉𝑧𝑅𝑑 ), the
lateral torsional reduction factor 𝜒𝐿𝑇 and all interaction values (𝑘𝑦𝑦 , 𝑘𝑧𝑧 , 𝑘𝑦𝑧 , 𝑘𝑧𝑦 ) of the cross beam,
with adjusted buckling length, at position f from Karamba 3D and FEM-Design are presented.

Table 4.15: Results of the normal resistance (𝑁𝑅𝑑 ), critical normal force capacity (𝑁𝑐𝑟 ), moment
resistance (𝑀𝑦𝑅𝑑 ), critical moment capacity (𝑀𝑐𝑟 ), shear resistance (𝑉𝑦𝑅𝑑 , 𝑉𝑧𝑅𝑑 ), the lateral torsional
reduction factor (𝜒𝐿𝑇 ) and all interaction values (𝑘𝑦𝑦 , 𝑘𝑧𝑧 , 𝑘𝑦𝑧 , 𝑘𝑧𝑦 ) of the cross beam at position f
from Karamba 3D and FEM-Design after adjustment of the buckling lengths.
Karamba 3D FEM - Design Difference
𝑁𝑅𝑑 [kN] 591.5 583.1 1%
𝑁𝑐𝑟 [kN] 127.9 127.7 0%
𝑀𝑦𝑅𝑑 [kNm] 31.8 31.4 1%
𝑀𝑐𝑟 [kNm] 18.0 16.1 11 %
𝑉𝑦𝑅𝑑 [kN] 209.4 215.1 -3 %
𝑉𝑧𝑅𝑑 [kN] 123.3 156.6 -21 %
𝜒𝐿𝑇 0.51 0.42 21 %
𝑘𝑦𝑦 1 1 0%
𝑘𝑧𝑧 1.00 1.03 -3 %
𝑘𝑦𝑧 0.6 0.8 -27 %
𝑘𝑧𝑦 1.0 0.5 92 %
As can be seen in Table 4.15 there are still some differences, considering the critical moment capacity
(𝑀𝑐𝑟 ), shear resistance (𝑉𝑧𝑅𝑑 ), the lateral torsional reduction factor 𝜒𝐿𝑇 and the interaction values 𝑘𝑦𝑧
and 𝑘𝑧𝑦 . Nevertheless, the maximum utilization values for the cross beams and the end cross beams
result in somewhat better ratios after all, see Table 4.16. Also, the buckling length of the bottom
chord was detected to be wrong and changed from 0.8 m to 4 m in z-direction, which also resulted in
somewhat better results than before the adjustment, see Table 4.16.
Table 4.16: Comparison of maximum utilization ratios for the bottom chord, cross beams and end
cross beams in Karamba and FEM-Design with adjusted buckling lengths.
Element Karamba 3D FEM-Design Difference
Bottom chord 64.0 % 67.0 % -3 %
Cross beams 35.3 % 47.0 % -12 %
End cross beams 14.9 % 15.0 % -0 %
To be able to have a look at the natural frequency, the bridge deck needs to be stiff, i.e. the weakening
factor needs to be 1 - only for the verification of the natural frequency. As can be seen in Table 4.17, the
results from Karamba 3D are compared with the ones from FEM-Design and with hand calculations,
which can be found in Appendix E. The results seem to cohere well to each other, especially the
results from Karamba 3D compared to the hand calculations. The shapes of the first modes from
Karamba 3D and FEM-Design are shown in Figure 4.10 and look the same.

Table 4.17: Natural frequency for mode 1 from Karamba 3D compared with the results from FEM-
Design and hand calculations.
Natural frequency Natural frequency
mode 1 [Hz] mode 1 [Hz]
Element Karamba 3D 1* FEM-Design Difference Karamba 3D 1* Hand calculations Difference
All 6.7 7.0 -3 % 6.7 6.7 0%
(a)
(b)
Figure 4.10: The natural frequency (mode 1) of the Warren sub model in Karamba (a) and FEM-
Design (b).
4.4 Development and verification of the Pratt and Howe

sub models
To give the designer a variety of truss bridge designs to chose from, two additional geometries are
added to the script in Grasshopper. As the two most common trusses, the geometries of the Pratt and
Howe truss, constituting two additional sub models in Karamba 3D, are implemented. The process
of implementing these two extra geometries is similar to the one for the Warren truss explained in
Section 4.2.1. At the beginning of the Grasshopper script there is an option to choose which of
the three trusses to do the calculations on. Both the Pratt and Howe truss have been verified with
a created FEM-Design model in the same way as for the Warren truss, see Section 4.2, having a
weakening factor 10 000 for the bridge deck and using load case 6.10𝑏1 . Additionally the same
perquisites as for the case study are used, i.e. the length is 20 m, the width 2.7 m and the height 1.4
m. The cross-sections of the members are the same as in Table 4.5 and n△ is 5 and n𝑐𝑏 is 5. For the
Pratt and Howe truss sub models n△ can only be an even number. This implies, that if an uneven
number is set, it is rounded up to the next number.

Starting with the verification of the Pratt sub model, the sectional forces, the maximum displacement
of the bridge and utilization ratio of the elements of the bridge in Karamba 3D are - accordingly to
the verification of the Warren truss bridge - compared to the created FEM-Design Pratt model. In
Figure 4.11 the shape of the Pratt truss is presented, followed by results for the reaction force of one
support and the maximum displacement of the bridge from both Karamba 3D and FEM-Design, see
Table 4.18. The difference of both results seem to coincide well to each other.
Figure 4.11: The Pratt truss bridge with position identifications.
Table 4.18: Reaction force of one support and the maximum displacement of the Pratt truss sub
model.
Description Load Karamba 3D FEM-Design Difference
Reaction force (1 support) [kN] Self-weight 20.4 20.7 -2 %
Max. displacement [mm] Load case 6.10𝑏1 57.9 57.9 0%
In Table 4.19, the maximum utilization ratios for the different members of the bridge are presented
for both Karamba 3D and FEM-Design. All the results of the sectional forces of the members can
be found in Appendix F and are not shown in this chapter as the results are very satisfying. The
results in Table 4.19 seem to correspond well to each other, having the largest difference of 6 % for
the utilization ratio of the cross beams.
Table 4.19: Maximum utilization ratios for each member with the perquisites of the case study and
load case 6.10𝑏1 for the Pratt truss sub model.
Top chord 47 % 48 % -1 %
Diagonals (a-b) 52 % 53 % -1 %
Vertical end posts (a) 52 % 53 % -1 %
Cross beam (f) 31 % 37 % -6 %
End cross beams 13 % 14 % -1 %
In a next step, the Howe truss sub model in Karamba 3D is verified with the created FEM-Design
Howe model. For this model all the results including the sectional forces of the different members
are presented in Appendix G. The sectional forces proved also in this case to correspond well to each
other. The Howe truss is shown in Figure 4.12 and the results of the reaction force in one support and
the maximum displacement of the bridge is displayed in Table 4.20. The maximum displacement
and reaction force from both software agree well to each other.

Figure 4.12: Position identifications of the truss elements for the Howe truss bridge.
Table 4.20: Reaction force of one support and the maximum displacement of the Howe truss sub
model.
Description Load Karamba 3D FEM-Design Difference
Reaction force (1 support) [kN] Self-weight 20.4 20.9 -2 %
Max. displacement [mm] Load case 6.10𝑏1 61.0 60.7 1%
As for the Pratt truss sub model, the maximum utilization ratio of all elements for the Howe truss
sub model is shown in Table 4.21. Also for this model, the differences between the results between
the two software cohere well to each other, having the largest difference again for the cross beams.
It needs to be added that the buckling length of the cross beams is adjusted to 2.7 m in both x- and
z-direction for all results in this section, as was previously done for the Warren truss.
Table 4.21: Maximum utilization ratios for each member with the perquisites of the case study and
load case 6.10𝑏1 for the Howe truss sub model.
Top chord 43 % 45 % -2 %
Diagonals (a-b) 229 % 226 % 3%
Vertical end posts (a) 36 % 37 % -1 %
Cross beam (f) 31 % 37 % -6 %
End cross beams 13 % 13 % 0%
4.5 Discussion of the verification results for the Warren, Pratt

and Howe truss sub models
The reaction forces from Karamba 3D and FEM-Design differ insignificantly, that is 2 % considering
the Warren, Pratt and Howe truss sub models. The reason for the deviation is probably due to the
small differences in the input data of the two software.
The normal forces in the bottom chord are the ones that are the most affected by the stiffness of the
bridge deck. The influence of the stiffness of the bridge deck on the normal force distribution on
the bottom chord of the Warren truss sub model can be seen by comparing Figures 4.5 and 4.7. The
normal forces in the bottom chord, having a weakening factor 1, vary a lot between Karamba 3D
and FEM-Design. However, results with a weakening factor 10 000, see Table 4.8, show a good
correspondence to each other between Karamba 3D and FEM-Design for the Warren truss sub model.

Considering the normal force in the bottom chord for the Warren, Pratt and Howe truss sub models
with a weakening factor 10 000, the results from Karamba 3D and FEM-Design coincide well with
each other, having the largest difference of 4 % at position a for the Warren truss bridge sub model,
which can be accepted as the values are rather small compared to the other results in Table 4.8.
The results of the bending moment My from Karamba 3D and FEM-Design at the positions a to f
correspond well to each other, see Table 4.9, thus having a difference of 6 % in the Warren truss sub
model at position a in the bottom chord. Yet again the results in the end span can be accepted since
the values are low compared to the rest of the bending moments in the bottom chord.
The results of the shear forces Vz between Karamba 3D and FEM-Design are also in good accordance
to each other, see Table 4.10. The largest difference between the two software considering the shear
force of the bottom chord is 5 % at position b (left). The difference of the results can be explained by
small differences in effective shear area for Karamba 3D and FEM-Design, due to the facts mentioned
in Section 4.2.2. Also more decimals of the input data of the cross-sections are implemented into
FEM-Design. This can explain some of the differences of the shear force along the bottom chord.
Another explanation for the small deviations of all the sectional forces between the results in
Karamba 3D and FEM-Design could be the difference in calculation points on a member.
Karamba 3D defines an element to be between two points. Having points created on for instance
the bottom chord, reassembling connection points for the cross beams with the bottom chord or the
connection points for the diagonals with the bottom chord, many small elements represent the bottom
chord as one member. For each element of a member only the results of the start and end points are
displayed in Karamba 3D, which are then connected linearly to each other, which can be seen in
for instance Figure 4.8 or Figure 4.9. This does not give accurate results in between the start and
end points of an element, only an average. FEM-Design on the other side calculates the sectional
forces in much more points, giving a more accurate picture of many more results of a member. Due
to this fact, for instance the results of the bending moment from Karamba 3D at position c in the
bottom chord, see Table 4.9, is actually not exactly at position c, but the largest value of the element
at position c. Knowing this, the values from Karamba 3D are already smaller than the one from
FEM-Design, which means that an exact calculation of the mean value at the exact position c would
not help for the better for Karamba 3D.
Further, looking at the normal force in the bottom chord with a weakening factor 1, the average value
of each small element at the given positions are compared to the peak points in FEM-Design, which
seems quite unfair to compare with. The overall shape of the normal forces of the bottom chord can
be found in Appendix D and are approximately the same in both Karamba 3D and FEM-Design,
which indicates that the same conditions are applied for the bridge deck in both software. The results,
as stated before, are differing though, due to the average values in Karamba 3D.
Looking at the displacement, the Warren truss sub model in Karamba 3D shows a larger maximum
displacement compared to FEM-Design. All in all it only represents a difference of 2 % corresponding
to 1 mm which is a very small value and insignificant. Considering the Pratt and Howe truss sub
models, the maximum displacement is 0 % and 1 % respectively and also negligible. One explanation
for the small deviations is that the self-weight differs about 2 % between Karamba 3D and FEM-
Design, which makes it acceptable that the displacements differ about 2 % as well.

The maximum utilization ratio for different members, see Table 4.13, first showed large differences
between Karamba 3D and FEM-Design, especially considering the cross beams. Investigating the
cross beams further, the utilization of the moment My turned out to be the issue. The utilization of
the moment My is dependent on among others the factor 𝜒𝐿𝑇 , which in turn is dependent on critical
moment capacity 𝑀𝑐𝑟 and the moment and interaction factors. The crucial cause why the results
differ is the way how the interaction values and moment factors are calculated. In FEM-Design
method 1 in Annex A in EN 1993-1-1 is used, while Karamba 3D uses method 2 in Annex B in EN
1993-1-1. Additionally the critical moment capacity relies on the critical normal capacity 𝑁𝑐𝑟 which
is dependent on the buckling length of the cross beams. Having the buckling length of the cross beams
between the start and end points of all elements, as mentioned before, the buckling length of the cross
beams in Karamba 3D was too short. Adjusting the buckling length to 2.7 m in both x- and z-direction,
the results in Table 4.16 showed somewhat better results. Still, the difference for the cross beams is
12 % and not to forget, there still is a difference of 5 % considering the maximum utilization ratio
of the diagonals. The differences can at this state only be explained by the fundamental difference
between Karamba 3D and FEM-Design when calculating the moment and interaction factors and is
only stated and not to be considered further in this thesis, even though some of the results are not
quite acceptable and on the unsafe side compared to the results in FEM-Design.
Regarding the natural frequency, the results show great accordance looking at the Warren sub model,
both compared to the results in FEM-Design and the hand calculations with a difference of 3 % and
0 % respectively. Unfortunately, this thesis does not comprise a further investigation of the natural
frequencies of the Pratt and Howe sub models and gives the opportunity to be developed further in
the future. Due to the fact that the weakening factor needs to be changed to 1 to retrieve the results
for the natural frequency, it also constitutes an issue when optimizing, as the factor needs to remain
one number, either 1 or 10 000. As the majority of the results rely on a weakening factor of 10 000,
this factor is used and the natural frequency is left behind, but can be checked separately, changing
the weakening factor to 1.

5 Development of the optimization features
Multi-objective optimization is advanced and it can be difficult to confirm what solution is the most
optimal one since the solution is a compromise of several solutions. To get a grip on what solutions
are reasonable for a multi-objective optimization, some optimization runs are executed using single-
objective optimization. This is why in the first phase of the optimization Galapagos is used, since it
is a plug-in designed for single-objective optimization. Once the optimization process works for one
fitness function several fitness functions can be introduced and then Octopus is used. This section
starts with describing the constraints used when optimizing and how they are implemented into
Karamba 3D and concludes in specific settings and features used when optimizing with Galapagos
and Octopus.
5.1 Constraints when optimizing

To avoid getting unreasonable solutions from a optimization, constraints are introduced. The con-
straints, see Table 5.1, are the maximum allowable displacement of the whole bridge (SLS), the
maximum utilization of the cross-sections (ULS) and a global lateral torsional buckling criteria for
the top chord (ULS), see Table 5.1.
Table 5.1: These are the used constraints when optimizing the Karamba 3D models.
Maximum displacement (SLS) L/400
Maximum utilization ratio for each the cross-section (ULS) 85 %
Maximum normal force in top chord (ULS) N𝐸𝐷 <N𝑏,𝑅𝐷
The displacement criterion is limited to L/400 according to Krav Brobyggande TDOK 2016:0204
section B.3.4.2.2. According to Krav TDOK 2016:0204 section B.3.4.2.1 only the frequent values
should be used when calculating the maximum displacement, implying that the self-weight is ignored
for this constraint. Only the frequent load, being the crowd load, is considered and multiplied with
the 𝜓1 -factor of 0.4, see Table 4.3. This also presumes that the bridge is precambered with the height
being the displacement that the self-weight alone would cause.
The utilization constraint is set to 85 % of the cross-section capacity, not 100 %, having a 15 % safety
buffer considering the somewhat unsatisfying results in Section 4.2.2 for the utilization ratio. Since
the biggest difference between the Karamba 3D model and the FEM-Design model in utilization ratio
is around 12 % the safety buffer of 15 % seems warranted.
As mentioned in Section 3.2.3 Karamba 3D only considers local buckling of elements. For through
truss bridges without top lateral bracing, global buckling of the top chord could be a limiting factor.
To include the phenomenon of lateral torsional buckling of the top chord into the model, which should
be considered according to the statement SS-EN 1993-2 section 6.3.4.2 equation (2), calculations are
executed according to (Wahlström, 1971), which are based on empirical results.
The calculations are included in the script in Grasshopper according to Appendix H, where the top
chord is modeled as a column loaded in axial compression with the maximum normal force N𝐸𝑑 of
the top chord, supported on a spring bed. The spring stiffness of the springs is calculated by applying

a unit load of 1 000 kN onto the truss system, leaving the top chord out, see Figure 5.1, which shows
the Warren sub model. The supports can also be seen in Figure 5.1, where the arrows reassemble
the directions which are locked. All joints are defined as rigid, except at the joint where the load is
applied. At this location the rotational condition is defined as free. The verification of the global
LT-buckling of the top chord is only executed on the Warren sub model. The corresponding figures
of Figure 5.1 of the Pratt and Howe sub models can be found in Appendix H.
Figure 5.1: The Warren sub model without the top chords. A unit load of 1 000 kN is applied to
calculated the displacement in y-direction.
To verify the displacement (𝛿) received after applying the unit load, the same is executed in FEM-
Design. The Warren sub models in Karamba 3D and FEM-Design have the same perquisites as in
the case study, see Section 4.2. The displacement in y-direction for the Warren truss bridge from
Karamba 3D and FEM-Design are presented in Table 5.2. The displacement for the Warren truss
in Karamba 3D seems to correspond well to the results from FEM-Design. The maximum reduced
normal capacity (𝑁𝑏𝑅𝑑 ), which is based on 𝛿, is also presented in Table 5.2. The maximum normal
force in the top chord, 𝑁𝐸𝐷 , must not exceed 𝑁𝑏𝑅𝑑 .
Table 5.2: The displacement in y-direction after applying a unit load of 1 000 kN and the maximum
normal capacity 𝑁𝑏𝑅𝑑 of the top chord.
Karamba 3D FEM-Design Difference
𝛿 in y-direction [m] 7.1 7.0 0%
𝑁𝑏𝑅𝑑 [kN] 507.0 507.5 0%
Table 5.2 shows that the results from the Warren sub model in Karamba 3D corresponds well to
the results from FEM-Design. The constraint considering the global lateral torsional buckling is
implemented in Karamba 3D as an additional model, which is continuously updated each time a
parameter changes during optimization.
5.2 Settings for the optimization features

The model can be optimized according to different fitness functions. In this thesis the mass and
transparency are chosen to be fitness functions. The mass of all elements can easily be retrieved from
an implemented component in Karamba 3D. The side view area on the other hand is calculated, using
the dimensions of the cross-sectional profiles. The elements affecting the side view area are the top
chord, the bottom chord, the diagonals, the verticals and the vertical end posts.

The two fitness functions are depending on other parameters such as the choice of cross-section and
the topology of the bridge. For the topology two parameters are changeable, n△ as well as n𝑐𝑏 . The
minimum of n△ is set to 1 and the maximum to 20, seeTable 5.3. The value of n𝑐𝑏 can vary between
1 and 10 within a diagonal triangle. Considering the cross-sections the sections are restricted to
VKR-, KCKR- and IPE-profiles, as mentioned in Section 4.2.2, which gives the availability to chose
between 163 different cross-sections. All variable parameters are shown in Table 5.3; the span length,
width and height of the footbridges are kept constant during one optimization. Since the bridge deck
has been weakened with a factor 10 000, the cross-section of the bridge deck is not of interest and
therefore not optimized.
Table 5.3: Variable parameters when optimizing. The span length, width and height of the bridge are
kept constant.
Variable parameters Min Max
n△ 0 20
n𝑐𝑏 0 10
Cross sections - -
When optimizing, Galapagos and Octopus aims to maximize the fitness of the design space. In the
case of this thesis, to maximize the fitness, the mass should be minimized, whereas the transparency
should be maximized, i.e. the side view area is minimized. Even though there may be solutions
that violate the constraints, theses solutions can give high fitness. For instance if the goal is to
minimize the weight of the construction, Galapagos or Octopus may choose a solution where there
are no structural elements, making the weight of the structure zero, which is a unreasonable solution
but the one with highest fitness. To avoid this problem, the changeable parameters are restricted
to solutions which generate a complete geometry. Furthermore, if Galapagos or Octopus chooses
a solution which violates at least one constraint, the solution is punished. To do this, a penalty
function, as mentioned in Section 3.1.2, is introduced, see Figure 5.2a. In short meaning that if
violating any of the shown constraints, the fitness function of the solution is redirected and multi-
plied with a high value, giving the solution low fitness, to facilitate the elimination of the bad solutions.
Since the purpose of the optimization is to get an optimal design of the main load bearing structure,
the truss sub models are optimized with the worst load combination only including the permanent
load and the evenly distributed crowd load, which is load case 6.10𝑏1 , see Table 4.4. The constraint
for the maximum displacement relies on load combination 6.10𝑎𝑏1 (SLS), see Table 4.4, not including
the self-weight, as explained in Section 5.1.
To summarize the procedure of optimization, it can be stated that it consists of many loops, which is
shown in Figure 5.2b. The loops start with the variable parameters which are tested, which form a
geometry and finally a structural model in Karamba 3D. The solution is then checked if it fulfills
all the constraints. If the solution violates at least one of the three constraints, it is multiplied with
the penalty function and receives a very low fitness value and is ranked low in a list of solutions.
If the solution is acceptable, it gets a high fitness value and is ranked high. Before starting a new
optimization loop, the evolutionary algorithm evaluates the solutions and the good ones are crossbred
to then create even better starting values and solutions for the next loop, as explained in Section 3.1.2.

(a) (b)
Figure 5.2: (a) Grasshopper script that penalize the fitness functions if the solutions violates any of
the implemented constraints. (b) Optimization procedure consisting of many loops.
In the optimization plug-ins, Galapagos and Octopus, specific initial conditions are defined before
starting. The settings for the single-objective plug-in Galapagos are presented in Table 5.4a and the
settings for the multi-objective plug-in Octopus are presented in Table 5.4b.
Description Galapagos Description Octopus

Max. stagnant 50 Elitism 0.5
Maintain 5% Mutation probability 0.1
Inbreeding (+) 75 % Mutation rate 0.5
Population 100 Cross over rate 0.8
Initial boost 2 Population 100
Initial boost 2
(a) (b)
Table 5.4: Settings in: Single-objective optimization plug-in Galapagos (a), multi-objective
optimization plug-in Octopus (b).
All values in Table 5.4a and Table 5.4b are default values, except the ones for the population and
initial boost for Galapagos. Technically having the default values of 50 and 2 for the population and
initial boost respectively, it would not be exactly comparable with the results from Octopus. Based
on this, the same values are preset for Galapagos as for Octopus, also giving the opportunity to avoid
local maximums and minimums.
5.3 Optimization procedures

In this thesis the aim is to first investigate how the results between the single- and multi-objective
optimization plug-ins Galapagos and Octopus differ, henceforth called part 1. This is executed by

using the case study as an initial condition and then optimizing it with the two plug-ins. The procedure
when optimizing is shown in Table 5.5 and starts with Galapagos, which first aims to minimize the
mass of the footbridges (run A to C). In the first run (A), only the topology is changed, i.e. n△ as
well as n𝑐𝑏 are variable parameters, and the cross-sections are kept constant as in Table 4.5. In a
second run the topology, as can be seen in Figure 4.3, is fixed and the cross-sections are variable to
detect if the mass can be reduced once the topology is fixed, seen as run B in Table 5.5. In a third
run (C), both the cross-sections of the elements and the topology of the bridge can be changed, see
Table 5.5. The optimization procedure for the transparency will undergo the same three steps (run D
to F in Table 5.5). To increase the chance of Galapagos finding a global optimum, run C and F in
Table 5.5 are executed three times. First, the optimization starts with parameters from the case study.
Thereafter, the optimization starts with the results from the previous optimization.
Table 5.5: Procedure of the optimization with Galapagos and Octopus, called part 1. For part 1 only
the Warren sub model with a span length of 20 m is considered.
Plug-in Optimization run Cross-section Topology Fitness function Repetitions
A x Mass 1
B x Mass 1
C x x Mass 3
Galapagos
D x Transparency 1
E x Transparency 1
F x x Transparency 3
Octopus G x x Mass and transparency 3
After using the single-objective optimization plug-in Galapagos, the multi-objective optimization
plug-in Octopus is used, optimizing the parameters for both the cross-sections and the topology,
optimizing both mass and area, run G in Table 5.5. Also in this case the optimization is executed
three times.
Secondly, the aim is to investigate which solution is the most optimal one for a given span length,
having the bridge width and height constant with the values 2.7 m and 1.4 m respectively, which
is henceforth referred to as part 2. Now all the three sub models, Warren, Pratt and Howe, are
considered and the sub models are optimized with Octopus for the span length 10 m, 20 m and
30 m, see Table 5.6. These sub models are evaluated by optimizing the mass and transparency while
changing both topology and cross-sections at the same time. Also in this case, the optimization is
execute 3 times for each span length, starting with the initial condition of the case study, see Table 4.5
and with n△ and n𝑐𝑏 equal to 5 respectively. The repetitions are then based on the optimization
solutions from before. The reason why only Octopus is used, is the more efficient way of optimizing,
considering the number of fitness function which can be optimized at the same time.
Table 5.6: Procedure for the second part of the optimization, where the most optimal solution for a
given span length is sought, referred to as part 2.
Plug-in Span length Geometry Cross-section Topology Fitness function Repetitions
10 m All 3 x x Mass and transparency 3
Octopus 20 m All 3 x x Mass and transparency 3
30 m All 3 x x Mass and transparency 3

6 Results of optimization of Warren, Pratt and
Howe trusses
In this chapter the results from the optimization of part 1 and part 2, mentioned in Section 5.1, are
presented. Part 1 includes the comparison of the results from Galapagos and Octopus for the Warren
sub model with a span length 20 m and part 2 reveals the best optimization results for the span length
of 10 m, 20 m and 30 m for all three sub models.
6.1 Part 1 - Galapagos versus Octopus

As explained in Section 5.1, the Warren truss bridge is first optimized with the same preconditions as
in the case study, i.e. load case 6.10𝑏1 and the bridge span of 20 m, the height of 1.4 m and the width
of 2.7 m is used. In Table 6.1, the results from run A, B, C and G are presented, see Table 5.5. Run
A, B and C were optimized in Galapagos and run G is retrieved from Octopus, all results having the
mass as fitness value.
Table 6.1: Optimization results of the Warren truss bridge with the mass as fitness value.
Input values Galapagos results Octopus result
Optimization run Case study A B C G
Utilization of displacement 41 % 23 % 41 % 35% 29 %
Utilization of elements 77 % 88 % 99 % 96 % 92 %
Utilization of N (top chord) 86 % 70 % 94 % 87 % 98 %
Side view area [m2 ] 7.74 7.85 7.14 7.05 6.25
Fitness value: Mass [kg] 8 060 7 600 7 010 6 470 6 290
The results, from the optimization of the Warren truss sub model with the mass as fitness value,
show that all the new solutions have lower mass than the case study. Changing only the topology
(run A) generated a slight decrease in weight, i.e. 6 % compared to the case study. Keeping the
original topology but changing the cross-sections (run B) generated a solution which gave a reduction
in mass by 13 %. Changing both topology and cross-sections (run C), the weight of the bridge is
reduced by 20 %. The fitness value from Octopus is reduced by 22 % compared to the mass of the
case study.
In Table 6.2 the results from run D, E, F and G of the Warren truss bridge, with the transparency as
fitness values, are presented.
Table 6.2: Optimization results of the Warren truss bridge with the transparency as fitness value.
Input values Galapagos results Octopus result
Optimization run Case study D E F G
Utilization of displacement 41 % 21 % 22 % 24% 25 %
Utilization of elements 77 % 84 % 87 % 92 % 97 %
Utilization of N (top chord) 86 % 90 % 99 % 100 % 98 %
Mass [kg] 8 060 9 970 8 790 8 200 6 690
Fitness value: Side view area [m2 ] 7.74 7.74 6.60 6.08 5.75

Looking at the results in Table 6.2, it can be stated that in run D, when only changing the topology,
the side view area could not be decreased, instead the value is the same and the mass even larger than
for the case study. However, only changing the cross-sections (run E) generated a side view area that
is 15 % less than the one from the case study. Run F shows a solution having 21 % less side view
area and Octopus, finding the best fitness value, decreased the side view area by 26 % compared to
the case study.
Putting all solutions from Galapagos and Octopus together, including the case study, Figure 6.1
gives an overview. In the multi-objective optimization plug-in Octopus, the number of possible
solutions after one optimization can vary. The distribution of solutions shows the best considering
only mass or transparency or possible options in between, where both fitness functions are considered
(Pareto-front). However, for Galapagos only one solution is retrieved after each run. As mentioned
in Table 5.5, run A, B, D and E have either the topology or the cross-sections as variable parameters,
while run C, F and G have both as changeable parameters.
Figure 6.1: Results from the case study and the solutions for Warren 20 m optimized with Galapagos
and Octopus with specified fitness values.
To summarize, all the extreme solutions in Octopus, meaning the solutions with lowest mass or
highest transparency, have better values than Galapagos. Further, since Octopus generates more than
one solution, Octopus contains solutions which are compromises between mass and transparency.
These solutions form the non-dominated front or Pareto-front, see Figure 6.1. Even though these
solutions are compromises between mass and transparency they are still optimal solutions. Thus
Octopus gives the possibility to chose a solution which is between the extremes of optimizing mass
and transparency, which gives more flexibility than in Galapagos. Another observation from the
optimization in both Galapagos and Octopus is that the optimization in Octopus is faster than in Gala-
pagos, also since Octopus optimizes both transparency and mass at the same time, the computation
time needed, to get new results, is drastically lowered.
In Table 6.3 the two best solutions, from the optimization of the Warren sub model with a span length
of 20 m, are presented. Both solutions are retrieved from optimization runs with Octopus. Comparing

the new solutions with the case study, it is apparent that the optimized results use the material more
efficiently since the utilization ratios are higher. The natural frequency for mode 1 is 3.5 Hz and
5.0 Hz for the solution with the highest transparency and lowest mass respectively.
Table 6.3: The best results from the optimization of the Warren truss sub model compared with the
case study.
Initial condition Lowest mass Lowest side view area
Plug-in Case study Octopus Octopus
n△ 5 6 5
n𝑐𝑏 5 1 1
Bottom chord VKR 150x100x6.3 VKR 100x100x4 VKR 100x100x4
Top chord VKR 150x150x6.3 VKR 120x120x4.5 VKR 100x100x10
Diagonals VKR 100x60x5 VKR 100x60x4 VKR 120x60x3.6
Vertical end post VKR 100x100x5 VKR 100x60x3.6 VKR 100x60x3.6
Cross beams IPE 140 KCKR 219.1/12.5 KCKR 219.1/12.5
End cross beams IPE 180 VKR 100x60x3.6 VKR 100x60x3.6
Utilization of displacement 41 % 29 % 25 %
Utilization of elements 77 % 92 % 97 %
Utilization of N (top chord) 86 % 98 % 98 %
Mass [kg] 8 060 6 290 6 690
Side view area [m2 ] 7.74 6.25 5.75
In Figure 6.2 a part of the solution with lowest mass from Table 6.3 is shown with the different
cross-sections. Worth noting is the height of the cross beams compared to the height of the bottom
chords.
Figure 6.2: Zoom-in of the solution with lowest mass in Table 6.3.

6.2 Part 2 - Optimization of the Warren, Pratt and Howe sub
models for given span lengths
In this section the results from optimization of the Warren, Pratt and Howe truss sub models with
different span lengths are presented. For the results that follow, load case 6.10𝑏1 is used.
For the span length of 10 m, all the results from Octopus are shown in Figure 6.3. It can be stated
that the Pratt truss sub model shows the best results, regarding both the transparency and the mass
looking at the clear Pareto-front shown in Figure 6.3. The topology of the Pratt truss for the best
solutions concerning mass and transparency is shown in Figure 6.4, since both solutions has the same
topology.
Figure 6.3: All solutions from Octopus, optimizing the Warren, Pratt and Howe sub models with a
span length of 10 m.
The topology of the best results from Octopus considering both the mass is shown in Table 6.4. For a
span length of 10 m, the Pratt sub model showed the best results.
Figure 6.4: The most optimal topology for a span length of 10 m.

It can be concluded, as already seen in Figure 6.3, that the Pratt sub model has the lowest mass with
a value of 2 750 kg, also having the most utilized elements and normal force capacity on the top
chord. Second place is the Howe sub model differing in mass with 4 % and third place the Warren
sub model, differing with 6 % compared to the Howe sub model.
Table 6.4: The best solutions of the Warren, Pratt and Howe sub models considering mass for 10 m.
Warren Pratt Howe
n△ 3 3 (=4) 4
n𝑐𝑏 1 1 1
Bottom chord VKR 100x60x3.6 KCKR 40.4/3 IPE 80
Top chord KCKR 88.9/4 KCKR 88.9/4 KCKR 76.1/4
Diagonals VKR 100x60x3.6 KCKR 60.3/4 VKR 100x60x3.6
Vertical end post KCKR 48.3/3 KCKR 60.3/3 KCKR 40.4/3
Cross beams VKR 100x100x5 KCKR 139.7/4 VKR 120x80x4
End cross beams VKR 100x60x3.6 KCKR 101.6/4 VKR 100x60x3.6
Side view area [m2 ] 2.81 2.43 2.62
Fitness value: Mass [kg] 2 920 2 750 2 860
Looking at the transparency as fitness value, the best results for each sub model is presented in
Table 6.5. As concluded before, the Pratt sub model shows the best results, one reason being the
almost fully utilization of the normal force capacity in the top chord and the utilization. The next
best solution is the Howe sub model, differing with 15 % compared to the Pratt sub model. The least
suitable solution shows the Warren sub model, differing with 20 %.
Table 6.5: The best solutions of the Warren, Pratt and Howe sub models considering transparency for
10 m.
Warren Pratt Howe
n△ 3 4 4
n𝑐𝑏 1 1 1
Bottom chord VKR 100x60x3.6 KCKR 40.4/3 IPE 80
Diagonals VKR 100x60x3.6 KCKR 40.4/3 VKR 100x60x3.6
Vertical end post KCKR 40.4/4 KCKR 48.3/4 KCKR 40.4/3
Cross beams VKR 100x100x5 KCKR 139.7/4 VKR 120x80x4
Mass [kg] 2 920 2 790 2 860
Fitness value: Side view area [m2 ] 2.79 2.24 2.62

Now, analyzing the sub bridge for span length 20 m, all the results from Octopus are shown in
Figure 6.5, where three quite distinct Pareto-fronts are displayed. It can be stated that the Warren sub
model seems to be the best choice looking at both the mass and the transparency, followed by the
Pratt sub model. The Howe sub model seems the least suitable for span length 20 m considering both
mass and transparency.
The topology of the best optimization results, regarding both mass and transparency, for a span length
of 20 m is presented in Figure 6.6.
In Table 6.6 the best solution for each sub model, considering the lowest mass, is presented. The
Warren sub model shows the best result with 6 290 kg. However, the results for the Warren and Pratt
sub models are very similar and only differ with about 2 %, corresponding to 10 kg. The Howe truss
bridge seems to be the least suitable choice with a mass of 6 570 kg, having a larger mass even though
the utilization of the normal force capacity on the top chord and the utilization is higher than for the
Warren and Pratt sub models. The Warren sub model, proving to have the best results regarding the
mass, has a natural frequency (mode 1) of 5.0 Hz as mentioned in Section 6.1.

Warren Pratt Howe
n△ 6 5 (=6) 12
n𝑐𝑏 1 1 1
Bottom chord VKR 100x100x4 VKR 120x60x4 VKR 100x100x4
Top chord VKR 120x120x4.5 VKR 120x120x5 VKR 100x100x5
Diagonals VKR 100x60x4 VKR 100x60x3.6 VKR 120x60x3.6
Vertical end post VKR 100x60x3.6 KCKR 76.1/4 KCKR 40.4/3
Cross beams KCKR 219.1/12.5 KCKR 168.3/4 VKR 120x60x3.6
Side view area [m2 ] 6.25 6.73 6.61
In Table 6.7, the best results considering the transparency for all three sub models are presented. As
mentioned before, the Warren sub model is the best choice with an area of 5.75 m2 . Second best
is the Pratt sub model, differing with about 3 %, followed by the Howe truss model with 6.34 m2 ,
differing with about 9 % compared to the Warren sub model. The Warren sub model, proving to have
the best results regarding the transparency, has a natural frequency (mode 1) of 3.5 Hz as mentioned
in Section 6.1.
20 m.
Warren Pratt Howe
n△ 5 5 (=6) 10
n𝑐𝑏 1 1 1
Top chord VKR 100x100x10 VKR 100x100x8 VKR 100x100x5
Diagonals VKR 120x60x3.6 VKR 100x60x3.6 VKR 120x60x5
Vertical end post VKR 100x60x3.6 KCKR 76.1/4 KCKR 40.4/3
Cross beams KCKR 219.1/12.5 VKR 160x80x4 VKR 120x60x3.6
End cross beams VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Mass [kg] 6 690 6 510 6 690
All the solutions obtained in Octopus for the span length of 30 m are presented in Figure 6.7. Also
this time, the results show clear Pareto-fronts for each sub model. For 30 m the Warren sub model

performed the best considering mass and area. The topology of the best solutions for mass and
transparency is shown in Figure 6.8, since both solutions has the same topology.
The topology of the best optimization results, regarding both mass and transparency, for a span length
of 30 m is presented in Figure 6.8.
Presenting the best solution for each sub model with a span length of 30 m with the lowest mass, see
Table 6.8, it can be stated that the Warren and Pratt sub models show a quite similar solution when
looking at the mass. However, the Warren sub model gives the best solution, followed by the Pratt
sub model differing with only 1 %. The solution for the Howe sub model differs with 4 % compared
to the Warren sub model. All three solutions are very utilized looking at the utilization ratio and the
normal force capacity of the top chord. The sub model which has the lowest mass for 30 m has a
natural frequency (mode 1) of 4.2 Hz.

Warren Pratt Howe
n△ 11 9 (=10) 14
n𝑐𝑏 1 1 1
Bottom chord VKR 140x140x6.3 VKR 140x140x6.3 VKR 100x100x10
Top chord VKR 180x180x6.3 VKR 180x180x6.3 VKR 160x160x6.3
Diagonals VKR 100x60x4 VKR 120x60x4 VKR 160x80x4
Vertical end post KCKR 76.1/5 VKR 120x60x3.6 KCKR 40.4/3
Cross beams KCKR 139.7/4 VKR 140x70x4 VKR 120x60x4
End cross beams KCKR 88.9/4 VKR 100x100x4 120x60x3.6
Side view area [m2 ] 12.39 12.62 12.24
Looking at the best solutions considering the transparency, the Warren sub model performs with
the highest transparency, followed by the Pratt sub model with 1 % difference. The solution of the
Howe sub models differs with 4 % compared to the Warren sub model, representing the least suitable
solution considering the transparency for a span length of 30 m. Still, the results do not differ a lot.
The sub model with the highest transparency has a natural frequency (mode 1) of 4.4 Hz.
30 m.
Warren Pratt Howe
n△ 12 10 14
n𝑐𝑏 1 1 1
Top chord VKR 140x140x10 VKR 140x140x10 VKR 120x120x10
Diagonals VKR 120x60x3.6 VKR 120x60x4 VKR 140x70x6.3
Vertical end post KCKR 76.1/5 VKR 100x60x6.3 KCKR 40.4/3
Cross beams KCKR 139.7/4 VKR 160x80x6.3 KCKR 139.7/4
End cross beams KCKR 101.6/4 VKR 100x100x4 VKR 140x70x4
Mass [kg] 12 050 12 280 12 910

7 Discussion
This chapter starts off with a discussion about the results in part 1, i.e. the comparison of the results
between Galapagos and Octopus. Thereafter a discussion of the results from part 2, where the most
optimal solutions for the different span lengths of 10 m, 20 m and 30 m are debated. Finally, a
discussion of the model in Karamba 3D and its verification results is introduced, concluding in an
analysis of the optimization results.
7.1 Part 1 - Galapagos versus Octopus

Starting with the comparison of the results with load case 6.10𝑏1 in Galapagos and Octopus, see
Table 6.1 and Table 6.2, it can clearly be stated that Octopus concluded in the best result both having
the mass and area as fitness values. In Table 7.1 the different runs are presented again to facilitate the
understanding of the discussion which follows.
Table 7.1: Procedure of the optimization with Galapagos and Octopus, called part 1. For part 1 only
the Warren sub model with a span length of 20 m is considered.
Plug-in Optimization run Cross-section Topology Fitness function Repetitions
A x Mass 1
B x Mass 1
C x x Mass 3
Galapagos
D x Transparency 1
E x Transparency 1
Investigating the results from Galapagos, optimization run A/D show the worst fitness values.
Looking at run D, the side view area could not even be decreased, moreover resulting in an even
larger mass than the case study. For run A/D the topology was a variable parameter, only offering
two values to be changed resulting in few possibilities to optimize with. Therefore it is difficult for
Galapagos to find a good fitness value for the solutions, leading to a longer computation time than for
run B/E. Still, the best result from run A has a better fitness value than the initial condition being the
case study.
Run B and E in Galapagos, having the cross-sections only as variable parameter, gives better fitness
values than run A and D, which is due to the fact that the topology from the case study is already quite
optimal, but not the cross-sections with the given constraints implemented in Karamba 3D. Also, the
range of cross-sections is large, giving the plug-in more options and combinations to optimize with,
which ends up in a shorter computation time as for run A/D.
Run C and F in Galapagos, having both the topology and the cross-sections as a variable parameters,
offer the plug-in a lot of different possible combinations to optimize with, which results in the best
fitness value looking at the results from Galapagos. Still, the computation time for Octopus is much
faster and the best result even better. This can be explained by the optimization algorithm used
in Octopus, SPEA-2, which is more powerful than the GA used in Galapagos. Since Octopus, as

mentioned in Section 3.1.2, stores the best solutions in an archive and uses this to mutate the popula-
tion, the population is less likely to be trapped in a local optimum than Galapagos, which relies only
on the current population to find the optimal solution.
Having a more detailed look at the best results with the mass as fitness value, see Table 6.1, it is
interesting to analyze the utilization of the three constraints. Run G from Octopus shows a 98 %
utilization for the normal capacity of the top chord, while run C from Galapagos has a value of
87 %. Both values are high, but the result from Octopus is more utilized, having this as the limiting
factor, as it is the highest number of all utilization ratios for run G. Run C on the other hand seems
to have the utilization of the elements as the limiting value, reaching 96 %. The best result from
Octopus shows a value of 92 %. The utilization of the displacement seems not to be limiting at all,
both considering the best solution in Galapagos and Octopus, with a utilization of 35 % and 29 %
respectively.
Analyzing the results with the side view area as fitness value, see Table 6.2, the results from Octopus
are almost fully utilized both considering the normal capacity of the top chord and the utilization with
the values 98 % and 97 % respectively. The best result from Galapagos, run F, is fully utilized looking at
the normal capacity of the top chord, 100 %, and quite utilized looking at the utilization ratio, 92 %. For
both Galapagos and Octopus, the utilization of the normal capacity seems to be the limiting value. The
utilization of the displacement is again not at all decisive with 24 % and 25 % for run F and G respec-
tively.
Investigating the details of the best results for the lowest mass and highest transparency, see Table 6.3,
the first reflection is that both solutions are very similar. Looking at n△ , the best solution, having
the mass as fitness value, has n△ equal to 6, while the best solution with the transparency as fitness
value and the case study both have n△ equal to 5. The difference compared to the case study lies
though in n𝑐𝑏 , which is equal to 1 for both optimized solutions, whereas it is 5 for the case study. It
can also be stated that the dimensions of the cross-section of the cross beams for both optimization
solutions are rather large, having KCKR 219.1/12.5 as a suggestion. The result of n𝑐𝑏 being equal to
1 gives rise to several issues that will be further discussed in Section 7.4.
It is rather understandable that Octopus, having the transparency as fitness value, tries to increase the
dimensions of the cross beams as much as possible while decreasing the side view dimensions for
the cross-sections for the top and bottom chord, the diagonals and the vertical end posts as much as
possible, knowing that the dimensions of the cross beams, including the end cross beams, do not
affect the side view area at all.
Worth mentioning is the fact why the focus did not lie in how the case study could be optimized.
The two main differences are two details which were not implemented in the optimization, which
are considered in the case though. The first difference is the calculation of the end cross beams,
which was constructed to withstand a change of the bearings used for the supports in the case over
Genevadsån, but not in the model in Karamba 3D. This resulted in a somewhat larger cross-section
of the end cross beam compared to the optimization results from Galapagos and Octopus. The
second difference is the c-c distance of the cross beams, which in the case over Genevadsån considers
the deflection of the bridge deck, which is not included in the optimization results. This is further
discussed in Section 7.3.

7.2 Part 2 - The Warren, Pratt and Howe sub models for given
span lengths
Looking at the best solutions for each span length, the optimization results show that the Pratt truss
gives the best results for a span length of 10 m and the Warren truss bridge for span length of 20 m and
30 m. However, when analyzing the results more in detail, it is not so obvious which sub model
actually is the best one.
From Tables 6.4 and 6.5, the utilization ratios of the normal force capacity in the top chord for the
Warren and Howe sub models for span length 10 m seems to be low, with the corresponding values
of 77 % and 64 % for both mass and area. For the Pratt sub model, these values are 94 % and 98
% respectively. Looking at the utilization of the elements, the numbers are quite high for all three
sub models, both having the mass and transparency as fitness values. Based on this, the possibility
arises that the Warren and Howe sub models have potential to be further optimized, while the Pratt
sub model is almost fully utilized considering the normal force capacity on the top chord and the
utilization and therefore gives the best results. Looking at the final fitness values of a span length
of 10 m, the Howe sub model gives the second best results, which are very close to the ones from
the Pratt sub model. This, even though the solution is not as fully utilized as the Pratt sub model.
Therefore it is difficult to say, which sub model is the most suitable one for a span length of 10 m, as
they are not quite comparable.
Analyzing the results for span length 20 m, the Warren sub model gives the best results. For span
length 20 m, all three sub models are almost fully utilized considering the normal force capacity on
the top chord and the utilization. This makes the comparing process more equitable, showing very
similar results between the Warren and Pratt sub model. In fact, The Warren sub model only differs
with less than 1 %, corresponding to 10 kg, looking at the best solution considering the mass and
3 %, corresponding to 0.18 m2 , considering the best solution for the transparency, compared to the
Pratt sub model.
As mentioned in Section 4.4, n△ can only be an even number for the Pratt and Howe trusses.
Therefore, as shown in Table 6.6, for a span length of 20 m, the Warren and the Pratt sub models
have the same topology with n△ equal to 6 and n𝑐𝑏 equal to 1, whereas the Howe truss has the double
amount of n△ , 12, than the other two sub models. However, this allows the Howe truss to have more
slender beams even though it has more diagonals, which in the end still ends up in a larger mass and
side view area.
Investigating the results with a span length of 30 m, the Warren sub model shows the best results
again. Also for these results, the utilization ratios are almost maximized to 100 % for all three sub
models, except the one considering the displacement. The utilization of the displacement did increase
to about 50 % for all three sub models, which seems reasonable as the span length increased.
Looking at the results with the mass as fitness value, the Warren and Pratt sub models showed quite
similar results, 11 500 kg and 11 590 kg respectively, with almost the same topology. The best
solution for the Warren sub model proposes n△ to be equal to 11, whereas the Pratt sub model has
n△ equal to 10. The Howe bridge has some additional verticals and diagonals with n△ equal to 14.
For all three cases n𝑐𝑏 equal to 1 is proposed.

The results having the transparency as fitness value are very similar for all three cases with
10.12 m2 , 10.22 m2 and 10.50 m2 for the Warren, Pratt and Howe sub model respectively. The
Warren sub model has in this case n△ equal to 12, whereas the Pratt sub model has n△ equal to 10,
giving the second best solution. The cross-section for the elements that influence the side view area
are exactly the same, except the vertical end posts, being the decisive factor for the Warren sub model,
together with two more n△ .
To summarize, studying the results from the best solutions for 10, 20 and 30 m, it is apparent that
the deflection criterion was never the limiting constraint for the optimization. Rather the utilization
ratio of the elements or normal force capacity in the top chord are factors that limit the design. Since
the deflection demand according to Krav TDOK 2016:0204 section B.3.4.2.1 states that only the
frequent values of the load combinations are needed to check the deflection in SLS, the self-weight
of the structure was ignored. Thus, even for larger span lengths, as for instance 30 m, the utilization
of the displacement lies around 50 %, whereas the utilization of the elements and utilization of the
normal force capacity are closer to 100 %.
All in all, considering the results for all studied span lengths in Section 6.2, there is no clear winner
between the sub models, rather there are clear losers. Having the worst results, for all but one of the
studied span lengths, the Howe truss is indeed the worst sub model considering the chosen fitness
values. Whereas for the winner, it could be argued that both the Warren and the Pratt work equally
good for these studied span lengths, with the Pratt being better at 10 m, both being approximately
equal at 20 m and the Warren being a better solution at 30 m.
7.3 The model in Karamba 3D and its verification results

Considering the verification as mentioned in Section 4.3, there is a fundamental difference between
how Karamba 3D and FEM-Design calculates the utilization of structural elements. The main
problem is that there are two possible methods to calculate the interaction factors in EC3, which
explains some of the differences between the verification results in Karamba 3D and FEM-Design.
The possibility to investigate changes in the calculation procedure in Karamba 3D could be initiated
in further studies if desired, resulting in better verification results with FEM-Design.
To get sectional forces for the sub models in Karamba 3D that correspond to the sub models in
FEM-Design, the bridge deck had to be weakened with a factor of 10 000. This, as the behaviour
and optimization of the bridge deck was out of interest in this thesis and the focus only relied on the
truss footbridge structure and its behaviour. Naturally, weakening the bridge deck means that it has
no stiffness and therefore can take no load. Although the sub models in Karamba 3D have sectional
forces that correspond well to the sectional forces in FEM-Design, the weakened bridge deck causes
problems in other areas of the structural analysis.
First, as mentioned in Section 7.1, the deflection of the bridge deck is large, since it has no stiffness
and therefore needs to be ignored when looking at the maximum displacement of the bridge. Even
if the bride deck would have a weakening factor 1, the displacements would be rather big as the
distances between the cross beams are large in the optimization results, see Chapter 6. Secondly,
the natural frequencies of the sub models cannot be introduced in the optimization runs of the sub

models with a weakened bridge deck, since this leads to really low natural frequencies as it is very
flexible. The verification results of the natural frequency with a weakening factor 1 for the Warren
sub model showed very good results though, indicating that it can be relied on but cannot be a part of
the optimization as the bridge deck is weakened with a factor 10 000.
Discussing the topic of Karamba 3D, there are mainly three issues with the software. First, the global
treatment of all analysis, meaning that Karamba 3D performs well looking at the structural behaviour
locally. Globally, the answers are hard or even impossible to find. This was the reason why the extra
script with the calculations for the normal force restriction on the top chord considering LT-buckling
was implemented, as this check did not exist globally in the calculations in Karamba 3D, only locally.
Also, the issue with the buckling length comes into this subject, as Karamba 3D finds it hard to
define the length and behaviour of a beam, which needs to be defined manually to be sure of the right
interpretation from Karamba 3D.
The second issue is the post processing in Karamba 3D, which is very time consuming. All solutions,
looking for instance at the sectional forces, can be found in lists, which are hard to understand and
locate onto the elements in the bridge. There is also an alternative to see the solutions for the sectional
forces in a visual picture with numbers, which has the disadvantage though to put several numbers
on top of each other, making it hard to read. Furthermore, the results need to be found and typed into
an external sheet manually, which opens up and allows the possibility of writing errors. There are
ways in Karamba 3D to store the results in excel sheets, which was tried and neglected as the update
of the numbers in the excel sheet was unreliable and inaccurate.
The third problem is the definition of the behaviour of all joints in the structure. According to the
definition in Karamba 3D, if the elements are connected to each other with no further input, the
connection is rigid. If the properties of the joint should be changed, the specific node is to be found
and the properties changed. This was executed when applying a unit load of 1 000 kN, retrieving
the 𝛿 to calculate the spring stiffness. The issue is though, that finding this exact node needs to be
generalized, as the optimization brings a change of topology and cross-sections with it. The process
of generalizing the search of a specific node, in list that consists of a lot of nodes, with varying orders
in the system is very time consuming.
7.4 Overall optimization results

Having a look at the constraints used in this thesis, which are the maximum displacement, the maxi-
mum utilization ratio of an element and 𝑁𝑏𝑅𝑑 being the maximum reduced normal capacity of the
top chord, the subject can be more expanded. For instance, continuing on the discussion from before,
Section 7.2 states that all solutions obtained with Octopus have n𝑐𝑏 equal to 1. This gives rise to the
issue of the displacement between the cross beams, should there be a bridge deck. As the bridge
deck is not considered in the optimization in this thesis, there is no problem, as both solutions from
Octopus fulfill the implemented constraints. In reality though, when there actually is a bridge deck,
the issue can be solved by implementing a constraint. This could be executed locking the maximum
c-c distance of the cross beams, which is adapted to geometry and properties of the bridge deck and
its maximum allowable displacement. Further, the first mode of the natural frequency could be added
as a constraint, which would restrict the solutions to be above 5 Hz, to avoid further investigations of
the natural frequency.

It is of course of importance to adapt the sub models and optimization results to the reality and
aim for the actual possibility that the footbridges can be constructed. The results obtained from the
optimization runs using Galapagos and Octopus do for instance not consider the production of the
footbridges, meaning that the possibility to weld the different connections of the cross-sections is
not considered. Since no constraint was introduced restricting the maximum height of for instance
the cross beams to the the height of the bottom chord, unreasonable solutions were obtained, which
can be seen in Figure 6.2. This can also be solved, implementing additional constraints adapting the
connection of the elements. Practically this would mean restricting the height of the cross beams to
be less or equal to the height of the bottom chord. The diagonals and vertical end posts cannot be
wider than the width of the bottom chord and top chord. These restrictions would ensure that the
bridge could actually be constructed.
Further, as welding is rather expensive, this could be implemented as a constraint, only allowing as
few connections of elements as possible. Also, the number of different dimensions and cross-sections
within one footbridge could be minimized, to facilitate the production of the footbridges. Additionally,
most optimization results seem to have KCKR cross-sections for the cross beams, which might not
be preferable considering the application of a bridge deck for instance. This could also easily be
implemented as a constraint, where the allowable cross-sections for each element can be restricted or
further expanded. The possibilities for the designer to specify the optimization and its constraints are
endless.
A final constraint could be to implement the calculations of a replacement of the bearings of the
supports. If this load case and restriction were to be implemented, as it was in the case study which
is discussed in Section 7.1, the final optimization results of the structure would contain larger end
cross beams.
The fitness functions chosen in this thesis were the mass and the transparency of the footbridges.
Starting with the mass, it might not always be good to reduce the mass as much as possible, for
instance looking at the dynamics of the sub models, since reducing the mass leads to lower natural
frequency, which is not looked into in detail in this thesis. Further, obvious but still important, only
using transparency as a fitness function showed to be sub-optimal since the mass of the sub-models
was increased, which is an unwanted side effect. Thus, only using transparency as a fitness function
serves no purpose. However, in combination with mass as a fitness function it works to give the
bridge a more slender expression.
In this thesis only the extremes of the solutions in Octopus were presented. However, Octopus
generates several Pareto-optimal solutions, the best trade offs between the fitness functions. These
solutions give the designer the choice to weigh the preferences in mass or transparency. There is no
actual optimal solution, it all depends on if the designer is more concerned about the mass or the
transparency of the footbridge.
The variable parameters in this thesis were n△ , n𝑐𝑏 and the cross-sections. These allowed the
optimization plug-ins, Galapagos and Octopus, to choose from quite a large range; turning out to be
almost too large, as it affected the optimization time. A recommended restriction is for instance to
limit the number of cross-sections to the ones longed for, looking at the design, and the ones most
likely to be used. Also, additional cross-sections and properties could be added or replace the ones

used in this thesis. Additionally, when optimizing with the cross-sections as variable parameters, the
number of different cross-sections which are the same should be rewarded. This, to maybe match the
expectancy of the reality, where 5 different cross-sections in one footbridge might not be appreciated.
Based on the experience when optimizing in Galapagos and Octopus, it can be stated that one
optimization run can take time, for instance an optimization run in Galapagos takes between 1.5 - 4
hours. As has been stated in Section 7.1, Octopus is faster than Galapagos considering the computa-
tion time and also gives better results due to the more advanced algorithm used. An optimization run
in Octopus often took less than 1 hour to complete, which could be reduced with a few adjustments
to the script.
One reason for the long optimization time could be the way the structure of the model is built up.
Having quite a lot of variable parameters plus an inexperienced background when starting this project,
the structure of the model could probably be more efficiently rebuilt, thus reducing the computation
time. Another way to reduce the computation time could be to adapt the settings in Galapagos
and Octopus, depending on if the designer prioritizes to find the most optimal solution or a short
computation time.
In this thesis the settings in Galapagos and Octopus were set to a population of 100 and a initial boost
of 2. This choice is based on the default values in Octopus, which was adapted in Galapagos. To make
sure how the setting should be preset, among others depending on the number of fitness functions and
variable parameters, a sensitivity analysis should be executed to find the optimal presetting conditions.
Moreover, the number of repetitions of an optimization until the best utilized result is met, can be
debated and investigated.
All in all, the possibilities for the designer are endless. A lot of more aspects regarding cost and
environmental analyses could be implemented, including the whole process of the development of
steel truss footbridges. For instance the life cycle assessment of the footbridges could be imple-
mented, introducing the choice of material, its transportation, the procedure of construction until the
disassembling of the bridges at the end of the lifespan. Further, the optimization results in this thesis
only consider steel elements with standard dimensions, which probably is most accurate and useful as
the demand looks like today. In the future though, when optimized and unique shapes become more
available and the belonging standards are introduced, this gives the possibility to deviate from the
standard dimensions. This would probably result in even better results, regarding the chosen fitness
functions. This is an amazing opportunity that is handed over to the designers and Civil engineers in
the future though.

8 Conclusion
In this thesis a computational script of the superstructure of steel truss footbridges was created in
Grasshopper and translated into a parametric model with structural elements in Karamba 3D, based
on Swedish norms and regulations. The model was optimized with the single- and multi-objective
optimization plug-ins Galapagos and Octopus, to minimize the mass and maximize the transparency
of the structures.
First, it can be concluded that the script, the way it is built, and its results with respect to sectional
forces and maximum displacement are reliable, if the bridge deck is weakened with a factor 10 000.
There are some limitations that the user of Karamba 3D needs to be aware of, for instance the results
considering the utilization of the elements. The difference lies in the calculation methods used in
Eurocode; FEM-Design uses method 1 and Karamba 3D uses method 2 to calculate the interaction
factors. The second limitation is the global LT-buckling, which is not automatically included in
the calculations in Karamba 3D, but was implemented as an additional constraint and works for the
intended purpose. The natural frequency of mode 1 works for the Warren sub model and gives satisfy-
ing results, if the weakening factor is changed to 1, i.e. the stiffness of the bridge deck is not weakened.
Secondly, the conclusion considering the optimization plug-ins Galapagos and Octopus is the fact
that Octopus is more beneficial to use than Galapagos. This mainly due to the fact that problems in
structural engineering seldom can be condensed into one simple fitness function and thus there are
needs to express the problem using multiple fitness functions, as was the case in this thesis. Also,
Octopus needed less computation time and gave more solutions to choose from than Galapagos which
only gives one solution, often at the cost of other potential fitness functions. In combination with
the fact that all solutions from Octopus are optimal, it gives more options which are possible for a
preliminary design. Furthermore, the best solutions in Galapagos were worse than the best solutions
obtained with Octopus. The reason for Octopus finding better solutions is due to the more advanced
optimization algorithm used.
Thirdly, it can be concluded that the Howe sub model preformed the worst out of all trusses. Also, it
can be said that for 10 m the Pratt sub model is the best geometry and for 20 m as well as for 30 m the
Warren sub model is the best. It needs to be pointed out though, that both Warren and the Pratt sub
model are viable results for 20 m and 30 m, since the differences between them are almost negligible.
Finally, the developed model can be used in preliminary design to give a good initial estimation of
cross-sections and topology for the three included sub models. Since Octopus gives a variety of
results in a short time the preliminary design process could be reduced. This information could then
be used further as initial values for the detailed design. At the same time the user should be aware
of all the possibilities that the model now offers, regarding development of constraints for instance,
which is discussed in Section 7.4. Now as the basis script is created, there is an opportunity to develop
it further and adapt all the constraints and preferences that the designer desires. Some examples
of further studies, which could be implemented to develop the script and model, are explained in
Section 8.1.

8.1 Further studies
In this section suggested improvements to the model as well as topics for further studies are proposed.
• Additional geometries, not only truss bridges could be added, verified and developed to broaden
the span of design choices.
• The possibility of adding more loads, for instance wind and temperature loads, should be
investigated further. Additionally, the opportunity of adding extra supports as well as the
new load combinations that come with it could be introduced and analyzed. Also, the service
vehicle, or moving concentrated loads in general, could be implemented.
• The issue of the bridge deck could be solved to allow the possibility to introduce constraints
regarding natural frequencies and deflections of the bridge deck.
• The calculation procedure of the utilization of the elements in Karamba 3D could be

reprogrammed to achieve a better coherence with the utilization ratios in FEM-Design.
• More fitness functions and constraints could be introduced to steer the optimization process
towards designs that are more feasible, which is mentioned in Section 7.4.
• Galapagos and especially Octopus are powerful tools regarding optimization. Nevertheless, it
would be beneficial to conduct a sensitivity analysis for Octopus and Galapagos to determine
how large the population and other initial values need to be to get fast and/or accurate results.
It would be of interest to see how the initial values should be set for any amount of parameters
and fitness functions.
• Other materials than steel could be introduced - for instance fibre reinforced polymers, rein-
forced concrete and timber - and their behaviour in Karamba 3D investigated, which sets the
basis for further and a more broadened opportunities to optimize.
• Finally, economical and environmental assessments and aspects could be implemented in the
optimization.

References
Ahston, C., Aldin, N., Andersson, J.-O., Gustafsson, Å., Hedgren, E., Lidegran, G., . . . Slunga, K.
(2017). Trafikverkets strategiska inköpsarbete samt arbete med att arbetsrättsliga krav efterlevs
i upphandlad verksamhet. Borlänge. (Cited on pages i, 1, 7).
Alexandersson, S. (2014). Halmstads kommun , Teknik- och fritidsförvaltningen Ny gång- och
cykelbro över Genevadsån Bakgrund. 2, 1–7. (Cited on page 19).
Bader, J., Zitzler, E., & Rudolph, G. (2010). Hypervolume-based search for multiobjective optimiza-
tion: theory and methods. Optimization, 1–310. Retrieved from https://sop.tik.ee.ethz.ch/
publicationListFiles/bade2010a.pdf. (Cited on page 15)
Bader, J. & Zitzler, E. (2011). HypE: An Algorithm for Fast Hypervolume-Based Many-Objective
Optimization. Evolutionary Computation, 19(1), 45–76. doi:10.1162/EVCO_a_00009. (Cited
on page 15)
Balling, R. J., Briggs, R. R., & Gillman, K. (2006). Multiple Optimum Size/Shape/Topology Designs
for Skeletal Structures Using a Genetic Algorithm. Journal of Structural Engineering, 132(7),
1158–1165. doi:10.1061/(ASCE)0733-9445(2006)132:7(1158). (Cited on page 1)
Bates, W. & Fistructe, C. (1991). HISTORICAL STRUCTURAL STEELWORK HANDBOOK
Properties of U .K . and European Cast Iron, Wrought Iron and Steel Sections including Design,
Load and Stress Data since the Mid 19th Century. (Cited on page 8).
Beck, A. T. & Gomes, W. J. D. S. (2012). A comparison of deterministic, reliability-based and
risk-based structural optimization under uncertainty. Probabilistic Engineering Mechanics, 28,
18–29. doi:10.1016/j.probengmech.2011.08.007. (Cited on page 10)
Berns, H. & Theisen, W. (2008). Ferrous materials: Steel and cast iron. doi:10.1007/978-3-540-
71848-2. (Cited on page 6)
Bhadeshia, H. & Honeycombe, S. R. (2006). Iron and its Interstitial solid solutions. In H. Bhadeshia &
S. R. Honeycombe (Editors), Steels: Microstructure and properties (Third, Chapter 1, Pages 1–
16). ASM International. doi:https://doi-org.proxy.lib.chalmers.se/10.1016/B978-075068084-
4/50003-0. (Cited on page 6)
Black, J., Hashimzade, N., & Myles, G. (2017). A Dictionary of Economics. Retrieved February
16, 2018, from http://www.oxfordreference.com.proxy.lib.chalmers.se/view/10.1093/acref/
9780198759430.001.0001/acref-9780198759430-e-837. (Cited on page 10)
Camp, C., Pezeshk, S., & Cao, G. (1998). Optimized Design of Two-Dimensional Structures Using a
Genetic Algorithm. Journal of Structural Engineering, 124(5), 551–559. doi:10.1061/(ASCE)
0733-9445(1998)124:5(551). (Cited on pages 1, 13)
Chinneck, J. W. (2001). Chapter 17 : Pattern Search for Unconstrained NLP. Practical Optimization:
a Gentle Introduction, (0), 1–7. (Cited on page 65).
Coello Coello, C. A. (2006). Evolutionary Multi-Objective Optimization: A Historical View of the
Field. IEEE Comput. Intell. Mag. 1(1), 28–36. doi:10.1109/MCI.2006.1597059. (Cited on
pages 1, 10, 15)
Colman, A. M. (2015). A Dictionary of Psychology. Retrieved February 16, 2018, from http://www.
oxfordreference.com.proxy.lib.chalmers.se/view/10.1093/acref/9780199657681.001.0001/
acref-9780199657681-e-1855%7B%5C#%7D. (Cited on page 11)
COX, H. (1965). Introduction: Theory in Design. The Design of Structures of Least Weight, 1–5.
doi:10.1016/B978-1-4831-9755-5.50005-8. (Cited on page 10)
Craigh, R. & Kurdila, A. (2006). Fundamentals of structural dynamics (Second edi). John Wiley &
Sons, Inc. (Cited on page 6).

Davidson, S. (2018). Grasshopper - Algorithmic modeling for Rhino. Retrieved February 26, 2018,
from http://www.grasshopper3d.com/. (Cited on page 16)
Dominik Holzer, Richard Hough, M. B. (2007). Parametric Design and Structural Optimisation for
Early Design Exploration. International Journal of Architectural Computing, 5(4), 625–643.
doi:10.1260/147807707783600780. (Cited on page 15)
Domone, P. & Illston, J. (2010). Construction Materials: Their Nature And Behaviour (4th) (P.
Domone & J. Illston, Editors). Abdingdon: Spon press. (Cited on pages 6, 7).
Duan, L. (2017). Truss bridge. Retrieved March 2, 2018, from http://www.accessscience.com.proxy.
lib.chalmers.se/content/712890. (Cited on pages 4, 5)
Dumitrescu, G. C. & Tanase, I. A. (2016). 3D Printing - A New Industrial Revolution. Knowledge
Horizons. Economics, 8(1), 32–39. Retrieved from http://proxy.lib.chalmers.se/login?url=https:
//search-proquest-com.proxy.lib.chalmers.se/docview/1793552762?accountid=10041. (Cited
on page 9)
Dunning, H. (2017). The world’s first 3D printed steel bridge will be a ’living laboratory’. Retrieved
February 13, 2018, from http://www3.imperial.ac.uk/newsandeventspggrp/imperialcollege/
newssummary/news%7B%5C_%7D20-10-2017-16-21-38%7B%5C%%7D0A. (Cited on
page 9)
Eindhoven University of Technology. (2017). The world’s first 3D printed reinforced concrete bridge
starts to take shape. Retrieved February 13, 2018, from https://www.tue.nl/en/university/news-
and-press/news/16-06-2017-the-worlds-first-3d-printed-reinforced-concrete-bridge-starts-
to-take-shape/. (Cited on page 9)
Gandomi, A. H., Yang, X. S., Talatahari, S., & Alavi, A. H. (2013a). Metaheuristic Algorithms
in Modeling and Optimization (First Edit). Elsevier Inc. doi:10.1016/B978-0-12-398364-
0.00001-2. (Cited on pages 12, 13)
Gandomi, A. H., Yang, X. S., Talatahari, S., & Alavi, A. H. (2013b). Metaheuristic Applications in
Structures and Infrastructures. doi:10.1016/C2011-0-08778-1. (Cited on page 12)
Gro, C., Oltean, M., & Oltean, M. (no date). The role of elitism in Multiobjective Optimization with
Evolutionary Algorithms, 83–90. (Cited on page 15).
Gu, D. (2015). Laser Additive Manufacturing (AM): Classification, Processing Philosophy, and
Metallurgical Mechanisms. doi:10.1007/978-3-662-46089-4_2. arXiv: arXiv:1011.1669v3.
(Cited on page 8)
Gupta, M. (2017). 3D Printing of Metals (1st edition) (M. Gupta, Editor). Basel, Switzerland: MDPI.
Retrieved from http://www.mdpi.com/books/pdfview/book/384. (Cited on page 9)
Hasançebi, O., Çarbaş, S., Doǧan, E., Erdal, F., & Saka, M. P. (2009). Performance evaluation
of metaheuristic search techniques in the optimum design of real size pin jointed structures.
Computers and Structures, 87(5-6), 284–302. doi:10.1016/j.compstruc.2009.01.002. (Cited on
page 10)
infra, B. (2017). Rondweg Gemert Noord-Om - Van 3D geprinte brug tot ingeschoven rotonde.
Retrieved February 13, 2018, from https://www.baminfra.nl/projecten/rondweg- gemert-
noord-om-van-3d-geprinte-brug-tot-ingeschoven-rotonde?position=3%7B%5C&%7Dlist=
x5OsDiWF7aNwjkAvPyn67NyLsKkaMsdr02jHI1CYOps. (Cited on page 9)
Kassimali, A. (1993). Pratt truss, 1. Retrieved from http://www.areforum.org/up/GeneralStructures/
must%20read/CommonTrusses.pdf. (Cited on page 5)
Kianian, B., Tavassoli, S., Larsson, T. C., & Diegel, O. (2016). The Adoption of Additive Manufac-
turing Technology in Sweden. Procedia CIRP, 40, 7–12. doi:10.1016/j.procir.2016.01.036.
(Cited on page 9)

Krauss, G. (2005). Introduction: Purpose of Text, Steel Definitions, and Specifications. In G. Krauss
(Editor), Steels - processing, structure, and performance (Chapter 1, Pages 1–5). ASM Interna-
tional. (Cited on page 6).
Larsson, R. (2016). MÖJLIGHETER & HINDER MED 3D-PRINT AV BETONG. (Cited on page 9).
LKAB. (2018). SSAB, LKAB och Vattenfall planerar bygga världsunik pilotanläggning för fossilfritt
stål. Retrieved February 14, 2018, from https://www.lkab.com. (Cited on page 8)
Mandal, S. K. (2015). Steel Metallurgy: Properties, Specifications and Applications. New York,
Chicago, San Francisco, Athens, London, Madrid, Mexico City, Milan, New Delhi, Singapore,
Sydney, Toronto: McGraw-Hill Education. Retrieved from http://accessengineeringlibrary.com.
proxy.lib.chalmers.se/browse/steel-metallurgy-properties-specifications-and-applications%
7B%5C#%7DfullDetails. (Cited on page 6)
Martini, K. (2015). Multiobjective Optimization of Structure and Visual Qualities. Proceedings of
the Symposium on Simulation for Architecture & Urban Design, 182–188. Retrieved from
http://dl.acm.org/citation.cfm?id=2873021.2873046. (Cited on pages 1, 12, 14, 65, 66)
McGuire, M. F. (2008). Stainless Steels for Design Engineers. A S M International. Retrieved from
https://ebookcentral.proquest.com/lib/chalmers/detail.action?docID=3002326v. (Cited on
page 6)
McNeel, R. (2018). What are NURBS? Retrieved February 7, 2018, from https://www.rhino3d.com/
nurbs. (Cited on page 16)
Mehrjoo, M., Khaji, N., Moharrami, H., & Bahreininejad, A. (2008). Damage detection of truss bridge
joints using Artificial Neural Networks. Expert Systems with Applications, 35(3), 1122–1131.
doi:10.1016/j.eswa.2007.08.008. (Cited on page 5)
Miljönytta. (2018). Stålindustrin: Stål formar en bättre framtid. Retrieved February 7, 2018, from
http://miljonytta.se/branscher/stalindustrin/. (Cited on pages 6, 8)
Mitropoulou, C. C., Fourkiotis, Y., Lagaros, N. D., & Karlaftis, M. G. (2013). Evolution Strategies-
Based Metaheuristics in Structural Design Optimization (First Edit). Elsevier Inc. doi:10.1016/
B978-0-12-398364-0.00004-8. (Cited on pages 13, 14)
Nanakorn, P. & Nimtawat, A. (2013). Layout Design of Beam-Slab Floors by a Genetic Algorithm
(First Edit). Elsevier Inc. doi:10.1016/B978-0-12-398364-0.00007-3. (Cited on pages 1, 10)
Potschka, A. (2014). A Direct Method for Parabolic PDE Constrained Optimization Problems.
doi:10.1007/978-3-658-04476-3. arXiv: arXiv:1011.1669v3. (Cited on page 12)
Preisinger, C. (2016). PARAMETRIC STRUCTURAL MODELING User Manual for Karamba
Version 1.2.2. Karamba3D, 142. Retrieved from https://www.karamba3d.com/news/new-
manual/. (Cited on page 17)
Preisinger, C. & Bollinger-Grohmann-Schneider. (2018). Karamba 3D - Parametric engineering.
Retrieved February 26, 2018, from https://www.karamba3d.com/. (Cited on page 17)
Rennie, R. (2016). A Dictionary of Chemistry (7th edition). Oxford University Press. Retrieved from
http://www.oxfordreference.com.proxy.lib.chalmers.se/view/10.1093/acref/9780198722823.
001.0001/acref-9780198722823-e-1468. (Cited on page 7)
Ronnenbrant, R. & Glans, L.-Å. (1996). Broprojektering - En handbok (First edit). Borlänge: Enheten
för statlig väghållning. (Cited on page 6).
Rothwell, A. (2017). Optimization Methods in Structural Design. doi:10 . 1061 / (ASCE ) 0733 -
9445(1983)109:10(2420). (Cited on pages 1, 10, 12, 65–68)
Ruiz-Geli, E. (2017). What is Rhino?: Rhinoceros 5 – Modeling Tools for Designers and Architects.
Retrieved from http://www.visualarq.com/info/what-is-rhino/. (Cited on page 16)

Rutgersson, B. (2008). Kodförteckning och beskrivning av Brotyper: BaTMan Bro och Tunnel
Managment. Borlänge: Trafikverket. doi:10.1007/s11947-009-0181-3. arXiv: arXiv:1011.
1669v3. (Cited on page 4)
Sahab, M. G., Toropov, V. V., & Gandomi, A. H. (2013). A Review on Traditional and Modern
Structural Optimization: Problems and Techniques (First Edit). Elsevier Inc. doi:10.1016/B978-
0-12-398364-0.00002-4. (Cited on pages 10–13)
Sharafi, P., Hadi, M. N., & Teh, L. H. (2013). Cost Optimization of Column Layout Design of
Reinforced Concrete Buildings (First Edit). Elsevier Inc. doi:10.1016/B978-0-12-398364-
0.00006-1. (Cited on page 1)
Simon, D. (2013). Evolutionary Algorithm Variations. In Evolutionary optimization algorithms.
(Cited on pages 13, 14).
Smith, A. E. & Coit, D. W. (1995). Penalty functions. Handbook of Evolutionary Computation, 97(1),
C5. (Cited on page 13).
StruSoft. (2018). FEM-Design FEM-Design applications. StruSoft. (Cited on page 20).
Tapabrata, R. (2018). CHAPTER 2 APPLICATIONS OF MULTI-OBJECTIVE EVOLUTIONARY
ALGORITHMS IN ENGINEERING DESIGN. In Applications of multi-objective evolutionary
algorithms (Chapter 2, Pages 29–52). Singapore: World Scientific Publishing Co Pte Ltd. (Cited
on pages 14, 15).
TIBNOR. (2011). Konstruktionstabeller. Tibnor. (Cited on pages 21, 70).
Troitsky, M. S. (1994). PLANNING and DESIGN of BRIDGES. Montreal: John Wiley & Sons, Inc.
(Cited on pages 4, 5).
United Nations. (2018). Sustainable development goals: Goal 11: Make cities inclusive, safe, resilient
and sustainable. Retrieved from http://www.un.org/sustainabledevelopment/cities/page/2/.
(Cited on pages 1, 7)
Wagner, D. B. (1996). Iron and Steel in Ancient China (second). Leiden: E.J Brill. (Cited on page 6).
Wahlström, B. (1971). Bygg handbok för hus-, väg och vattenbyggnad Huvuddel. 1A, Allmänna
grunder; avd. 11-15 (Tredje upp). Stockholm: Stockholm. (Cited on pages 36, 97).
World Steel Association. (2017). World Steel in Figures 2017. World Steel Association, 3–30. doi:https:
//www.worldsteel.org/en/dam/jcr:0474d208-9108-4927-ace8-4ac5445c5df8/World+Steel+
in+Figures+2017.pdf. (Cited on page 8)
World Steel Association. (2018a). About steel. Retrieved February 8, 2018, from https : / / www.
worldsteel.org/faq/about-steel---to-delete-if-ok.html. (Cited on page 7)
World Steel Association. (2018b). Sustainability indicators. Retrieved February 7, 2018, from https:
//www.worldsteel.org/steel-by-topic/sustainability/sustainability-indicators0.html. (Cited on
page 7)
World Steel Association (WSA). (2015). Energy use in the steel industry. (51), 1–3. Retrieved from
https://www.worldsteel.org/en/dam/jcr:f07b864c-908e-4229-9f92-669f1c3abf4c/fact%7B%
5C_%7Denergy%7B%5C_%7D2016.pdf. (Cited on page 7)
Xylia, M., Silveira, S., Duerinck, J., & Meinke-Hubeny, F. (2017). Weighing regional scrap availability
in global pathways for steel production processes. Energy Efficiency, 1–25. doi:10.1007/s12053-
017-9583-7. (Cited on pages 7, 8)
Yang, X. S. (2010). Engineering Optimization: An Introduction with Metaheuristic Applications.
doi:10.1002/9780470640425. (Cited on pages 1, 12)
Yang, X.-S. (2010). Nature-inspired Metaheuristic Algorithms (Second Edi). Cambridge, United King-
dom: Luniver Press. Retrieved from https://www.researchgate.net/publication/235979455%
7B%5C_%7DNature-Inspired%7B%5C_%7DMetaheuristic%7B%5C_%7DAlgorithms. (Cited
on pages 10–13)

Zietala, M., Durejko, T., Polański, M., Kunce, I., Płociński, T., Zieliński, W., . . . Bojar, Z. (2016).
The microstructure, mechanical properties and corrosion resistance of 316 L stainless steel
fabricated using laser engineered net shaping. doi:10.1016/j.msea.2016.09.028. (Cited on
page 8)

A Gradient-free and gradient-based methods
Depending on the Taylor polynomial used to achieve an approximation of the solution of the objective
functions, see Equations (3.1) to (3.4), the methods are called gradient-free, first-order or second-order
methods, which are explained more below (Martini, 2015).
A.1 Gradient-free method

The gradient-free method, also called zero-order or derivative-free method, has as the name already
indicates no need of Taylor’s expansion to approximate the solution (Martini, 2015). The Hook and
Jeeves method is an example of the gradient-free method and uses an initial point as the start of
the search, shown as green in Figure A.1, for a maximum or minimum solution (Rothwell, 2017).
From this point initial small steps (Δ𝑥) are taken in the directions of the design variable, which is
represented in the vector shown in Equation (A.1) (Chinneck, 2001).
ℙ0 = (Δ𝑥1 , Δ𝑥2 , ⋯ , Δ𝑥𝑛 ) (A.1)
Figure A.1: Gradient-free method search process, based on (Chinneck, 2001).
After each "move", the new values of the objective functions are analyzed and compared with the
value from just before, see Figure A.1 (Chinneck, 2001). In the case of a minimization problem,
where a reduction of the objective function is sought-after, the function value is said to be improved
if it decreases. If so, this becomes the new starting point for the next step. If the function value has
increased, in the case of a minimization problem, the step is neglected and the next step is taken into
another direction. In case of a maximization problem it is the other way around. To get close to the
local minimum or maximum, the step size is reduced at each iteration. To accelerate the process the
Hooke and Jeeves method uses the search direction from the starting point 𝑥(0) to first point 𝑥(1) ,
which implies that the next point 𝑥(2) is calculated using movement form 𝑥(0) through 𝑥(1) , as can be
seen in Equation (A.2).
𝑥(2) = 2𝑥(1) − 𝑥(0) (A.2)

If the function values have been improved in this direction between 𝑥(0) and 𝑥(1) , it can be assumed
that the local optimum lies in that direction. These methods have the benefit to be simple and robust,
but are impractical for more complex problems.
A.2 Gradient-based method

Gradient-based methods are dependent on the expansion of the Taylor series and need at least a
first-order polynomial to achieve an approximation of the solution (Martini, 2015).
A.2.1 First-order methods

In the first-order method the solution, which is a linear approximation of the objective function,
is found with the first-order Taylor polynomial (Martini, 2015). One example of the first-order
method is the steepest decent method, which can be used to compute more advanced numerical
problems, than with the gradient-free methods (Rothwell, 2017). The derivative function, being a
linear approximation, can indicate in which direction the function decreases the quickest, hence the
name "steepest descent method". The formulation of the derivative of the objective function can be
seen in Equation (A.3) and can be elaborated into Equation (A.4). 𝛿𝑓 (𝑥) represents the reduction,
which needs to be as large as possible to achieve the fastest search direction, which results in a
quick convergence towards a solution. Equation (A.4) can be rewritten into Equation (A.5), with the
substitution into the derivative functions 𝑓1 , 𝑓2 and so forth and the defined search directions, which
is described in Equation (A.6).
𝑑𝑓 (𝑥) [ ]
𝑑𝑓 (𝑥) 𝑑𝑓 (𝑥) 𝑑𝑓 (𝑥)
= ∇𝑓 (𝑥) = 𝑑𝑥 , 𝑑𝑥2
, ⋯, 𝑑𝑥𝑛
(A.3)
𝑑𝕩 1
𝑑𝑓 (𝑥) 𝑑𝑓 (𝑥) 𝑑𝑓 (𝑥)

𝛿𝑓 (𝑥) = 𝛿𝑥1 + 𝛿𝑥2 + ⋯ + 𝛿𝑥𝑛 (A.4)
𝑑𝑥1 𝑑𝑥2 𝑑𝑥𝑛
∑
𝑛
= 𝑓1 𝑠1 + 𝑓2 𝑠2 + ⋯ + 𝑓𝑛 𝑠𝑛 = 𝑓𝑖 𝑠𝑖 (A.5)
𝑖=1
𝑤ℎ𝑒𝑟𝑒 𝑠1 = 𝛿𝑥1 , 𝑠2 = 𝛿𝑥2 , ⋯ , 𝑠𝑛 = 𝛿𝑥𝑛 (A.6)

As the maximum reduction of f(x) is sought, in the case of a minimization problem, 𝑠𝑖 is to be solved
and a negative derivative of the objective function defined. Equation (A.7) is the final equation, which
describes the search direction of the steepest descent method, where 𝕤 is a vector with all the search
directions, see Equation (A.8).
𝕤 = −∇𝕗 (𝕩) (A.7)
[ ]
𝑤ℎ𝑒𝑟𝑒 𝕤 = 𝑠1 , 𝑠2 , ..., 𝑠𝑛 (A.8)
Once a minimum point is found along the direction of the derivative line, see blue lines in Figure A.2,
the derivative at the minimum point, in the case of a minimization problem, is analyzed and a
new search direction found. This process is repeated until a minimum of the objective function is
found, which represents point D in Figure A.2. The steepest decent method is an iterative method

and takes several iterations to get a converging result (linear convergence). Figure A.2 shows an
example of the iterative process, with the starting point at A, continuing to B and finally ending up in
D, which is the minimum of the objective function. The circular lines represent the objective functions.
Figure A.2: Example of a search process used in the steepest decent method, based on
(Rothwell, 2017).
A.2.2 Second-order methods
There are also second-order methods, which use the second-order derivatives to improve the search
direction at each iteration (Rothwell, 2017). The Quasi-newton method is one example of such
method and it also continuously uses the previous iterations to improve the search direction for the
next iteration. The Quasi-Newton method gets faster convergence than the steepest descent method
since the second-order derivatives, see Equation (A.9), are used (quadratic convergence).
2 𝜕2 𝑓 𝜕2 𝑓
⎡ 𝜕 𝑓2 ⋯ ⎤
⎢ 𝜕𝜕𝑥2 𝑓1 𝜕𝑥1 𝜕𝑥2
𝜕2 𝑓
𝜕𝑥1 𝜕𝑥𝑛
𝜕2 𝑓
⎥
⎢ ⋯ ⎥
𝐻(𝑥) = ⎢ 𝜕𝑥22 𝜕𝑥22
⎥
𝜕𝑥2 𝜕𝑥𝑛 (A.9)
⎢ ⋮ ⋮ ⋱ ⋮ ⎥
⎢ 𝜕2 𝑓 𝜕𝑑 2 𝑓
⋯ 𝜕2 𝑓 ⎥
⎣ 𝜕𝑥𝑛 𝜕𝑥1 𝜕𝑥𝑛 𝜕𝑥2 𝜕𝑥2𝑛 ⎦

In Equation (A.10) the position of a point 𝕩′ can be calculated using the "Hessian" matrix ℍ which
contains the second-order derivatives of the objective function (Rothwell, 2017). When moving from
a point 𝕩 to another point 𝕩′ the gradient will change and the Hessian-matrix will be evaluated at
each point. Equation (A.10) can be used to calculate the minimum point if the gradient, ∇𝑓 (𝕩), is zero.
𝕩′ = 𝕩 − ℍ(𝕩)−1 ∇𝑓 (𝕩) (A.10)
𝕤 = −𝔹∇𝑓 (𝕩) (A.11)
ℍ𝑘+1 = 𝔹−1
𝑘+1
(A.12)
The new search direction is stated in Equation (A.11) and is continuously updated with the result
obtained from the previous iteration (Rothwell, 2017). The subscripts (k+1) and (k) are referencing
to the next iteration and the current iteration. One way to update the 𝔹 matrix is the Broyden-Fletcher-
Goldfarb-Shannon (BFGS)-method, see Equation (A.13). Where the new 𝔹-matrix is updated using
the results from the previous iterations. For the first iteration 𝔹 = 𝕀 which is the unit-matrix.
( )
𝕪 𝑇 𝔹(𝕜) 𝕪 𝕡𝕡𝕋 𝕡𝕪 𝕋 𝔹(𝐾) 𝔹(𝑘) 𝕪𝕡𝑇
𝑘+1
𝔹 =𝔹 + 1+ 𝑘
− − (A.13)
𝕡𝑇 𝕪 𝕡𝑇 𝕪 𝕡𝑇 𝕪 𝕡𝑇 𝕪
Equation (A.14) are the gradient differences which are used to estimate the new 𝔹-matrix and Equa-
tion (A.15) is the position difference (Rothwell, 2017).
𝕪 = ∇𝑓 (𝕩)(𝑘+1) − ∇𝑓 (𝕩)(𝑘) (A.14)
𝕡 = 𝕩(𝑘+1) − 𝕩(𝑘) (A.15)

B
A-2321
KNUTPUNKSDETALJER
Rv 400 HÄNVISNINGAR
SE RITNING -2322
1 4000 4000 4000 4000 4000
2
ANVISNINGAR OCH FÖRESRIFTER,
SE RITNING 101K2021
6 5 3 5 5 5 5 800 * 4 5 5 5 6
BETECKNINGAR
7 8 FL = FAST LAGER
8 7 EL = ENKELSIDIGT RÖRLIGT LAGER
** ** AL = ALLSIDIGT RÖRLIGT LAGER
800 * 2
1
2000 4000 4000 4000 4000 2000
KNUTPUNKSDETALJER
A-2321 SE RITNING -2322
ELEVATION 1:50
DETALJER SE RITNINGAR -2322, -2323, -2324
*) SKARVAR I ÖVRE OCH UNDRE RAMSTÄNGER
**) AVVÄGNINGSDUBB PLACERAS I BROMITT OCH BROÄNDAR
660 670 670 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 800 670 670 660 9 TVÄRBALKAR
C
EL AL
12 ÄNDTVÄRBALKAR
12
ÖVERGÅNGSKONSTRUKTION ÖVERGÅNGSKONSTRUKTION
TYP MAURER-FOG TYP MAURER-FOG
SE DETALJER RITNING -2324 SE DETALJER RITNING -2324
A
FL EL
1380 1470 1600 1600 1600 1600 1600 1600 1600 1600 1600 1470 1380 10 GÅNGBANEPLÅT
PLAN 1:50
B
C
2500 A
3 4
FRIBROBREDD
DIAGONALER
6 5
POS BENÄMNING DIMENSION ANTAL KVALITET
0041
0011
0551
ÄNDVERTIKALER
0021
8 7 1 HÅLPROFIL VKR 150x100x6.3 2 S355J2H
2 HÅLPROFIL VKR 150x100x6.3 2 S355J2H

10 BET ANT ÄNDRINGEN AVSER DATUM SIGN
2,5% 2 1 3 HÅLPROFIL VKR 150x150x6.3 2 S355J2H
KLABMAR TOM TTÄRLEKNIV TTÅM

11
GRANSKNINGSHANDLING
021
4 HÅLPROFIL VKR 150x150x6.3 2 S355J2H
66
+2,200 12 +2,200 5 HÅLPROFIL VKR 100x60x5 16 S355J2H GC-BRO ÖVER GENEVADSÅN
P LYFT P LYFT 6 HÅLPROFIL VKR 100x60x5 4 S355J2H
Ramböll Sverige AB
850 850 Vädursgatan 6
7 HÅLPROFIL VKR 100x100x5 2 S355J2H Box 5343
402 27 Göteborg
Figure B.1: Used plans from the case study over Genevadsån.
8 HÅLPROFIL VKR 100x100x5 2 S355J2H
Plans of the case study over Genevadsån
2722
Tfn: 010 615 60 00
9 I-PROFIL IPE 140 25 S355J2 www.ramboll.se
alkeT\lledoM\K\kinkeT\1731300231\7102\gyB\2toG\buP\krowten.labolg.puorg-llobmar.esmar\\
UPPDRAG.NR RITAD/KONSTR AV HANDLÄGGARE

SEKTION A-2321 1:20 1320031371 VIC/LVE L. VIBE
DETALJER SE RITNINGAR -2322, -2323, -2324 10 GÅNGBANEPLÅT t= 10 mm - S355J2
DATUM ANSVARIG
P = 125kN MOT ÄNDTVÄRBALK 2018-01-25 M. Nummedal
LYFT 11 AVVÄGNINGSDUBB ø20 L=30 6 S235JRG2
ÖVERBYGGNAD
12 I-PROFIL IPE 180 2 S355J2 PLAN, ELEVATION OCH SEKTION
FÖRESKRIFT
1 0 1 2 3 4 5 6 7 8 9 10 m
SKALA SKALA NUMMER BET
1:100
XREF: Plottad:
SE FIG. 101K2321
, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

69
C Cross-sections inserted in Karamba 3D
In this appendix first the VKR, then the KCKR and at last the IPE profiles, which were implemented
into Karamba 3D, are presented. The values are taken from a construction table based on SS-EN10210
(TIBNOR, 2011) and the imperfection values are calculated according to EN 1993-1-1 chapter 6.3.1.

VKR-profiles
Second Elastic Second Imperfection
moment modulus Plastic Radius of moment Elastic Plastic Radius of Imperfection Imperfection value for
Upper Thickness Position Effective Effective of area upper modulus gyration of area modulus modulus gyration Torsional Elastic value for value for lateral Type of
Depth of Thickness width of of upper of Area of shear shear (strong (strong (Strong (strong (weak (weak (weak (weak constant modulus Warping buckling buckling torsional product
section of web section flange centroid section area Vy area Vz axis axis) axis) axis) axis) axis) axis) axis) inertia torsion factor (strong axis) (weak axis) buckling (*)
h t_web b_upper t_upper zs A Ay Az Iy Wy Wply i_y Iz Wz Wplz i_z It Wt Cw alpha_y alpha_z alpha_LT Product
Contry Family Name Shape mm mm mm mm cm cm2 cm2 cm2 cm4 cm3 cm3 cm cm4 cm3 cm3 cm cm4 cm3 cm6 - - - -
Sweden VKR VKR100x100x4 [] 100 4 100 4 5 15,2 7,6 7,6 232 46,4 54,4 3,91 232 46,4 54,4 3,91 361 68,2 0 0,21 0,21 0,76 3
Sweden VKR VKR100x100x5 [] 100 5 100 5 5 18,7 9,35 9,35 279 55,9 66,4 3,86 279 55,9 66,4 3,86 439 81,8 0 0,21 0,21 0,76 3
Sweden VKR VKR100x100x6,3 [] 100 6,3 100 6,3 5 23,2 11,6 11,6 336 67,1 80,9 3,8 336 67,1 80,9 3,8 534 97,8 0 0,21 0,21 0,76 3
Sweden VKR VKR100x100x8 [] 100 8 100 8 5 28,8 14,4 14,4 400 79,9 99,2 3,73 400 79,9 99,2 3,73 646 116 0 0,21 0,21 0,76 3
Sweden VKR VKR100x100x10 [] 100 10 100 10 5 34,9 17,45 17,45 462 92,4 116 3,64 462 92,4 116 3,64 761 133 0 0,21 0,21 0,76 3
Sweden VKR VKR120x120x4,5 [] 120 4,5 120 4,5 6 20,5 10,25 10,25 452 75,3 88,2 4,7 452 75,3 88,2 4,7 707 112 0 0,21 0,21 0,76 3
Sweden VKR VKR120x120x5 [] 120 5 120 5 6 22,9 11,45 11,45 503 83,8 98,4 4,68 503 83,8 98,4 4,68 777 122 0 0,21 0,21 0,76 3
Sweden VKR VKR120x120x6,3 [] 120 6,3 120 6,3 6 28,5 14,25 14,25 610 102 121 4,62 610 102 121 4,62 950 147 0 0,21 0,21 0,76 3
Sweden VKR VKR120x120x8 [] 120 8 120 8 6 35,5 17,75 17,75 738 123 149 4,55 738 123 149 4,55 1160 176 0 0,21 0,21 0,76 3
Sweden VKR VKR120x120x10 [] 120 10 120 10 6 43,5 21,75 21,75 870 145 178 4,46 870 145 178 4,46 1382 206 0 0,21 0,21 0,76 3
Sweden VKR VKR140x140x5 [] 140 5 140 5 7 26,9 13,45 13,45 814 116 136 5,5 814 116 136 5,5 1253 170 0 0,21 0,21 0,76 3
Sweden VKR VKR140x140x6,3 [] 140 6,3 140 6,3 7 33,5 16,75 16,75 994 142 168 5,44 994 142 168 5,44 1540 206 0 0,21 0,21 0,76 3
Sweden VKR VKR140x140x8 [] 140 8 140 8 7 41,9 20,95 20,95 1212 173 207 5,36 1212 173 207 5,36 1892 249 0 0,21 0,21 0,76 3
Sweden VKR VKR140x140x10 [] 140 10 140 10 7 51,5 25,75 25,75 1441 206 250 5,27 1441 206 250 5,27 2272 294 0 0,21 0,21 0,76 3
Sweden VKR VKR150x150x5 [] 150 5 150 5 7,5 28,9 14,45 14,45 1009 135 157 5,9 1009 135 157 5,9 1550 197 0 0,21 0,21 0,76 3
Sweden VKR VKR150x150x6,3 [] 150 6,3 150 6,3 7,5 36 18 18 1236 165 194 5,85 1236 165 194 5,85 1909 240 0 0,21 0,21 0,76 3
Sweden VKR VKR150x150x8 [] 150 8 150 8 7,5 45,1 22,55 22,55 1510 201 240 5,77 1510 201 240 5,77 2351 291 0 0,21 0,21 0,76 3
Sweden VKR VKR150x150x10 [] 150 10 150 10 7,5 55,5 27,75 27,75 1803 240 290 5,68 1803 240 290 5,68 2832 344 0 0,21 0,21 0,76 3
Sweden VKR VKR160x160x5,6 [] 160 5,6 160 5,6 8 33,8 16,9 16,9 1330 166 194 6,26 1330 166 194 6,26 2333 275 0 0,21 0,21 0,76 3
Sweden VKR VKR160x160x6,3 [] 160 6,3 160 6,3 8 38,3 19,15 19,15 1499 187 220 6,18 1499 187 220 6,18 2880 335 0 0,21 0,21 0,76 3
Sweden VKR VKR160x160x8 [] 160 8 160 8 8 48 24 24 1831 229 272 6,09 1831 229 272 6,09 3478 398 0 0,21 0,21 0,76 3
Sweden VKR VKR160x160x10 [] 160 10 160 10 8 58,9 29,45 29,45 2186 273 329 7,07 2186 273 329 7,07 3361 355 0 0,21 0,21 0,76 3
Sweden VKR VKR180x180x6,3 [] 180 6,3 180 6,3 9 43,6 21,8 21,8 2186 243 283 7 2186 243 283 7 4162 434 0 0,21 0,21 0,76 3
Sweden VKR VKR180x180x8 [] 180 8 180 8 9 54,7 27,35 27,35 2689 299 352 6,91 2689 299 352 6,91 5048 518 0 0,21 0,21 0,76 3
Sweden VKR VKR180x180x10 [] 180 10 180 10 9 67,5 33,75 33,75 3237 360 429 7,89 3237 360 429 7,89 4653 444 0 0,21 0,21 0,76 3
Sweden VKR VKR200x200x6,3 [] 200 6,3 200 6,3 10 48,6 24,3 24,3 3033 303 353 7,81 3033 303 353 7,81 5778 545 0 0,21 0,21 0,76 3
Sweden VKR VKR200x200x8 [] 200 8 200 8 10 61,1 30,55 30,55 3744 374 439 7,72 3744 374 439 7,72 7031 655 0 0,21 0,21 0,76 3
Sweden VKR VKR200x200x10 [] 200 10 200 10 10 75,5 37,75 37,75 4525 452 536 7,61 4525 452 536 7,61 8491 778 0 0,21 0,21 0,76 3
Sweden VKR VKR200x200x12,5 [] 200 12,5 200 12,5 10 93 46,5 46,5 5419 542 651 7,46 5419 542 651 7,46 10340 927 0 0,21 0,21 0,76 3
Sweden VKR VKR200x200x16 [] 200 16 200 16 10 117 58,5 58,5 6524 652 799 8,71 6524 652 799 8,71 6240 544 0 0,21 0,21 0,76 3
Sweden VKR VKR220x220x6,3 [] 220 6,3 220 6,3 11 52,8 26,4 26,4 3980 362 421 8,54 3980 362 421 8,54 9473 807 0 0,21 0,21 0,76 3
Sweden VKR VKR220x220x10 [] 220 10 220 10 11 81,4 40,7 40,7 5890 535 634 9,93 5890 535 634 9,93 9238 712 0 0,21 0,21 0,76 3
Sweden VKR VKR250x250x6,3 [] 250 6,3 250 6,3 12,5 61,2 30,6 30,6 6049 484 559 9,86 6049 484 559 9,86 11525 880 0 0,21 0,21 0,76 3
Sweden VKR VKR250x250x8 [] 250 8 250 8 12,5 77,1 38,55 38,55 7510 601 699 9,77 7510 601 699 9,77 14106 1065 0 0,21 0,21 0,76 3
Sweden VKR VKR250x250x10 [] 250 10 250 10 12,5 95,5 47,75 47,75 9141 731 858 9,66 9141 731 858 9,66 17164 1279 0 0,21 0,21 0,76 3
Sweden VKR VKR250x250x12,5 [] 250 12,5 250 12,5 12,5 118 59 59 11050 884 1048 9,5 11050 884 1048 9,5 21138 1546 0 0,21 0,21 0,76 3
Sweden VKR VKR250x250x16 [] 250 16 250 16 12,5 149 74,5 74,5 13480 1078 1298 11,8 13480 1078 1298 11,8 24807 1575 0 0,21 0,21 0,76 3
Sweden VKR VKR300x300x10 [] 300 10 300 10 15 116 58 58 16150 1077 1254 11,7 16150 1077 1254 11,7 30333 1904 0 0,21 0,21 0,76 3
Sweden VKR VKR300x300x12,5 [] 300 12,5 300 12,5 15 143 71,5 71,5 19630 1309 1538 11,5 19630 1309 1538 11,5 37622 2325 0 0,21 0,21 0,76 3
Sweden VKR VKR300x300x16 [] 300 16 300 16 15 181 90,5 90,5 24160 1610 1916 13,9 24160 1610 1916 13,9 39886 2185 0 0,21 0,21 0,76 3
Sweden VKR VKR350x350x10 [] 350 10 350 10 17,5 136 68 68 26050 1489 1725 13,7 26050 1489 1725 13,7 48934 2654 0 0,21 0,21 0,76 3
Sweden VKR VKR350x350x12,5 [] 350 12,5 350 12,5 17,5 168 84 84 31810 1817 2122 13,6 31810 1817 2122 13,6 60990 3264 0 0,21 0,21 0,76 3
Sweden VKR VKR400x400x10 [] 400 10 400 10 20 156 78 78 39350 1968 2272 15,9 39350 1968 2272 15,9 60092 2895 0 0,21 0,21 0,76 3
Sweden VKR VKR400x400x12,5 [] 400 12,5 400 12,5 20 193 96,5 96,5 48190 2409 2800 15,8 48190 2409 2800 15,8 73906 3530 0 0,21 0,21 0,76 3
Sweden VKR VKR400x400x16 [] 400 16 400 16 20 245 122,5 122,5 59905 2995 3514 15,6 59905 2995 3514 15,6 92442 4362 0 0,21 0,21 0,76 3
Sweden VKR VKR100x60x3,6 [] 100 3,6 60 3,6 5 10,9 4,1 6,8 145 28,9 35,6 3,58 64,4 21,6 24,9 2,44 142 35,6 0 0,21 0,21 0,76 3
Sweden VKR VKR100x60x4 [] 100 4 60 4 5 12 4,5 7,5 158 31,6 39,1 3,53 70,5 23,5 27,3 2,43 156 38,7 0 0,21 0,21 0,76 3
Sweden VKR VKR100x60x5 [] 100 5 60 5 5 14,7 5,5 9,2 189 37,8 47,4 3,48 83,6 27,9 32,9 2,38 188 45,9 0 0,21 0,21 0,76 3
Sweden VKR VKR100x60x5,6 [] 100 5,6 60 5,6 5 16,3 6,1 10,2 206 41,2 52 3,45 90,6 30,2 36,1 2,36 205 49,7 0 0,21 0,21 0,76 3
Sweden VKR VKR100x60x6,3 [] 100 6,3 60 6,3 5 18,1 6,8 11,3 225 45 57,3 3,42 98,1 32,7 39,5 2,33 224 53,8 0 0,21 0,21 0,76 3
Sweden VKR VKR120x60x3,6 [] 120 3,6 60 3,6 6 12,3 4,1 8,2 227 37,9 47,2 3,33 76,3 25,4 28,9 2,49 183 43,3 0 0,21 0,21 0,76 3
Sweden VKR VKR120x60x4 [] 120 4 60 4 6 13,6 4,5 9,1 249 41,5 51,9 3,65 83,1 27,7 31,7 2,47 201 47,1 0 0,21 0,21 0,76 3
Sweden VKR VKR120x60x5 [] 120 5 60 5 6 16,7 5,6 11,1 299 49,9 63,1 3,63 98,9 32,9 38,4 2,43 242 56 0 0,21 0,21 0,76 3
Sweden VKR VKR120x60x6,3 [] 120 6,3 60 6,3 6 20,7 6,9 13,8 358 59,7 76,7 3,58 116 38,8 46,3 2,37 290 65,9 0 0,21 0,21 0,76 3
Sweden VKR VKR120x80x4 [] 120 4 80 4 6 15,2 6,08 9,1 303 50,4 61,2 3,55 161 40,2 46,1 3,25 330 65 0 0,21 0,21 0,76 3
Sweden VKR VKR120x80x5 [] 120 5 80 5 6 18,7 7,48 11,2 365 60,9 74,6 3,52 193 48,2 56,1 3,21 401 77,9 0 0,21 0,21 0,76 3
Sweden VKR VKR120x80x6,3 [] 120 6,3 80 6,3 6 23,2 9,28 13,9 440 73,3 91 4,3 230 57,6 68,2 3,15 487 92,9 0 0,21 0,21 0,76 3
Sweden VKR VKR120x80x8 [] 120 8 80 8 6 28,8 11,52 17,3 525 87,5 111 4,28 273 68,1 82,6 3,08 587 110 0 0,21 0,21 0,76 3
Sweden VKR VKR140x70x4 [] 140 4 70 4 7 16 5,3 10,7 404 57,7 71,7 4,23 136 38,8 44 2,91 325 66 0 0,21 0,21 0,76 3
Sweden VKR VKR140x70x5 [] 140 5 70 5 7 19,7 6,6 13,1 488 69,8 87,6 4,16 163 46,5 53,5 2,87 394 79 0 0,21 0,21 0,76 3
Sweden VKR VKR140x70x6,3 [] 140 6,3 70 6,3 7 24,4 8,1 16,3 589 84,2 107 4,46 194 55,3 65 2,81 477 94 0 0,21 0,21 0,76 3
Sweden VKR VKR140x80x4 [] 140 4 80 4 7 16,8 6,1 10,7 441 62,9 77,1 4,42 184 46 52,2 3,31 411 76,5 0 0,21 0,21 0,76 3
Sweden VKR VKR140x80x6,3 [] 140 6,3 80 6,3 7 25,7 9,3 16,4 646 92,3 115 4,36 265 66,2 77,5 3,21 607 110 0 0,21 0,21 0,76 3
Sweden VKR VKR150x100x5 [] 150 5 100 5 7,5 23,7 9,48 14,2 739 98,5 119 4,27 392 78,5 90,1 4,07 807 127 0 0,21 0,21 0,76 3
Sweden VKR VKR150x100x6,3 [] 150 6,3 100 6,3 7,5 29,5 11,8 17,7 898 120 147 5,5 474 94,8 110 4,01 986 153 0 0,21 0,21 0,76 3
Sweden VKR VKR150x100x8 [] 150 8 100 8 7,5 36,8 14,72 22,1 1087 145 180 4,98 569 114 135 3,94 1203 183 0 0,21 0,21 0,76 3

Sweden VKR VKR150x100x10 [] 150 10 100 10 7,5 44,9 17,96 26,9 1282 171 216 4,91 665 133 161 3,85 1432 214 0 0,21 0,21 0,76 3
Sweden VKR VKR160x80x4 [] 160 4 80 4 8 18,4 6,1 12,3 612 76,5 94,7 5,77 207 51,7 58,3 3,35 493 88,1 0 0,21 0,21 0,76 3
(*) 0:cold formed; 1: hot finished; 2: welded; 3: rolled;
71
72
VKR-profiles
Upper Thickness Position Effective Effective of area upper modulus gyration of area modulus modulus gyration Torsional Elastic value for value for lateral Type of
Depth of Thickness width of of upper of Area of shear shear (strong (strong (Strong (strong (weak (weak (weak (weak constant modulus Warping buckling buckling torsional product
section of web section flange centroid section area Vy area Vz axis axis) axis) axis) axis) axis) axis) axis) inertia torsion factor (strong axis) (weak axis) buckling (*)
h t_web b_upper t_upper zs A Ay Az Iy Wy Wply i_y Iz Wz Wplz i_z It Wt Cw alpha_y alpha_z alpha_LT Product
Contry Family Name Shape mm mm mm mm cm cm2 cm2 cm2 cm4 cm3 cm3 cm cm4 cm3 cm3 cm cm4 cm3 cm6 - - - -
Sweden VKR VKR160x80x5 [] 160 5 80 5 8 22,7 7,6 15,1 744 93 116 5,72 249 62,3 71,1 3,31 600 106 0 0,21 0,21 0,76 3
Sweden VKR VKR160x80x6,3 [] 160 6,3 80 6,3 8 28,2 9,4 18,8 903 113 142 5,66 299 74,8 86,8 3,26 730 127 0 0,21 0,21 0,76 3
Sweden VKR VKR160x80x8 [] 160 8 80 8 8 35,2 11,7 23,5 1091 136 175 5,57 356 89 106 3,18 883 151 0 0,21 0,21 0,76 3
Sweden VKR VKR160x80x10 [] 160 10 80 10 8 42,9 14,3 28,6 1284 161 209 5,47 411 103 125 3,1 1041 175 0 0,21 0,21 0,76 3
Sweden VKR VKR160x90x5 [] 160 5 90 5 8 23,7 8,532 15,2 804 101 124 5,82 326 72,5 82,7 3,71 738 121 0 0,21 0,21 0,76 3
Sweden VKR VKR160x90x7,1 [] 160 7,1 90 7,1 8 32,9 11,844 21,1 1080 135 169 5,72 431 95,7 112 3,62 995 160 0 0,21 0,21 0,76 3
Sweden VKR VKR160x90x8 [] 160 8 90 8 8 36,8 13,248 23,6 1180 148 187 5,68 470 105 124 3,58 1100 174 0 0,21 0,21 0,76 3
Sweden VKR VKR180x100x5,6 [] 180 5,6 100 5,6 9 29,8 10,6 19,2 1270 141 174 6,54 506 101 115 4,12 1150 169 0 0,21 0,21 0,76 3
Sweden VKR VKR180x100x6,3 [] 180 6,3 100 6,3 9 33,3 11,9 21,4 1407 156 194 6,5 557 111 128 4,09 1277 186 0 0,21 0,21 0,76 3
Sweden VKR VKR180x100x8 [] 180 8 100 8 9 41,6 14,9 26,7 1713 190 239 6,42 671 134 157 4,02 1560 224 0 0,21 0,21 0,76 3
Sweden VKR VKR180x100x10 [] 180 10 100 10 9 50,9 18,2 32,7 2036 226 288 6,32 787 157 188 3,93 1862 263 0 0,21 0,21 0,76 3
Sweden VKR VKR200x100x5 [] 200 5 100 5 10 28,7 9,6 19,1 1495 149 185 7,21 505 101 114 4,19 1204 172 0 0,21 0,21 0,76 3
Sweden VKR VKR200x100x6,3 [] 200 6,3 100 6,3 10 35,8 11,9 23,9 1829 183 228 7,15 613 123 140 4,14 1475 208 0 0,21 0,21 0,76 3
Sweden VKR VKR200x100x8 [] 200 8 100 8 10 44,8 14,9 29,9 2234 223 282 7,06 739 148 172 4,06 1804 251 0 0,21 0,21 0,76 3
Sweden VKR VKR200x100x10 [] 200 10 100 10 10 54,9 18,3 36,6 2664 266 341 6,96 869 174 206 3,98 2156 295 0 0,21 0,21 0,76 3
Sweden VKR VKR200x100x12,5 [] 200 12,5 100 12,5 10 67,1 22,4 44,7 3136 314 408 6,84 1004 201 245 3,87 2541 341 0 0,21 0,21 0,76 3
Sweden VKR VKR200x120x6,3 [] 200 6,3 100 6,3 10 38,3 12,8 25,5 2065 207 253 7,34 929 155 177 4,92 2028 255 0 0,21 0,21 0,76 3
Sweden VKR VKR200x120x8 [] 200 8 120 8 10 48 18 30 2529 253 313 7,26 1128 188 218 4,85 2495 310 0 0,21 0,21 0,76 3
Sweden VKR VKR200x120x10 [] 200 10 120 10 10 58,9 22,1 36,8 3026 303 379 7,17 1337 223 263 4,76 3001 367 0 0,21 0,21 0,76 3
Sweden VKR VKR220x120x6,3 [] 220 6,3 120 6,3 11 40,8 14,4 26,4 2610 237 292 8 1010 168 191 4,98 2320 283 0 0,21 0,21 0,76 3
Sweden VKR VKR220x120x8 [] 220 8 120 8 11 51,2 18,1 33,1 3200 291 362 7,91 1230 205 236 4,9 2850 343 0 0,21 0,21 0,76 3
Sweden VKR VKR220x120x10 [] 220 10 120 10 11 62,9 22,2 40,7 3840 349 440 7,82 1460 243 285 4,81 3430 407 0 0,21 0,21 0,76 3
Sweden VKR VKR250x150x6,3 [] 250 6,3 150 6,3 12,5 48,4 18,15 30,3 4143 331 402 9,25 1874 250 283 6,22 4054 413 0 0,21 0,21 0,76 3
Sweden VKR VKR250x150x8 [] 250 8 150 8 12,5 60,8 22,8 38 5111 409 501 9,17 2298 306 350 6,15 5021 506 0 0,21 0,21 0,76 3
Sweden VKR VKR250x150x10 [] 250 10 150 10 12,5 74,9 28,1 46,8 6174 494 611 9,08 2755 367 426 6,06 6090 605 0 0,21 0,21 0,76 3
Sweden VKR VKR250x150x12,5 [] 250 12,5 150 12,5 12,5 92,1 34,5 57,6 7387 591 740 8,96 3265 435 514 5,96 7326 717 0 0,21 0,21 0,76 3
Sweden VKR VKR260x140x6,3 [] 260 6,3 140 6,3 13 48,4 16,94 31,5 4350 335 411 9,49 1660 237 267 5,86 3800 399 0 0,21 0,21 0,76 3
Sweden VKR VKR260x140x8 [] 260 8 140 8 13 60,8 21,28 39,5 5370 413 511 9,4 2030 290 331 5,78 4700 488 0 0,21 0,21 0,76 3
Sweden VKR VKR300x200x6,3 [] 300 6,3 200 6,3 15 61 24,4 36,6 7829 522 624 11,3 4193 419 472 8,29 8476 681 0 0,21 0,21 0,76 3
Sweden VKR VKR300x200x8 [] 300 8 200 8 15 76,8 30,72 46,1 9717 648 779 11,3 5184 518 589 8,22 10562 840 0 0,21 0,21 0,76 3
Sweden VKR VKR300x200x10 [] 300 10 200 10 15 94,9 37,96 56,9 11819 788 956 11,2 6278 628 721 8,13 12908 1015 0 0,21 0,21 0,76 3
Sweden VKR VKR300x200x12,5 [] 300 12,5 200 12,5 15 117 46,8 70,2 14273 952 1165 11 7537 754 877 8,02 15677 1217 0 0,21 0,21 0,76 3
Sweden VKR VKR300x200x16 [] 300 16 200 16 15 147 58,8 88,2 17390 1159 1441 10,9 9109 911 1080 7,87 19252 1468 0 0,21 0,21 0,76 3
Sweden VKR VKR400x200x10 [] 400 10 200 10 20 115 38,3 76,7 23914 1196 1480 14,4 8084 808 911 8,38 19259 1376 0 0,21 0,21 0,76 3
Sweden VKR VKR400x200x12,5 [] 400 12,5 200 12,5 20 142 47,3 94,7 29063 1453 1813 14,3 9738 974 1111 8,28 23438 1656 0 0,21 0,21 0,76 3
Sweden VKR VKR400x200x16 [] 400 16 200 16 20 179 59,7 119,3 35738 1787 2256 14,1 11824 1182 1374 8,13 28871 2010 0 0,21 0,21 0,76 3
Sweden VKR VKR450x250x10 [] 450 10 250 10 22,5 135 48,2 86,8 36895 1640 2000 16,5 14819 1185 1331 10,5 33284 1986 0 0,21 0,21 0,76 3
Sweden VKR VKR450x250x12,5 [] 450 12,5 250 12,5 22,5 167 59,6 107,4 45026 2001 2458 16,4 17973 1438 1631 10,4 40719 2406 0 0,21 0,21 0,76 3
Sweden VKR VKR450x250x16 [] 450 16 250 16 22,5 211 75,35 135,6 55705 2476 3070 16,2 22041 1763 2029 10,2 50545 2947 0 0,21 0,21 0,76 3

Circular cross sections (KCKR)

Position Effective Effective of area upper modulus gyration of area modulus modulus gyration Torsional Elastic value for value for lateral Type of
Depth of Thickness of Area of shear shear (strong (strong (Strong (strong (weak (weak (weak (weak constant modulus Warping buckling buckling torsional product
section of web centroid section area Vy area Vz axis axis) axis) axis) axis) axis) axis) axis) inertia torsion factor (strong axis) (weak axis) buckling (*)
h t_web zs A Ay Az Iy Wy Wply i_y Iz Wz Wplz i_z It Wt Cw alpha_y alpha_z alpha_LT Product
Contry Family Name Shape mm mm cm cm2 cm2 cm2 cm4 cm3 cm3 cm cm4 cm3 cm3 cm cm4 cm3 cm6 ‐ ‐ ‐ ‐
Sweden KCKR KCKR40,4/3 O 42.4 3 2.12 3.71 2.36 2.36 7.25 3.42 4.67 1.4 7.25 3.42 4.67 1.4 14.5 6.84 0 0.49 0.49 0.76 0
Sweden KCKR KCKR40,4/4 O 42.4 4 2.12 4.83 3.07 3.07 8.99 4.24 5.92 1.36 8.99 4.24 5.92 1.36 18 8.48 0 0.49 0.49 0.76 0
Sweden KCKR KCKR48,3/3 O 48.3 3 2.415 4.27 2.72 2.72 11 4.55 6.17 1.61 11 4.55 6.17 1.61 22 9.11 0 0.49 0.49 0.76 0
Sweden KCKR KCKR60,3/4 O 60.3 4 3.015 7.07 4.50 4.50 28.2 9.34 12.7 2 28.2 9.34 12.7 2 56.3 18.7 0 0.49 0.49 0.76 0
Sweden KCKR KCKR76,1/4 O 76.1 4 3.805 9.06 5.77 5.77 59.1 5.15 20.8 2.55 59.1 5.15 20.8 2.55 118 31 0 0.49 0.49 0.76 0
Sweden KCKR KCKR76,1/5 O 76.1 5 3.805 11.2 7.13 7.13 70.9 18.6 25.3 2.52 70.9 18.6 25.3 2.52 142 37.3 0 0.49 0.49 0.76 0
Sweden KCKR KCKR88,9/4 O 88.9 4 4.445 10.7 6.81 6.81 96.3 21.7 28.9 3 96.3 21.7 28.9 3 193 43.3 0 0.49 0.49 0.76 0
Sweden KCKR KCKR101,6/6 O 101.6 6 5.08 18 11.46 11.46 207 40.7 54.9 3.39 207 40.7 54.9 3.39 413 81.4 0 0.49 0.49 0.76 0
Sweden KCKR KCKR114,3/5 O 114.3 5 5.715 17.2 10.95 10.95 257 45 59.8 3.87 257 45 59.8 3.87 514 89.9 0 0.49 0.49 0.76 0
Sweden KCKR KCKR139,7/4 O 139.7 4 6.985 17.1 10.89 10.89 393 56.2 73.7 4.8 393 56.2 73.7 4.8 786 112 0 0.49 0.49 0.76 0
Sweden KCKR KCKR139,7/6 O 139.7 6 6.985 25.2 16.04 16.04 564 80.8 107 4.73 564 80.8 107 4.73 1129 162 0 0.49 0.49 0.76 0
Sweden KCKR KCKR139,7/8 O 139.7 8 6.985 33.1 21.07 21.07 720 103 139 4.66 720 103 139 4.66 1441 206 0 0.49 0.49 0.76 0
Sweden KCKR KCKR168,3/4 O 168.3 4 8.415 20.6 13.11 13.11 697 82.8 108 5.81 697 82.8 108 5.81 1394 166 0 0.49 0.49 0.76 0
Sweden KCKR KCKR193,7/12,5 O 193.7 12.5 9.685 71.2 45.33 45.33 2934 303 411 6.42 2934 303 411 6.42 5869 606 0 0.49 0.49 0.76 0
Sweden KCKR KCKR244,5/6 O 244.5 6 12.225 45 28.65 28.65 3199 262 341 8.43 3199 262 341 8.43 6397 523 0 0.49 0.49 0.76 0
Sweden KCKR KCKR273/8 O 273 8 13.65 66.6 42.40 42.40 5852 429 562 9.37 5852 429 562 9.37 11703 857 0 0.49 0.49 0.76 0
Sweden KCKR KCKR273/10 O 273 10 13.65 82.6 52.58 52.58 7174 524 692 9.31 7174 524 692 9.31 14308 1048 0 0.49 0.49 0.76 0
Sweden KCKR KCKR273/12,5 O 273 12.5 13.65 102 64.94 64.94 8697 637 849 9.22 8697 637 849 9.22 17395 1274 0 0.49 0.49 0.76 0
Sweden KCKR KCKR323,9/12,5 O 323.9 12.5 16.195 122 77.67 77.67 14847 917 1213 11 14847 917 1213 11 29693 1833 0 0.49 0.49 0.76 0
(*) 0:cold formed; 1: hot finished; 2: welded; 3: rolled

73
74
European I beams
Upper Position Effective Effective of area upper modulus gyration of area modulus modulus gyration Torsional Elastic value for value for lateral Type of
Depth of Thickness width of Thickness of Root of Area of shear shear (strong (strong (Strong (strong (weak (weak (weak (weak constant modulus Warping buckling buckling torsional product
section of web section upper flange radius centroid section area Vy area Vz axis axis) axis) axis) axis) axis) axis) axis) inertia torsion factor (strong axis) (weak axis) buckling (*)
h t_web b_upper t_upper r zs A Ay Az Iy Wy Wply i_y Iz Wz Wplz i_z It Wt Cw alpha_y alpha_z alpha_LT Product
Contry Family Name Shape mm mm mm mm mm cm cm2 cm2 cm2 cm4 cm3 cm3 cm cm4 cm3 cm3 cm cm4 cm3 cm6 ‐ ‐ ‐ ‐
EU IPE IPE80 I 80 3.8 46 5.2 5 4 7.64 4.78 2.64 80 20.0 23.2 3.24 8.5 3.69 5.82 1.05 0.7 1.35 118 0.21 0.34 0.34 3
EU IPE IPE100 I 100 4.1 55 5.7 7 5 10.32 6.27 3.63 171 34.2 39.4 4.07 15.9 5.79 9.15 1.24 1.21 2.12 351 0.21 0.34 0.34 3
EU IPE IPE120 I 120 4.4 64 6.3 7 6 13.21 8.06 4.73 318 53 60.7 4.9 27.7 8.65 13.6 1.45 1.74 2.76 890 0.21 0.34 0.34 3
EU IPE IPE140 I 140 4.7 73 6.9 7 7 16.43 10.07 5.93 541 77.3 88.3 5.74 45 12.3 19.2 1.65 2.45 3.55 1900 0.21 0.34 0.34 3
EU IPE IPE160 I 160 5 82 7.4 9 8 20.09 12.14 7.26 869 109 124 6.58 68 16.7 26.1 1.84 3.62 4.89 3960 0.21 0.34 0.34 3
EU IPE IPE180 I 180 5.3 91 8 9 9 23.95 14.56 8.69 1317 146 166 7.42 101 22.2 34.6 2.05 4.8 6 7430 0.21 0.34 0.34 3
EU IPE IPE200 I 200 5.6 100 8.5 12 10 28.48 17.00 10.25 1943 194 221 8.26 142 28.5 44.6 2.24 7.02 8.26 13000 0.21 0.34 0.34 3
EU IPE IPE220 I 220 5.9 110 9.2 12 11 33.37 20.24 11.89 2772 252 285 9.11 205 37.3 58.1 2.48 9.1 9.89 22700 0.21 0.34 0.34 3
EU IPE IPE240 I 240 6.2 120 9.8 15 12 39.12 23.52 13.66 3892 324 367 9.97 284 47.3 73.9 2.69 12.9 13.2 37400 0.21 0.34 0.34 3
EU IPE IPE270 I 270 6.6 135 10.2 15 13.5 45.94 27.54 16.47 5790 429 484 11.2 420 62.2 97 3.02 16 15.7 70600 0.21 0.34 0.34 3
EU IPE IPE300 I 300 7.1 150 10.7 15 15 53.81 32.10 19.78 8356 557 628 12.5 604 80.5 125 3.35 20.2 18.9 126000 0.21 0.34 0.34 3
EU IPE IPE330 I 330 7.5 160 11.5 18 16.5 62.61 36.80 23.03 11770 713 804 13.7 788 98.5 154 3.55 28.3 24.6 199000 0.21 0.34 0.49 3
EU IPE IPE360 I 360 8 170 12.7 18 18 72.73 43.18 26.77 16270 904 1020 15.0 1043 123 191 3.79 37.5 29.5 314000 0.21 0.34 0.49 3
EU IPE IPE400 I 400 8.6 180 13.5 21 20 84.46 48.60 32.08 23130 1160 1310 16.5 1318 146 229 3.95 51.4 38.1 490000 0.21 0.34 0.49 3
EU IPE IPE450 I 450 9.4 190 14.6 21 22.5 98.82 55.48 39.56 33740 1500 1700 18.5 1676 176 276 4.12 67.1 46 791000 0.21 0.34 0.49 3
EU IPE IPE500 I 500 10.2 200 16 21 25 115.5 64.00 47.74 48200 1930 2190 20.4 2142 214 336 4.31 89.7 56.1 1250000 0.21 0.34 0.49 3
EU IPE IPE550 I 550 11.1 210 17.2 24 27.5 134.4 72.24 57.23 67120 2440 2790 22.3 2668 254 401 4.45 124 72.1 1880000 0.21 0.34 0.49 3
EU IPE IPE600 I 600 12 220 19 24 30 156 83.60 67.44 92080 3070 3510 24.3 3387 308 486 4.66 166 87.4 2850000 0.21 0.34 0.49 3

D Verification of the Warren sub model in Karamba
3D with FEM-Design
This chapter presents the results from the Warren truss sub model in Karamba 3D and FEM-Design
considering the sectional forces, the reaction forces and the maximum displacement of the bridge.
The chapter is concluding with an extensive study of the maximum utilization ratio of the cross beams
of the Warren truss sub model.
D.1 Verification of the sectional forces and the maximum dis-

placement
In this section the verification results are presented, comparing the Warren truss sub model in Karamba
3D with FEM-Design. First, Table D.1 shows all the factors used in the different load combinations
with given names, which are referred to later in this appendix. The load combination used for the
verification is 6.10𝑏1 , including the permanent load and the crowd load.
Table D.1: All load combinations used in the created model in Karamba 3D.
Load combination Type Name Self-weight Crowd load
6.10𝑎1 1.35 0.6
6.10a ULS
6.10𝑎2 1 0.6
6.10𝑏1 1.2 1.5
6.10𝑏2 1.2 0.6
6.10b ULS
6.10𝑏3 1 1.5
6.10𝑏4 1 0.6
6.10𝑎𝑏1 1 0.4
6.10a-b SLS
6.10𝑎𝑏2 1 0
Secondly, the positions (a to f ) considering the x-direction only, which are used in both models and
are presented in Figure D.1 from a side view to facilitate the understanding of the results later.
Figure D.1: Identification of the positions of the Warren sub model.
On the following pages first the convergence study of the weakening factor of the bridge deck and its
influence on the normal force of the bottom chord is presented. Following, the results of sectional
forces of the different members, i.e. the bottom chord, top chord, diagonals and the end posts, are
presented. Also, the maximum displacements downwards and reaction forces of the bridge from the

two software are compared. The tables are followed by diagrams of the normal force (𝑁), the shear
forces in z-direction (𝑉 𝑧 and 𝑇 𝑧′ in Karamba 3D and FEM-Design respectively) and the moment
about the y-axis (𝑀𝑦′ ) of the top chord and bottom chord.

Load case 6.10b1
Karamba 3D
Weakening
1 10 50 100 1000 5000 10 000 1 000 000
factor
a 1,3 1,1 1,2 1,8 2,5 2,5 2,5 2,5
31,4 21,8 4,7 1,2 2,2 2,5 2,5 2,5
b
77,8 203,2 249,5 257,6 265,6 266,5 266,6 266,7
Bottom chords
32,4 169,1 239,7 252,5 265,2 266,4 266,5 266,7
d
86,7 284,1 370,5 385,9 401,1 402,6 402,8 402,9
f 71,6 272,0 367,1 384,1 401,0 402,6 402,8 402,9
b 162,2 162,0 161,9 161,9 161,9 161,9

Normal force [kN] Top chord d 369,4 368,9 368,7 368,6 368,5 368,5
(absolute values) f 436,4 435,8 435,5 435,4 435,4 435,4
a-b 194,7 194,3 194,2 194,2 194,2 194,2

b-c 128,1 127,7 137,4 127,4 127,3 127,3
Diagonals c-d 124,4 124,4 124,4 124,4 124,4 124,4
d-e 42,8 42,4 42,2 42,2 42,2 42,2
e-f 39,3 39,5 39,6 39,7 39,7 39,7
End posts All 4 112,8 122,7 112,7 112,7 112,7 112,7
450
400
350
300
Normal force 250
at position f
in the bottom chord
200
[kN]
150
100
50
0
1 10 100 1 000 10 000 100 000 1 000 000
Weakening factor

Load case 6.10b1
Normal force [kN] Normal force [kN]

(absolute values) (absolute values)
Karamba 3D FEM-Design
Description Member Difference Karamba 3D 1* FEM-Design 1* Difference
10 000* 10 000*
a 3,2 3,0 4,3% 1,0 1,8 -42%
b (left) 31,1 91,3 -66%
b (right) 77,6 141,5 -45%
Bottom chords c 266,3 270,1 -1,4%
d (left) 32,0 0,5 6709%
d (right) 86,7 122,7 -29%
f 403,0 411,2 -2,0% 71,6 73,9 -3%
a-b 162,3 164,1 -1,1% 162,7 164,6 -1%

Top chord b-d 368,6 376,0 -2,0% 396,6 377,2 5%
d-f 435,5 444,5 -2,0% 436,6 445,9 -2%
Normal force N a-b 194,1 196,5 -1,2% 194,7 197,1 -1%

b-c 126,8 129,6 -2,1% 127,5 130,3 -2%
Diagonals c-d 124,7 129,0 -3,3% 124,9 129,2 -3%
d-e 42,2 43,3 -2,5% 42,6 43,7 -3%
e-f 39,5 40,5 -2,5% 39,3 40,4 -3%
End posts All 4 113,2 114,5 -1,1% 113,3 114,7 -1%
a 0,13 0,12
b 0,25 0,25
Cross beams
d 0,02 0
f 0,07 0,08
1* Bridge deck is weakened with a factor of 1

10 000* Bridge deck is weakened with a factor of 10 000
Moment My [kNm] Moment My [kNm]

Description Member Difference Karamba 3D 1* FEM-Design 1* Difference
10 000* 10 000*
a 2,5 2,3 5,6% 2,4 2,3 3%
b 10,3 10,4 -1,3% 10,2 10,6 -4%
c 10,2 10,5 -3,2% 9,7 10,0 -3%
Bottom chord
d 14,0 14,4 -3,2% 14,1 14,7 -4%
e 9,1 9,3 -2,3% 8,3 8,5 -3%
f 12,2 12,6 -2,9% 12,5 13,1 -5%
a 2,1 2,1
b - -
c 2,3 2,3
Top chord
d 2,7 2,7
e 2,2 2,3
f 2,8 2,9
0,1 0,1
a-b
0,5 0,5
0,1 0,1
b-c
Moment My 0,1 0,1
0,1 0,1
Diagonals c-d
0,1 0,2
0,1 0,0
d-e
0,1 0,1
0,1 0,1
e-f
0,0 0,1
2,5 2,3
End posts All 4
2,0 1,9
Description Member
10 000* 10 000*
a 5,1 4,6
b 5,3 5,4
Cross beams
d 5,6 6,1
f 5,7 6,2


Shear Vz [kN] Shear Vz [kN]
Description Member Differemce Karamba 3D 1* FEM-Design 1* Difference
10 000* 10 000*
a 6,0 5,7 4,2% 5,6 5,4 2,9%
b (left) 12,3 11,7 5,0% 11,9 11,5 4,0%
b (right) 17,7 18,3 -3,0% 16,8 18,6 -9,6%
Bottom chord
d (left) 19,6 20,3 -3,6% 18,7 20,1 -6,8%
d (right) 19,0 19,7 -3,5% 18,1 19,4 -6,9%
f 18,2 18,9 -3,3% 17,4 18,6 -6,7%
a 1,8 1,8
c 0,8 0,8
Top chord e (left) 0,6 0,5
e (right) 0,7 0,7
f 0,0 0,0
Shear force Vz
a-b 0,3 0,3
b-c 0,1 0,1
Diagonals c-d 0,0 0,0
d-e 0,0 0,0
e-f 0,0 0,0
End posts All 4 3,2 3,0
a 4,3 2,9
b 8,3 7,8
Cross beams
d 8,3 8,5
f 8,3 8,5

10'000* Bridge deck is weakened with a factor of 10'000
Reaction force of
one support [kN]
Load Member Difference
10 000* 10 000*
Reaction forces
Self-weight All 24,8 24,4 1,6%
Crowd load All 62,5 62,5 0,0%
Max. displacement ULS [m]
Description Member Difference
10 000* 10 000*
Displacement
Max. displacement
All 52,2 53,2 -2,0%
[mm]
10 000* Bridge deck is weakened with a factor of 10'000

Figure D.2: Normal force, moment and shear force diagram of the top chord from FEM-Design (left)
and Karamba 3D (right) with a bridge deck which is weakened with a factor 1.
Figure D.3: Normal force, moment and shear force diagram of the top chord from FEM-Design (left)
and Karamba 3D (right) with a bridge deck which is weakened with a factor 10 000.

Figure D.4: Normal force, moment and shear force diagram of the bottom chord from FEM-Design
(left) and Karamba 3D (right) with a bridge deck which is weakened with a factor 1.
Figure D.5: Normal force, moment and shear force diagram of the bottom chord from FEM-Design
(left) and Karamba 3D (right) with a bridge deck which is weakened with a factor 10 000.

D.2 Verification of the utilization ratios of the elements
In this section the comparison of the utilization ratios from Karamba 3D and FEM-Design of load
case 6.10𝑏1 , see Table D.1, are presented. First, all the controls executed in both software are
displayed, followed by the worst utilization ratios for each member in the Warren truss sub model for
Karamba 3D and FEM-Design. After that, the cross beam at position f, see Figure D.1, is investigated
further without and with adjustment of the buckling length of the cross beam. With adjustment of
the buckling length indicates, that the buckling length now is 2.7 m in x- and z-direction, i.e. the
length of the member. Following pages show the calculation results from FEM-Design for the cross
beam at position f with load case 6.10𝑏1 .

Controls which are executed in Input values of the cross section
FEM-Design and Karamba 3D 6.10b1 cross beam (IPE 140) at position f
Chapter in EN
Control Karamba 3D FEM-Design Difference
1993-1-1
Shear resistance 6.2.6,6.2.8 A [mm2] 1643 1643 0,0%
Torsional resistance 6.2.7 Iy [mm4] 5410000 5412000 0,0%
Shear stress 6.2.6 Iz [mm4] 450000 449200 0,2%
Normal stress 6.2.1 Wply [mm3] 88300 88340 0,0%
Normal capacity 6,2 Wplz [mm3] 19200 19260 -0,3%
Flexural buckling 6.3.1 iy [mm] 57,4 57 0,7%
Torsional-flexural
6.3.1 iz [mm] 16,5 17 -2,9%
buckling
Lateral torsional
6.3.2.2 It [mm4] 24500 24010 2,0%
buckling
Interaction between
normal force and 6.3.3 Iw [mm6] 1900000000 1951000000 -2,6%
bending
Interaction between
normal force and 6.3.3 fy [N/mm2] 355 355 0,0%
bending, 2nd order
ε 0,81 0,81 0,0%
Preconditions: Load case 6.10b1, cross sections and geometry are the same as in the case study.
Comparison of the maximum utilization ratios of each element between Karamba 3D and FEM-Design.
Without adjustments for buckling

With adjustments for buckling lengths
lengths
Karamba 3D FEM-Design Karamba 3D 10'000, FEM-Design

10'000, maximum 10'000, maximum Difference maximum utilization 10'000, maximum Difference
utilization ratio utilization ratio ratio utilization ratio
Bottom chord 61,5% 67,0% -5,5% 64,0% 67,0% -3,0%

Top chord 51,2% 53,0% -1,8% 51,2% 53,0% -1,8%
Diagonals 65,3% 60,0% 5,3% 65,0% 60,0% 5,0%
Vertical end post 30,6% 30,0% 0,6% 30,6% 30,0% 0,6%
Cross beams 17,7% 47,0% -29,3% 35,3% 47,0% -11,7%
End cross beams 8,5% 15,0% -6,5% 14,9% 15,0% -0,2%
Maximum utilization of the cross beams at position f - Maximum utilization of the cross beams at position f - with
without adjustment of the buckling length adjustment of the buckling length
Description Karamba 3D FEM-Design Difference Karamba 3D FEM-Design Difference
Shear resistance Vy 0% 0% 0% 0% 0% 0%
Shear resistance Vz 6,8% 5,0% 1,8% 5,6% 5,0% 0,6%

Moment resistance
17,7% 47,0% -29,3% 35,2% 47,0% -11,8%
My
Moment resistance
0,6% 0,0% 0,6% 0,3% 0,0% 0,3%
Mz
Normal capacity N 0% 0% 0% 0,2% 0,0% 0,2%
Max utilization ratio 17,7% 47,0% -29,3% 0,35 47,0% -11,8%
Comparison of details about input and output data between Karamba 3D and FEM-Design, without and with adjustment of the buckling lengths
Without adjustment of buckling

With adjustment of buckling lengths
lengths
Karamba 3D, Karamba 3D,
FEM-Design Difference FEM-Design Difference Comment
1st order 1st order
CSC 1 1 0% 1 1 0% OK
NRd [kN] 591,5 583,1 1,4% 591,5 583,1 1,4% OK
Ncr [kN] 127,9 127,7 0,2% 127,9 127,7 0,2% OK
Due to different shear

VyRd [kN] 209,4 215,1 -2,7% 209,4 215,1 -2,7%
areas
Due to different shear
VzRd [kN] 123,3 156,6 -21,3% 123,3 156,6 -21,3%
areas
MyRd [kNm] (No
31,8 31,4 1,2% 31,8 31,4 1,2% OK
X_LT reduction)
Dependent on the k-
Mcr [kNm] 846,4 16,1 5147,5% 18,0 16,1 11,3%
factors below
Cross beam
Cmy 0,9 1,0 -10,0% 1 1 0% OK
at position f
Cmz 0,9 1,0 -10,0% 1 1 0% OK
CmLT 0,9 1,0 -10,0% 1 1 0% OK
X_LT 1 0,42 138,1% 0,51 0,42 20,8% Dependent on Mcr
kyy* 0,90 1 -10,0% 1 1 0% OK
kzz* 0,90 1,03 -12,6% 1,00 1,03 -2,6% OK
kyz* 0,54 0,82 -34,1% 0,60 0,82 -26,6% *
kzy* 0,815 0,52 56,7% 1,00 0,52 92,2% *
* Karamba 3D FEM-Design
Interaction values
Interaction values
(kyy, kzz etc) from
(kyy, kzz etc) from
EN1991-1-1
EN1991-1-1 appendix
appendix B, method
A, method 1
2

B.37.1
Maximum of load combinations
S 355
2
E = 210000 N/mm
2
G = 80769 N/mm
γM0,ult = 1.00 γM0,acc/seis = 1.00
IPE 140
2 2
A = 1643 mm fy = 355 N/mm
4
2 Iy = 5.412e+06 mm ε = 0.81
4
z Iz = 4.492e+05 mm λ1 = 76.40
4
I1 = 5.412e+06 mm
4
I2 = 4.492e+05 mm
3
y 1 Wpl,1 = 8.834e+04 mm
3
Wpl,2 = 1.926e+04 mm
3
Wel,min,1 = 7.732e+04 mm
3
Wel,min,2 = 1.231e+04 mm
i1 = 57 mm
i2 = 17 mm
4
It = 2.401e+04 mm
6
Iw = 1.951e+09 mm
Shear resistance, 1-1 - Part 1-1: 6.2.6, 6.2.8

LC: '6.10b_1', x = 2700 mm
ClassN = 1, ClassM1 = 1, ClassM2 = 1
A1,v . fy 1049 . 355
V1,pl,Rd = = = 215.10 kN (6.18)
3 . γM0 3 . 1.00
τt,Ed
V1,pl,T,Rd = 1- .V =
1.25 ( fy / 3 ) / γM0
1,pl,Rd
0.00 . 215.10 = 215.10 kN

= 1- (6.26)
1.25 ( 355 / 3 ) / 1.00
V1,Ed 0.00
= = 0.00  1.00 (6.25) - OK
V1,pl,T,Rd 215.10
Shear resistance, 2-2 - Part 1-1: 6.2.6, 6.2.8

LC: '6.10b_1', x = 2700 mm
A2,v . fy 764 . 355
V2,pl,Rd = = = 156.64 kN (6.18)
3 . γM0 3 . 1.00
τt,Ed
V2,pl,T,Rd = 1- .V =
1.25 ( fy / 3 ) / γM0
2,pl,Rd
0.00 . 156.64 = 156.64 kN

= 1- (6.26)
1.25 ( 355 / 3 ) / 1.00
V2,Ed 8.49
= = 0.05  1.00 (6.25) - OK
V2,pl,T,Rd 156.64

Torsional resistance - Part 1-1: 6.2.7
LC: '6.10a_1', x = 0 mm
2
N/mm
τmax,unit = 439.20 is calculated by FEM analysis.
kNm
fy 355
TRd = = = 0.47 kNm
.
3 τ 3 . 439.20
max,unit
TEd 0.00
= = 0.00  1.00 (6.23) - OK
TRd 0.47
Shear stress - Part 1-1: 6.2.6

Not relevant
Normal stress - Part 1-1: 6.2.1

Not relevant
Normal capacity - Part 1-1: 6.2

LC: '6.10b_1', x = 1350 mm
V1,Ed = 0.00 kN  0.5 . V1,pl,T,Rd = 107.55 kN --> ρ1 = 0.00
V2,Ed = 0.00 kN  0.5 . V = 78.32 kN --> ρ = 0.00
2,pl,T,Rd 1
NEd M1,Ed M2,Ed 0.08 6.18 0.00
+ + = + + = 0.20  1.00 (6.2) - OK
NRd M1,Rd M2,Rd 583.12 31.36 6.84
Flexural buckling, 1-1 - Part 1-1: 6.3.1

LC: '6.10b_1', x = 1800 mm
Lcr,1 2700
λ1 = = = 0.62 (6.50)
i1 . λ1 57 . 76.40
α1 = 0.21 (Buckling curve: a)
2
φ1 = 0.5 1 + α1 . ( λ1 - 0.2 ) + λ1 = 0.5 1 + 0.21 . ( 0.62 - 0.2) + 0.62 = 0.73
2
( ) ( )
1 1
χ1 = min , 1.0 = min , 1.0 = 0.88 (6.49)
2 2 2 2
φ1 + φ1 - λ1 0.73 + 0.73 - 0.62
χ .A.f
1 y 0.88 . 1643 . 355
Nb,Rd,1 = = = 515.45 kN (6.47)
γM1 1.00
NEd 0.08
= = 0.00  1.00 (6.46) - OK
Nb,Rd,1 515.45

Flexural buckling, 2-2 - Part 1-1: 6.3.1
LC: '6.10b_1', x = 1800 mm
Lcr,2 2700
λ2 = = = 2.14 (6.50)
i2 . λ1 17 . 76.40
α2 = 0.34 (Buckling curve: b)
2
φ2 = 0.5 1 + α2 . ( λ2 - 0.2 ) + λ2 = 0.5 1 + 0.34 . ( 2.14 - 0.2) + 2.14 = 3.11
2
( ) ( )
1 1
χ2 = min , 1.0 = min , 1.0 = 0.19 (6.49)
2 2 2 2
φ2 + φ2 - λ2 3.11 + 3.11 - 2.14
χ .A.f
2 y 0.19 . 1643 . 355
Nb,Rd,2 = = = 108.46 kN (6.47)
γM1 1.00
NEd 0.08
= = 0.00  1.00 (6.46) - OK
Nb,Rd,2 108.46
Torsional-flexural buckling - Part 1-1: 6.3.1

LC: '6.10b_1', x = 1800 mm
2 2 2 2 2 2 2 2
i0 = i1 + i2 + y0 + z0 = 57 + 17 + 0 + 0 = 60 mm
2
π . E . I1 π . 210000 . 5412243
2
Ncr,1 = 2 = 2 = 1538.75 kN
Lcr,1 2700
π . E . I2
2
π . 210000 . 449178
2
Ncr,2 = 2 = 2 = 127.71 kN
Lcr,2 2700
( )
2
1 π . E . Iw
Ncr,T = 2 G . It + 2 =
i0 lT
( )
2. .
1 . 2.401e+04 + π 210000 1.951e+09 = 698.91 kN
= 2 80769 2
60 2700
2 2 2 2 2
i0 ( N - Ncr,1) ( N - Ncr,2) ( N - Ncr,T) - N y0 ( N - Ncr,2) - N z0 ( N - Ncr,1) = 0
Smallest root of the above equation:
Ncr,TF = 698.91 kN
Ncr = min( Ncr,T, Ncr,TF) = min( 698.91, 698.91) = 698.91 kN
A . fy 1643 . 355
λT = = = 0.91 (6.53)
Ncr 698.91
αT = 0.34 (Buckling curve: b)
2
φT = 0.5 1 + αT . ( λT - 0.2 ) + λT = 0.5 1 + 0.34 . ( 0.91 - 0.2) + 0.91 = 1.04
2
( ) ( )
1 1
χT = min , 1.0 = min
, 1.0 = 0.65 (6.49)
2 22 2
φT + φT - λT 1.04 + 1.04 - 0.91
χT . A . fy 0.65 . 1643 . 355
Nb,Rd,T = = = 380.52 kN (6.47)
γM1 1.00
NEd 0.08
= = 0.00  1.00 - OK
Nb,Rd,T 380.52

Lateral torsional buckling - Part 1-1: 6.3.2.2
LC: '6.10b_1', x = 1350 mm
π . E . Iz π2 . 2.100e+05 . 4.492e+05
2
Ncr,LT = 2= = 127.71 kN
(k . L ) ( 1.00 . 2700)
2
z cr
Loaded on top edge.
Z = ( C2 . zg - C3 . zj) = ( 0.45 . 70 - 0.52 . 0) = 31.50 mm
0.5
( ) Iw G . It
2
Mcr = C1 . Ncr,LT . kz . 2 -Z =
+ +Z
kw Iz Ncr,LT
0.5
( )
2 .
= 1.13 . 1.277e+05 . 1.00 . 1.951e+09 + 8.077e+04 2.401e+04 + 31.502 - 31.50 =
1.00 4.492e+05 1.277e+05
= 16.13 kNm
Wy . fy 88344 . 355
λLT = = = 1.39
Mcr 1.613e+07
αLT = 0.21 (Buckling curve: a)
2
φLT = 0.5 1 + αLT . ( λLT - 0.2 ) + λLT =
2
= 0.5 1 + 0.21 . ( 1.39 - 0.2) + 1.39 = 1.60
( ) ( )
1 1
χLT = min , 1.0 = min , 1.0 = 0.42 (6.56)
2 2 2 2
φLT + φLT - λLT 1.60 + 1.60 - 1.39
fy 355
My,b,Rd = χLT . Wy = 0.42 . 88344 = 13.19 kNm (6.55)
γM1 1.00
M1,Ed 6.18
= = 0.47  1.00 (6.54) - OK
My,b,Rd 13.19
Interaction between normal force and bending 1. - Part 1-1: 6.3.3

LC: '6.10b_1', x = 1350 mm
kij factors are calculated according to Method 1
Cmy = 1.00 Cyy = 1.00
Cmz = 1.00 Cyz = 0.84
CmLT = 1.00 Czy = 1.00
Czz = 0.97
M2,Rk = fy . Wpl,2 = 355 . 19260 = 6.84 kNm
comp
NEd M1,Ed M2,Ed
+ k11 . + k12 . =
Nb,Rd,1 My,b,Rd M2,Rk
γM1
0.08 6.18 0.00
= + 1.00 . + 0.82 . = 0.47  1.00 (6.61) - OK
515.45 13.19 6.84
1.00

Interaction between normal force and bending 2. - Part 1-1: 6.3.3
LC: '6.10b_1', x = 1350 mm
kij factors are calculated according to Method 1
Cmy = 1.00 Cyy = 1.00
Cmz = 1.00 Cyz = 0.84
CmLT = 1.00 Czy = 1.00
Czz = 0.97
M2,Rk = fy . Wpl,2 = 355 . 19260 = 6.84 kNm
comp
NEd M1,Ed M2,Ed
+ k21 . + k22 . =
Nb,Rd,2 My,b,Rd M2,Rk
γM1
0.08 6.18 0.00
= + 0.52 . + 1.03 . = 0.25  1.00 (6.62) - OK
108.46 13.19 6.84
1.00
Interaction between normal force and bending, 2nd order - Part 1-1: 6.3.3
Not relevant
Shear buckling - Part 1-5: 5

hw 126 72 . 72 .
= = 26.9  ε= 0.81 = 48.8 --> Not relevant
t 5 η 1.20
Summary
Utilization [%]
100 Cross-section resistance
Flexural buckling
80 Torsional-flexural buckling
Lateral torsional buckling
60 Interaction
Shear buckling
40
20

E Verification of the natural frequency (mode 1) of
the Warren sub model
In this chapter, the first mode of the natural frequency of the Warren truss bridge is presented and
compared with the results from FEM-Design and hand calculations. In Table E.1 the comparison of
the results is presented, showing very similar results from Karamba 3D compared to FEM-Design
and the hand calculations, having a weakening factor 1.
Table E.1: Natural frequency of mode 1 compared to FEM-Design and hand calculations.
Natural frequency Natural frequency
mode 1 [Hz] mode 1 [Hz]
Member Karamba 3D 1* FEM-Design Difference Karamba 3D 1* Hand calculations Difference
All 6.74 6.95 -3.2 % 6.74 6.71 0.4 %
*Weakening factor of the bridge deck is 1
The shapes of mode 1 in Karamba 3D and FEM-Design are presented below in Figure E.1 and
Figure E.2 respectively and cohere well to each other. On the following pages, the hand calculations
for the natural frequency of mode 1 for the Warren sub model is presented.
Figure E.1: Natural frequency (mode 1) for the Warren sub model in Karamba 3D.
Figure E.2: Natural frequency (mode 1) for the Warren sub model in FEM-Design.

Calculation of natural frequency for Warren truss (mode 1)
Indata
L1  20 m Span length of the bridge
m1  8058.5kg
 Mass of the Warren truss footbridge
kN
q  5 Distributed load on the bridge (crowd load)
2
m
b  2.5 m Width of the bridge deck
δ1  0.014236m
 Displacement of the Warren truss footbridge
with weakening factor 1
Stiffness calculation
 q  L1 b 
   m1 g
K1 
 2  7 N
 1.4  10 
Stiffness of the structure
δ1 m
K1 Natural frequency in rad/s (equation from:

rad
ω1   42.2 Fundamentals of Structural Dynamics 2006)
m1 s
ω1 Natural frequency of the Warren

f1   6.7 Hz
2 π truss (mode 1)

F Verification of the Pratt sub model in Karamba 3D
with FEM-Design
In this chapter the results from the verification of the Pratt truss sub model in Karamba 3D and
FEM-Design are presented. In Figure F.1 the FEM-model for the Pratt truss bridge is presented,
having the distances shown in meters.
Figure F.1: The Pratt truss bridge modeled in FEM-Design. The distances are shown in meters.
In Figure F.2 the positions of the bridge are presented. The sectional forces, the reaction forces, the
maximum displacement and the maximum utilization ratios are presented on the following pages,
referring to the positions shown in Figure F.2. Also in this verification, load case 6.10𝑏1 is applied.
Figure F.2: Position identifications for the Pratt truss sub model.

Load case 6.10b1
Normal force [kN]

(absolute values)
10 000* 10 000*
a-b 9,7 9,7 -0,9%
Bottom chord B.1/B.3 b-c 251,1 252,1 -0,4%
c-d 397,3 398,5 -0,3%
a-b 249,2 250,1 -0,4%

Top chord B.2/B.4 b-c 396,6 397,7 -0,3%
c-d 446,3 447,3 -0,2%
Normal force N
a-b 259,0 260,7 -0,7%
b (verticals) 59,8 60,1 -0,5%
Diagonals b-c 157,5 158,1 -0,3%
c (verticals) 21,9 22,0 -0,6%
c-d 53,1 53,0 0,1%
End posts All 4 104,6 104,9 -0,3%
b 0,2 0,2
Cross beams
d 0,1 0,0
Moment My [kNm]
(absolute values)
10 000* 10 000*
a/ k 8,1 8,2 -1,7%
a-b 7,4 7,5 -1,7%
b 10,2 10,3 -0,9%
Bottom chord b-c 7,1 7,2 -0,4%
c 8,3 8,4 -0,8%
c-d 7,4 7,5 -1,3%
d 8,4 8,5 -0,6%
a/ k 5,1 5,1 -0,4%

b/ j 4,1 4,0 1,8%
Top chord
c/ i 2,9 3,0 -2,0%
Moment My
d/ h 2,8 2,8 0,4%
a-b 0,2 0,2

b (verticals) 0,3 0,3
Diagonals b-c 0,1 0,1
c (verticals) 0,1 0,1
c-d 0,0 0,0
8,1 8,2 -1,7%

End posts All 4
5,5 5,5 -0,4%
b 4,8 4,7 1,9%

Cross beams
d 4,9 4,9 0,2%

Shear Vz [kN]
(absolute values)
Description Member Differemce
10 000* 10 000*
a 15,8 16,8 -5,9%
16,4 16,6 -1,5%
b
16,9 17,1 -1,1%
Bottom chord
15,3 15,5 -1,5%
c
16,1 16,1 -0,6%
d 16,1 16,2 -0,9%
Top chord a-b 3,3 3,3 0,3%

Shear force Vz
a-b 0,0 0,0
Diagonals b-c 0,0 0,0
c-d 0,0 0,0
End posts All 4 9,7 9,8 -1,1%
b 7,24 7,17 1,0%

Cross beams
d 7,22 7,17 0,7%
Reaction force of one support

[kN]
Reaction forces 10 000* 10 000*
Self-weight All 20,38 20,73 -1,7%
10 000* 10 000*
Displacement
Max. displacement
All 57,89 57,85 0,1%
[mm]
10 000* 10 000*
Bottom chord 51,7% 54,0% -2,3%
Maximum Top chord 47,2% 48,0% -0,8%
With adjustment of the
utilization ratio Vertical end post 51,6% 53,0% -1,4%
buckling length of the
Diagonal (a-b) 52,1% 53,0% -0,9%
cross beams
Cross beams (f) 30,6% 37,0% -6,4%
End cross beams 13,1% 14,0% -0,9%

G Verification of the Howe sub model in Karamba
3D with FEM-Design
In this section the results from the Howe truss sub model in Karamba 3D and the corresponding
FEM-Design sub model, shown in Figure G.1, considering the sectional forces are presented and
compared. The load cases used for verification are 6.10𝑏1 . In Figure G.2 the positions, where
sectional forces are studied on the next page, are shown.
Figure G.1: The Howe truss bridge modeled in FEM-Design. The distances are shown in meters.
Figure G.2: Position identifications for the Howe truss sub model.

Load case 6.10b1
Normal force [kN]

(absolute values)
10'000* 10'000*
a-b 249,24 249,8 -0,2%
Bottom chord b-c 396,4 397,0 -0,2%
c-d 444,9 445,6 -0,2%
a-b 9,41 9,53 -1,3%

Top chord b-c 251,6 252,1 -0,2%
c-d 397,8 398,5 -0,2%
Normal force N a-b 260,3 260,7 -0,1%

b (verticals) 101,6 102,0 -0,4%
b-c 157,3 157,4 -0,1%
Diagonals
c (verticals) 59,1 59,4 -0,5%
c-d 51,4 51,5 -0,2%
d 38,9 39,1 -0,5%
End posts All 4 3,6 3,6 0,8%
b 0,19 0,25
Cross beams
d 0,06 0,02
Moment My [kNm]
(absolute values)
10'000* 10'000*
a/ k 8,2 8,06 1,9%
a-b 7,3 7,49 -2,1%
b 10,3 10,3 -0,5%
Bottom chord b-c 7,1 7,01 1,1%
c 8,6 8,62 0,0%
c-d 7,7 7,67 0,3%
d 7,7 7,75 -0,1%
a/ k 5,2 5,3 -0,4%

b/ j 4,0 4,0 0,2%
Top chord
c/ i 2,9 3,0 -1,3%
Moment My d/ h 3,8 3,7 0,3%
a-b 0,2 0,2

b-c 0,0 0,0
Diagonals
c-d 0,0 0,0
d 0,1 0,1
7,9 8,2 -2,7%

End posts All 4
5,2 5,4 -2,8%
b 4,8 4,79 -0,6%

Cross beams
d 4,7 4,9 -3,9%

Shear Vz [kN]
(absolute values)
Description Member Differemce
10 000* 10 000*
a 15,9 16,11 -1,3%
16,3 16,70 -2,7%
b
16,8 17,10 -1,5%
Bottom chord
15,3 15,66 -2,3%
c
16,3 16,62 -1,7%
d 15,8 16,14 -2,0%
Top chord a-b 3,3 3,33 0,3%

Shear force Vz
a-b 0,1 0,06
b-c 0,0 0,01
Diagonals
c-d 0,0 0,01
d 0,03 0,01
End posts All 4 9,4 9,68 -2,8%
b 7,24 6,78 6,8%

Cross beams
d 7,2 6,75 6,7%
Reaction force of one support

[kN]
Reaction forces 10 000* 10 000*
Self-weight All 20,38 20,85 -2,2%

10 000* 10 000*
Displacement
Max. displacement
All 61,0 60,7 0,5%
[mm]
10 000* 10 000*
Bottom chord 56,6% 60,0% -3,4%
Maximum Top chord 43,3% 45,0% -1,7%
With adjustment of the
utilization ratio Vertical end post 35,9% 37,0% -1,1%
buckling length of the
Diagonal (a-b) 229,2% 226,0% 3,2%
cross beams
Cross beams (f) 30,6% 37,0% -6,4%
End cross beams 13,2% 13,0% 0,2%

H Calculation of the global lateral torsional buckling
of the top chord
In this chapter the hand calculations of the global lateral torsional buckling of the top chord of
the Warren sub model, according to SS-EN 1993-2 chapter 6.3.4.2 (2) and (Wahlström, 1971), are
presented. In Figure H.1 it can be seen how the unit load of 1 000 kN is applied on the different sub
models. The purpose is to calculated the displacement of the truss where the unit load is applied,
which concludes in the results of the stiffness of the springs that the top chord lies on. Finally,
the reduced normal capacity of the top chord, 𝑁𝑏𝑅𝑑 , can be calculated, which corresponds to the
maximum value that the maximum normal force in the top chord, 𝑁𝐸𝐷 , must not exceed.
(a)
(b)
(c)
Figure H.1: The three sub model without the top chords. A unit load of 1 000 kN is applied to
calculated the displacement in y-direction. (a) The Warren sub model (b) The Pratt sub model
(c) The Howe sub model

Global control of buckling of the top chord
According to SS-EN 1993-2 chapter 6.3.4.2 (2) and
Bygg Handbok för hus-, väg- och vattenbyggnad (Allmänna grunder del 1A, main author:
Civilengineer SVR Bengt Wahlström, publisher: AB Byggmästarens förlag, Stockholm
1971).
To check the utilization of the top chord globally when it comes to buckling sideways,
the top chord is modelled as column with a normal compression force N.ED, getting
some support from the truss system, the cross beams and the bottom chord, which are
connected to each other and give some stiffness. This contribution of stiffness is
simulated as a bed of springs with a certain spring stiffness.
Indata
NED  435kN Maximum normal force from Karamba 3D in the top chord
fy  355  MPa
E  210  GPa
2
Atopchord  3580mm
Ltopchord  20m
n triangles  5
γM1  1
4 4
Itopchord.z  1223 10 ( mm) (weak axis)
Calculation procedure
Nunit  1000kN
δunitload  7.061m The value is taken from the Karamba 3D model of a Warren
truss bridge of 20 m length and without a top chord. When
appying a unit load of 1000 kN, the displacement sideways
becomes δ.unitload.
1000kN kN
Cunitload   141.6  spring stiffness (sideways)
δunitload m
Ltopchord
a   4m Distance between the connections of the diagonals to the top
n triangles chord.

The figure is taken from Bygg (Huvuddel 1A), page 664.
With the quote calculated in equation (1) below, the
beta-value can be retrieved.
3
Cunitload  a
 3.5 equation (1)
E Itopchord.z
β  1.61 From figure 362i
2
π  E Itopchord.z
Ncr   611.2  kN According to (Wahlstrom, 1971)
2
( β a)

1
2
 Atopchord fy 
λLT     1.4 Accroding to SS-EN 1993-1-1 6.3.1.4 (2)
 Ncr

Lcr  β a  6.4 m
Buckling curve a, varm calsed S355 steel
αLT  0.21
Accroding to SS-EN 1993-1-1 Table 6.2
ϕLT  0.5 1  αLT λLT  0.2  λLT

2
     1.7
Accroding to SS-EN 1993-1-1 6.3.2.2 (1)
1
χLT   0.4 Accroding to SS-EN 1993-1-1 6.3.2.2 (1)
1
2
ϕLT   ϕLT  λLT
2 2
 
χLT fy  Atopchord
NbRd   505.8  kN Accroding to SS-EN 1993-1-1 6.3.1.1 (3)
γM1
NED
Utilization   0.9 Smaller than 1 --> OK
NbRd

I All optimization results for part 1 and part 2
In this section all results obtained from the optimization simulations for Warren, Pratt and Howe
trusses for part 1 and part 2 are presented. Below, see Table I.1, the procedure of part 1 is presented
followed by the procedure of part 2, see Table I.2 to get an overview. On the following pages, all
optimization results from both part 1 and part 2 are presented.
Table I.1: Procedure of the optimization of part 1.

Plug-in Optimization run Cross section Topology Fitness function Repetitions
A x Mass 1
B x Mass 1
C x x Mass 3
Galapagos
D x Transparency 1
E x Transparency 1
Table I.2: Procedure of the optimization of part 2.

Plug-in Span length Geometry Cross section Topology Fitness function Repetitions
10 m all 3 x x Mass and transparency 3
Octopus 20 m all 3 x x Mass and transparency 3
30 m all 3 x x Mass and transparency 3

Settings for Galapagos Settings for Octopus
102
Fitness minimize Elitism 0,5
Max. Stagnant 50 Mutation probabilty 0,1
Maintain 5% Mutation rate 0,5
Inbreeding (+) 75% Cross over rate 0,8
Population 100 Population 100
Initial boost 2 Initial boost 2
Min Max
nΔ 1 20
ncb 1 10
Height and width of the truss bridges are kept constant. First order analysis is used in Karamba 3D.
The start of every optimization run is with the perquisites from the case study.
*when optimizing only topology the cross sections are fixed to the values from the case study
**Irrelevant as the side view area is not considering the area of the cross beams
Best solution
Plug-in Case study 20 m Galapagos

Geometry Warren Warren
Load case 6.10b1 (ULS) 6.10b1 (ULS)
Fitness function - Mass Mass Mass Mass Mass Transparency Transparency Transparency Transparency Transparency
Optimization runs - A B C C C D E F F F
Cross section - - x x x x - x x x x
Topology - x - x x x x - x x x
nΔ 5 6 5 6 6 6 5 * 5 5 5
ncb 5 2 5 1 1 1 16 * 6 6 6
Bottom chord VKR 150x100x6.3 * VKR 120x120x5 VKR120x60x4 VKR 120x60x4 VKR 120x60x4 * VKR 120x80x8 VKR 100x100x8 VKR 100x100x8 VKR 100x100x10
Top chord VKR 150x150x6.3 * VKR 150x150x5 VKR 140x140x5 VKR 140x140x6 VKR 140x140x7 * VKR 120x120x8 VKR 120x120x10
Diagonals VKR 100x60x5 * VKR 120x60x3.6 VKR 100x60x3.6 VKR 100x60x3.7 VKR 100x60x3.8 * VKR 100x60x5.6 VKR 120x60x3.6 VKR 100x60x5.6 VKR 100x60x5
Vertical end posts VKR 100x100x5 * VKR 100x60x3.6 VKR 120x120x4.5 VKR 100x100x4 VKR 100x100x5 * IPE 120 KCKR 76.1/4 KCKR 76.1/5
Cross beams IPE 140 * KCKR 76.1/4 VKR 120x120x4.5 VKR 120x120x4.6 VKR 120x120x4.7 * VKR 150x100x6.3 VKR 120x80x6.3 VKR 120x60x3.6 VKR 100x60x5.6
End cross beams IPE 180 * KCKR 40.4/3 VKR 100x100x4 VKR 100x100x5 VKR 100x100x6 * VKR 150x150x5 VKR 180x100x8 VKR 180x100x5.6 VKR 160x80x10
20 m (Galapagos) Displacement 0,021 0,012 0,021 0,017 0,017 0,017 0,010 0,011 0,012 0,012 0,012
Max. displacement 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05
Utilization displacement 41% 23% 41% 34% 35% 35% 21% 22% 24% 24% 23%
Element where the max
- Bottom chord Bottom chord Bottom chord Diagonal
utilization ratio occurs
Utilization ratio 65% 74% 84% 82% 82% 82% 71% 74% 82% 78% 74%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 77% 88% 99% 96% 96% 96% 84% 87% 97% 92% 88%
Ned [kN] (Max normal force in
435 425 426 371 371 371 467 445 455 450 454
top chord)
NbRd [kN] (Max allowable
507 607 454 426 426 426 517 449 506 448 458
normal force in top chord)
Utilization normal force (top
86% 70% 94% 87% 87% 87% 90% 99% 90% 100% 99%
chord)
Area [m2] 7,74 7,85 7,14 7,10 7,05 7,05 7,74 6,60 6,08 6,08 6,08
Mass [kg] 8058 7602 7007 6490 6466 6466 9973 8785 8781 8204 8653
Plug-in Octopus
Geometry Warren
Fitness function Mass and trasparency
Load case 6.10b1 (ULS)
Optimization runs G
Cross section x
Topology x
nΔ 6 6 5 6 6
ncb 2 2 1 1 1
Bottom chord VKR 100x100x6.3 VKR 100x100x6.3 VKR 100x100x4 VKR 100x100x4 VKR 100x100x4
Top chord VKR 100x100x10 VKR 120x120x5 VKR 100x100x10 VKR 100x100x8 VKR 120x120x4.5
Diagonals VKR 120x60x3.6 VKR 120x60x4 VKR 120x60x3.6 VKR 100x60x3.6 VKR 100x60x4
Vertical end posts KCKR 76.1/5 KCKR 76.1/4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Cross beams VKR 120x60x3.6 VKR 120x60x3.6 KCKR 219.1/12.5 KCKR 219.1/12.5 KCKR 219.1/12.5
End cross beams IPE 120 IPE 120 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,016 0,016 0,016 0,016 0,018 0,018 0,016 0,018 0,012 0,013 0,014 0,015
20 m (Octopus) Max allowable displacement [m]
0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05
SLS 6.15ab1
Utilization of displacement 31% 31% 33% 33% 35% 37% 33% 37% 25% 26% 28% 29%
Bottom chord Bottom chord End cross beam End cross beam End cross beam Top chord Top chord Top chord
Utilization ratio 80% 78% 83% 81% 77% 82% 82% 82% 83% 73% 75% 78%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 94% 91% 97% 95% 90% 97% 97% 97% 97% 86% 88% 92%
426 410 422 422 406 419 423 419 381 373 386 371
top chord)
433 429 429 429 410 420 440 420 390 374 395 379
98% 96% 98% 98% 99% 100% 96% 100% 98% 100% 98% 98%
chord)
Mass 7314 7209 7087 7087 6825 6748 7132 6748 6690 6509 6402 6290
Area 5,78 5,78 5,78 5,82 6,18 6,18 5,78 6,18 5,75 5,85 5,96 6,25
Solution 1 2 3 4 5 6 7 8 9 10 11 12
Repeated - - - - - - x (reset)

Optimization run G
Cross section x
Topology x
Geometry Warren
nΔ 3 3 3 3 3 3 3 3 3
ncb 2 1 1 2 1 1 2 1 1
Bottom chord VKR 100x60x4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x4 VKR 100x60x3.6 VKR 100x60x3.6
Top chord KCKR 88.9/5 KCKR 88.9/4 KCKR 88.9/4 KCKR 88.9/5 KCKR 88.9/4 KCKR 88.9/4 KCKR 88.9/5 KCKR 88.9/4 KCKR 88.9/4
Diagonals VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKr 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Vertical end posts KCKR 40.4/4 KCKR 40.4/4 KCKR 48.3/3 KCKR 40.4/4 KCKR 40.4/4 KCKR 48.3/3 KCKR 40.4/4 KCKR 40.4/4 KCKR 48.3/3
Cross beams IPE 160 VKR 100x100x5 VKR 100x100x5 IPE 160 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5
End cross beams VKR 140x70x4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 140x70x4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,006 0,009 0,009 0,006 0,009 0,009 0,008 0,009 0,009
Max allowable displacement [m] SLS 6.15ab1 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025
Utilization of displacement 23% 37% 37% 23% 37% 37% 33% 37% 37%
Element where the max utilization ratio occurs Vertical end post Cross beam Cross beam Vertical end post Cross beam Cross beam Vertical end post Cross beam Cross beam
Utilization ratio 84% 78% 78% 84% 78% 78% 84% 78% 78%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 99% 91% 91% 99% 91% 91% 99% 91% 91%
Ned [kN] (Max normal force in top chord) 111 91 91 111 91 91 111 91 91
NbRd [kN] (Max allowable normal force in top
140 118 118 140 118 118 140 118 118
chord)
Utilization normal force (top chord) 79% 77% 77% 79% 77% 77% 79% 77% 77%
Mass [kg] 3096 2921 2919 3096 2921 2919 3059 2921 2919
Transparency (side view area) [m2] 2,79 2,79 2,81 2,79 2,79 2,81 2,79 2,79 2,81
Solution 1 2 3 4 5 6 7 8 9
Repeated - - - x x x x x
Geometry Pratt
nΔ 4 3 4 3 4 3
ncb 1 1 1 1 1 1
Bottom chord KCKR 40.4/4 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3
Top chord KCKR 101.6/6 KCKR 88.9/4 KCKR 101.6/6 KCKR 88.9/4 KCKR 101.6/6 KCKR 88.9/4
Diagonals KCKR 40.4/3 KCKR 60.3/4 KCKR 40.4/3 KCKR 60.3/4 KCKR 40.4/3 KCKR 60.3/4
Vertical end posts KCKR 48.3/4 KCKR 60.3/3 KCKR 48.3/4 KCKR 60.3/3 KCKR 48.3/4 KCKR 60.3/3
Cross beams KCKR 139.7/4 KCKR 139.7/4 KCKR 139.7/4 KCKR 139.7/4 KCKR 139.7/4 KCKR 139.7/4
End cross beams KCKR 114.3/4 KCKR 101.6/4 KCKR114.3/4 KCKR 101.6/4 KCKR114.3/4 KCKR 101.6/4
Displacement [m] SLS 0,0075 0,0107 0,0076 0,0077 0,0077 0,0076 0,0077 0,0077 0,0080 0,0076 0,0077
Max allowable displacement [m] SLS 6.15ab1 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025 0,025
Span length 10 Utilization of displacement 30% 43% 30% 31% 31% 30% 31% 31% 32% 30% 31%
m (OCTOPUS) Element where the max utilization ratio occurs Vertical end post End cross beam Vertical end post Bottom chord Vertical end post Vertical end post Bottom chord Vertical end post Bottom chord Vertical end post Bottom chord
Utilization ratio 83% 82% 84% 75% 83% 84% 75% 83% 75% 84% 75%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 97% 97% 99% 88% 98% 99% 88% 98% 88% 99% 88%
Ned [kN] (Max normal force in top chord) 104 104 104 104 104 104 104 104 104 104 104
107 107 110 110 107 110 110 107 107 110 110
chord)
Utilization normal force (top chord) 98% 98% 94% 94% 98% 94% 94% 98% 98% 94% 94%
Mass [kg] 2812 2779 2755 2754 2794 2755 2754 2794 2786 2755 2754
Transparency (side view area) [m2] 2,24 2,27 2,39 2,43 2,24 2,39 2,43 2,24 2,27 2,39 2,43
Solution 1 2 3 4 5 6 7 8 9 10 11
Repeated - - - - x x x x x
Geometry Howe
nΔ 4 4 4
ncb 1 1 1
Bottom chord VKR 100x60x3.6 IPE 80 IPE 80
Diagonals VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Vertical end posts KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3
Cross beams VKR 120x80x4 VKR 120x80x4 VKR 120x80x4
End cross beams VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,0087 0,0088 0,0088
Max allowable displacement [m] SLS 6.15ab1 0,025 0,025 0,025
Utilization of displacement 35% 35% 35%
Element where the max utilization ratio occurs Cross beam Cross beam Cross beam
Utilization ratio 83% 83% 83%
Max allowable utilization ratio 85% 85% 85%
Utilizaiton of utilization 98% 98% 98%
Ned [kN] (Max normal force in top chord) 79 79 79
124 123 123
chord)
Utilization normal force (top chord) 64% 64% 64%
Mass [kg] 2912 2860 2860
Transparency (side view area) [m2] 2,82 2,62 2,62
Solution 1 2 3
Repeated - x

103
104
Optimization run G
Cross section x
Topology x
Geometry Warren
nΔ 6 6 5 6 6
ncb 2 2 1 1 1
Bottom chord VKR 100x100x6.3 VKR 100x100x6.3 VKR 100x100x4 VKR 100x100x4 VKR 100x100x4
Top chord VKR 100x100x10 VKR 120x120x5 VKR 100x100x10 VKR 100x100x8 VKR 120x120x4.5
Diagonals VKR 120x60x3.6 VKR 120x60x4 VKR 120x60x3.6 VKR 100x60x3.6 VKR 100x60x4
Vertical end posts KCKR 76.1/5 KCKR 76.1/4 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Cross beams VKR 120x60x3.6 VKR 120x60x3.6 KCKR 219.1/12.5 KCKR 219.1/12.5 KCKR 219.1/12.5
End cross beams IPE 120 IPE 120 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,016 0,016 0,016 0,016 0,018 0,018 0,016 0,018 0,012 0,013 0,014 0,015
Max allowable displacement [m]
0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05
SLS 6.15ab1
Utilization of displacement 31% 31% 33% 33% 35% 37% 33% 37% 25% 26% 28% 29%
Element where the max Bottom chord Bottom chord End cross beam End cross beam End cross beam Top chord Top chord Top chord
Utilization ratio 80% 78% 83% 81% 77% 82% 82% 82% 83% 73% 75% 78%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 94% 91% 97% 95% 90% 97% 97% 97% 97% 86% 88% 92%
426 410 422 422 406 419 423 419 381 373 386 371
top chord)
433 429 429 429 410 420 440 420 390 374 395 379
98% 96% 98% 98% 99% 100% 96% 100% 98% 100% 98% 98%
chord)
Mass 7314 7209 7087 7087 6825 6748 7132 6748 6690 6509 6402 6290
Area 5,78 5,78 5,78 5,82 6,18 6,18 5,78 6,18 5,75 5,85 5,96 6,25
Solution 1 2 3 4 5 6 7 8 9 10 11 12
Repeated - - - - - - x (reset)
Natural frequency [Hz] 3,5 5,0
With IPE 180 as end cross beam
6345
[kg]
Geometry Pratt
nΔ 6 6 5 5 5
ncb 1 1 1 1 1
Bottom chord VKR 100x60x5 VKR 100x60x5 VKR 120x60x4 VKR 100x100x4 VKR 100x100x4
Top chord VKR 100x100x8 VKR 120x120x5 VKR 120x120x5 VKR 100x100x8 VKR 120x120x5
Diagonals VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6
Vertical end posts KCKR 76.1/4 KCKR 76.1/4 KCKR 76.1/4 KCKR 76.1/4 KCKR 76.1/4
Cross beams KCKR 168.3/4 KCKR 168.3/4 KCKR 168.3/4 VKR 160x80x4 VKR 160x80x4
End cross beams KCKR 101.6/4 KCKR 101.6/4 KCKR 101.6/4 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,019 0,020 0,021 0,019 0,020 0,021 0,019 0,021
0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05
Span length 20 SLS 6.15ab1
m (OCTOPUS) Utilization of displacement 37% 41% 42% 37% 41% 42% 38% 41%
Top chord Vertical end post Top chord Vertical end post Top chord Top chord
Utilization ratio 82% 82% 83% 82% 82% 83% 82% 81%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 96% 96% 98% 96% 96% 98% 96% 95%
425 422 422 425 422 422 425 422
top chord)
442 455 455 442 455 455 438 452
96% 93% 93% 96% 93% 93% 97% 93%
chord)
Mass 6524 6335 6300 6524 6335 6300 6510 6321
Area 5,94 6,33 6,73 5,94 6,33 6,73 5,93 6,33
Solution 1 2 3 4 5 6 7 8
Repeated - - - x x x (reset)
Geometry Howe
nΔ 7 7 7 7 10 12
ncb 2 2 2 2 1 1
Bottom chord VKR 100x100x6.3 VKR 120x120x4.5 VKR 100x100x6.3 VKR 120x120x4.5 VKR 100x100x4 VKR 100x100x4
Top chord VKR 100x100x5 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5 VKR 100x100x5
Diagonals VKR 140x70x5 VKR 140x80x4 VKR 140x70x5 VKR 140x80x4 VKR 120x60x5 VKR 120x60x3.6
Vertical end posts KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3 KCKR 40.4/3
Cross beams VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 100x60x3.6 VKR 120x60x3.6 VKR 120x60x3.6
End cross beams KCKR 60.3/3 KCKR 60.3/3 KCKR 60.3/3 KCKR 60.3/3 VKR 100x60x3.6 VKR 100x60x3.6
Displacement [m] SLS 0,019 0,019 0,020 0,019 0,019 0,020 0,023 0,022
0,05 0,05 0,05 0,05 0,05 0,05 0,05 0,05
SLS 6.15ab1
Utilization of displacement 38% 39% 41% 38% 39% 41% 45% 44%
Top chord Top chord Top chord Top chord Top chord Top chord Cross beam Bottom chord
Utilization ratio 85% 85% 85% 85% 85% 85% 85% 83%
Max allowable utilization ratio 85% 85% 85% 85% 85% 85% 85% 85%
Utilizaiton of utilization 100% 100% 100% 100% 100% 100% 100% 98%
406 404 403 406 404 403 409 412
top chord)
423 418 424 423 418 424 432 432
96% 97% 95% 96% 97% 95% 95% 95%
chord)
Mass 7080 6928 6842 7080 6928 6842 6692 6573
Area 6,41 6,74 7,14 6,41 6,74 7,14 6,34 6,61
Solution 1 2 3 4 5 6 7 8
Repeated - - - x x x (reset) x x

Optimization run G
Cross section x
Topology x
Geometry Warren
nΔ 12 12 11 11
ncb 1 1 1 1
Bottom chord VKR 100x100x10 VKR 100x100x10 VKR 140x140x6.3
Top chord VKR 140x140x10 VKR 180x180x6.3 VKR 180x180x6.3
Diagonals VKR 120x60x3.6 VKR 100x60x4 VKR 100x60x4
Cross beams KCKR 139.7/4 KCKR 139.7/4 KCKR 139.7/4
End cross beams KCKR 101.6/4 KCKR 88.9/4 KCKR 88.9/4
Displacement [m] SLS 0,035 0,035 0,036 0,036 0,036 0,036 0,037 0,037 0,038 0,035 0,035 0,036 0,036 0,037 0,037 0,038 0,035 0,035 0,036 0,036 0,037 0,037 0,038
Max allowable
displacement [m] SLS 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075
6.15ab1
Utilization of displacement 46% 46% 48% 49% 48% 48% 50% 50% 51% 46% 46% 48% 48% 50% 50% 51% 46% 46% 48% 48% 50% 50% 51%
Vertical end post Vertical end post Vertical end post Vertical end post Diagonals Diagonals Bottom chord Vertical end post Vertical end post Vertical end post Diagonals Diagonals Diagonals Vertical end post Vertical end post Vertical end post Diagonals Diagonals Diagonals
Utilization ratio 79% 78% 81% 78% 82% 83% 84% 84% 81% 79% 78% 82% 83% 84% 84% 84% 79% 78% 82% 83% 84% 84% 84%
Max allowable utilization
85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
ratio
Utilizaiton of utilization 94% 92% 96% 92% 97% 97% 99% 99% 96% 94% 92% 97% 97% 99% 99% 99% 94% 92% 97% 97% 99% 99% 99%
Ned [kN] (Max normal
926,5 926,5 922,7 932,3 924,2 924,2 919,4 919,4 917,1 926,5 926,5 924,2 924,2 919,4 919,4 916,8 926,5 926,5 924,2 924,2 919,4 919,4 916,8
force in top chord)
927,2 927,3 923,2 974,9 930,0 928,8 926,6 926,0 978,3 927,2 927,3 930,0 928,8 926,6 926,0 932,2 927,2 927,3 930,0 928,8 926,6 926,0 932,2
Utilization normal force
100% 100% 100% 96% 99% 100% 99% 99% 94% 100% 100% 99% 100% 99% 99% 98% 100% 100% 99% 100% 99% 99% 98%
(top chord)
Mass [kg] 12051 12049 11831 11830 11799 11790 11567 11558 11521 12051 12049 11799 11790 11567 11558 11500 12051 12049 11799 11790 11567 11558 11500
Transparency (side view
10,12 10,15 10,42 10,54 10,59 10,63 11,19 11,23 12,39 10,12 10,15 10,59 10,63 11,19 11,23 12,39 10,12 10,15 10,59 10,63 11,19 11,23 12,39
area) [m2]
Solution 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 10 11 12 13 14 15 16
Repeated - - - - - - - - - x x x x x x x x x x x x x
Natural frequency [Hz] 4,4 4,2
Geometry Pratt
nΔ 10 9 10 9 10 9
ncb 1 1 1 1 1 1
Bottom chord VKR 100x100x10 VKR 140x140x6.3 VKR 100x100x10 VKR 140x140x6.3 VKR 100x100x10 VKR 140x140x6.3
Top chord VKR 140x140x10 VKR 180x180x6.3 VKR 140x140x10 VKR 180x180x6.3 VKR 140x140x10 VKR 180x180x6.3
Diagonals VKR 120x60x4 VKR 120x60x4 VKR 120x60x4 VKR 120x60x4 VKR 120x60x4 VKR 120x60x4
Vertical end posts VKR 100x60x6.3 VKR 120x60x3.6 VKR 100x60x6.3 VKR 120x60x3.6 VKR 100x60x6.3 VKR 120x60x3.6
Cross beams VKR 160x80x6.3 VKR 140x70x4 VKR 160x80x6.3 VKR 140x70x4 VKR 160x80x6.3 VKR 140x70x4
End cross beams VKR 100x100x4 VKR 100x100x4 VKR 100x100x4 VKR 100x100x4 VKR 100x100x4 VKR 100x100x4
Displacement [m] SLS 0,038 0,039 0,040 0,040 0,040 0,041 0,043 0,044 0,038 0,039 0,039 0,040 0,040 0,043 0,044 0,038 0,040 0,396 0,043 0,044
Max allowable
displacement [m] SLS 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075
6.15ab1
Utilization of displacement 51% 52% 53% 53% 54% 55% 58% 58% 51% 52% 52% 53% 53% 58% 58% 51% 53% 527% 58% 58%
Span length 30 Element where the max
Diagonals Bottom chord Diagonals Diagonals Bottom chord Bottom chord Bottom chord Cross beam Bottom chord Diagonals Top chord Vertical end post Cross beam Bottom chord
m (OCTOPUS) utilization ratio occurs
Utilization ratio 82% 81% 81% 82% 82% 83% 84% 85% 82% 83% 81% 81% 80% 84% 85% 82% 81% 81% 84% 85%
85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
ratio
Utilizaiton of utilization 97% 95% 95% 96% 96% 98% 99% 100% 97% 98% 95% 95% 94% 99% 100% 97% 95% 96% 99% 100%
992,2 990,2 988,8 988,8 984,2 983,9 979,2 967,9 992,2 990,2 990,2 988,8 985,0 979,2 976,9 992,2 988,8 986,4 979,2 976,9
force in top chord)
1009,7 990,4 991,7 991,7 993,8 988,0 981,1 999,1 1009,7 990,8 990,4 992,2 1014,2 981,1 999,1 1009,7 993,4 1034,8 981,1 999,1
98% 100% 100% 100% 99% 100% 100% 97% 98% 100% 100% 100% 97% 100% 98% 98% 100% 95% 100% 98%
(top chord)
Mass [kg] 12280 12236 12121 12116 11941 11939 11659 11592 12280 12234 12231 12131 11922 11659 11592 12280 12139 12035 11659 11592
10,22 10,47 10,52 10,58 11,07 11,42 11,42 12,62 10,22 10,47 10,52 10,58 10,82 11,42 12,62 10,22 10,58 10,82 11,42 12,62
area) [m2]
Solution 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Repeated - - - - - - - - x x x x x x
Geometry Howe
nΔ 14 14 13
ncb 1 1 1
Top chord VKR 120x120x10 VKR 160x160x6.3 VKR 120x120x10
Diagonals VKR 140x70x6.3 VKR 160x80x4 VKR 160x80x4
Cross beams KCKR 139.7/4 VKR 120x60x4 KCKR 139.7/4
End cross beams VKR 140x70x4 VKR 120x60x3.6 VKR 120x60x3.6
Displacement [m] SLS 0,040 0,039 0,040 0,040 0,041 0,043 0,040 0,039 0,040 0,041 0,044 0,040 0,040 0,041 0,044
Max allowable
displacement [m] SLS 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075 0,075
6.15ab1
Utilization of displacement 53% 52% 54% 53% 55% 57% 53% 52% 54% 55% 59% 53% 54% 55% 59%
Bottom chord Top chord Bottom chord Bottom chord Top chord
Utilization ratio 85% 84% 84% 84% 84% 85% 85% 84% 84% 84% 85% 85% 84% 84% 85%
85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85% 85%
ratio
Utilizaiton of utilization 100% 99% 99% 99% 99% 100% 100% 99% 99% 99% 99% 100% 99% 99% 99%
982,2 976,4 974,0 972,4 971,7 967,0 982,2 976,4 974,0 971,7 966,2 982,2 974,0 971,7 966,2
force in top chord)
986,7 993,0 975,8 977,4 1035,2 1007,7 986,7 999,3 975,8 1035,2 997,4 986,7 975,8 1035,2 997,4
100% 98% 100% 99% 94% 96% 100% 98% 100% 94% 97% 100% 100% 94% 97%
(top chord)
Mass [kg] 12909 12555 12374 12304 12298 12089 12909 12569 12374 12298 12027 12909 12374 12298 12027
10,50 10,63 11,04 11,17 11,64 12,24 10,50 10,60 11,04 11,64 12,24 10,50 11,04 11,64 12,24
area) [m2]
Solution 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Repeated - - - - - - x x x x x x x

105

Truss

Uploaded by

Copyright:

Available Formats

Truss

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Truss

Uploaded by

Copyright:

Available Formats

What is the thesis investigating?

What is the thesis investigating?

What are the objectives of the thesis?

What are the objectives of the thesis?

Evolutionary optimization of steel truss

Evolutionary optimization of steel truss footbridges

Master’s thesis in Structural Engineering and Building Technology

Department of Architecture and Civil Engineering

© VICTOR ANDERSSON , CECILIA HILLBERG, 2018

Master’s thesis ACEX30-18-35

Gothenburg, Sweden 2018

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 i

ii , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

2 Steel truss footbridges 4

3 Optimization and parametric design 10

4 Development and verification of the model in

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 iii

6 Results of optimization of Warren, Pratt and

Appendix A Gradient-free and gradient-based methods 65

Appendix B Plans of the case study over Genevadsån 69

Appendix C Cross-sections inserted in Karamba 3D 70

Appendix D Verification of the Warren sub model in Karamba 3D with FEM-Design 75

Appendix F Verification of the Pratt sub model in Karamba 3D with FEM-Design 91

Appendix G Verification of the Howe sub model in Karamba 3D with FEM-Design 94

Appendix I All optimization results for part 1 and part 2 101

iv , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 v

vi , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 vii

viii , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 ix

BOF Blast oxygen furnace.

CAD Computer aided design.

DLF Direct laser fabrication.

EAF Electric arc furnace.

FEM Finite element modeling.

HypE Hypervolume estimation algorithm for multi-objective optimization.

LENS Laser engineered net shaping.

LSF Laser solid forming.

MOA Multi-objective optimization algorithm.

MOEA Multi-objective evolutionary algorithms.

MOP Multi-objective optimization problem.

NURBS Non-uniform rational b-spline.

SLS Serviceability limit state.

SPEA-2 Strength Pareto evolutionary algorithm 2.

ULS Ultimate limit state.

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 xi

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 1

2 , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

• Maximum allowable deflection of the footbridge (SLS)

• Cross-sectional capacity of the elements (ULS)

• Global lateral torsional buckling of the top chord (ULS)

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 3

Figure 2.1: Explanation of the different parts in a truss bridge.

4 , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 5

2.1 Steel as a construction material

Starting at a microscopic plane of description, metals consist of an atomic structure, where a

6 , Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35

, Department of Architecture and Civil Engineering, Master’s thesis, ACEX30-18-35 7