You are on page 1of 92

Fatigue Life Analysis of Weld Ends

Comparison between finite element analysis and testing

Marcus Ramström

Mechanical Engineering / Solid Mechanics


Division of Solid Mechanics

Master thesis
Department of Management and Engineering
LIU-IEI-TEK-A--15/02300—SE
Fatigue Life Analysis of Weld Ends
Comparison between finite element analysis and testing

Marcus Ramström

Supervisor: Maria Nygren, Toyota Material Handling, Mjölby


Richard Fyhr, Toyota Material Handling, Mjölby
Daniel Leidermark, Linköping University, Linköping

Examiner: Bo Torstenfelt, Linköping University, Linköping

Student reviewer: Hugo Inglehammar

LIU-IEI-TEK-A--15/02300—SE
Linköping University Electronic Press

M. Ramström ǀ Fatigue life analysis with respect to weld end modelling

Upphovsrätt
Detta dokument hålls tillgängligt på Internet – eller dess framtida ersättare – från publiceringsdatum under
förutsättning att inga extraordinära omständigheter uppstår.
Tillgång till dokumentet innebär tillstånd för var och en att läsa, ladda ner, skriva ut enstaka kopior för
enskilt bruk och att använda det oförändrat för ickekommersiell forskning och för undervisning.
Överföring av upphovsrätten vid en senare tidpunkt kan inte upphäva detta tillstånd. All annan användning
av dokumentet kräver upphovsmannens medgivande. För att garantera äktheten, säkerheten och
tillgängligheten finns lösningar av teknisk och administrativ art.
Upphovsmannens ideella rätt innefattar rätt att bli nämnd som upphovsman i den omfattning som god sed
kräver vid användning av dokumentet på ovan beskrivna sätt samt skydd mot att dokumentet ändras eller
presenteras i sådan form eller i sådant sammanhang som är kränkande för upphovsmannens litterära eller
konstnärliga anseende eller egenart.
För ytterligare information om Linköping University Electronic Press se förlagets hemsida
http://www.ep.liu.se/

Copyright
The publishers will keep this document online on the Internet – or its possible replacement –from the date
of publication barring exceptional circumstances.
The online availability of the document implies permanent permission for anyone to read, to
download, or to print out single copies for his/hers own use and to use it unchanged for non-commercial
research and educational purpose. Subsequent transfers of copyright cannot revoke this permission. All
other uses of the document are conditional upon the consent of the copyright owner. The publisher has
taken technical and administrative measures to assure authenticity, security and accessibility.
According to intellectual property law the author has the right to be mentioned when his/her work is
accessed as described above and to be protected against infringement.
For additional information about the Linköping University Electronic Press and its procedures for
publication and for assurance of document integrity, please refer to its www home page:
http://www.ep.liu.se/.

© Marcus Ramström, 2015

I
I
M. Ramström ǀ Fatigue life analysis with respect to weld end modelling

ABSTRACT

This master thesis investigates the fatigue life of weld ends. Fatigue life analyses have been an important
subject in many industries but the research for weld ends is limited, and often the weld ends are the most
critical parts of the weld.

The main purpose of this thesis is to find a calculation method for prediction of fatigue life of weld ends.
Today’s computational methods for fatigue life calculations are based on continuous welds. The methods
assume that the weld does not have any start or end. Therefore, focus is on how to model the weld ends
accurate in a finite element (FE) analysis, since it affects the prediction of the fatigue life. The Effective
notch approach is the calculation method at focus for this research in order to develop a new value of the
fatigue class (FAT) adjusted for weld ends. The Effective notch method is only applicable for a
continuous weld. Therefore, a comparison between the continuous weld and discontinuous weld end is
made.

The project is divided into several parts, the main parts are; fatigue testing between a continuous and
discontinuous weld end, validation against strain gauge measurements, calibration of the FE-models,
comparison between the computational methods and develop a new FAT-value for weld end based on the
Effective notch approach.

The tests result indicates higher fatigue strength for the discontinuous weld end compared with the
continuous weld. The fatigue testing is compared with FE-analysis, with focus on the Effective notch
approach. The Effective notch method with standard FAT225 MPa, estimates the fatigue life close to the
discontinuous test results. The adjusted FAT-value to match the discontinuous weld end is in the interval
194-269 MPa, dependent on actual test results or International Institute of Welding (IIW) standard norm.

The conclusion is that it is possible to simulate the discontinuous weld end with a continuous FE-model.
The continuous model can be evaluated with the developed FAT-value for the Effective notch approach.
The FE-model should include the notch radius and the standards IIW recommend.

II
III
M. Ramström ǀ Fatigue life analysis with respect to weld end modelling

PREFACE AND ACKNOWLEDGEMENT

The work presented in this master thesis has been carried out at Toyota Material Handling, Mjölby and at
the Division of Solid Mechanics, Linköping University, during the spring semester 2015. The project has
been performed at the CAE-group. This thesis is a research study of fatigue life analysis of weld ends. By
this thesis, I will conclude my Master of Science in Mechanical Engineering at Linköping University,
Sweden.

First, I would like to thank my supervisors, for all their support and guidance during the project.

Maria Nygren Specialist, Finite element analysis and fatigue analysis.


Toyota Material Handling

Richard Fyhr Manager CAE-group


Toyota Material Handling

Daniel Leidermark Associate, Division of Solid Mechanics.


Linköping University. Linköping Institute of Technology.

I would also like to thank the CEA-group and the coworkers I meet at Toyota Material Handling for their
support and interesting discussions and my opponent, Hugo Ingelhammar, for rewarding opinions and
discussions.

Mjölby 2015-06-04

Marcus Ramström

IV
V
M. Ramström ǀ Chapter 1 - Introduction

NOMENCLATURE
𝐸 = 𝑌𝑜𝑢𝑛𝑔′ 𝑠 𝑚𝑜𝑑𝑢𝑙𝑢𝑠 [𝐺𝑃𝑎]
𝜈 = 𝑃𝑜𝑠𝑠𝑖𝑜𝑛′ 𝑠 𝑟𝑎𝑡𝑖𝑜 [−]
𝜌 = 𝐷𝑒𝑛𝑠𝑖𝑡𝑦 [𝑘𝑔/𝑚3 ]
𝑅𝑝0.2 = 0.2 % 𝑜𝑓𝑓𝑠𝑒𝑡 𝑦𝑖𝑒𝑙𝑑 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ [𝑀𝑃𝑎]
𝑅𝑚 = 𝑈𝑙𝑡𝑖𝑚𝑎𝑡𝑒 𝑡𝑒𝑛𝑠𝑖𝑙𝑒 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ [𝑀𝑃𝑎]
𝜎𝑟 = 𝐶𝑜𝑚𝑝𝑢𝑡𝑒𝑑 𝑠𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑛𝑔𝑒 [𝑀𝑃𝑎]
∆𝜎𝑟𝑑 = 𝐷𝑒𝑠𝑖𝑔𝑛 𝑛𝑜𝑟𝑚𝑎𝑙 𝑠𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑛𝑔𝑒 [𝑀𝑃𝑎]
𝜎𝑛𝑜𝑚 = 𝑁𝑜𝑚𝑖𝑛𝑎𝑙 𝑠𝑡𝑟𝑒𝑠𝑠 [𝑀𝑃𝑎]
𝜎ℎ𝑠 = 𝐻𝑜𝑡 𝑠𝑝𝑜𝑡 𝑠𝑡𝑟𝑒𝑠𝑠 [𝑀𝑃𝑎]
∆𝜎𝑚𝑎𝑥 = 𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑠𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑛𝑔𝑒 [𝑀𝑃𝑎]
𝜎𝑚𝑎𝑥 = 𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑠𝑡𝑟𝑒𝑠𝑠 [𝑀𝑃𝐴]
𝜎𝑚𝑖𝑛 = 𝑀𝑖𝑛𝑖𝑚𝑢𝑚 𝑠𝑡𝑟𝑒𝑠𝑠 [𝑀𝑃𝐴]
𝐹𝐴𝑇 = 𝐹𝑎𝑡𝑖𝑔𝑢𝑒 𝑐𝑙𝑎𝑠𝑠 [𝑀𝑃𝑎]
𝑅 = 𝑆𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑡𝑖𝑜
𝑎 = 𝑇ℎ𝑟𝑜𝑎𝑡 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [𝑚]
𝑁 = 𝐸𝑠𝑡𝑖𝑚𝑎𝑡𝑒𝑑 𝑓𝑎𝑡𝑖𝑔𝑢𝑒 𝑙𝑖𝑓𝑒 [𝑐𝑦𝑐𝑙𝑒𝑠]
𝑁𝑡 = 𝐷𝑒𝑠𝑖𝑔𝑛 𝑙𝑖𝑓𝑒, 𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑦𝑐𝑙𝑒𝑠 [𝑐𝑦𝑐𝑙𝑒𝑠]
𝑛𝑡 = 𝑇𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑦𝑐𝑙𝑒𝑠 [𝑐𝑦𝑐𝑙𝑒𝑠]
𝑛𝑖 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑐𝑦𝑐𝑙𝑒𝑠 𝑎𝑡 𝑙𝑜𝑎𝑑 𝑙𝑒𝑣𝑒𝑙 𝑖 [𝑐𝑦𝑐𝑙𝑒𝑠]
𝜑𝑡 = 𝑇ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑓𝑎𝑐𝑡𝑜𝑟
𝜑𝑚 = 𝑀𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝑓𝑎𝑐𝑡𝑜𝑟
𝜑𝑒 = 𝑆𝑡𝑟𝑒𝑠𝑠 𝑣𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟
𝜑𝑄 = 𝑅𝑖𝑠𝑘 𝑜𝑓 𝑓𝑎𝑖𝑙𝑢𝑟𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡
𝛾𝑚 = 𝐹𝑎𝑖𝑙𝑢𝑟𝑒 𝑐𝑜𝑛𝑠𝑒𝑞𝑢𝑛𝑐𝑒 𝑓𝑎𝑐𝑡𝑜𝑟
𝑠𝑚 = 𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑠𝑡𝑟𝑒𝑠𝑠 𝑝𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟
∆𝜎𝑖 = 𝑆𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑛𝑔𝑒 𝑎𝑡 𝑙𝑜𝑎𝑑 𝑙𝑒𝑣𝑒𝑙 𝑖 [𝑀𝑃𝑎]
∆𝜎𝑟𝑒𝑓 = 𝑅𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑣𝑎𝑙𝑢𝑒 𝑓𝑜𝑟 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑠𝑡𝑟𝑒𝑠𝑠 𝑟𝑎𝑛𝑔𝑒 [𝑀𝑃𝑎]
𝑘𝑚 = 𝐶𝑜𝑙𝑙𝑒𝑐𝑡𝑖𝑣𝑒 𝑓𝑎𝑐𝑡𝑜𝑟
𝑚 = 𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡, 𝑔𝑟𝑎𝑑𝑖𝑒𝑛𝑡 𝑜𝑓 𝑡ℎ𝑒 𝑆/𝑁 𝑐𝑢𝑟𝑣𝑒
𝑓 = 𝐸𝑥𝑝𝑜𝑛𝑒𝑛𝑡 𝑢𝑠𝑒𝑑 𝑓𝑜𝑟 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑓𝑎𝑐𝑡𝑜𝑟
∆𝐿 = 𝐷𝑒𝑙𝑡𝑎 𝑙𝑒𝑛𝑔𝑡ℎ, 𝑙𝑒𝑛𝑔𝑡ℎ 𝑐ℎ𝑎𝑛𝑔𝑒 𝑓𝑟𝑜𝑚 𝑖𝑛𝑖𝑡𝑖𝑎𝑙 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑙𝑒𝑛𝑔𝑡ℎ. [𝑚]
𝐿 = 𝐼𝑛𝑖𝑡𝑖𝑎𝑙 𝑙𝑒𝑛𝑔𝑡ℎ [𝑚]
𝛼 = 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑒𝑥𝑝𝑎𝑛𝑠𝑖𝑜𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 [1/°C]
∆𝑇 = 𝐷𝑒𝑙𝑡𝑎 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒, 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑐ℎ𝑎𝑛𝑔𝑒 𝑓𝑟𝑜𝑚 𝑖𝑛𝑖𝑡𝑖𝑎𝑙 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 [°𝐶]
𝐷 = 𝑃𝑎𝑟𝑖𝑠 ′ 𝑙𝑎𝑤 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡
𝑔 = 𝑃𝑎𝑟𝑖𝑠 ′ 𝑙𝑎𝑤 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡

ABBREVIATIONS
𝐶𝐴𝐸 = 𝐶𝑜𝑚𝑝𝑢𝑡𝑒𝑟 𝑎𝑖𝑑𝑒𝑑 𝑒𝑛𝑔𝑖𝑛𝑒𝑒𝑟𝑖𝑛𝑔
𝐹𝐸 = 𝐹𝑖𝑛𝑖𝑡𝑒 𝐸𝑙𝑒𝑚𝑒𝑛𝑡
𝐹𝐸𝑀 = 𝐹𝑖𝑛𝑖𝑡𝑒 𝐸𝑙𝑒𝑚𝑒𝑛𝑡 𝑀𝑒𝑡ℎ𝑜𝑑
𝐼𝐼𝑊 = 𝐼𝑛𝑡𝑒𝑟𝑛𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝐼𝑛𝑠𝑡𝑖𝑡𝑢𝑡𝑒 𝑜𝑓 𝑊𝑒𝑙𝑑𝑖𝑛𝑔
𝑑. 𝑜. 𝑓 = 𝐷𝑒𝑔𝑟𝑒𝑒 𝑜𝑓 𝑓𝑟𝑒𝑒𝑑𝑜𝑚
𝑠𝑡𝑑𝑣. = 𝑆𝑡𝑎𝑛𝑑𝑎𝑟𝑑 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛

VI
M. Ramström ǀ Chapter 1 - Introduction

Table of Contents

CHAPTER 1 ................................................................................................................................................ 1
INTRODUCTION ....................................................................................................................................... 1
1.1 Background ................................................................................................................................... 1
1.2 Toyota Material Handling ............................................................................................................. 2
1.3 Problem description ....................................................................................................................... 3
1.4 Purpose .......................................................................................................................................... 4
1.5 Goal ............................................................................................................................................... 4
1.6 Question at issue ............................................................................................................................ 4
1.7 Delimitations ................................................................................................................................. 5
1.8 Method........................................................................................................................................... 7
1.8.1 Literature study...................................................................................................................... 7
1.8.2 Finite element analysis .......................................................................................................... 7
1.8.3 Fatigue testing ....................................................................................................................... 7
1.9 Finite element software ................................................................................................................. 8
1.10 Other considerations ...................................................................................................................... 8
CHAPTER 2 ................................................................................................................................................ 9
THEORY...................................................................................................................................................... 9
2.1 Introduction of fatigue ................................................................................................................... 9
2.1.1 Fatigue class ........................................................................................................................ 10
2.1.2 S/N Curve ............................................................................................................................ 10
2.2 Welds ........................................................................................................................................... 13
2.2.1 Welding classes ................................................................................................................... 14
2.3 How the material is affected by welding ..................................................................................... 14
2.3.1 Residual stresses .................................................................................................................. 14
2.3.2 Deformation due to weld shrinkage..................................................................................... 15
2.3.3 Crack initiation .................................................................................................................... 15
2.4 Welds affect in structures ............................................................................................................ 16
2.5 Welds effect on fatigue life ......................................................................................................... 16
2.6 Fatigue design of welded joints and components ........................................................................ 17
2.6.1 Nominal stress method ........................................................................................................ 18
2.6.2 Geometrical stress / Hot-spot method ................................................................................. 20
2.6.3 Effective notch method ........................................................................................................ 21
2.6.4 Generally for the methods ................................................................................................... 23
2.6.5 Fracture mechanics .............................................................................................................. 24
2.6.6 Evaluation of test data ......................................................................................................... 26

VII
M. Ramström ǀ Chapter 1 - Introduction

CHAPTER 3 .............................................................................................................................................. 27
FATIGUE TESTING ................................................................................................................................ 27
3.1 Fatigue testing ............................................................................................................................. 27
3.1.1 Weld properties ................................................................................................................... 27
3.1.2 Measuring before testing ..................................................................................................... 29
3.1.3 Test setup ............................................................................................................................. 30
3.2 Strain gauge test .......................................................................................................................... 30
3.2.1 Strain gauge test results ....................................................................................................... 32
Test program............................................................................................................................................ 35
3.3 Results from testing ..................................................................................................................... 35
CHAPTER 4 .............................................................................................................................................. 39
FINITE ELEMENT ANALYSIS ............................................................................................................. 39
4.1 FE-model ..................................................................................................................................... 39
4.2 Model Description ....................................................................................................................... 39
4.2.1 Material ............................................................................................................................... 40
4.2.2 Analysis setup...................................................................................................................... 40
4.3 Mesh convergence study ............................................................................................................. 40
4.3.1 Elements .............................................................................................................................. 40
4.3.2 Convergence study .............................................................................................................. 40
4.4 Boundary conditions and Load cases study................................................................................. 43
4.4.1 Comparison study for nonlinear geometric ......................................................................... 45
4.5 Validation of FE-model ............................................................................................................... 47
4.5.1 Boundary condition, Load case and load level .................................................................... 47
4.5.2 Simulation of the pre-stress ................................................................................................. 47
4.6 Numerical analysis ...................................................................................................................... 49
4.6.1 Effective notch model.......................................................................................................... 49
4.6.2 Hot-spot model .................................................................................................................... 51
4.6.3 Nominal stress method ........................................................................................................ 53
CHAPTER 5 .............................................................................................................................................. 54
RESULTS ................................................................................................................................................... 54
5.1 Test results................................................................................................................................... 54
5.2 Comparison of the computational methods ................................................................................. 55
5.3 Comparison study between FE-analysis and tests ....................................................................... 56
5.4 Developed FAT-value ................................................................................................................. 57
5.4.1 Discontinuous weld end ...................................................................................................... 58
5.4.2 Continuous weld end ........................................................................................................... 59

VIII
M. Ramström ǀ Chapter 1 - Introduction

CHAPTER 6 .............................................................................................................................................. 61
CONCLUSIONS........................................................................................................................................ 61
6.1 Tests ............................................................................................................................................ 61
6.2 Comparison of computational methods and tests ........................................................................ 61
6.3 Comparison of the Effective notch method and tests .................................................................. 61
6.4 Developed new FAT-value.......................................................................................................... 62
CHAPTER 7 .............................................................................................................................................. 63
DISCUSSION............................................................................................................................................. 63
CHAPTER 8 .............................................................................................................................................. 65
FUTURE WORK ...................................................................................................................................... 65
BIBLIOGRAPHY ..................................................................................................................................... 67
APPENDIX ................................................................................................................................................ 69
APPENDIX A ......................................................................................................................................... 69
Drawings for the specimens ................................................................................................................ 69
APPENDIX B.......................................................................................................................................... 74
Strain Gauges placement ..................................................................................................................... 74
APPENDIX C.......................................................................................................................................... 76
Nominal stress method ........................................................................................................................ 76

IX
M. Ramström ǀ Chapter 1 - Introduction

List of Figures
Figure 1. a. Continuous T-joint fillet weld. b. Zoom in on continuous fillet weld. c. Discontinuous T-joint
fillet weld. d. Zoom in on discontinuous fillet weld...................................................................................... 5
Figure 2. Specimens dimension. a) Discontinuous weld end. b) Continuous weld end. ............................... 6
Figure 3. Flowchart that describes the projects procedure. ........................................................................... 7
Figure 4. Abaqus work procedure. ................................................................................................................ 8
Figure 5. S/N curve with key positions. [13]............................................................................................... 11
Figure 6. Weld zones. [15] .......................................................................................................................... 13
Figure 7. To the left is a real fillet weld profile, [7] to the right is a simplified fillet weld profile, with
marked throat thickness, referred as a, and 45 degree angles...................................................................... 14
Figure 8. Fatigue fracture modes in a fillet weld. [7] .................................................................................. 16
Figure 9. Comparison between the different methods used for fatigue life design. .................................... 17
Figure 10. To the left, points for linear extrapolation and to the right, points for quadratic extrapolation. 21
Figure 11. Fictitious notch radii. ................................................................................................................. 22
Figure 12. FE-model with notch radius. ...................................................................................................... 22
Figure 13. Illustrates the crack propagation rate as a function of stress intensity range. [12] ..................... 25
Figure 14. Robot welding. To the left a discontinuous string. .................................................................... 27
Figure 15. Weld drawing for the remade continuous weld end................................................................... 28
Figure 16. End sides for the continuous weld design. To the left side A, and to the right side B. .............. 28
Figure 17. Continuous weld, to the left the first string, to the right the second string. ............................... 28
Figure 18. End sides for the discontinuous weld end design....................................................................... 29
Figure 19. Specimens during measurements and markup. .......................................................................... 29
Figure 20. The heat affects. The above is the discontinuous end, and the below is the continuous end. .... 29
Figure 21. Tensile test machine. .................................................................................................................. 30
Figure 22. Red lines indicate strain gauge placements. Left, continuous weld end. Right, discontinuous
weld end. ..................................................................................................................................................... 31
Figure 23. Strain gauge placements for the discontinuous weld end. ......................................................... 32
Figure 24. Strain gauge placements for the continuous weld end. To the left side A and to the right side B.
..................................................................................................................................................................... 32
Figure 25. The recorded pre-stresses for the discontinuous weld end. Four gauges in the figure; OS_AH_3,
OS_AV_3, OS_BH_3 and OS_BV_3. ........................................................................................................ 34
Figure 26. The start of the load level. Four gauges in the figure; OS_AH_3, OS_AV_3, OS_BH_3 and
OS_BV_3. ................................................................................................................................................... 34
Figure 27. Overview for the discontinuous weld end, all four load levels. Four gauges in the figure;
OS_AH_3, OS_AV_3, OS_BH_3 and OS_BV_3. ..................................................................................... 35
Figure 28. Discontinuous weld end. ............................................................................................................ 36
Figure 29. Continuous weld end.................................................................................................................. 36
Figure 30. Example of cracks. To the left a continuous, and to the right a discontinuous. ......................... 37
Figure 31. a. T-joint fillet weld with continuous weld end b. T-joint fillet weld with discontinuous weld
end. .............................................................................................................................................................. 39
Figure 32. The path starts in the weld toe and continuous 14 mm perpendicular against the weld. ........... 40
Figure 33. Illustrates Mesh 1, Mesh2, Mesh 3 and Mesh 4. The grey area indicates stresses above 225
MPa. ............................................................................................................................................................ 41
Figure 34. Result for the different mesh qualities. ...................................................................................... 42
Figure 35. a. B.C. and load case 1. b. B.C. and load case 2. c. B.C. and load case 3. d. B.C. and load case
4. e. B.C. and load case 5. ........................................................................................................................... 44
Figure 36. Results for B.C. and load case study. (F = 120 kN) ................................................................... 45
Figure 37. Path for Nlgeom study. (F= 240kN, R=0.1, mesh size = 0.5mm) ............................................. 46
Figure 38. Comparison study between Nlgeom and strain gauge measurements. ...................................... 46
Figure 39. Base plate with curvature. .......................................................................................................... 47

X
M. Ramström ǀ Chapter 1 - Introduction

Figure 40. Boundary condition for simulating mounting in the test machine. ............................................ 48
Figure 41. Pre-stress due to mounting in the test machine. ......................................................................... 48
Figure 42. Deformation scale factor is set to 25. ......................................................................................... 48
Figure 43. Sub-model dimensions and notch refinement. Notch radius = 1 mm, (toes and root), Green
radius = 2 mm. ............................................................................................................................................. 50
Figure 44. Sub-model dimensions. .............................................................................................................. 50
Figure 45. Sub-model mesh, element size 0.25 mm in the notches. ........................................................... 50
Figure 46. Effective notch result. Grey area indicates stress above 225 MPa. (F=120, Mesh size=0.25
mm) ............................................................................................................................................................. 51
Figure 47. Mesh Hot-spot model. ................................................................................................................ 52
Figure 48. Hot-spot results. Result points are marked with red dots, distance 4, 9, 10 and 14 mm from the
weld toe. ...................................................................................................................................................... 52
Figure 49. Comparison between the continuous and discontinuous test results.......................................... 54
Figure 50. Comparison between the computational methods for the continuous weld end. ....................... 55
Figure 51. Comparison between the Effective notch approach and the tests. ............................................. 56
Figure 52. Developed S/N curve discontinuous, characteristic curve with slope 3 and stdv. 0.17. ............ 58
Figure 53. Developed S/N curve discontinuous, characteristic curve with slope 3 and stdv. 0.034. .......... 58
Figure 54. Developed S/N curve discontinuous, characteristic curve with slope 3.13 and stdv. 0.034. ..... 59
Figure 55. Developed S/N curve continuous, characteristic curve with slope 3 and stdv. 0.17. ................. 59
Figure 56. Developed S/N curve continuous, characteristic curve with slope 3 and stdv. 0.052. ............... 60
Figure 57. Developed S/N curve continuous, characteristic curve with slope 3.42 and stdv. 0.052. .......... 60

List of Tables
Table 1. Risk of failure coefficient with slope, m = 3. (Konstruktionshandbok, 2014, [9]) – Table 9-6. ... 12
Table 2. Recommended values of n according to SSAB's research. (SSAB, 2011) – Table 5.9. .............. 19
Table 3. Stress variation factor. (SSAB, 2011) – Eq. 5.12 -5.14................................................................ 19
Table 4. IIW's proposal of values for the partial coefficient in different situations. [3] ............................. 19
Table 5. k-value dependent of number of test results. [3] ........................................................................... 26
Table 6. Stresses from the strain gauge tests. .............................................................................................. 33
Table 7. Test program. ................................................................................................................................ 35
Table 8. Discontinuous test data.................................................................................................................. 36
Table 9. Continuous test data. ..................................................................................................................... 36
Table 10. Test results................................................................................................................................... 37
Table 11. Material properties. ..................................................................................................................... 40
Table 12. Mesh levels for the mesh study. .................................................................................................. 41
Table 13. Load levels. (B.C. and load case 3) ............................................................................................. 47
Table 14. Effective notch results. ................................................................................................................ 51
Table 15. Hot-spot results. Linear and Quadratic........................................................................................ 52
Table 16. Nominal stress method results, Continuous weld end. ................................................................ 53
Table 17. Nominal stress results, Discontinuous weld end. ........................................................................ 53
Table 18. Test data. ..................................................................................................................................... 55
Table 19. FAT-value. .................................................................................................................................. 55
Table 20. Cycles to failure. Effective notch calculations and test data. ...................................................... 56
Table 21. Data for new developed FAT-value, discontinuous. ................................................................... 58
Table 22. Data for new developed FAT-value, continuous. ........................................................................ 59

XI
M. Ramström ǀ Chapter 1 - Introduction

CHAPTER 1
INTRODUCTION

This chapter is an introduction to the project, describing the background, problem, purpose and method.

1.1 Background
Most failures in load carrying structures designed with welds are caused by fatigue failure. This is due to
the strength against cyclic loading, particularly around the welds, which is significantly lower than the
static strength. This ratio is difficult to consider but important when designing welded structures.

Fatigue is a process where the material gradually breaks down by cyclic loading and eventually fails. The
process starts with crack initiation and further crack propagation in areas with high local stress. Fatigue
cracks can be difficult to detect, because it is a local phenomenon characterized by small plastic
deformations. [1]

A welded design’s fatigue life is affected by varying factors whose effects are difficult to predict. Fatigue
strength is therefore regarded as an empirical science and knowledge in the subject has been built through
extensive testing. The factors with the greatest impact on the fatigue strength are; number of cycles,
distribution and intensity of loads and the degree of notch action. Other factors are the steels material’s
static strength, surface quality, weld defects, mean stress, residual stresses, plate thickness, temperature
and corrosion. [1] Fatigue failure in welded designs is usually initiated in the transition zone between the
base material and the weld, but also in the weld root, weld end and by defects in the weld. [2]

For fatigue design of steel structures there are some developed standards. These are primarily for
structures, such as steel building constructions, cranes, ships and pressure vessels. The standards are
developed by the International Institute of Welding, IIW and American Welding Society, AWS. These
standards can only be used to a certain extent, and can for instance not be used to calculate fatigue life of
weld ends. Many industries lack standards for fatigue design of welded structures. [1] This is a major
research area which needs to be investigated more deeply, especially regarding weld ends.

It has been shown in research that a weld with a strict perpendicular end, a so called discontinuous end,
provide approximately equivalent fatigue life as a continuous end. This similarity is investigated deeper in
this thesis through comparison studies between finite element (FE) analysis and tests. [3]

1
M. Ramström ǀ Chapter 1 - Introduction

1.2 Toyota Material Handling


Toyota Material Handling Europe (TMHE) is the European part of the world wide concern Toyota
Material Handling Group (TMHG). TMHG is today the world leader in material handling. [4]

BT Products was founded in 1946 by Ivan Lundqvist in Stockholm, with the main focus to simplify the
movement of the raw material at workplaces. The workplaces in focus were then new production of
houses, apartments and production facilities. The company’s name was originally Bygg- och
Transportekonomi, BT, (Construction and Transport economics). In the beginning the company started to
import forklifts from United States (Clark counterbalanced forklifts) for the Swedish construction
industry. The market needed these new products and the production started. The first BT hand pallet
forklift left the factory in 1948. The market was in need of this new product and the company continued to
grow. In 1949 BT Products created the BT-pallet, together with Statens Järnvägar, SJ, due to the big
amount of material moved by railway. The purpose of the cooperation was to construct a design that could
carry as much material as possible and fit in the freight trains. Then the same pallet standard begun to be
used all over Europe and the BT-pallet were renamed to Euro-pallet. [4] In 1952 BT Products moved their
organization to Mjölby, Sweden. Larger facilities were needed and good connection to the railway was
important. New products were developed to satisfy customer’s demands. This good development of the
company opened up for new opportunities for an increasing market. BT Products were then established all
over Europe, but it was still difficult to reach the market outside Europe, especially in United States. [4] A
major step in the development was in 1978 when BT Products introduced the first driverless forklift. [4]

BT Products bought up Raymond in the United States and BT thereby expanded its market.

In 2000 BT Products became a part of the Toyota family after a two years long cooperation with Toyota.
BT Products have since then belonged to the Toyota Industry Corporation (TICO). Today TICO has four
primary areas; Automotive, Material handling, Electronics and Textile machines. The material handling
organization is mainly composed of the Toyota Material Handling Group (TMHG), where Toyota
Material Handling Europe represents TMHG in Europe.

BT Products also bought up CESAB and their main suppliers LTE in Italy and further expanded its
market.

TMHG has development and research in Japan, Sweden, USA and Italy. TMHE has three production
centers in Europe, placed in Sweden, France and Italy. The main office for TMHE is in Mjölby, Sweden.

Toyota Material Handling Europe is the new name of the organization in Mjölby, but the name BT
Products lives on, largely because of the reputation BT Products has built up. [5]

2
M. Ramström ǀ Chapter 1 - Introduction

1.3 Problem description


Welded joints are commonly used in BT Product's forklifts. The welded joints are designed against both
fatigue failure and static failure. BT Product works with accurate fatigue analysis including IIW standards,
FE-analysis, load analysis and fatigue testing.
There are mainly four methodologies for calculating the fatigue life; Nominal stress method, Geometrical
method also called Hot-spot method, Effective notch method and Fracture mechanics. The Nominal
method and the Hot-spot method are used for simple weld geometry and are only applicable for estimating
the fatigue life of the weld toe. The Effective notch method can manage more complex geometries and
evaluates both the weld toe and root. The Effective notch method is classified as a more accurate method
then the Nominal method and Hot-spot method. The problem with the Effective notch method is that
singular points should be avoided in the FE-model, singularities are avoided by a fictitious radius in the
toe and root. However, there is no method for avoiding singularities in the weld ends. [6] The fracture
mechanics approach is often the most accurate method but it requires significant effort and knowledge,
such as crack length, weld defects and material properties [7]. The weld can be modelled in different ways
in FE-analysis, for example with different detail accuracy such as triangle or convex/concave weld profile,
with/without notch radius and rounded end. The most critical areas in the FE-model are the weld’s start
and end, due to the singularities. The four methods handle this problem differently, but no method is
suitable for weld ends. The Nominal stress method only considers the nominal stress and has different
factors depending on the case. The Hot-spot method considers the strain on two or three points in front of
the weld toe, usually calculated through a FE-analysis.The Effective notch method considers the stress in
the transition zone between the weld and the base material. The transition is rounded with a fictitious
radius recommended by International Institute of Welding, IIW. The Effective notch and the Hot-spot
require that the max principle stress should be perpendicular to the weld toe, but it gives valid results for
max principle stress direction oriented ± 60°, illustrated in Figure 1. Since, this project is a research for
weld ends with the stress direction 90° to the weld toe the original method cannot be used.

Figure 1. The stress direction oriented ± 60° perpendicular to the weld toe.

Today at BT Products the weld is modelled with an isosceles triangle profile with a strict perpendicular
end, so called discontinuous end. No modifications are made at the start or end of the weld. The Effective
notch method with fictitious radius is mainly used at BT Products. They have developed a method to
evaluate the result, where the crucial maximum principle stress is taken in the notch radius at the third
corner node from the welds start or end, notice the node distance is fixed. The problem is that it does not
always provide an accurate stress level, depending on weld geometry, load direction etc. Therefore, the
aim of this project is to develop a modeling technique and evaluation method for fatigue life prediction of
weld ends.

3
M. Ramström ǀ Chapter 1 - Introduction

1.4 Purpose
The main purpose of this thesis is to develop a method for prediction of fatigue life of weld ends. There
are accurate methods and recommendations from IIW and AWS, for how to compute fatigue life of welds,
but for weld ends these methods are not applicable [1].

1.5 Goal
The project’s objective is to develop a modeling technique and evaluation method for fatigue life
prediction of weld ends. The focus is on how to model the weld ends correct in a FE-analysis, since the
modeling affects the prediction of the fatigue life, and to find the right FAT-value for the weld ends, FAT-
value is described in chapter 2.2.1 Fatigue class,. The aim is to guarantee that the calculation method used
to estimate fatigue life match the welds produced in BT Product’s production. This enables more accurate
fatigue life calculations at the dimensioning level in the design process. Overall, this leads to better
product quality, lower development cost, lower claim cost, shorter development time and increased safety
in design.

1.6 Question at issue


Q1 - How exact are the methods estimating the fatigue life of the weld ends?
Q2 - What affect the differences of the results from the methods?
Q3 - Is there a similarity between discontinuous and continuous weld ends regarding fatigue
strength?
Q4 - Can the similarity be used to calculate the fatigue life of weld ends?
Q5 - Does the FAT-value need to be adjusted for dimensioning weld ends?

4
M. Ramström ǀ Chapter 1 - Introduction

1.7 Limitations
All the specimens investigated in this project have the same welded geometry, T-joint fillet weld, shown
in Figure 2. The difference is two types of weld design; discontinuous and continuous weld ends. The first
welding case is a continuous weld end, shown in Figure 2 a. and Figure 2 b. The second case is a
discontinuous weld end, shown in Figure 2 c. and Figure 2 d. That is the original weld design used in BT
Products’.

Load case is limited to alternating tensile load in the base plate’s horizontal direction, z-direction in Figure
2 a. and c. The tests are limited to one load case at four different load levels.

a. b.

d.
c.

Figure 2. a. Continuous T-joint fillet weld. b. Zoom in on continuous fillet weld. c. Discontinuous T-joint fillet weld.
d. Zoom in on discontinuous fillet weld.

5
M. Ramström ǀ Chapter 1 - Introduction

Geometry
The dimensions of the specimen are shown in Figure 3. The fillet weld is made as two types, continuous
respectively discontinuous weld ends. The welds throat thickness is 5 mm and welding class C.

a.

b.

Figure 3. Specimens dimension. a) Discontinuous weld end. b) Continuous weld end.

6
M. Ramström ǀ Chapter 1 - Introduction

Assumptions
- The material properties are assumed isotropic and linear elastic in the FE-analysis. Thus, no
plasticity is considered. This is also how the evaluation methods are designed. Since it is the stress
range that gives fatigue life and not the stress level.

- The weld and the base material have matching material properties in the FE-analysis. The applied
material properties are for the flat iron steel used in the specimens. Material data is taken from BT
Product’s material data base.

- The FE-models mesh convergence is not investigated. The element size is according to IIW
recommendations for each method.

- No properties from the welding operation are considered in the FE-models.

- The residual stresses and the pre-stresses due to the welding process and deformed test specimens
are not considered, since they have very little influence at fatigue testing. The pre-stresses can be
seen in Figure 27.

1.8 Method
This project includes three main parts; literature study, FE-analysis and fatigue testing. The thesis
procedure is described in the flowchart shown in Figure 4.

Calibration
Literature FE- Fatigue
/ Evaluation
study analysis testing Validation

Figure 4. Flowchart that describes the projects procedure.

1.8.1 Literature study


The project starts with a pre-study of literature, articles and further research in the subject. It is performed
to see where the development is today and how the problem could be tackled. It will also provide a
theoretical knowledge and understanding of the subject. The theories used for validation of the FE-model
are described in chapter 2.6 Fatigue design of welded joints and components.

1.8.2 Finite element analysis


The simulations are performed with a FE-program. Different weld designs are analyzed with varying load
levels. The FE-models are analyzed with both ideal geometrical conditions and with misalignments. The
FE results are used in the different evaluation methods to calculate fatigue life of the weld. The focus has
been on the Effective notch method, but the Nominal stress method and Hot-spot method are also evaluated
to receive a comparison. The simulation results are then compared with the test results plotted in an S/N
curve.

1.8.3 Fatigue testing


The testing is performed by clamping the welded specimens in a fatigue test machine. The specimens are
loaded with a repeating tensile load until fracture. The machine is set up with a maximum force and a fix
stress ratio. The specimens are tested at four different load levels and several specimens are tested at each
load level. The test results are used for a comparison study with the FE-analysis plotted in an S/N curve. A

7
M. Ramström ǀ Chapter 1 - Introduction

specimen for each weld design, continuous and discontinuous weld end, sets up with strain gauges. This
test results are used to validate the FE-model.

1.9 Finite element software


In general, the finite element simulation procedure can be divided in three main stages; a pre-processor
where the model’s geometry and setup is defined, a solver where the calculations are performed and a
post-processor where the results can be visualized and studied.

The software used in this project is Abaqus CAE version 6.14-1. Abaqus CAE procedure includes these
three stages, as illustrated in Figure 5. First geometry is created in, or imported to, the pre-processor. In
this stage different conditions are assigned to the model, for example; initial conditions, material
properties, boundary conditions, mesh, etc.

After the pre-processor is done, the job is submitted for analysis and an input file is created. The input file
is named; jobname.inp. In the input file all data are defined in option blocks, option blocks include two
main types of data; Model data and History data. Model data describes specific aspects of the problem
definition, such as initial conditions, element and boundary conditions, etc. History data describes the
option blocks; analysis procedure, loading, output request. Abaqus solver reads the input file and performs
the analysis. The solver sends back an Abaqus Output Database-file, jobname.odb. The output can then be
presented in Abaqus post-processor, named Abaqus Viewer. In the Abaqus Viewer the OBD-file can be
processed with different plug-ins. Furthermore, results can also be directly extracted using Python scripts
from the output-file.

Pre-processor
Abaqus/CAE-file (jobname.cae)

INP-file (jobname.inp)

Solver
(Abaqus)

ODB-file (jobname.odb)

Post-processor
(Abaqus Viewer or Abaqus CAE)

Figure 5. Abaqus work procedure.

1.10 Other considerations


No ethical or gender specific issues are raised or discussed in the project. Nor is it directly related to
questions concerning the environment or sustainable development.

8
M. Ramström ǀ Chapter 2 - Theory

CHAPTER 2
THEORY

This chapter describes the basic theory of fatigue, welds and calculation methods.

2.1 Introduction of fatigue


Fatigue is a localized process when cracks initiate and grow to failure under cyclic and fluctuating loads.
Fatigue begins at stress concentrations. In general, a design is exposed for fatigue if the number of load
cycles is above 1000 repetitions. [8]

A design can be exposed to several million cycles before failure eventually occurs. Fatigue of this kind is
a significant problem in areas around rotating machines such as propellers, and other areas exposed to
constant vibration. Fatigue is also a problem in structures exposed to dynamic and fluctuating loads over a
long period, which is often the case for components in construction machines such as forklifts. [9] 70 %
of all machine failure is due to fatigue failure [10].

High cycle fatigue includes analyzes of infinite and finite fatigue life, typically with number of load cycles
over 100 000. The loading only gives elastic responses. Low cycle fatigue includes analyzes for only finite
failure, typically with number of load cycles lower than 100 000. Cyclic stresses are close to or at the
materials yield limit, thus significant degree of global and local plasticity. [11]

The fatigue process contains three main phases: crack initiation, crack propagation and residue fracture.
Crack initiation and further crack propagation begins when the initiated crack is subjected to dynamic
loadings. The first phase is crack initiation, where microscopic cracks occur in the material’s microscopic
structure. Cracks can be initiated from loads, damages and/or machining etc. High cycle fatigue has two
main initiation causes; stress concentrations and pile-up of dislocations, both are connected to the
metallurgical microstructure. Stress concentrations are caused by inclusions, initial cracks, pores, grains
and corrosion pits etc. This leads to increased stress level locally. The mechanism pile-up of dislocations
is because of persistent slip bands that become the nucleus of cracks. This results in a local decrease in
fatigue strength. The dominating initiation cause depend on the purity of the material, loading, surface
roughness and surface defects/scratches etc. [12]

When the crack is initiated the fatigue process will continue into the next phase, crack propagation. The
three most important factors for crack propagation are: stress intensity range, stress intensity ratio and
stress history. Fracture mechanics analysis can be performed to predict the crack propagation, for
example with Paris’ law. With Paris’ law number of cycles before failure, maximum crack length for the
corresponding load or the specific threshold stress intensity can be calculated. More information about
crack propagation and calculation methods can be found in the chapter 0

9
M. Ramström ǀ Chapter 2 - Theory

Fracture mechanics. [12]

Last phase is residue fracture, which is a type of ultimate failure/fracture. Fracture occurs when a crack
propagates through the material or the weld. This will result in one or several break points.

Fatigue life is affected of environmental conditions, primarily by corrosion, known as corrosion fatigue,
and temperature. Corrosion fatigue will initiate cracks due to corrosion pits because of stress
concentrations. Corrosion fatigue tests indicate a significant reduction of fatigue strength. High service
temperature also reduces the fatigue strength. [8]

Fatigue cracks are sensitive and occur mostly during tensile stress or alternating stresses, but fatigue
cracks have occurred during compressive loads due to positive residual stresses [12]. The fatigue process
is cumulative, the material do not recover when rested. [9]

2.1.1 Fatigue class


Fatigue class (FAT) also known as property classes, is defined as the characteristic fatigue strength at
constant stress range at 2∙106 load cycles. The characteristic fatigue strength has the unit N/mm2. The
fatigue class value can be referred to as FAT or C, dependent on literature. Property classes have been
developed through fatigue testing of various welded joints. Test results have been complied over time,
which has formed the property classes. Property classes are mainly used for fatigue life calculations with
the Nominal stress method. [1]

2.1.2 S/N Curve


The S/N curve, also called Wöhler curve, is a diagram that illustrates fatigue strength, see Figure 6. The
S/N curve shows the number of load cycles to failure as a function of certain stress range. The stress range
is more appropriate to use for welded joints than the stress amplitude, and therefore, stress range is used
hereafter. The S/N curve is created based on test results plotted in a logarithmic diagram. In order to
obtain reliable values, at least five identical specimens should be tested at each stress range. [9]

The assumption of using stress range instead of stress amplitude for welds is because of the residual
stresses in the weld. Residual stresses can occur in both tension and compression, which leads to difficult
calculations of how the loading and the residual stress counteract or contribute to each other. Calculations
made with the stress range give a more conservative fatigue life result. [13]

10
M. Ramström ǀ Chapter 2 - Theory

Figure 6. S/N curve with key positions. [13]

The quantities in Figure 6 are, the number of load cycles to failure, 𝑁𝑓 , the fatigue limit, 𝜎𝐹𝐿 , the ultimate
strength, 𝜎𝑈𝑇𝑆 , the stress range, ∆𝜎 , and the curve slope, m.

The S/N curve is used to design the fatigue life of a certain component, either dependent on dimensioning
for a certain fatigue life, e.g. infinite life, or for a given stress range. [1]

The fatigue limit is a stress limit where the detail has a theoretical infinite life at constant stress range, thus
no fatigue failure occurs. When the fatigue limit is reached, this varies between different standards, but is
usually around 107 cycles for welded joints and 108 cycles for base material not affected by welding.
These values are correct according to IIW recommendations. [9]

Characteristic fatigue strength is used to take into account some scatter in the test samples. It is often used
when dimensioning with FAT-values. If a large number of tests are carried out, the characteristic fatigue
strength can be determined by the risk of failure coefficient, 𝜑𝑄 , see Table 1.

The risk of failure coefficient, 𝜑𝑄 , in Table 1 is only valid in the area where the S/N curve has the gradient
-1/3, m=3. See values for gradient -1/5, m=5 in the book; Konstruktionshandbok för svetsade produkter i
stål, 2014, page 291. [9]

For example, the stress range for the mean fatigue strength minus two standard deviations, stdv.,
calculated in log (N) will correspond to a probability of failure of 2.3%, see Table 1.

11
M. Ramström ǀ Chapter 2 - Theory

Table 1. Risk of failure coefficient with slope, m = 3. (Konstruktionshandbok, 2014, [9]) – Table 9-6.
Approximated risk of failure 50 % 15.9 % 2% 1% 1.35 ‰ 10⁻³
Number standard deviations 0 1 2 2.33 3 3.1
Risk of failure coefficient 1.32-1.40 1.15-1.19 1 0.95 0.84-0.87 0.83-0.86

A correlation of the characteristic curves gradient has been found, both for welds and base material. This
correlation is a simplification of the curve to a straight line with the gradient -1/m in a logarithmic
diagram. For welds loaded in normal stress the characteristic curve has often a gradient of -1/3, m=3, but it
may vary for different cases. For welds affected in shear stress and base material not affected of welding
the gradient is usually -1/5, m=5. This value is used in most standards and applies to constant stress range.
There are different perceptions of the fatigue limit and some argue that there is no absolute fatigue limit,
and consider instead a change in the assumed fatigue limit point to a new gradient of -1/22, m=22. This
slope is shown in Figure 6 marked as Slope for varying amplitude. [9]

The S/N curves slope can be described by Equation (1).

∆𝜎 𝑚 𝑁 = 𝐶 (1)

Where 𝑁 is number of cycles, ∆𝜎 is the stress range, 𝑚 and 𝐶 are material parameters.

For calculating number of load cycles to failure Equation (1) is rewritten by Equation (2).

𝐶
𝑁=
∆𝜎 𝑚 (2)

If the stress range versus number of load cycles to failure are plotted with logarithmic scale a linear
relationship is received. The S/N curves linear relationship is described in Equation (3).

log 𝑁 = −𝑚 log ∆𝜎 + log 𝐶 (3)

The stress amplitude is defined in Equation (4).


𝜎𝑚𝑎𝑥 − 𝜎𝑚𝑖𝑛
𝜎𝑎 = (4)
2

The stress range is defined in Equation (5).

∆𝜎 = 𝜎𝑚𝑎𝑥 − 𝜎𝑚𝑖𝑛 (5)

The mean stress is defined in Equation (6).

𝜎𝑚𝑎𝑥 + 𝜎𝑚𝑖𝑛
𝜎𝑚 = (6)
2

The stress ratio defined in Equation (7).


𝜎𝑚𝑖𝑛
𝑅= (7)
𝜎𝑚𝑎𝑥

12
M. Ramström ǀ Chapter 2 - Theory

The parameter with the highest impact on the fatigue life calculations is the stress amplitude and stress
range. The load variation can be pulsating-, alternating- or fully reversed loading. Pulsating loading occur
if the stress ratio is positive. Alternating loading leads to a negative stress ratio corresponding to both
tensile and compressive stresses. Fully reversed loading includes both tension and compression, and it has
a mean stress of zero. The loading variation does not affect the number of cycles to failure significantly
and it is often assumed to be a sinusoidal loading. [14]

There are some general effects regarding an S/N curve for a welded joint, these are stated below:

- The sharper notch, the steeper is the slope of the curve.


- The sharper notch, lead to less scatter in the curve, less scatter of the test results.
- The fewer cycles to failure, lead to less scatter of the test results.

The S/N curve is limited and mainly valid for uniaxial loads, primary normal and shear stress. The curve
is also only valid for loads with a certain R-ratio. [9]

There is usually a spread of test results in the S/N diagram, due to the welding process which cause
different residual stresses, initial cracks, discontinuities etc. When testing, the failure probability is 50 %.
Therefore, a lower quartile is normally selected when dimensioning for the characteristic fatigue strength,
often a quartile corresponding to 2.3% failure probability. The quartiles are calculated by the standard
deviation. [1]

2.2 Welds
Welding is a process for bonding parts together. This is performed by melting the base material and
adding a filler material to form a pool of molten material. When it cools down a strong bond between the
materials is formed. Welding can be performed by robot or manually.
Welds consists of three zones, illustrated in Figure 7. The three zones are; weld metal, heat affected zone
and base material. Each zone has different metallurgical structure and the heat affected zone can be
subdivided into several zones with different metallurgical structures. There are essential load transferring
and non-load-transferring welds. [15]

Figure 7. Weld zones. [15]

13
M. Ramström ǀ Chapter 2 - Theory

The throat thickness, referred as a, is a measure of the welds thickness. The dimensions of a weld are
usually defined with the throat thickness and the length of the weld. The throat thickness depends also on
the penetration into the base material. The nominal throat thickness is the largest triangle that can be
formed inside the weld’s profile.

In reality welds have a complex geometry, illustrated in Figure 8. Both the profile and the start and end
have complex shapes. In the FE-model the fillet weld is assumed to be an isosceles triangle.

Figure 8. To the left is a real fillet weld profile, [7] to the right is a simplified fillet weld profile, with marked throat
thickness, referred as a, and 45 degree angles.

2.2.1 Welding classes


General welding classes available are B, C, and D. The weld class indicates how much defects the weld
contains and also misalignments of the base material components, both internal and external. Welding
Class B is the class with minimum amount of misalignments and requires, in principle, post-treatment.
Class B is used for high fatigue stresses, high risk zones and risk of brittle fracture. Class C does not
require post-treatment after the welding. It is less time consuming than class B. Welding class C is
standard at BT Products. Class D is used for welds not subjected to high loads, for example, mounts for
cables and hoses. The cost and time for processing a weld increase with quality.

2.3 How the material is affected by welding


2.3.1 Residual stresses
Residual stresses in welds, also called initial-, rest- or internal stresses, always occurs after the welding
process. The reason for the residual stresses is that of the melted material wants to shrink during cooling,
while the adjacent materials prevent shrinking. When producing a welded joint high temperature is
applied, in general so high that the adjacent materials begin to melt. When a metal is heated it will expand
due to its metallurgy. The expansion is dependent of the materials’ thermal expansion coefficient and
temperature difference. Heat expansion is defined in Equation (8). [9]

(8)

14
M. Ramström ǀ Chapter 2 - Theory

∆𝐿 = 𝐿 ∗ 𝛼 ∗ ∆𝑇

Where ∆𝐿 is the length change from initial reference length, 𝐿 is the initial length, 𝛼 is the thermal
expansion coefficient and ∆𝑇 is the temperature change. For steel the thermal expansion coefficient is
about 13x10-6 / (°C) at 20 degrees. [16]

Since the heating is so high that plasticity occurs, tensile stresses will remain in the material. The result
leads to longitudinal tensile stresses in and around the weld. Equilibrium in the section must be fulfilled
and compression residual stresses arise some distance from the weld. The magnitude of the residual
stresses also depends on the material’s mechanical properties. A high strength material receives higher
residual stresses then a low strength material. This relationship is not linear. For a low strength material
with low yield strength the residual stresses usually exceed the yield stress, but not for a high strength
material. The area affected with residual stresses is dependent of the applied heat expressed in [MJ/m]
when welding. The applied energy depends on the welding method. The residual stresses, especially the
longitudinal tensile stress in and around the welded joints have a negative effect on the fatigue resistance.
It cause fatigue cracks for both tensile- and compression stress. Usually fatigue failure only occurs under
tensile stresses. Residual stresses increase the possibility for crack initiation and propagation both in and
near the weld. It is assumed in many standards that the longitudinal tensile stress in and around all welds
is equal to the material’s yield stress. If it is not assumed to be equal then a correction factor is used.

2.3.2 Deformation due to weld shrinkage


During the welding process deformation occurs due to the cooling process. Generally, the interaction is;
smaller deformation leads to higher stresses and vice versa. Different deformations that may occur are;
cross shrinkage, deformation angle, length contraction, bending and buckling, depending on the base
material’s geometry, weld design and weld sequence.

An increase in throat thickness or in joint volume results in increased deformations, due to increased
amount of heat. The size of the deformation is important when tight tolerances are needed. Effective
welding techniques such as symmetrical welding, welding sequence of components and placement near
the center of gravity can reduce deformations.

2.3.3 Crack initiation


After welding, the crack initiation phase is already fully or partially passed due to the residual stresses.
When calculating the weld’s fatigue life the crack initiation phase is usually ignored. This is a
conservative assumption for the fatigue life of the weld toe. An initial crack will occur between the
bonded plates. The cracks length and placement are dependent on the welded design etc. Fatigue fracture
possibilities are shown in Figure 9. Statistics indicate that the fractures occur more often in the weld toe
than in the root. [7]

15
M. Ramström ǀ Chapter 2 - Theory

Figure 9. Fatigue fracture modes in a fillet weld. [7]

2.4 Welds affect in structures


Welded joints strength at static load is generally regarded and assumed as the base materials’ properties.
One important exception is if the weld is exposed to an instability phenomenon, e.g. buckling. The
problem is the residual stresses in the welded area. It causes decreased stress limits for an instability
phenomenon. Dynamic loads have a significant effect on the welds fatigue life and it needs to be
considered while dimensioning. [1]

After the welding process the base material’s properties have been changed and cracks have been initiated
in the welded area. Different types of defects and discontinuities are also formed in and around the weld.
Welds have often naturally sharp notch radii, which cause high notch stresses in the transition zone
between the weld and the base materials. Therefore, a welded design will never achieve the same fatigue
life as the unprocessed base material. [1]

2.5 Welds effect on fatigue life


It is in reality hard and almost impossible to precisely determine the fatigue life of a welded joint. The
main reason for this is that there are never two welds that are equal. The welded joints never obtain
exactly the same geometry and properties. The weld receives custom characteristics due to various factors
that affect each other in a complex way and affecting the fatigue life of the weld.

The five factors with the greatest impact on the fatigue life are:

- Stress range
- Mean stress and residual stresses
- Material properties (Crack propagation/Crack initiation)
- Geometric stress concentrations
- Size and location of weld discontinuities

The factors effect depending on welded geometry, weld dimensions, the residual stresses and
discontinuities size and type. [15]

The weld has geometric details that cause stress concentrations, especially at the weld ends, toes, root and
internal defects. These stress concentrations leads to crack propagation and eventually fracture [1]. Tensile
residual stresses can occur even in compressive loads due to micro cracking. Therefore, crack propagation
and fracture also occur in compression due to positive residual stresses [12]. The crack propagation rate is
basically independent of the material’s strength. Thus, the weld’s fatigue life is almost independent of the
material’s strength [1].

16
M. Ramström ǀ Chapter 2 - Theory

2.6 Fatigue design of welded joints and components


It is complicated to calculate the fatigue life of a weld. Fatigue life is affected by a number of factors,
where many of them are difficult to derive physically. [1]

There are mainly four methods for calculating the fatigue life; Nominal stress method, Geometrical
method also called Hot-spot method, Effective notch method and Fracture mechanics. Fracture mechanics
also includes different evaluation methods. [6]

The methods are appropriate for different types of problem, depending on the methods limitations and
assumptions. The methods have different pros and cons, and the results of the methods do not always
correspond to each other.

The Nominal stress method is used for basic geometry, where the nominal stress is defined. The Hot-spot
method can be used for basic and a bit more advanced geometries, but it loses accuracy for more complex
geometries. The method evaluates results from tests or field measurements or from a FE-analysis. The
Effective notch method and Fracture mechanics are appropriate for more complex problems and are
primary used with FE-analysis. The methods require more effort and give a result with higher accuracy.
Generally, the methods requiring less effort are applicable for basic geometries and vice versa, illustrated
in Figure 6. Today Fracture mechanics and Nominal stress method are often based on FE-analysis,
although the methods originally were based on analytical calculations. [1]

Based on a research project investigating the different evaluation methods made by Volvo VCE, 2006, the
conclusion was that the Effective notch method should be used for the weld toe analysis and Fracture
mechanics for the root analysis. [7]

Calculation methods for fatige life of welds


4,5

3,5 Fracture mechanics

3
Effective notch
Accuracy

2,5

2 Hot spot

1,5
Effort Nominal stress
1

0,5

0
Problem complexity

Figure 10. Comparison between the different methods used for fatigue life design.

Calculations for welds are normally made with the assumption of elastic conditions, and that the throat
thickness includes the weld’s penetration into the base material.

17
M. Ramström ǀ Chapter 2 - Theory

2.6.1 Nominal stress method


The first developed method for fatigue life calculations of welds was the Nominal stress method, the
method was developed through testing of different weld designs. The method is based on and interacts
with fatigue classes. More about fatigue classes can be found in chapter 2.1.1 Fatigue class. The Nominal
stress method only considers the nominal stress in the calculation. Due to this, the method is suitable for
calculations without FE-analysis. The notch stresses and local stress concentrations are considered in the
FAT-value. The FAT-value depends mainly on the welded geometry, weld’s design and the welding class,
where a higher FAT-value indicates higher fatigue strength. Tables with FAT-values can be found in
literature about the field, i.e. Konstruktionshandbok för svetsade produkter i stål, Claes Olsson [9]. The
method is limited to only evaluate the weld toe, so it has no capability of evaluating the weld root.
Calculation of the fatigue life is defined by Equation (9) . (SSAB, 2011) – Eq. 5.22.

∆𝜎𝑟𝑑 𝑚
𝑁 = 𝑁𝑡 ( ) (9)
∆𝜎

Where 𝑁 is calculated fatigue life, 𝑁𝑡 is the designed life, ∆𝜎𝑟𝑑 is the allowed stress range, ∆𝜎 is stress
range, 𝑚 is the gradient of the S/N curve.

The maximal stress range is calculated with Equation (10). Note that if the minimum stress is negative, the
negative sign is taken into account. (SSAB, 2011) – Eq. 5.19.

∆𝜎𝑚𝑎𝑥 = 𝜎𝑚𝑎𝑥 − 𝜎𝑚𝑖𝑛 (10)

Equation (9) can be rewritten if the number of cycles is given and the stress range is sought.
Then, the allowed stress range can be determined by Equation (11). (SSAB, 2011) – Eq. 5.15.

𝐹𝐴𝑇 𝜑𝑡 𝜑𝑚 𝜑𝑒
∆𝜎𝑟𝑑 = (11)
𝛾𝑚 𝑚√𝑆𝑚

Where 𝜑𝑡 is the thickness factor, 𝜑𝑚 is the material factor, 𝜑𝑒 is the stress variation factor (residual
stresses), 𝛾𝑚 is a failure impact factor and 𝑆𝑚 is the cumulative stress parameter.

Equation (11) contains of partial factors and the FAT-value. Additionally, other correction factors can be
used, for example for elevated temperature or for corrosion in air or seawater. See IIW recommendations
for more information.

The base material’s thickness has an impact on the fatigue strength. Reduced plate thickness increases
fatigue strength (for the same stress level over the cross-section). The reason for this is that a smaller
volume will be exposed for higher stresses, and the probability of failure then decrease which results in a
higher fatigue strength. The thickness factor is applied when welds are loaded perpendicular to the weld
direction, and when a fracture is assumed to occur in the weld toe and not in the weld root. The thickness
factor is defined in Equation (12). (SSAB, 2011) – Eq. 5.11.

𝑡0 𝑓
𝜑𝑡 = ( ) (12)
𝑡

Where 𝑡 is the material thickness, 𝑡0 is the reference thickness; 15 mm. The exponent 𝑓 is dependent on
the joint type and varies between 0 and 0.15. The 𝑛-value is selected according to Table 2. [1]

18
M. Ramström ǀ Chapter 2 - Theory

Table 2. Recommended values of n according to SSAB's research. (SSAB, 2011) – Table 5.9.
Joint type Class f
Fillet weld, transverse T-weld, sheets with transverse junction, longitudinal Untreated 0.15
stiffeners weld
Fillet weld, transverse T-weld, sheets with transverse junction, longitudinal Treated weld 0.1
stiffeners
Transverse butt weld Untreated 0.1
weld
Treated butt, rye, transverse welds or weld junctions All 0
Non welded material All 0

The material factor, 𝜑𝑚 , is dependent on the surface roughness and the yield stress. For welds the fatigue
strength is almost independent of the static yield stress. Hence the material factor is assumed to equal to
unity.

The stress variation factor, 𝜑𝑒 , is dependent of the stress variation and the stress ratio. The stress variation
factor is used because of the residual stresses occurring in and around the weld. The factor is defined in
Table 3.

Table 3. Stress variation factor. (SSAB, 2011) – Eq. 5.12 -5.14.


Stress variation factor Stress ratio Max stress
Weld 𝜑𝑒 = 1 0 ≤ R ≤ 0.5 𝜎𝑚𝑎𝑥 > 0
Base material 𝜑𝑒 = 1 − 0.3𝑅 0 ≤ R ≤ 0.5 𝜎𝑚𝑎𝑥 > 0

Weld 𝜑𝑒 = 1 − 0.2𝑅 -1 ≤ R ≤ 0
Base material 𝜑𝑒 = 1 − 0.25𝑅 -1 ≤ R ≤ 0

𝜑𝑒 = 1.3 𝜎𝑚𝑎𝑥 < 0

When the applied stress variation is unknown the factor is set to unity, which gives a conservative result.
A safety factor, 𝛾𝑚 , is included in Equation (11). The safety factor is often connected to fracture
consequences. Safety factors are either the partial coefficient, 𝛾𝑚 , or the risk of failure coefficient, 𝜑𝑄 .
IIW (24) gives the following recommendations of the factor in Table 4. [9]

Table 4. IIW's proposal of values for the partial coefficient in different situations. [3]
Consequence Fail safe, damage resistance design Safe life, infinite life
Failure in secondary design details 1.00 1.15
Loss of vital structures 1.15 1.30
Loss of human life 1.30 1.45

In certain cases the mean fatigue strength is more interesting than the characteristic fatigue strength. For
example, if the calculations should be compared with testing or when the purpose is to design with a
specific risk of failure. Then it is useful to take advantage of the risk of failure coefficient rather than
partial coefficient. [9] More information about risk of failure coefficient is in chapter 2.1.2 S/N Curve. The
Cumulative stress parameter, 𝑆𝑚 , defines in Equation (13). (SSAB, 2011) – Eq. 5.10.

19
M. Ramström ǀ Chapter 2 - Theory

𝑁𝑡 𝑘𝑚
𝑆𝑚 = (13)
2 ∗ 106
Where 𝑘𝑚 is the collective factor, defined in Equation (14). (SSAB, 2011) – Eq. 5.8.

𝑚
∆𝜎𝑖 𝑛𝑖
𝑘𝑚 = ∑ ( ) ∗ (14)
∆𝜎𝑟𝑒𝑓 𝑛𝑡
𝑖

Where ∆𝜎𝑖 is the stress range on level i, 𝑛𝑖 is number of cycles at level i, 𝑛𝑡 is total number of cycles, and
∆𝜎𝑟𝑒𝑓 is the reference value for the stress range. With constant stress range, 𝑘𝑚 = 1 and with varying
stress range, 𝑘𝑚 < 1.

2.6.2 Geometrical stress / Hot-spot method


The Geometrical stress also known as Hot-spot approach was developed for estimate fatigue life through
tests and field measurements. The method calculates a geometric stress, also known as Hot-spot stress, hs,
by extrapolating strain/stress measurements at the weld toe. The measurements are usually performed with
strain gauges and are picked in specific reference points depending the method; linear extrapolation or
quadratic extrapolation. With the development of computers and availability of FE-software, FE-analysis
became more frequently used for Hot-spot calculations. The method has high accuracy for simple
geometries, but for complex geometries the accuracy decreases.

The method is only suitable for evaluation of fracture at the weld toe for fillet and butt welds. Therefore,
one limitation is that there is no capability of evaluating the root. Ideally, the max principle stress should
be perpendicular to the weld, but it gives valid results for max principle stress direction oriented ± 60°
perpendicular to the weld toe.

The Hot-spot method requires fewer S/N curves than the nominal stress method. Since the Hot-spot stress
itself takes account of the welded geometry. For example, the same S/N curve can be used for all fillet
welds without post-treatment or large misalignments, regardless the welds geometry. The method has no
simple way to consider geometrical misalignments and eccentricities in the weld design. One way is to
model the misalignments and eccentricities in the FE-model. [6]

The fatigue life is calculated with the Hot-spot stress together with a FAT-value suitable for the Hot-spot
approach, see Equation (15). For example, recommends IIW FAT-values in the report; “Fatigue design of
welded joints and components”. [3]

𝐹𝐴𝑇 𝜑𝑄 𝑚
𝑁 = 𝑁𝑡 ( ) (15)
𝜎ℎ𝑠

Where 𝑁 is calculated fatigue life, 𝑁𝑡 is the required life, FAT is the FAT-value, 𝜑𝑄 is the risk of failure
coefficient, 𝜎ℎ𝑠 is the Hot-spot stress and 𝑚 is the gradient of the S/N curve.

The Hot-spot stress is calculated through Hooke’s law, since the stress is approximated to a linear elastic
uniaxial stress state. The uniaxial orientation is perpendicular to the weld. Equation (16) is Hooke’s law
with the Hot-spot stress and strain.

𝜎ℎ𝑠 = 𝐸 𝜀ℎ𝑠 (16)

20
M. Ramström ǀ Chapter 2 - Theory

The stress/strain is calculated in specific points dependent on linear or quadratic extrapolation. The
specific points are recommendations from IIW. The linear extrapolation is calculated through two points,
at distance 0.4t and 1.0t in front of the weld toe, see Equation (17) and Figure 11. The factor t is the
thickness of the base material. The quadratic extrapolation is calculated through three points, at distance
0.4t, 0.9t and 1.4t, see Equation (18) and Figure 11.

𝜎ℎ𝑠 = 1.67 𝜎0.4𝑡 − 0.67 𝜎1.0𝑡 (17)

𝜎ℎ𝑠 = 2.52 𝜎0.4𝑡 − 2.24 𝜎0.9𝑡 + 0.72 𝜎1.4𝑡 (18)

Figure 11. To the left, points for linear extrapolation and to the right, points for quadratic extrapolation.

Linear extrapolation is suitable in cases where there is no significant nonlinear behavior (e.g. low stress
gradients) or the mesh is coarse (i.e. element size > 0.2t), otherwise quadratic extrapolation is
recommended. [6]

2.6.3 Effective notch method


Calculations with the Effective notch method are based on the stresses in the local weld radius, referred as
notch stresses. Notch stress is considered the max principle stress in a certain notch, under the assumption
of linear elastic material. This notch stress has been used successfully in fatigue calculations of non-
welded structures. [6]

The notch stress is obtained by FE-analysis. In the FE-model the entire design's form is considered, with
transition radii, fillets etc. A fictitious radius is used at the weld toes and/or root, see Figure 12. In more
comprehensive FE-models sub models are effective to use to get a refined element size. The element size
should be analyzed with a mesh convergence study or follow standards for the method. This results in
accurate solutions.

When the notch stress is determined a dense mesh is required, because the fictitious notch radius is small
and the stress gradient is usually very large near the notch. An example of fictitious notch radii is
illustrated in Figure 12. For base material thickness greater or equal to 5 mm, IIW [4] recommend using a
notch radius of 1 mm and a FAT-value of 225 MPa. The previous recommendations are connected with an
element size. The recommended element size around the notch is; notch radius divided by four. [1] For
example, notch radius 1 mm should have element size 0.25 mm.

21
M. Ramström ǀ Chapter 2 - Theory

Figure 12. Fictitious notch radii.

The effective notch approach is an effective method when a number of weld geometries or penetration
depth should be compared with each other. Besides evaluating the weld toes, the weld root can also be
evaluated, this makes it possible to dimension the weld for both toe and root fractures.

The method has the ability to evaluate complex designs where the nominal stress is not possible to use.
The method can be applied to major misalignments, such as angular errors and eccentricity. These errors
can be modelled in the current FE-model or applied with a scale factor. Figure 13 illustrates a FE-model
with notch radius 1 mm in toe and root.

Figure 13. FE-model with notch radius.

Disadvantages and limitations


- For valid results the principle stress direction needs to be oriented ± 60° to the weld toe and root.
- The method may require an advanced FE-model with a large number of elements and it may need
sub models. This lead to increased work effort and calculation time.
- The problem with Effective notch method is that singular points should be avoided in the FE-
model.

22
M. Ramström ǀ Chapter 2 - Theory

Calculation system
The fatigue life is calculated by Equation (19).

∆𝜎𝑟𝑑 𝑚
𝑁 = 𝑁𝑡 ( ) (19)
𝜎𝑟

Where ∆𝜎𝑟𝑑 is the design normal stress range, also called allowed stress range, 𝑁𝑡 is the required life, ∆𝜎
is the stress range and 𝑚 is the gradient of the S/N curve.

The stress range is calculated with a FE-model. It is the largest principle stress value during the load cycle.

𝜑𝑄 𝐹𝐴𝑇
∆𝜎𝑟𝑑 =
𝑚 𝑁𝑡 (20)

2 ∗ 106

Equation (19) combined with Equation (20) can be rewritten to Equation (21).

𝐹𝐴𝑇 𝜑𝑄 𝑚
𝑁 = 𝑁𝑡 ( ) (21)
𝜎𝑟

Risk of failure is 2.3% and the design life is 2 million cycles correspond the safety factor, 𝜑𝑄 = 1.
𝐹𝐴𝑇 3
𝑁 = 2 ∗ 106 ( ) (22)
𝜎𝑟

Median fatigue life: FAT-value = 225 MPa, 𝜑𝑄 = 1.3, 𝑁𝑡 = 2*106 (IIW). Use max principal stress for 𝜎𝑟 .

225 ∗ 1.3 3
6
𝑁 = 2 ∗ 10 ( ) (23)
𝜎𝑟

2.6.4 Generally for the methods


The design is approved if Equation (24) and Equation (25) are satisfied.

𝑁 ≥ 𝑁𝑡 (24)

Where 𝑁 is the fatigue life and 𝑁𝑡 is the designed fatigue life.

∆𝜎𝑟 ∗ 𝛾𝑓 ≤ ∆𝜎𝑟𝑑 (25)

1
Where 𝛾𝑓 is a safety factor, 𝛾𝑓 = 𝜑𝑄
,

23
M. Ramström ǀ Chapter 2 - Theory

2.6.5 Fracture mechanics


Fracture mechanical is mainly used in areas where the crack propagation dominates, for example fatigue
calculations of welds.

2.6.5.1 Stress intensity factor


One way to calculate the crack propagation and failure state is with the use of the stress intensity factor.
Equation (26) defines the stress intensity factor for a Mode I condition. 𝐾𝐼 is the predominant failure
mode.

𝐾𝐼 = 𝜎𝑛𝑜𝑚 𝑓 √𝜋𝑎 (26)

Where 𝑎 is the crack length, 𝑓 depends of geometry and 𝜎𝑛𝑜𝑚 is the nominal stress.

The 𝑓-function is computed through derivation of numerous analytical cases or numerically through FE-
analyses. Derived cases can be found in literature regarding fatigue and fracture mechanics, e.g. in the
book: Failure, Fracture, Fatigue by Tore Dahlberg [17].

Equation (27) defines the range of the stress intensity factor and Equation (28) defines the stress intensity
ratio. The stress intensity range can be used in Paris’ law when estimating the number of cycles to failure.

∆𝐾𝐼 = 𝐾𝐼 𝑚𝑎𝑥 − 𝐾𝐼 𝑚𝑖𝑛 (27)

𝐾𝐼 𝑚𝑖𝑛
𝑅= (28)
𝐾𝐼 𝑚𝑎𝑥

Where 𝐾𝐼 𝑚𝑎𝑥 is the maximum intensity factor and 𝐾𝐼 𝑚𝑖𝑛 is the minimum intensity factor.

There are also other methods to calculate the stress intensity factor.

Fracture toughness is a mechanical property that describes the material’s point of fracture and is denoted
as KIC.

2.6.5.2 Paris’ law, also known as Paris-Erdogan law


Paris’ law is a method developed for calculations of crack propagation. It is the most commonly used
method. Tests have shown how crack propagations depend on the stress intensity factor [12]. Paris’ law is
defined in Equation (29).

𝑑𝑎
= 𝐷(∆𝐾𝐼 ) 𝑔 (29)
𝑑𝑁

Where 𝑁 is the number of load cycles, 𝑎 is the crack length, ∆𝐾𝐼 is the stress intensity range and 𝐷 and 𝑔
are material parameters describing the crack propagation behavior. The left hand side in Equation (29) is
known as the crack propagation rate.

24
M. Ramström ǀ Chapter 2 - Theory

Figure 14. Illustrates the crack propagation rate as a function of stress intensity range. [12]

Paris’ law describes the crack propagation rate as a function of the stress intensity factor. With logarithmic
scale it is a linear relationship between the variables. The linear part is illustrated in zone II in Figure 14.

There is a threshold value, ∆Kth, located as zone I in Figure 14. If the stress intensity range is below the
threshold value the initial crack will not start to propagate. For example, if the crack is short, see Equation
(26).

Fracture toughness, KIC, is set as an above limit in Paris’ laws diagram, located as zone III in Figure 14.
Fracture criteria at plane strain condition is KIC, fracture occur when KI = KIC. [13]

25
M. Ramström ǀ Chapter 2 - Theory

2.6.6 Evaluation of test data


For evaluating the fatigue tests with the FE-analysis, the test results characteristic fatigue strength is
calculated. The aim is to compare the tests with the FE-analysis in an S/N curve. The results validate the
use of the methods for weld ends. The fallowing calculation method for determined the S/N curve is taken
from IIW recommendations [3].

From the fatigue tests, the number of load cycles to fracture is received for a specific stress range. With
logarithmic scale there is a linear relationship defined in Equation (30). This equation forms the S/N
curve.

log 𝑁 = log 𝐶 − 𝑚 log ∆𝜎 (30)

Where 𝑁 is the number of cycles, 𝐶 is the constant, 𝑚 is the slope of the S/N curve and ∆𝜎 is the stress
range.

The slope, m, and the constant C can be determined from Equation (30). The characteristic value is
defined in Equation (31).
𝐶 = 𝑥𝑚 − 𝑘 ∗ 𝑠𝑡𝑑𝑣 (31)

Where 𝐶 is the characteristic fatigue strength, 𝑥𝑚 is the mean stress value, 𝑘 is dependent of specimens
and taken from Table 5 and 𝑠𝑡𝑑𝑣 is the standard deviation.

To investigate the characteristic fatigue strength and the scatter in the test results, the standard deviation is
used. The standard deviation is defined in Equation (32).

∑(𝑥𝑚 − 𝑥𝑖 )2
𝑠𝑡𝑑𝑣 = √ (32)
𝑛−1

Where 𝑥𝑚 is the mean value, 𝑥𝑖 is the test samples and 𝑛 is the number of test results.

The mean value is defined in Equation (33).

∑ 𝑥𝑖
𝑥𝑚 = (33)
𝑛

Table 5. k-value dependent of number of test results. [3]


Number of test results, 𝑛 10 15 20 25 30 40 50 100
𝑘 2.7 2.4 2.3 2.2 2.15 2.05 2.0 1.9

To produce the characteristic curve in this project, two standard deviations have been applied (i.e. 2.3 %
failure probability).

26
M. Ramström ǀ Chapter 3 – Fatigue testing

CHAPTER 3
FATIGUE TESTING

This chapter describes all facts about the testing including the strain gauge measurements.

3.1 Fatigue testing


The specimens have plate thicknesses 10 mm and throat thickness 5 mm. The choice of plate and throat
thickness is based on common material that is used in the forklifts at BT Products. Dimensions are
illustrated in APPENDIX A - Drawings for the specimens.

3.1.1 Weld properties


The specimens were robot welded, with welding class C and no post-treatment. The weld power was set to
295 ampere and the feeding was constant. Robotic welding was chosen since it is the most common
method used in the production at BT Products. Robot welding gives less scatter and an increased
credibility to the results compared with manual welding. The welding process is shown in Figure 15.

Figure 15. Robot welding. To the left a discontinuous string.

The continuous weld design where remade after the machine operators suggestions. The problem in the
first design was the demand of no stop and start around the corners. According to the operator it could not
be welded continuously around the corner, because of the burn and melting in the base material. The
suggested design was two continuous strings, the first string contained the long side and both the short
sides, and the second string were the other long side, see drawing in Figure 16. This influences the
continuous welds short ends, a visual difference is noted and shown in Figure 17. It also affects the
penetration at the weld end. In the weld start, the penetration is less than in the weld ending. The end sides
are called side A and side B. Side A is the start side. The discontinuous weld design is made with
continuous strings along each side of the vertical plate. The strings ends are similar with one start and stop
at each end side, see Figure 19.

27
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 16. Weld drawing for the remade continuous weld end.

Figure 17. End sides for the continuous weld design. To the left side A, and to the right side B.

Figure 18. Continuous weld, to the left the first string, to the right the second string.

28
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 19. End sides for the discontinuous weld end design.

3.1.2 Measuring before testing


Markup and measuring of the specimen’s misalignments and deformation caused by the welding was
performed before testing, see Figure 16. The welding process showed a significant effect on the base plate,
especially for the specimens with continuous weld. For the continuous weld, the outer plate edge’s was
raised up about 1.2-1.5 mm on each side, and for the discontinuous welded specimens, the outer edge
raised about 0.15-0.30 mm on each side. In Figure 21 the heat effect on the underside is shown, where the
continuous welding shows a larger heat affected area due to longer welding strings. The measuring data
was registered in a document to interpret the test results.

Figure 20. Specimens during measurements and markup.

Figure 21. The heat affects. The above is the discontinuous end, and the below is the continuous end.

29
M. Ramström ǀ Chapter 3 – Fatigue testing

3.1.3 Test setup


The testing was performed in a tensile test machine, shown in Figure 22. The specimens were mounted in
the machine’s fixture with hydraulic grips, the grip fixture dimensions were width 50 mm and depth 65
mm. First, one side was mounted in the hydraulic grip, and then the other side was steered into the other
grip. During mounting the specimen were straightened out from the curvature produced by the welding
process. The criterion for when the tests are finished is when the deformation increased 5 % with respect
to the undeformed state, according to BT products test standard.

Figure 22. Fatigue test machine.

3.2 Strain gauge test


In order to validate the FE-model strain gauges where placed on one specimen of each weld end design,
19 gauges on the discontinuous respectively 18 on the continuous. The strain gauge placements depend on
the welds geometry. Gauges are placed around the area of interest, which is around the weld toes and the
center of the plate, see Figure 23. Figure 23 shows only half of the specimens. Gauges are placed
symmetrical on both sides. Also three gauges are placed on the underside of the base plate. Figure 24
show the gauge placement for the discontinuous weld end. For the continuous end, 5 respectively 6 were
mounted on each side around the weld. The difference in gauges are caused by the weld ends local
geometry, therefore, one extra gauge were mounted on one side, see Figure 25. The used strain gauges
model is Quarter Bridge, which evaluate a single axis in tension or compression. The gauge measures the
changes in resistance, which unbalances the bridge and produces an output voltage. More information can
be found in APPENDIX B - Strain Gauges placement.

30
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 23. Red lines indicate strain gauge placements. Left, continuous weld end. Right, discontinuous weld end.

Each specimen were loaded at the four load levels, see Table 7, for a short period. The period was long
enough to stabilize the cycles, approximately 150-200 cycles, see Figure 27. The strain gauges were set to
zero before the specimen was mounted in the machine. During the mounting, the specimen was
straightened out which cause pre-stresses. The pre-stresses were measured, but not part of the fatigue test.

31
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 24. Strain gauge placements for the discontinuous weld end.

Figure 25. Strain gauge placements for the continuous weld end. To the left side A and to the right side B.

3.2.1 Strain gauge test results


The results from the strain gauge measurements contained pre-stresses and fatigue stresses for the
different load levels. The results for the pre-stresses and the load level, 200 kN, are summarized in Table
6. The occurred pre-stresses where different for the weld designs, the continuous weld end received higher
stresses mainly affected by the significant curvature and misalignments in the baseplate. The pre-stresses
indicate both compressive and tension stresses, which leads to a stress ratio below zero at some locations.
The software used for the test data is FAMOS. The strain measurements were transformed to stresses
through Hooke’s law in one dimension.

From the test results, the magnitude of the stresses for load case, 200 kN, was about 330-338 MPa for the
continuous end, and about 266-314 MPa for the discontinuous end, measured in z-direction in the center
gauges 2.5 mm from the weld toe. The pre-stress was 58 and 15 MPa for the continuous end, and 15 and
-18 MPa for the discontinuous end, measured on the upper surface of the baseplate in z-direction in the
center gauges 100 mm away from the weld toe. The stress ranges for load case 200 kN and the pre-
stresses are shown in Table 6.

32
M. Ramström ǀ Chapter 3 – Fatigue testing

Table 6. Stresses from the strain gauge tests.


Continuous weld end Discontinuous weld end
Strain Pre-stress Stress range Strain Pre-stress Stress range
gauge [MPa] [MPa] gauge [MPa] [MPa]
200 kN 200 kN
OS_A_1 -6 53 OS_AC_1 15 200
OS_A_2 50 140 OS_AH_1 4 49
OS_A_3 112 330 OS_AH_2 13 96
OS_A_4 38 141 OS_AH_3 6 266
OS_A_5 -8 51 OS_AH_4 3 220
OS_AC_1 54 230 OS_AV_1 -6 48
OS_AC_2 58 200 OS_AV_2 -15 101
OS_B_1 1 50 OS_AV_3 -2 308
OS_B_2 42 203 OS_AV_4 3 217
OS_B_3 55 338 OS_BC_1 -18 223
OS_B_4 18 138 OS_BH_1 -3 49
OS_B_5 6 42 OS_BH_2 -22 139
OS_B_6 -5 59 OS_BH_3 -20 314
OS_BC_1 28 237 OS_BH_4 -12 230
OS_BC_2 15 226 OS_BV_1 6 56
US_AC_1 -48 222 OS_BV_2 8 120
US_BC_1 -5 197 OS_BV_3 -9 282
US_CC_1 2 163 OS_BV_4 -11 220
US_AC_1 -2 207
US_BC_1 30 190
US_CC_1 4 163

The measured pre-stresses for some gauges are shown in Figure 26. The start of the first load level,
120 kN, including the pre-stresses and the stabilization of the gauges are shown in Figure 27. The four
load levels and the total measurements for the discontinuous weld end are shown in Figure 28.

33
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 26. The recorded pre-stresses for the discontinuous weld end. Four gauges in the figure; OS_AH_3, OS_AV_3,
OS_BH_3 and OS_BV_3.

Figure 27. The start of the load level. Four gauges in the figure; OS_AH_3, OS_AV_3, OS_BH_3 and OS_BV_3.

34
M. Ramström ǀ Chapter 3 – Fatigue testing

Figure 28. Overview for the discontinuous weld end, all four load levels. Four gauges in the figure; OS_AH_3, OS_AV_3,
OS_BH_3 and OS_BV_3.

Test program
The testing was performed externally at Volvo trucks in Göteborg. The external testing included 16
specimens, eight with continuous and eight with discontinuous weld end, shown in Table 7. The frequency
doesn’t affect the fatigue result but reduce testing time.

Table 7. Test program.


Specimen Specimen Force max Force min Stress range Frequency
R
Continuous Discontinuous [kN] [kN] [MPa] [Hz]
2 2 240 24 0.1 216 8
2 2 200 20 0.1 180 20
2 2 160 16 0.1 144 20
2 2 120 12 0.1 108 20

The stress ratio, R, is set to 0.1, since the test object should never be exposed to compression and be more
stabilized during the test.

3.3 Results from testing


The testing includes the two different weld ends. Each weld end design has been tested on four different
load levels. Every load level is tested for two specimens. The test results are marked with crosses in
Figure 29 for discontinuous and Figure 30 for continuous. The three lines in the figure below represent,
mean curve, results mean curve and the characteristic curve. The results mean curve is determined with
the least square method and receives a specific slope shown in Table 8 and Table 9 and 50 % risk of
failure. The mean curve and the characteristic curve have slope 3 and 50 % risk of failure respectively
2.3 % risk of failure. The test results are summarized in Table 10. The results indicate higher fatigue
strength for the discontinuous weld end.

35
M. Ramström ǀ Chapter 3 – Fatigue testing

S/N curve - Discontinuous weld end


1 000

Mean curve
Stress range σr [MPa]

Characteristic
100 curve

Results mean
curve

Results
10
10 000 100 000 1 000 000
Cycles
Figure 29. Discontinuous weld end.

Table 8. Discontinuous test data.


Slope Standard deviation
Slope mean curve and characteristic curve 3.00 0.170
Slope results mean curve 3.13 0.034

S/N curve - Continuous weld end


1 000

Mean curve
Stress range σr [MPa]

Characteristic
curve
100
Results mean
curve
Results

10
10 000 100 000 1 000 000
Cycles
Figure 30. Continuous weld end.

Table 9. Continuous test data.


Slope Standard deviation
Slope mean curve and characteristic curve 3.00 0.170
Slope results mean curve 3.42 0.052

36
M. Ramström ǀ Chapter 3 – Fatigue testing

Table 10. Test results.


Mean cycles Mean cycles
F amp. [kN] Fmax [kN] Fmin [kN]
Continuous Discontinuous
Load case 1 54 12 120 776,000 1,452,000
Load case 2 72 16 160 289,000 598,000
Load case 3 90 20 200 133,000 326,000
Load case 4 108 24 240 73,000 159,000

The occurred crack after the fatigue test for the different weld designs are shown in Figure 31. The results
for the continuous weld end occurs always at the same side, side A. This is an interesting notification
which is assumed to be caused by the less penetration compared with the other side. For the discontinuous
weld no similar pattern can be seen, since the sides are similar. Unfortunately no note where made of
which side of the specimens where pressed in the test machine which may have had an increased effect of
the pre-stresses.

Figure 31. Example of cracks. To the left a continuous, and to the right a discontinuous.

37
38
M. Ramström ǀ Chapter 4 - Finite Element Analysis

CHAPTER 4
FINITE ELEMENT ANALYSIS

This chapter includes a mesh convergence study, a boundary condition study, a nonlinear geometric
study, a Pre-stress study and FE-models for the different methods. The results from the strain gauge test
are used for calibration and validation.

4.1 FE-model
The aim with the FE-analysis is to simulate reality to provide accurate estimations of the failure
probability. FEM calculate a numerical solution to the problem defined in the pre-processor and are
mainly dependent on geometry, boundary conditions, mesh, etc. The more accurate the boundary
conditions are, the more accurate and comparable the FE-analysis will be.

4.2 Model Description


The model investigated is a T-joint fillet weld with two different weld ends, continuous and discontinuous,
illustrated in Figure 32. The dimensions are equal in the specimens and can be seen in APPENDIX A -
Drawings for the specimens.

b.
a.

Figure 32. a. T-joint fillet weld with continuous weld end b. T-joint fillet weld with discontinuous weld end.

39
M. Ramström ǀ Chapter 4 - Finite Element Analysis

4.2.1 Material
The material used for the test objects are SS2134 with properties according to Table 11 from BT Product’s
material library. The material in the FE-simulation is assumed to have isotropic and linear elastic
properties.

Table 11. Material properties.


General structural steel, SS2134
Young’s modulus, E 210 GPa
Poissions’s ratio, ʋ 0.30
Density, ρ 7850 kg/m3
Yield strength, Rp0.2 355 MPa
Ultimate tensile strength, Rm 490 MPa

4.2.2 Analysis setup


All simulations are made in Abaqus 6.14-1. The FE-analysis is a general static analysis. The solution
technique is Full Newton analysis.

4.3 Mesh convergence study


A mesh convergence study is performed to make sure that the solution is independent of the mesh
resolution.

4.3.1 Elements
The element type used is ten-node quadratic tetrahedral elements, known as C3D10. Each face of the
element is defined by six nodes and this element has curved faces. The elements have 4 integration points.
This quadratic element type will represent the displacement and boundary conditions better than linear
elements, with the same amount of elements. [18]

4.3.2 Convergence study


The models main aim is to produce accurate stresses at the region around the weld. The elements at larger
distance from the region of interest is only representing geometry and transmitting load. Therefore, the
convergence study is made through a local mesh refinement, performed through successive level of local
mesh refinement. The study is made for the continuous weld end model. The convergence is investigated
for a path started from the weld toe and along 14 mm perpendicular to the weld. The investigated output is
the maximum principal stress, and the paths node values include intersections from the connected nodes,
see Figure 33. Four different mesh levels are compared and described in Table 12 and illustrated in Figure
34. The convergence study is made for Load level 1, 120 kN, and B.C and load case 1, simple support
condition. B.C. and load case 1 are described in chapter 4.4 Boundary conditions and Load cases study.

Figure 33. The path starts in the weld toe and continuous 14 mm perpendicular to the weld.

40
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Table 12. Mesh levels for the mesh study.


Mesh1 Mesh2 Mesh3 Mesh4
Number of nodes: 122,000 290,000 567,000 798,000
Number of elements: 81,000 201,000 400,000 564,000
Element types: C3D10 C3D10 C3D10 C3D10
Element size ca: 2 mm 1-2 mm 0.5-1 mm 0.5 mm

Figure 34. Illustrates Mesh 1, Mesh2, Mesh 3 and Mesh 4. The grey area indicates stresses above 225 MPa.

41
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Mesh convergence study


400

350
Stress, Max. Principel [MPa]

300

Mesh 1
250
Mesh2

200 Mesh 3
Mesh4
150

100
0 1 2 3 4 5
Distance along path [mm]

Figure 35. Result for the different mesh qualities.

The results from the mesh convergence study are shown in Figure 35. It shows that convergence occurs at
about 400,000 elements with element size of about 1 mm close to the weld. The difference is about 1.5 %
between mesh 3 and mesh 4, which indicates that convergence is reached and the solution is independent
of further refinements. This element size, mesh 3, will be used in verification of the B.C and calibration
analysis for the specimens and strain gauge measurements.

Generally, every problem needs a convergence study to ensure that the solution is independent of the
mesh. Already made mesh studies can also be used for similar problems. There already exists mesh
standards for the Effective notch and the Hot-spot method. Therefore, no mesh studies are performed for
these models and the mesh size are taken from IIW [3], when these methods are used.

42
M. Ramström ǀ Chapter 4 - Finite Element Analysis

4.4 Boundary conditions and Load cases study


A comparison study of the models boundary conditions and the applied loads are performed due to
investigation of the effect on the solution and understanding how to simulate reality in a reliable way. The
solutions are compared with the strain gauge measurements. The investigation includes five different
variants of B.C. and load cases, named case 1-5, where case 1-4 includes a symmetry-plane condition and
case 5 includes no symmetric condition. The load magnitude used in the study is 120 kN.

 B.C. and Load case 1 – Simple support condition at the outer surface and z-symmetry condition,
see Figure 36 a. The load is applied as a negative pressure on the surface. The pressure is evenly
distributed over the surface nodes and oriented in z-direction in Figure 36 a.

 B.C. and Load case 2 – Clamped condition at the outer surface and z-symmetry condition, see
Figure 36 b. The clamped condition is applied on a reference point connected through a rigid body
with tie constraint to the outer surface and locked in all directions and rotations except the load
direction, z-direction. The load is applied as a negative pressure on the outer surface. The
pressure is even distributed over the surface nodes and oriented in z-direction in Figure 36 b.

 B.C. and Load case 3 – Clamped condition at the upper and under surfaces with similar size as the
grip fixture in the test machine and z-symmetry condition, see Figure 36 c. The clamped condition
is applied on a reference point connected through a rigid body with tie constrain to the grip fixture
surface and locked in all directions and rotations without the load direction, z-direction. The load
is applied as a concentrated force in a reference point connected to the upper and under surfaces
with similar size as the grip fixture with a rigid body tie constraint. The force are evenly
distributed over the surface nodes and orientated in z-direction in Figure 36 c.

 B.C. and Load case 4 – Simple support condition at the centrum edges and z-symmetry condition,
see Figure 36 d. The load is applied as a concentrated force in a reference point connected to the
upper and under surfaces with similar size as the grip fixture with a rigid body tie constraint. The
force are even distributed over the surface nodes and orientated in z-direction in Figure 36 d.

 B.C. and Load case 5 – No symmetry plane, se Figure 36 e. Clamped condition at upper and
under surface with similar size as grip fixture at one side where the surfaces are locked in all
directions and rotations. At the other side is a concentrated force in a reference point connected to
the upper and under surfaces with similar size as the grip fixture with a rigid body tie constraint.
The force are even distributed over the surface nodes and orientated in z-direction in Figure 36 e.

These different boundary conditions restrain the model to avoid rigid body motions.

43
M. Ramström ǀ Chapter 4 - Finite Element Analysis

a. b.

c. d.

e.

Figure 36. a. B.C. and load case 1. b. B.C. and load case 2. c. B.C. and load case 3. d. B.C. and load case 4. e. B.C.
and load case 5.

44
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Figure 37. Results for B.C. and load case study. (F = 120 kN)

The result shows different maximum principle stress around the weld toe. The stress varies in the range of
420 MPa to 470 MPa.

B.C and load case 3 and 5 simulate reality and the strain gauge measurements best. Therefore, they are
chosen to be used in further fatigue life analysis. The maximum principal stress between them differs
about 1.3 %. For the Effective notch and the Hot-spot models, B.C. and load case 3 is used, due to the
reduced number of elements.

4.4.1 Comparison study for nonlinear geometric analysis


Nonlinear geometric, called Nlgeom in Abaqus. Nlgeom is a setup alternative that considers nonlinear
effects for large deformation. A comparison study has been performed to consider this effect in the results.
It has been done for the continuous fillet weld along the path perpendicular to the weld toe, and the
measured points are located at 2.5 mm and 14 mm from the weld, illustrated in Figure 38.

The applied boundary condition and load case is B.C. and load case 3 from chapter 4.4 Boundary
conditions and Load cases study. The nonlinearity affects the result more with larger load level. The used
mesh size, mesh 3, is similar to the mesh size used in the mesh convergence study.

45
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Figure 38. Path for Nlgeom study. (F= 240kN, R=0.1, mesh size = 0.5mm)

The Figure 39 shows that analysis with Nlgeom is more close to the strain gauges measurements.
Therefore, FE-analyses compared with the strain gauges measurements and test results are done with
Nlgeom.

Nlgeom study
650

600
Stress range, Max Principel [MPa]

550

500
Nlgeom on
450 Nlgeom off

400 Strain gauge 2,5 side A


Strain gauge 2,5 side b
350
Strain gauge 14 side A
300 Strain gauge 14 side B
250

200
0 2 4 6 8 10 12 14
Z-distance along path [mm]

Figure 39. Comparison study between Nlgeom and strain gauge measurements.

46
M. Ramström ǀ Chapter 4 - Finite Element Analysis

4.5 Validation of FE-model


4.5.1 Boundary condition, Load case and load level
B.C. and load case 3 is applied on the two different weld ends. The load levels match the testing and load
data can be found in Table 13.

Table 13. Load levels. (B.C. and load case 3)


Load level Magnitude and Direction, max [kN] Magnitude and Direction, min [kN]
Load level 4 [0,0,240] [0,0,24]
Load level 3 [0,0,200] [0,0,20]
Load level 2 [0,0,160] [0,0,16]
Load level 1 [0,0,120] [0,0,12]

4.5.2 Simulation of the pre-stress


This FE-model simulates the pre-stress caused by the mounting in the test machine. The pre-stress
analysis is performed for the continuous weld end, where the base plates curvature were significant larger
than with the discontinuous weld end.

The geometry of the model includes the baseplate’s deformation caused by the welding. The deformation
is measured and modelled in twelve points around the baseplate, see Figure 40. All the points around the
base plates outer edges are bent up, the vertical distance varies between 0.3-1.2 mm.

The boundary conditions should simulate the tests mounting process in the test machine. One side is
clamped and the other side is exposed for a vertical displacement, seen in Figure 41. The clamped
condition is applied on upper and under surface with the same area as the grip fixture. The displacement
condition has the same surface size as the grip fixture and is applied on the upper surface. The applied
vertical displacement is the specific pre measured distance so the specimen is straightened out.

Figure 40. Base plate with curvature.

47
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Figure 41. Boundary condition for simulating mounting in the test machine.

The results from the pre-stress simulation are shown in Figure 42 and Figure 43. The FE-model and the
strain gauge measurements show similar stress field for the pre-stress. The clamped side’s upper surface is
exposed for a tensile stress, and the side with applied displacement is exposed for a compressive stress.
The base plate’s underside is exposed for compressive stress at the clamped side and tension stress at the
side with applied displacement. The result from this test corresponds to the results from the pre-stress in
the strain gauge measurements.

Ca: - 60 MPa

Ca: 60 MPa

Figure 42. Pre-stress due to mounting in the test machine.

Figure 43. Deformation scale factor is set to 25.

48
M. Ramström ǀ Chapter 4 - Finite Element Analysis

4.6 Numerical analysis


The numerical analysis is performed on the test objects geometry and compared with the test results. The
numerical analysis is made for the three commonly used calculation methods for fatigue life estimations.
The three methods are Effective notch, Hot-spot and Nominal stress.

Boundary condition and load case 3 is used and load levels are similar with tests. The FE-analysis is
performed with the min and max force in Table 13, the min forces are calculated with the stress ratio
factor 0.1. Then the stress range is calculated by the difference of the max and min results.

4.6.1 Effective notch model


The fatigue life estimated with the Effective notch approach is calculated from the FE-analysis. The
Effective notch approach evaluates the notch stress, and includes both the toe and the root. The largest
notch stress is used for calculating the fatigue life of the weld. The method has a standard FAT-value 225
MPa. [3]

The estimated number of cycles is calculated with equation (34). The number of cycles is calculated with
50 % risk of failure due to later comparison with the tests.

𝐹𝐴𝑇 𝜑𝑄 𝑚
𝑁 = 𝑁𝑡 ( ) (34)
𝜎𝑟

Where 𝑁𝑡 = 2,000,000, 𝜑𝑄 = 1.3, 𝑚 = 3.

The Effective notch model may need a global model and a local sub-model, where the global model has a
coarse mesh, due to computer limitations. The Effective notch approach can only be applied on the
continuous FE-model.

4.6.1.1 Sub-model
For the Effective notch method a sub-model is used to refine the mesh around the weld’s notches. With
notch radius 1 mm, the model needs an element size about 0.25 mm in the notch. [3] Therefore, the mesh
around the notches are refined in a radius of two 2 mm, see Figure 44, and with a smooth growth rate.
Dimensions of the sub-model are shown in Figure 44 and Figure 45.

49
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Figure 44. Sub-model dimensions and notch refinement. Notch radius = 1 mm, (toes and root), Green radius = 2 mm.

Figure 45. Sub-model dimensions.

Figure 46. Sub-model mesh, element size 0.25 mm in the notches.

The results from the FE-analysis are summarized in Table 14. Figure 47 shows the results for one specific
load level.

50
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Table 14. Effective notch results.


Force [kN] Stress range, σtoe [MPa] Cycles
12-120 313 1,630,000
16-160 424 656,000
20-200 534 329,000
24-240 644 187,000

Figure 47. Effective notch result. Grey area indicates stress above 225 MPa. (F=120, Mesh size=0.25 mm)

4.6.2 Hot-spot model


The fatigue life estimated with the Hot-spot approach is calculated from the FE-analysis. FAT-value 100
is recommended from IIW for a non-load transferring fillet weld. [3] Mesh size is from the same
recommendations, and the mesh size are 1 mm in a radius 14 mm around the weld, see Figure 48. The Hot
spot approach can only be applied on the continuous FE-model.

The estimated number of cycles is calculated with equation (35). The number of cycles is calculated with
50 % risk of failure.

𝐹𝐴𝑇 𝜑𝑄 𝑚
𝑁 = 𝑁𝑡 ( ) (35)
𝜎ℎ𝑠

Where 𝑁𝑡 = 2,000,000, 𝜑𝑄 = 1.3, 𝑚 = 3.

51
M. Ramström ǀ Chapter 4 - Finite Element Analysis

Figure 48. Mesh Hot-spot model.

The results from the FE-analysis are summarized in Table 15. The result from a specific load level is
shown in Figure 49. The Hot-spot measuring points are marked with red dots at distance 4, 9, 10 and 14
mm perpendicular to the weld toe.

Table 15. Hot-spot results. Linear and Quadratic.


Continuous weld end, FAT100
Stress range Linear Quadratic
Force σ0,4t σ0,9t σ1,0t σ1,4t 𝜎ℎ𝑠 Cycles 𝜎ℎ𝑠 Cycles
[kN] [MPa] [MPa] [MPa] [MPa] [MPa] [MPa]
12-120 149 135 133 128 159 1,092,000 164 997,000
16-160 200 181 178 172 214 449,000 221 410,000
20-200 251 227 223 215 269 226,000 277 206,000
24-240 302 274 270 259 324 130,000 335 117,000

Figure 49. Hot-spot result. Result points are marked with red dots, distance 4, 9, 10 and 14 mm from the weld toe.

52
M. Ramström ǀ Chapter 4 - Finite Element Analysis

4.6.3 Nominal stress method


The fatigue life estimated with the nominal stress approach is calculated with the analytical stress range.

The FAT-value 71 is used for a continuous fillet weld end and the FAT-value 80 is used for a
discontinuous fillet weld end. Both the values are from the book; Konstruktionshandbok, Claes Olsson,
Welded joint numbers: 57, weld class: Fillet weld, non-load transfer connections. [9]

The estimated number of cycles is calculated with equation (36). The number of cycles is calculated with
50 % risk of failure.

𝐹𝐴𝑇 𝜑𝑄 𝜑𝑡 𝑚
𝑁 = 𝑁𝑡 ( ) (36)
∆𝜎

Where 𝑁𝑡 = 2,000,000, 𝜑𝑄 = 1.3, 𝜑𝑡 = 1.0627, 𝑚 = 3.

Table 16. Nominal stress method results, Continuous weld end.


Continuous weld end, FAT71
Force [kN] Stress range, ∆σ [MPa] Cycles
120 112 1,343,000
160 144 632,000
200 180 323,000
240 216 187,000

Table 17. Nominal stress results, Discontinuous weld end.


Discontinuous weld end, FAT80
Force [kN] Stress range, ∆σ [MPa] Cycles
120 112 1,922,000
160 144 904,000
200 180 463,000
240 216 268,000

53
M. Ramström ǀ Chapter 5 - Results

CHAPTER 5
RESULTS

This chapter presents the testing and the FE-analysis results and comparison studies.

5.1 Test results


The results for the comparison between continuous and discontinuous weld end from the tests are
presented below. The S/N curves based on the test results are shown in Figure 50. Each curve in the S/N
diagram is based on eight specimens. The testing include four load levels in the interval 120-240 kN and
for each load level two specimens have been tested. The curves equation is a linear correlation between
the test data according to the least square method. According to IIW, [3], since the number of test results
are less than 10 test objects, 0.17 standard deviation has been used instead of the actual 0.052 (Continuous
weld end) respectively 0.034 (Discontinuous weld end). To produce the characteristic curve, two standard
deviations have been applied (i.e. 2.3 % failure probability). The actual and characteristic slopes are
presented in Table 18. The results indicate that the discontinuous end obtains significant higher fatigue
strength. The both slopes are slightly higher than three, but the discontinuous end receives a slope closer
to three. The FAT-value determined from the tests are FAT66 MPa for discontinuous and FAT51 MPa for
the continuous weld end, see Table 19.

S/N-Curve
1000
Stress range σr [MPa]

100

10
10000 100000 1000000 10000000
Cycles
Continous - Characteristic curve - 2,3% Continuous - Test results- 50%
DisContinous - Characteristic curve - 2,3% DisContinous - Test results - 50%
Figure 50. Comparison between the continuous and discontinuous test results.

54
M. Ramström ǀ Chapter 5 - Results

Table 18. Test data.


Slope Standard deviation
Continuous test results 3.42 0.052
Continuous characteristic 3.00 0.170
Discontinuous test results 3.13 0.034
Discontinuous characteristic 3.00 0.170

Table 19. FAT-value.


Discontinuous Continuous
FAT-value [MPa] 66 51
Stdv. 0.17 0.17
Slope, m 3.00 3.00
Cycles [N] 2x106 2x106

5.2 Comparison of the computational methods


The results for the comparison between the computational methods are presented in Figure 51. The
comparison is between the Effective notch method, the Hot-spot method and the Nominal stress method
for the continuous weld end. The curves are characteristic curves (i.e. 2.3 % failure probability) and have
slope -1/3 in the log-log diagram, m=3.

S/N curve - Computational method


1000

Effective notch
Force [kN]

Hot Spot
100

Nominal stress
Continuous (FAT
71)

10
10000 100000 1000000 10000000
Cycles
Figure 51. Comparison between the computational methods for the continuous weld end.

55
M. Ramström ǀ Chapter 5 - Results

5.3 Comparison study between FE-analysis and tests


The aim of the project is to find a method to calculate the fatigue life of the discontinuous weld end with
the Effective notch method. The FE-model used for the Effective notch approach is the continuous weld
end, due to the methods limitations and avoid singularities. The results for the comparison between the
computational Effective notch method and the tests are presented in Figure 52. The comparison includes
the tests for both the continuous and discontinuous weld end. The curves are characteristic curves (i.e. 2.3
% failure probability). To produce the tests characteristic curves, two standard deviations of the actual test
results have been applied. The actual standard deviations are 0.052 for continuous and 0.034 for
discontinuous weld end. The slopes are also the actual slope determined by the test results, 3.13 for
continuous and 3.42 for discontinuous weld end. Slope 3 and 2.3 % failure probability is used for the
Effective notch curve. The data for the S/N curve are summarized in Table 20.

S/N curve - Effective notch compared with physical tests


1000

Effetive notch
Force [kN]

Continuous test
100 data

Discontinuous
test data

10
10000 100000 1000000 10000000
Cycles
Figure 52. Comparison between the Effective notch approach and the tests.

Table 20. Cycles to failure. Effective notch calculations and test data.
Cycles, Cycles, Cycles,
Stress range [MPa]
Effective notch Discontinuous weld end Continuous weld end
108 1,630,000 1,452,000 776,000
144 656,000 598,000 289,000
180 329,000 326,000 133,000
216 187,000 159,000 73,000

56
M. Ramström ǀ Chapter 5 - Results

5.4 Developed FAT-value


The results of the tests and the Effective Notch approach are analyzed further, in order to investigate and
develop a new FAT-value for weld ends which are simulated continuous. The develop FAT-value are
made as an S/N-curve determined by Equation (37). The S/N curves vertical axis is the notch stress from
the FE-model and the horizontal axis is the tests result, the method is illustrated in Figure 53. This forms
an Effective notch curve suited for discontinuous ends simulated with a continuous FE-model. The factors
that affect the new FAT-value are standard deviation and slope, shown in Table 21. The FAT-value is
calculated for 2,000,000 cycles. The constant C is determined by Equation (38) and to produce the
characteristic curve, where two standard deviations have been applied (i.e. 2.3 % failure probability). The
constant, m, is determined by the least square method.
The new FAT-value calculations are made for both the discontinuous and continuous test results.

log 𝑁 = log 𝐶 − 𝑚 log ∆𝜎 (37)

𝐶 = 𝑥𝑚 − 2 ∗ 𝑠𝑡𝑑𝑣 (38)

Figure 53. Method to develop FAT-value. The circular dot is a mark for the calculated FAT-value.

57
M. Ramström ǀ Chapter 5 - Results

5.4.1 Discontinuous weld end


For the discontinuous test results the new FAT-value is in the interval of 194-269 MPa, see Table 21. The
actual value from the test results indicates a FAT-value about 269 MPa. According to the norm that IIW
recommend the FAT-value 194 MPa. Figure 54, Figure 55 and Figure 56 show the develop S/N curves
with the new FAT-value at 2,000,000 cycles, marked with a dot.

Table 21. Data for new developed FAT-value, discontinuous.


FAT-value [MPa] 194 263 269
Stdv. 0.17 0.034 0.034
Slope, m 3 3 3.13
Cycles [N] 2x10 2x10 2x106
6 6

S/N curve - Discontinuous compared with Effective notch method

1 000
Stress range σr [MPa]

Results curve

Characteristic
100 curve
Results

10
10 000 100 000 1 000 000
Cycles
Figure 54. Developed S/N curve discontinuous, characteristic curve with slope 3 and stdv. 0.17.

S/N curve. Discontinuous compared with Effective notch method

1 000 Results curve


Stress range σr [MPa]

Characteristic
curve
100 Results

10
10 000 100 000 1 000 000
Cycles
Figure 55. Developed S/N curve discontinuous, characteristic curve with slope 3 and stdv. 0.034.

58
M. Ramström ǀ Chapter 5 - Results

S/N curve. Discontinuous compared with Effective notch method

1 000
Stress range σr [MPa]

Results curve

Characteristic
100 curve
Results

10
10 000 100 000 1 000 000
Cycles
Figure 56. Developed S/N curve discontinuous, characteristic curve with slope 3.13 and stdv. 0.034.

5.4.2 Continuous weld end


For the continuous test results the FAT-value is in the interval of 151-218 MPa, see Table 22. The actual
value from the test results indicates a FAT-value about 218 MPa. According to the norm that IIW
recommend the FAT-value is 151 MPa. Figure 57, Figure 58 and Figure 59 show the determined S/N
curves for continuous weld with the FAT-value at 2,000,000 cycles, marked with a dot.

Table 22. Data for new developed FAT-value, continuous.


FAT-value [MPa] 151 196 218
Stdv. 0.17 0.052 0.052
Slope, m 3 3 3.42
Cycles [N] 2x10 2x10 2x106
6 6

S/N curve - Continuous compared with Effective notch method

1 000
Stress range σr [MPa]

Results curve

Characteristic
100 curve
Results

10
10 000 100 000 1 000 000
Cycles
Figure 57. Developed S/N curve continuous, characteristic curve with slope 3 and stdv. 0.17.

59
M. Ramström ǀ Chapter 5 - Results

S/N curve - Continuous compared with Effective notch method

1 000
Results curve
Stress range σr [MPa]

Characteristic
curve
100
Results

10
10 000 100 000 1 000 000
Cycles
Figure 58. Developed S/N curve continuous, characteristic curve with slope 3 and stdv. 0.052.

S/N curve - Continuous compared with Effective notch method


1 000

Results curve
Stress range σr [MPa]

Characteristic
100 curve

Results

10
10 000 100 000 1 000 000
Cycles
Figure 59. Developed S/N curve continuous, characteristic curve with slope 3.42 and stdv. 0.052.

60
M. Ramström ǀ Chapter 6 - Conclusions

CHAPTER 6
CONCLUSIONS

In this chapter conclusions of the results are summarized. The conclusion involves the test results and the
different methods with focus on the Effective notch approach.

6.1 Tests
The testing is an important part of the project to carry out a comparison with the calculation methods.
Number of specimens was eight for each weld end design, distributed in four load levels. This caused
problems to statistically ensure the results. Recommended are 5-7 specimens on each load level, therefore,
more specimens are needed to ensure the results [9]. The calculated standard deviation of the test results is
small compared with IIW recommendations. The slopes is slightly over three for both designs, three are
standard for welded joints subjected to normal stress [3].

The results from the tests indicate significant higher fatigue strength for the discontinuous weld end. The
Nominal stress method, which can be used for both the continuous and discontinuous weld end, estimates
the same outcome, but with different results in fatigue life. The Effective notch and the Hot-spot method
cannot be applied on the discontinuous weld end.

6.2 Comparison of computational methods and tests


The three computational methods investigated are compared for the continuous weld end. The Effective
notch indicates highest fatigue strength, the nominal stress method indicates the second highest fatigue
strength. The Hot-spot approach differs most and indicates lower fatigue strength.

No method shows conservative results compared with the tests for the continuous weld end. The Hot-spot
approach estimates the continuous weld best and is the most accurate method for the continuous case.
However, the Nominal stress and the Hot-spot approach cannot be applied in complex load cases and for
that reason are the Effective notch method was used in further investigations.

The Effective notch and the Hot-spot approach cannot evaluate the discontinuous FE-model in a reliable
way. They are only validated for max principle stress oriented ± 60° perpendicular to the weld toe.

The nominal stress method can only give a coarse estimation of the weld fatigue life (if the nominal stress
is known). It can be a useful method for a coarse estimation between different welded joints.

6.3 Comparison of the Effective notch method and tests


The Effective notch method can only be applied on the continuous FE-model, the model overestimate the
fatigue life of the continuous and match relative well with the test results of the discontinuous weld end.
The results indicate that the method needs some adjustment to match the discontinuous weld end, and
don’t match the continuous weld end.

61
M. Ramström ǀ Chapter 6 - Conclusions

6.4 Developed new FAT-value


The developed FAT-value for the discontinuous weld end is in the interval 194-269 MPa dependent on
actual results or IIW standard norm. The interval is dependent of the standard deviation and the S/N
curves slope. These values can be determined from test results or taken from IIW recommendations. In
this case, the obtained interval is mainly dependent on the standard deviation, 0.034 or 0.17. FAT269 is
the value that matches BT Products production, but for a more reliable result are more specimens needed.

The overall conclusion is that it is possible to simulate the discontinuous weld end with a continuous FE-
model. The continuous model can be evaluated with the developed FAT-value for the Effective notch
approach. The FE-model should include the notch radius and the standards IIW recommend. Adjustments
for the new FAT-value need further investigations and cases to ensure a more general and reliable value.
The standard value, FAT225, are in the interval determined in this project, and the value for weld ends
simulated with a continuous weld is assumed to be close to this value.

62
M. Ramström ǀ Chapter 7 - Discussion

CHAPTER 7
DISCUSSION

In this chapter the test results and the different methods including the developed FAT-value are discussed.

The aim of this thesis was to find a calculation method to estimate fatigue life of weld ends. Since, the
Effective notch approach is the most common used and appropriate method, the focus was on developing
a new FAT-value for weld ends. The Effective notch approach is only applicable to the continuous FE-
model and the weld ends of interest are the discontinuous ends. Therefore, a comparison study between
the continuous FE-model and the test results for the continuous and discontinuous is made.

The test results indicate higher fatigue strength for the discontinuous than the continuous weld end. The
Nominal stress method which can be applied in both cases, estimates the same outcome, but with higher
fatigue life and different interrelationship. This is due to the coarse accuracy of the method and need of
scale factors, such as misalignment factor etc., see chapter 2.6.1 Nominal stress method. Also the two
cases used don’t match perfect with the tested specimens design and the cases contain parameters with
large intervals, see cases in APPENDIX C - Nominal stress method. In the calculations for the nominal
stresses was only the first term in equation (39) considered, and the effect of the moment was neglected.
This can be a reason for the overestimating of the calculations compared with the tests. In reality, there
was a moment, caused by the plate curvature, which increased the nominal stress. The pre-stresses from
the mounting in the test machine could also be an affecting factor. However, this method is useless if the
nominal stress is not defined and if the weld don’t match the cases available.

𝐹 𝑀
𝜎𝑛𝑜𝑚 = + 𝑧 (39)
𝐴 𝐼

The fatigue test results indicates that the continuous specimens where affected by factor that decrease
fatigue life, such as mean stress and residual stresses, geometrical stress concentrations and
discontinuities. Other affecting factors such as stress range and the material properties are the same.
The strain gauge measurements indicate higher stresses 2.5 mm from the weld toe for the continuous weld
compared with the discontinuous. This higher stress concentration is partially due to the pre-stresses and
may also depend on geometrical stress concentrations.

The Effective notch method with standard FAT225 MPa, overestimates the fatigue life for the continuous
and the discontinuous weld ends but the calculated result is close to the test result with the discontinuous
weld. The adjusted FAT-value to match the discontinuous weld end is in the interval 194-269 MPa
dependent on actual results or IIW standard norm. The interval is dependent of the standard deviation and
the slope. The actual result, determined from the test results indicates FAT269 MPa, but since only two
specimens were used at each load level, there are uncertainties about the actual standard deviation and
slope. If the number of specimens is less than ten the standard deviation should be estimated to 0.17 for

63
M. Ramström ǀ Chapter 7 - Discussion

geometrically simple structures according to IIW [3]. The difference between standard deviation 0.034
and 0.17 gives the large interval.

For the calculated result of the continuous weld compared with the test result of the continuous weld, FAT
225 should match. The results however indicate a FAT-value in the interval 151-218 MPa.

There is no mention of if the standard deviation is for robot or manually welded joints, which affect the
weld quality and standard deviation. More tests are required to ensure the actual scatter in BT Products
production. The actual standard deviation should be used in further adjustments for the FAT-value for
weld ends.

The slope of the characteristic curve is -1/m in a log-log diagram, were m is 3 for a welded joint affected
by normal tension according to IIW [3]. The two weld end designs received both an m-value above 3 and
the calculated FAT-values increase with a higher m-value. The m-value is also dependent on the number
of tests, and more tests will therefore ensure a correct m-value.

The occurred pre-stresses from the mounting in the test machine can be a factor that decreases the fatigue
life. The pre-stresses for the continuous where significant high compared with the discontinuous weld end.
The difference is caused from the curvature and misalignments in the baseplate, where the continuous was
affected considerably more. The stresses perpendicular to the continuous weld toe was 112 MPa at one
side and 55 MPa at the other side. The pre-stresses increase the mean stress, which affects the fatigue life.
A few of the strain gauges showed compressive strains (stresses). This leads to a stress ratio with changing
signs, R<0, in that locations. Fatigue and also welded joints fatigue, are more sensitive to tension than
compression stresses, but fatigue cracks can occur during compressive stresses due to positive residual
stresses. The occurred tensile stresses increase the risk of reaching the yield limit, due to the increased
mean stress. At that yield limit the material will plastically deform and the methods do not consider this.

The test results indicate interesting similarities and dissimilarities. The crack propagation for the
continuous weld end where at the same end side, side A. Side A where the start side for the weld string.
This similarity may be caused by the less penetration into the base material compared with the other side.
For the discontinuous weld end no similarities can be seen, since to the sides are similar. Unfortunately no
note where made of which side of the specimens where pressed in the test machine and this may have had
an increased effect of the pre-stresses. This could be an interesting parameter in further studies and testing.

The testing shows nothing about when, where and how the crack initiates and propagates to fracture, than
a crack initiation analysis is required. Another important factor is the local geometry that affects the stress
concentrations near the weld ends. For example, a larger radius and a smoother transition increase the
fatigue strength. The main affecting local geometry parameters are radius, undercut and lack of fusion,
illustrated in Figure 60. The effect of these parameters needs further investigations with microscopic
studies of the weld ends.

Figure 60. Weld profile, local geometry parameters.

64
M. Ramström ǀ Chapter 8 - Future work

CHAPTER 8
FUTURE WORK

In this chapter some suggestions of future work are presented. This thesis should be regarded as a
continuous study of weld ends calculations and further investigations are recommended.

This thesis contains the theory of fatigue life calculations of welds and an introduction to weld end
calculations. The most common method today, Effective notch, should be developed and adjusted to weld
ends. The conclusion is to simulate the discontinuous weld end with a continuous weld end with a
modified FAT-value. The aim is to develop and ensure a general FAT-value for weld ends.

The adjusted FAT-value is determined through tests compared with FE-analysis. Therefore, the number of
specimens is an important factor to statistically ensure the test results. To ensure the standard deviation
and the S/N curves slope, data from at least 5-7 specimens at each load level is required. [9] Therefore,
there is a need for more specimens.

Since the method should be adjusted for weld ends simulated with a continuous weld more cases is
interesting. In this thesis only one load case is investigated. The case is a non-load-transferring fillet weld.
More and varied load cases are also an interesting factor. For example, load-transmitting welds and
bending stresses are two interesting cases. Other parameters are throat thickness, plate thickness and other
types of welds. This first case showed results useful in further investigations.

It would also be interesting to investigate where and how the crack propagates through a visual crack
initiation investigation.

In the future, it is proposed that BT Products creates an internal database for fatigue testing and
determines their own standard deviation that matches their welded joints.

65
66
M. Ramström ǀ Bibliography

BIBLIOGRAPHY

[1] Svenskt Stål AB, SSAB, Plåthandbok, att konstruera och tillverka i höghållfast stål, 2 ed., Nyköping:
Österbergs, Sörmlandstryck AB, 2011.
[2] Boverkets handbok om stålkonstruktioner, BSK 99, Stålkonstruktioner, Karlskrona: Boverket,
Byggavdelningen, 2004.
[3] International institute of welding, IIW, "IIW-1823-07, Recommendations for fatigue design of
welded joints and components," IIW, 2008.
[4] Toyota Material Handling, "http://www.toyota-forklifts.eu," Toyota forklifts, [Online]. Available:
http://www.toyota-forklifts.eu/en/company/tmhephilosophy/Pages/History.aspx. [Accessed 11 02
2015].
[5] Toyota Material Handling, "http://www.toyota-forklifts.eu," Toyota forklift, [Online]. Available:
http://www.toyota-forklifts.eu/en/company/Pages/About-TICO.aspx. [Accessed 11 02 2015].
[6] Åsa Eriksson, Anna-Maria Lingnell, Claes Olsson, Hans Spennare, Svetsutvärdering med FEM, third
ed., Malmö: Liber AB, 2002.
[7] Bertil Jonsson, "Assessment of fatigue life," Volvo VCE division HL, Göteborg, 2006.
[8] F. V. Lawrence, "Mechanisms of Fatigue Crack Initiation and Growth," University of Illinois,
DEpartment of Mechanical Science and Engineering, 2014-07-02.
[9] C. Olsson, Konstruktionshandbok, för svetsade produkter i stål, 5 ed., Onsala: TechStrat Publishing,
2014.
[10] LIU, Ru Lin Peng, "http://www.iei.liu.se/kmt/education/undergraduatecourses-tmmi18?l=sv,"
[Online]. Available: http://www.iei.liu.se/kmt/education/tmmi18-intranet/coursemap-
tmmi18/forelasningar/1.612690/TMMI18_FL1_2015.pdf. [Accessed 17 02 2015].
[11] LIU, Johan Moverare, "Deformation and Fracture TMKM90. Chapter 9: Cyclic Stress and Strain
Fatigue," 2013. [Online]. Available: http://www.iei.liu.se/kmt/education/tmkm90-
intranet/coursemap-tmkm90/slides/1.512500/SlidesChapter9.pdf.
[12] A. Ekberg, "http://www.am.chalmers.se," [Online]. Available:
http://www.am.chalmers.se/~anek/teaching/fatfract/98-6,7,8,9,10,11.pdf. [Accessed 18 02 2015].
[13] Niklas Karlsson, Per-Henrik Lenander, "Analysis of Fatigue Life, LITH-IKP-EX--05/2302--SE,"
Division of Solid Mechanics, Linköping, 2005.
[14] A. Göransson, "Fatique analysis of weld ends - Comparison between testing and FEM-calculations,"
BT Products, Mjölby, 2014.
[15] S. Bergdahl, "An Examination of Welded Joints Regarding Weld Geometry, Weld Discontinuities
and the Effects of Shot Peening," LITH-IKP-EX--05/2315--SE, Linköping, 2004.
[16] Hyperphysics, [Online]. Available: http://hyperphysics.phy-astr.gsu.edu/hbase/thermo/thexp.html.
[Accessed 16 03 2015].
[17] Tore Dahlberg, Anders Ekberg, Failure, Fracture, Fatigue: An introduction, Lund: Studentlitteratur,
2002.
[18] Robert D. Cook, David S. Malkus, Michael E. Plesha, Robert J. Witt, Concept and Applications of
Finite Element Analysis, fourth ed., John Wiley Sons, 2002.

67
68
M. Ramström ǀ Appendix

APPENDIX

The appendix includes the documentation considered not relevant to present in the main report, but still
fills an important purpose.

APPENDIX A
Drawings for the specimens

69
M. Ramström ǀ Appendix

70
M. Ramström ǀ Appendix

71
M. Ramström ǀ Appendix

72
M. Ramström ǀ Appendix

73
M. Ramström ǀ Appendix

APPENDIX B
Strain Gauges placement

Strain gauges placement for continuous weld end

OS-CA-2

OS-A-1:OS-A-5
OS-CA-1

US-CA/B/C-1

2.5 mm

74
M. Ramström ǀ Appendix

Strain gauges placement for discontinuous weld end

OS-CA-1

OS-AV-1:OS-AV-3 and OS-AH-1:OS-AH-3


OS-AV-4 and OS-AH-4

US-CA/B/C-1

2,5 mm

75
M. Ramström ǀ Appendix

APPENDIX C
Nominal stress method

Discontinuous weld end.

FAT80 MPa is used.

Continuous weld end.

FAT71 MPa is used. (l=100 mm)

The Figures and the nominal stress data are from Internal Institute of welding, IIW. [3]

76