Residual Stress Analysis of Pipeline Girth Weld Joints

University of Wollongong
Research Online
University of Wollongong Thesis Collection
University of Wollongong Thesis Collections
2014
Residual stress analysis of pipeline girth weld joints

Djarot B. Darmadi
Recommended Citation
Darmadi, Djarot B., Residual stress analysis of pipeline girth weld joints, Doctor of Philosophy thesis, School of Mechanical, Materials
and Mechatronic Engineering, University of Wollongong, 2014. http://ro.uow.edu.au/theses/4160
Research Online is the open access institutional repository for the

University of Wollongong. For further information contact the UOW
Library: [email protected]
RESIDUAL STRESS ANALYSIS

OF
PIPELINE GIRTH WELD JOINTS
A thesis submitted in fulfillment of the requirements for the award of the degree
DOCTOR OF PHILOSOPHY
from
by
DJAROT B. DARMADI
Ir., MT.
Mechanical Engineering
Faculty of Engineering
2014
CERTIFICATION
I, Djarot B. Darmadi, declare that this thesis, submitted in fulfillment of the
requirements for the award of Doctor of Philosophy, in the Mechanical Engineering,
Faculty of Engineering at the University of Wollongong, Australia, is wholly my own
work unless otherwise referenced or acknowledged. The document has not been
submitted for qualification at any other academic institution.
Djarot B. Darmadi
22 July 2014
Acknowledgements
The author would like to express his gratitude to his supervisors, Professor John Norrish
and Professor Anh Kiet Tieu for their excellent guidance and continuous encouragement
because only under their patient supervisions could this thesis have been completed.
Many thanks also to Dr Murugananth Marimuthu for his discussion, training, and
permission to carry out Finite Element Analysis in his Weld Simulation and Modeling
Laboratory. Special appreciation is also expressed to Dr Dominic Cuiuri for his
comprehensive training, helpful supervisions and willingness to become involved in the
experimental work required to validate the model. The author very much appreciates the
friendly and conducive environment created by Joe Abbot in the Manufacturing and
Research Laboratory (MRL) building, including the use of his skilful hands. Fruitful
discussions that further improved this thesis were held with Prof. Huijun Li, Dr. Buyung
Kosasih, Dr Mark Callaghan, Dr Lenka Kusmikova, Dr Ajit Godbole and Dr Stephen
Pan.
Many thanks are also expressed to the Director General of High Education of the
Indonesian Government for the scholarship to support the author while taking this PhD
study. Thanks are also expressed to Prof. John Norrish on behalf of CRC pipelines for
their support and willingness to extend the period of study. An International
Postgraduate Tuition Award (IPTA) given by University of Wollongong is also
appreciated.
The author is deeply grateful to his wife Jendra and son Bima for their continuous
support and encouragement through the good and bad times. Thank you all indeed.
ii
Abstract
Residual stress is one of the important factors that should be considered when assessing
the integrity of welded structures because it is well known that residual stress may lead
to failure in the weld joint. Residual stress may lead to stress corrosion cracking in a
corrosive environment and has resulted in failures in energy pipelines.
Residual stress and stress corrosion cracking in the region of the girth welded joint in
pipelines was investigated in the current work. The emphasis was placed on residual
stress modeling in pipeline girth welds since it was judged to be an essential prerequisite
for stress corrosion cracking.
In this thesis, the residual stress around girth welds was studied with FEM using
ANSYS (APDL) to give a deep insight into the formation of residual stress. A coupled
Thermo-Metallurgical-Mechanical analysis was carried out because the metallurgical
analysis of the high strength steel used extensively in pipelines cannot be ignored. The
final residual stress prediction using ANSYS was verified with experimentally to test
the validity of the FEM model.
The FEM analysis is presented step by step which commenced with a simple thermal
analysis, followed by a thermo-mechanical analysis and finally a coupled thermometallurgical-mechanical analysis. The weld configuration also started from a simple
bead-on-plate leading on to multi-pass weld on plate and multi-pass girth weld
respectively. Various moving heat source models were used, commencing with a simple
point heat source, and then a uniform surface heat source and finally Goldaks
volumetric heat source.
The FEM results for the early simple model were verified with an analytic solution. It
should be noted that verification (comparing FEM results with analytic solution) can be
done only on a simple model since it is difficult to obtain an analytic solution (not
FEM) for a complex model. The simple model was represented by moving point heat
source in an infinite solid. To analyse transient state a new analytic solution using a non
integration technique was developed.
iii
For the complex model the FEM results were validated with the experimental method to
ensure that the model was correct. A new heat source model that combined Goldaks
volumetric heat source and uniform temperature load was used and it provided good
results. The Goldaks heat source model was used to represent the heat transferred to the
base metal and previous bead (for multi-pass welding), whilst the uniform temperature
load represents melted filler metal (droplets) that form the weld bead. A new
programming technique using a database which was composed using the ANSYS
standard files was also developed. It was found that using the database provided at least
three advantages: greater flexibility, faster computing time and the ability to interchange
data.
The role of solid state phase transformation (SSPT) was also studied in terms of
volumetric change due to the atomic packaging factor (APF), alteration of the
mechanical properties or transformation plasticity. To assess the contribution of each
aspect: volume, mechanical properties and transformation plasticity, FEM with
volumetric change (mechanical properties and transformation plasticity excluded), FEM
with mechanical properties change (volumetric change and transformation plasticity
excluded) and FEM with transformation plasticity (volumetric and mechanical
properties excluded) were carried out. These contributions were studied by comparing
the results of the FEM with and without SSPT consideration, SSPT with volumetric
change only, SSPT with alterations to the mechanical properties only and SSPT with
transformation plasticity only. The martensitic SSPT was divided into two classes,
namely prime martensite and aged martensite and these prime and aged martensite
modes were included in the FEM model which is a new feature. The concept of prime
and aged martensite is not new one, but the literature did not indicate that this degree of
detail had previously been included in FEM programs. Some papers do use SSPT but
only prime martensite is modeled. It was found that the results which considered SSPT
were closer to the experimental results for ferritic steel welding.
In the earlier chapters, all above aspects were simulated in plate welding and validated
using secondary data. Using the same logic an FEM model was programmed on a multi
pass girth weld joint. Phase transformations that were included in the model were
austenite and martensite (both: prime and aged martensite) SSPT. A reduction factor
(Kd) was used to model the yield stress of filler metal at elevated temperature. Using Kd
iv
in the FEM model of welding may be considered as a new feature. All of the modeled
aspects have been validated or verified in the plate welding moreover the temperature
history and residual stress distribution obtained from FEM prediction in the multi pass
girth welding were validated by the experimental results.
DC-LSND (Direct Cooling Low Stress No Distortion) control is one of suggested
stress mitigation techniques for welding. The DC-LSND typically applied on thin plate
welding and the main goal is to reduce distortions. There are no published papers found
that quantify the effect of DC-LSND on the residual stresses of multi pass thick pipe
girth welds. In this work the effect of DC-LSND on the residual stress in a multi pass
girth weld joint of pipe with 8mm thickness was studied through FEM analysis. From
FEM predictions the application of DC-LSND reduced maximum residual stress by
around 35%. From this result, it is suggested that DC-LSND could be applied to the
multi pass butt girth welding of thick pipe although the practicality of this procedure
would need to be assessed against its potential benefits.
Overall it can be said that the FEM model in this thesis provides an accurate prediction
of residual stress and also a deeper understanding of its development particularly in
pipeline girth welds in ferritic steels. It is in line with one of the recent research trends
in the welding area: improving the basic understanding of the welding process and the
resulting residual stresses.
COMPUTER MODELLING
The process of providing to a computer, usually in the form of
mathematical equation, a precise and unambiguous description of the
system under study, including the relationships between system inputs and
outputs, and using this description to simulate or model the described
system
(Academic Press Dictionary of Science Technology)
SIMULATION
The technique of imitating the behavior of some situation of process
(whether economic, military, mechanical, etc.) by means of a suitably
analogous situation or apparatus, especially for the purpose of study of
personnel training
(The Oxford English Dictionary)
The representation of a physical system usually by a computer or physical

model that imitates some aspects of the system for a given purpose,
eventually making approximations excluding some behaviors of the
physical system.
(J.L. Hansen PhD Thesis T.U. Denmark)
Authors own words: Modeling and simulation are a pair of powerful

tools to understand a certain physical phenomenon through tiered
validation steps
(Djarot B. Darmadi)
vi
Table of Contents
Certification .
i
Acknowledgements .
ii
Abstract
iii
Table of content .. viii
List of figures
x
List of tables .. xvii
List of publications . xviii
List of symbols and abbreviations .. xix
1. Introduction .
1.1. Background ..
1.2. Research objective ..
1.3. Research approach .
1.4. Thesis outlines
References
1
2
5
5
7
9
PART I BASIC CONCEPTS

2. A Review .
2.1. FEM simulation of welding processes .
2.2. Thermal analysis of welding FEM model .
2.2.1. Thermal properties of a material .
2.2.2. Heat transfer and heat conservation ..
2.2.3. Parametric study via FEM
2.3. Development of stress in welding .
2.3.1. A review of the three bars model .
2.4. Solid state phase transformation in the stress development .......
References ...
11
12
13
14
15
18
22
23
31
34
PART II FEM vs ANALYTIC SOLUTION

3. Moving Heat Source Simulation ...
3.1. Solution by Rosenthal .
3.2. Moving point heat source model ..
3.3. Solution using non-dimensional integral to compensate for transient
conditions ..
3.4. Element model (SOLID 70) ..
3.5. Finite Element model for moving point heat-source .
3.6. Analytical model vs FE model Quasi steady state ..
3.7. Analytical model, Rosenthal solution and FE model
quasi steady state ....
3.8. Analytical models vs FE model transient conditions .
3.9. Conclusion .
References
vii
37
39
40
41
47
47
49
52
53
55
57
PART III MODELING EXPLORATION BASED ON EXISTING DATA

4. Thermal Analysis of NeT TG1 Specimen ...
4.1. Welding procedure provided by NeT
4.2. Previous works in weld modeling of NeT welding procedure
4.3. Comparison of FE model to NeT data ..
4.3.1. Non linear material model ..
4.3.2. Inputted parameters .
4.3.3. Standard technique ..
4.3.4. Birth and Death technique ..
4.3.5. Results and discussion .
4.4. Conclusion .
References
60
61
64
66
70
72
73
75
79
87
88
5. Mechanical Analysis of NeT TG1 Specimen ....

5.1. Thermo mechanical analysis
5.2. Element model (SOLID45) ..
5.3. Mechanical properties .
5.4. Exploring the birth and death technique
5.5. FEM for the thermo-mechanical model .
5.5.1. Accommodating a growing weld bead.
5.5.2. Accommodating the melting phenomenon ..
5.6. Results and discussion ...
5.6.1. Using isotropic hardening model
5.6.2. Comparing to Masubuchis
5.7. Conclusions ..
References ..
91
93
93
94
97
99
101
101
102
109
110
111
112
6. Finite Element Model of Residual Stress Involving Metallurgical

Considerations .
6.1. Welding procedures ..
6.2. Material properties
6.3. Considerations for phase transformation
6.4. Finite element model
6.4.1. Coupled thermo-metallurgical model
6.4.2. Development of residual stress while considering phase
transformation ..
6.4.3. Solutions and strategies .
6.4.4. The ANSYS APDL modeling
6.5. Results and discussion ..
6.6. Conclusion .
References .
115
116
117
119
122
123
125
127
132
138
144
146
PART IV Pipeline Girth Welding

7. Residual Stress Analysis of The Girth Weld Joint .
7.1. Welding procedure and parameters
7.2.1. Coordinate transformation .
7.2.2. Filtering peak temperature
7.2.3. Material properties for X-70 base metal and ER70S-6
filler metal .
viii
149
149
159
161
165
168
7.2.4. Considering transformation plasticity

7.2.5. Considering cooling rate ..
7.3. Residual stress measurements
7.4. Analysis and discussions .
7.4.1. Validation of the thermal results ..
7.4.2. Residual stress validation .
7.5. Conclusions ..
References ..
175
176
178
179
183
188
202
205
8. Simulating DC-LSND on the girth butt welding ...

8.1. Heat sink modeling in ANSYS ..
8.2. Applying DC-LSND on the girth weld ..
8.3. Results and discussions ....
8.4. Conclusions
References
209
213
217
220
227
228
PART V Discussions
9. General Discussions ..
9.1. Overview .
9.2. TMM in plate welding ...
9.2.1 Change of material properties in SSPT .
9.2.2 Volumetric change in SSPT ..
9.2.3 Aspects involved in SSPT
9.3. Resetting plastic strain in the girth butt welding .
9.4. Considering lower annealing temperature
9.5. Transformation plasticity
9.6. Aged martensite considerations ..
References
230
230
232
233
235
237
238
242
247
250
254
PART VI Conclusions
10. General Conclusions ..
10.1. Conclusions
10.2. Future work ..
ix
255
255
258
List of Figures
Figure 1.1
Figure 1.2
Australias Pipelines .
Research aspects ...
3
5
Figure 2.1
Figure 2.2
Figure 2.3
Development of Residual stress in welding

Finite element model for parametric study
Transient isothermal lines for moving heat source with varied
heat; (a)10s, (b)50s, (c)100s and (d)200s ...
thermal conductivity; (a)10s, (b)50s, (c)100s and (d)200s
specific heat; (a)10s, (b)50s, (c)100s and (d)200s .
Thermal stress in the three bars model (modified from Sindo
Kou) ..
Typical distribution of residual stress in welding: (a) longitudinal
residual stress, (b) transversal residual stress
Thermal residual stress formation
Three bars model ..
Strain-stress diagram of the material model by ANSYS
Sketch of the strain-stress diagram ...
Flow-chart for evaluating thermal stress in elastic strain rate
Distribution of stress when the middle bar is heated to 400 C
(a) full model and (b) half model ..
Flow-chart for evaluating thermal stress in plastic strain rate
Distribution of stress at the three bars model heated to 1100 C
(a) when the middle bar is heated, (b) when the temperature load
is omitted ...
Flow-chart for evaluating residual stress
Solid state transformations in ferrite steels
12
19
FEM, analytic solution and experiment results .

Fixed coordinate system and moving coordinate system ..
Plotted value for
Plotted value for
Point heat source in infinite media

Integral values for varied and u .
Temperature profiles at a longitudinal path ....
Temperature profiles at a transversal path .
Solid70 3D-thermal element (Release 12.0 Documentation for
ANSYS .
FEM model for point heat source ..
Flow chart of the FEM simulation .
Temperature profiles on longitudinal lines with moving point
heat source model ..
Temperature profiles on transversal lines with moving point
heat source model .
38
40
42
42
44
45
46
46
Figure 2.4
Figure 2.5
Figure 2.6
Figure 2.7
Figure 2.8
Figure 2.9
Figure 2.10
Figure 2.11
Figure 2.12
Figure 2.13
Figure 2.14
Figure 2.15
Figure 2.16
Figure 2.17
Figure 3.1
Figure 3.2
Figure 3.3
Figure 3.4
Figure 3.5
Figure 3.6
Figure 3.7
Figure 3.8
Figure 3.9
Figure 3.10
Figure 3.11
Figure 3.12
Figure 3.13
20
21
21
22
23
24
25
26
26
28
28
29
29
30
32
47
48
49
50
51
Figure 3.14 Solutions from analytic, FEM and Rosenthal for temperature
Profiles at longitudinal paths (left) and transversal paths (right)
Figure 3.15 Enlarged solutions for far field longitudinal paths .
Figure 3.16 Isothermal profiles at plate surface for transient state
Figure 3.17 Isothermal profiles at plate cross section for transient state ..
Figure 4.1
Figure 4.2
Figure 4.3
Figure 4.4
Figure 4.5
Figure 4.6
Figure 4.7
Figure 4.8
Figure 4.9
Figure 4.10
Figure 4.11
Figure 4.12
Figure 4.13
Figure 4.14
Figure 4.15
Figure 4.16
Figure 4.17
Figure 4.18
Figure 4.19
Figure 4.20
Figure 4.21
Figure 4.22
Figure 4.23
Figure 4.24
Figure 4.25
Figure 4.26
Sketch of welding process and coordinate system .

Thermocouple position on weld on bead welded plate ..
Temperature dependence of thermal properties .
Flowchart for moving heat source model .
FEM mesh design ..
Direct method ..
Modified method ..
Newton Raphson iteration method ..
Standard technique .
Temperature distribution for (a) solid rectangle, (b) rectangle with
a hole and (c) rectangle with deactivated elements inside the same
hole as (b) ..
Temperature flux for (a) solid rectangle, (b) rectangle with a hole
and (c) rectangle with deactivated elements inside the same hole
as (b) ..
Birth and death technique .
Temperature histories using standard technique for location of
thermocouple Tc1 & Tc4 (refer to figure 4.2) .
thermocouple Tc2 & Tc5 (refer to figure 4.2) .
thermocouple Tc3 & Tc6 (refer to figure 4.2) ....
thermocouple Tc7 (refer to figure 4.2) ...
thermocouple Tc9 (refer to figure 4.2) ...
Temperature histories using standard and birth & death technique
for location of thermocouple Tc1 & Tc4 (refer to figure 4.2) ....
(a) Weld pool of actual weld and (b) temperature contour from
standard technique .
(a) Weld pool of actual weld and (b) temperature contour from
birth and death technique .
Transverse cross section of weld pool at weld-start and
weld-end points .
Comparison of temperature contour, (a) standard technique,
(b) birth and death technique ..
Heat source model
xi
52
53
54
55
62
62
63
66
69
71
71
72
74
76
77
78
80
80
81
81
82
82
83
83
84
84
85
85
86
87
Figure 5.1
Figure 5.2
Figure 5.3a
Figure 5.3b
Figure 5.4
Figure 5.5
Figure 5.6
Figure 5.7
Figure 5.8
Figure 5.9
Figure 5.10
Figure 5.11
Figure 5.12
Figure 5.13
Figure 5.14
Figure 5.15
Figure 5.16
Figure 5.17
Figure 5.18
Figure 5.19
Figure 6.1
Figure 6.2
Figure 6.3
Figure 6.4
Figure 6.5
Figure 6.6
Figure 6.7
Figure 6.8
Figure 6.9
Figure 6.10
Figure 6.11
Figure 6.12
Figure 6.13
Figure 6.14
Figure 6.15
Pattern of longitudinal distribution of residual stress .

SOLID45 3D-structural elements (Release 12.0 Documentation
for ANSYS ..
Temperature dependence stress-strain diagram for base metal ..
Temperature dependence stress-strain diagram for weld metal .
Displacement distribution for (a) solid rectangle,(b) rectangle with
a hole and (c) rectangle with deactivated elements inside the same
hole as (b) ..
Vector plot of displacement for (a) solid rectangle, (b) rectangle
with a hole and (c) a rectangle with deactivated elements inside
the same hole as (b) .
FEM procedures to predict residual stress distribution ..
FEM longitudinal residual stress (Sz) prediction of bead on
plate welding .
FEM transverse residual stress (Sx) prediction of bead on
plate welding .
FEM longitudinal residual stress (Sz) prediction of bead on
plate welding at the mid cross section area
FEM transverse residual stress (Sx) prediction of bead on
plate welding at the mid cross section area
Residual stresses along the 2mm path in the mid cross section
area .
FEM longitudinal residual stresses (Sz) prediction at symmetry
area in a normal view .
FEM transverse residual stresses (Sx) prediction at symmetry
area in a normal view .
Residual stresses on D2 line
Residual stresses on D3 line
Longitudinal residual stresses (Sz) on BD line
Transverse residual stresses (Sx) on BD line ..
Kinematic hardening vs Isotropic hardening model ..
FEM vs Masubuchis .
CCT diagram for high strength steel (X-70) ..
Geometry and pass sequences of the welding process
Temperature dependent thermal properties of materials .
Temperature dependent mechanical properties of materials .
Effect of martensite on thermal expansion coefficient ..
Effect of austenite to coefficient of thermal expansion ..
Thermo-Mechanical-Metallurgical analysis ..
Coupled thermo metallurgical analysis
Effects of prime and aged martensite to yield stress of base
metal ..
Effects of prime and aged martensite to yield stress of weld
metal ...
Procedure in TMM analysis
Parameter transferred in multi-pass TMM .
ANSYS APDL basic pattern .
TMM ANSYS program .
Saved parameters ....
xii
91
94
95
95
97
98
100
103
103
104
104
105
106
106
107
108
108
109
109
110
116
117
118
118
120
121
123
125
126
127
130
131
133
135
136
Figure 6.16
Figure 6.17
Figure 6.18
Figure 6.19
Figure 6.20
Figure 6.21
Figure 6.22
Figure 6.23
Predicted microstructure configurations .

Longitudinal residual stress distributed on the whole model ..
Longitudinal residual stress distributed on mid cross section area ..
Longitudinal residual stress distributed on the top path
of the mid cross section area
Longitudinal residual stresses without SSPT considerations
Peak temperatures on the top path at mid cross section area .
Peak temperature distributions on the path close to the weld line ...
Residual stress distributions on the path close to the weld line ..
Figure 7.1
Figure 7.2
Figure 7.3
Figure 7.4
Figure 7.5
The rig for girth welding .

Main menu of mechanised girth weld program .
The pendant controller
System Setup menus .
Centering the pipe: a) radial (left) and b) transverse position
(right) .
Figure 7.6 The pipe is tilted from it ideal position: a) ideal position, b) tilted
horizontally, c) tilted vertically and d) tilted arbitrarily
Figure 7.7 The pipe is shifted from it ideal position: a) ideal position,
b) shifted horizontally, c) shifted vertically and d) shifted
arbitrarily .
Figure 7.8 Presenting dial indicator measurements in a diagram: a) radial
position (left side) and transversal position (right) .
Figure 7.9 Using hydraulic jack to minimise hi-lo problem
Figure 7.10 End pipe preparation (gap) and passes design .
Figure 7.11 Camera and its box for welding torch transversal positioning ..
Figure 7.12 Welding arc and gap as seen on camera: a) before welding and
b) when welding is ongoing
Figure 7.13 AMV 4000 .
Figure 7.14 Current probe .
Figure 7.15 STT Lincoln GMAW machine ..
Figure 7.16 The data logger and thermocouples attached on pipes surface .
Figure 7.17 Experiment set up ..
Figure 7.18 Mesh configurations for the girth weld joint .
Figure 7.19 Predicted cross section area (left) and mesh design on the
cross section (right) .
Figure 7.20 Sketch of the girth welding process .
Figure 7.21 Cartesian axes with rotations .
Figure 7.22a A point at (y,z) observed from a coordinate Rx =
..
Figure 7.22b A point at (x,z) observed from a coordinate Ry =
..
Figure 7.22c A point at (x,y) observed from a coordinate Rz =
..
Figure 7.23 Goldaks heat source model in rotated coordinate system .
Figure 7.24 Elements configurations of the girth weld joint .
Figure 7.25 Distributions of temperature when heat source is on moving
Figure 7.26 Temperature history of an element at 6 oclock
Figure 7.27 Filtering identification of peak temperature
Figure 7.28 Microstructure configurations at area close to 6 oclock position
for first pass with (right side) and without (left side) filtering
peak temperature
xiii
139
140
141
141
142
143
144
144
150
151
151
152
153
153
154
154
155
155
156
156
157
157
158
159
159
160
160
161
162
162
163
163
164
164
165
166
167
167
Figure 7.29 Coefficient of thermal expansion for (a) base metal and
(b) filler metal ..
Figure 7.30 Yield strength at elevate temperature .
Figure 7.31 Normalised yield stress at elevated temperature
Figure 7.32 Yield stresses of base metal and filler metal .
Figure 7.33 Effects of prime and aged martensite on yield stress of
base metal .
Figure 7.34 Effects of prime and aged martensite on yield stress of
filler metal .
Figure 7.35 Youngs modulus (E) models for base metal and weld metal
Figure 7.36 Thermal conductivity models for base metal and weld metal
Figure 7.37 Specific heat models for base metal and weld metal .
Figure 7.38 Temperature histories of nodes at varied transverse distances ....
Figure 7.39 Final phase as t8/5 function ..
Figure 7.40 Welding parameters for girth weld for (a) first pass, (b) second
pass, (c) third pass, (d) fourth pass and (e) fifth pass
Figure 7.41 TMM analysis of the girth weld joint
Figure 7.42 Sketch of measuring transversal position of thermocouples (left)
and nodal numbers with transversal position close to the
thermocouples (right)
Figure 7.43 Validation of temperature history at node 47055 for all passes .
Figure 7.47 Mistakes due to thermocouple attachments
Figure 7.48 X70A 3 oclock residual stress chart
Figure 7.49 X70B 3 oclock residual stress chart
Figure 7.50 X70A 9 oclock residual stress chart
Figure 7.51 X70B 9 oclock residual stress chart
Figure 7.52 X70 residual stress unwelded chart ..
Figure 7.53 Welding residual stress at 3 oclock of sample A ...
Figure 7.54 Welding residual stress at 3 oclock of sample B ...
Figure 7.55 Welding residual stress at 9 oclock of sample A ...
Figure 7.56 Welding residual stress at 9 oclock of sample B ...
Figure 7.57 Mean value of welding residual stress at 3 oclock
Figure 7.58 Mean value of welding residual stress at 9 oclock
Figure 7.59 Component of stresses in local coordinate system .
Figure 7.60 Longitudinal residual stress profiles on full and half model ...
Figure 7.61 Longitudinal residual stress on enlarge portion of half model ...
Figure 7.62 Transversal residual stress profiles on full and half model ...
Figure 7.63 Transversal residual stress on enlarge portion of half model ...
Figure 7.64 Normal residual stress profiles on full and half model ...
Figure 7.65 Normal residual stress on enlarge portion of half model ..
Figure 7.66 Residual stress on 3 oclock path ..
Figure 7.67 Residual stress on 9 oclock path ..
Figure 8.1
Figure 8.2
Figure 8.3
Figure 8.4
DC-LSNDI of a thin plate welding ..

hsink at T = 0 ..
hsink at L = 84.904 mm ..
Temperature and hsink distribution at T = 0
xiv
169
170
170
171
172
172
173
173
174
176
178
180
182
183
185
185
186
186
187
189
189
190
190
191
192
192
193
193
194
194
196
197
197
197
198
198
198
199
199
210
212
212
212
Figure 8.5
Figure 8.6
Figure 8.7
Figure 8.8
Figure 8.9
Figure 8.10
Figure 8.11
Figure 8.12
Figure 8.13
Figure 8.14
Figure 8.15
Figure 8.16
Figure 9.1
Figure 9.2
Figure 9.3
Figure 9.4
Figure 9.5
Figure 9.6
Figure 9.7
Figure 9.8
Figure 9.9
Figure 9.10
Figure 9.11
Figure 9.12
Figure 9.13
Figure 9.14
Figure 9.15
Figure 9.16
Figure 9.17
Defines hsink in ANSYS APDL

hsink at the defined area .
Defines Tmedia in ANSYS APDL ...
Tmedia at the defined area
Distributed hsink (W.m-2K-1) of DC-LSND model applied on
the girth weld
Distributed Tmedia ( C) of DC-LSND model applied on the girth
weld
Typical temperature profile of girth welding with DC-LSND .

Residual stress profile at 3 and 9 oclock areas when DC-LSND
is applied
Residual stress at 3 (right) and 9 (left) oclock for girt weld with
and without DC-LSND ..
Temperature histories of nodes with varied transversal position of
girth weld without DC-LSND ..
Temperature histories of nodes with varied transversal position of
girth weld with DC-LSND
.
Evaluating hsink of equation 8.2
Welding phenomenons, a FEM model and an actual condition ..
Excluding volumetric change in SSPT
Residual stress as a result of SSPT by neglecting volumetric
change ..
Excluding mechanical properties change in SSPT .
Residual stress as a result of SSPT by neglecting material
properties change .
Evaluating aspects that affect residual stress ..
Longitudinal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Transversal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Normal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Longitudinal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Transversal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Normal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Longitudinal residual stress at 3 oclock (right) and 9 oclock (left)
areas for a model with annealing temperature equal to 1400 C .
Transversal residual stress at 3 oclock (right) and 9 oclock (left)
Normal residual stress at 3 oclock (right) and 9 oclock (left)
Longitudinal residual stress at 3 oclock (right) and 9 oclock (left)
paths for a model with varied annealing temperature .
Transversal residual stress at 3 oclock (right) and 9 oclock (left)
xv
214
215
215
216
219
220
220
222
223
224
225
226
231
234
235
236
237
238
240
240
240
241
241
241
243
243
243
244
244
Figure 9.18 Normal residual stress at 3 oclock (right) and 9 oclock (left)
Figure 9.19 Temperature history of nodes at layer 3 and layer 4 close to
the evaluated path .
Figure 9.20 Longitudinal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Figure 9.21 Transversal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Figure 9.22 Normal residual stress at 3 oclock (right) and 9 oclock
(left) areas ..
Figure 9.23 Longitudinal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Figure 9.24 Transversal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Figure 9.25 Normal residual stress at 3 oclock (right) and 9 oclock
(left) paths ..
Figure 9.26 Residual stress at 9 oclock (left) and 3 oclock (right) for
model with and without aged martensite SSPT .
Figure 9.27 Stress strain diagram for material model used in
Figure 2.15 (left) and Figure 9.27 (right) ..
Figure 9.28 Three bars model with higher yield stress of middle bar; when the
middle bar is heated (left) and temperature load is omitted (right) ..
Figure 9.29 Stress and strain for yield stress of middle bar equal to 150MPa,
when middle bar is heated (left) and when temperature load
is omitted (right)
Figure 9.30 Stress and strain for yield stress of middle bar equal to 160MPa,
when middle bar heated (left) and when temperature load
omitted (right)
xvi
244
246
247
248
248
249
249
250
250
251
251
252
252
List of Tables
Table 1.1
Table 4.1
Table 4.2
Table 4.3
Table 5.1a
Table 5.1b
Table 5.2
Table 6.1
Table 6.2
Table 6.3
Table 6.4
Table 7.1
Table 7.2
Table 7.3
Table 7.4
Table 7.5
Table 7.6
Table 7.7
Table 7.8
Table 8.1
Next Simdex pipeline project 3

Summary of welding procedure . 61
Thermocouple mean-position 62
Properties of materials as a function of temperature . 63
Base metal properties . 96
Weld metal properties 96
Instantaneous thermal expansion (x 10-6/ C) .. 96
Welding parameters for each passes .. 117
Materials content of base metal and filler metal . 118
Latent heat of base metal and weld metal . 122
Change in material properties due to SSPT .. 129
Suggested parameters for high strength carbon steels 158
Chemical composition of weld metal (wire) and base metal . 169
Mechanical properties of AWS ER706S-6 . 171
Latent heat of the base metal and weld metal 174
Cooling rate of nodes . 177
Parameters for moving heat source model . 181
Thermocouples and nodes transversal position 184
Thermocouples and corresponding FEM nodes . 184
Mathematic formulas for L . 218
xvii
List of Publications
PUBLISHED PAPERS:
1.
Djarot B. Darmadi, John Norrish and Anh Kiet Tieu, Analytic and finite element
solutions for temperature profiles in welding using varied heat source models,
world academy of science, Engineering and Technology (WASET) 81, Singapore,
2011.
2.
Djarot B. Darmadi, Validating the accuracy of heat source model via temperature
histories and temperature field in bead-on-plate welding, International Journal of
Engineering & Technology, Vol.11, No.5, pp. 12 20, 2011.
3.
Djarot B. Darmadi, Study of residual stress mechanism using three elasto-plastic

bars model, Asian Transactions on Engineering (ATE), Volume 01 Issue 5, pp. 76
80, 2011.
4.
Djarot B. Darmadi, Anh Kiet Tieu and John Norrish, A validated thermal model of
bead-on-plate welding, Heat and Mass Transfer, vol.48, pp.1219 1230, 2012.
5.
Djarot B. Darmadi, Muruganant Marimuthu, Kiet Tieu and John Norrish,

Numerical Prediction of Residual Stresses in 316L Bead on Plate Welds, 15th
International Conference on Advances in Material & Processing Technology
(AMPT), Wollongong Australia, 23-26 September 2012, paper id:11355.
6.
Djarot B. Darmadi, Anh Kiet-Tieu and John Norrish, A validated thermomechanical FEM model of bead-on-plate welding, International Journal of
Materials and Product Technology, vol 48. Nos. 1-4, pp.146 166, 2014.
SUBMITTED PAPERS:
1.
Djarot B. Darmadi, John Norrish and Anh Kiet Tieu, Analytic and finite element
solutions for temperature profiles in welding for transient state and quasi steady
state using point heat source model, International Journal of Pressure Vessels and
Piping.
xviii
List of symbols and abbreviations

In some cases, a symbol is used to express different description. Explanation in the text
where the symbol is used will make the meaning of all symbols unambiguous.
Symbol
A
AHsink
b
cp
d
E
gi
G
h
hmax
hmin
hr
hsink
I0
Kd
kmedia
m, n, p
l
lo
m, n, p
q
q
q
qgen
qin
qout
qstr
R, r
St and Sr
Sx, Sy and
Sz
re
Rh
Rx, Ry, Rz
rx,ry,rz
t,
Description
units
area
area of trailing heat sink
width
specific heat
distance between two lattice
young modulus
longitudinal parameters for trailing heat sink
irradiation
convection heat transfer coefficient
maximum coefficient of convection of the trailing heat
sink
minimum coefficient of convection of the trailing heat
sink
overall radiation heat transfer coefficient
convection coefficient of the trailing heat sink
Bessel function first kind, order zero
reduction factor
constant parameters of heat sink
m2
m2
m
J.kg-1.K-1
m
Pa
m
W
W.m-2.K-1
W.m-2.K-1
length
initial length
constant parameters of heat sink
heat
heat rate
heat generated for a certain area
heat generated for a certain volume
heat flux
volumetric heat rate
heat generated in the system
inputted heat to the system
heat come out from the system
heat stored in the system
convection heat transfer rate
maximum volumetric heat rate (at heat source center)
heat transfer rate due to radiation
conduction heat transfer rate in x direction
conduction heat transfer rate in y direction
conduction heat transfer rate in z direction
relative distance to heat source center
tangential and radial component of stresses in
cylindrical coordinate
component of stresses in x,y and z directions
respectively
equivalent radius
a parameter in non dimensional integral
Coordinate rotation
parameters of heat source model
time
m
m
Joule
Joule/s
Joule.m-2
Joule.m-3
W.m-2
W.m-3
Joule
Joule
Joule
Joule
W
W.m-3
W
W
W
W
m
N
xix
W.m-2.K-1
W.m-2.K-1
W.m-2.K-1
-
N
M
M
deg, rad
M
S
T
T0
TCO2
Tmedia
Ts
T ; Tsur
u
V
v
w
z0-DC
zj
Symbol
eff
pre
0
j
max
x; y;
or
L
T
temperature
initial temperature
temperature of CO2 snow in solid state
temperature of cooling media
surface temperature
temperature at position far away from observed
position;
ambient/surrounding temperature.
volume
welding speed
transversal length of heat sink
z global coordinate for DC-LSND artificial center
z global coordinate for an element
K;
K;
K;
K;
K;
K;
Description
Units
thermal diffusivity
coefficient of thermal expansion
absorbtivity
wave length
thermal conductivity
emissivity
strain
thermal efficiency
preliminary thermal efficiency
poisson ratio
angle
angle position of DC-LSND artificial center
angle position of elements
density
Stefan Boltzmann constant
stress
maximum stress
stresses at x, y, z directions
relative position to heat source centre in a longitudinal
direction
relative position to heat source centre in a transverse
direction
m2.s-1
K-1; C-1
nm, m
W.m-1.K-1
m/m
; rad
; rad
; rad
kg.m3
W.m-2.K-4
Pa
Pa
Pa
m
xx
C
C
C
C
C
C
m3
m-1
m.s-1 ; cm.min-1
M
m, mm
m, mm
Abbreviation
APDL
APF
BCC
BCT
BM
CCT
CHT
CTWD
DC-LSND
ERW
FCAW
FCC
FEM
GCS
GIGO
GMAW
GTAW
GUI
HAZ
LBW
Mf
MR
Ms
PWHT
SAW
SCC
SSPT
STT
TIG
TM
TMM
UCS
WFS
WM
WPS
XRD
Meanings
Ansys Parametric Design Language
Atomic Packaging Factor
Body Centred Cubic
Body Centred Tetragonal
Base Metal
Continues Cooling Transformations
Convective Heat Transfer
Contact Tip to Work Distance
Direct Cooling Low Stress No Distortion
Electric Resistance Welding
Flux Cored Arc Welding
Face Centred Cubic
Finite Element Method
Global Coordinate System
Garbage In Garbage Out
Gas Metal Arc Welding
Gas Tungsten Arc Welding
Graphics Users Interface
Heat Affected Zone
Laser Beam Welding
Martensite finish
Melting Rate
Martensite start
Post Weld Heat Treatment
Submerged Arc Welding
Stress Corrosion Cracking
Solid State Phase Transformation
Surface Tension Transfer
Tungsten Inert Gas
Thermo Mechanical
Thermo Metallurgical Mechanical
Users Coordinate System
Wire Feed Speed
Weld Metal
Welding Procedure Specifications
X-Ray Diffraction
xxi
Chapter 1
Introduction
Withers
[1]
discussed the role of residual stress on failure at various loads. Residual
stress can be induced over a range starting from below grain size scale to the complete
structure; it may affect failure mechanism at material level up to structural level. In the
failure criteria for plastic collapse, residual stress is considered as secondary stress and
has no significant effects on the failure. Residual stress has significant effect in fracture,
i.e., catastrophic propagation of a crack. Residual stress generally leads to fracture in
both regimes: linearly elastic fracture mechanics or linearly plastic fracture mechanic,
by altering the stress intensity factor (KI) and the total of integral J respectively.
However it should be noted that the increasing length of crack also alters the
distribution of residual stress.
In fatigue load (mechanical or thermal) the effect of residual stress is more detrimental
in high cycle fatigue (HCF) than low cycle fatigue (LCF). LCF involves fatigue at stress
above yield point and the structure usually fails at less than 10,000 cycles. HCF
involves a larger number of cycles at stress levels insufficient to induce plastic strain.
The role of residual stress in HCF is in shifting stress intensity factor threshold (Kth).
At low stress levels over long periods creep mode dominates the failure mechanism.
Creep strain can be driven by residual stress for instance in the case of thermal
relaxation of weld residual stress when the residual stress is induced in area of poor
material creep ductility at the operating temperature. This mechanism is found in
nuclear reactor structures.
For oil, petrochemical and power generation industries and pipelines, stress corrosion
cracking (SCC) may dominate the failure mechanism. This mode involves stress,
corrosive environment and material susceptibility to the corrosion. Residual stress has
significant effects by changing the stress factor in SCC mode which in turn determines
the failure.
1.1 Background
Welding is a process that is almost universally applied for joining structures, and the
welding of almost all materials can be carried out with recently developed technology.
Controlling the integrity of a structure with a welded joint is extremely important
because the welding process may reduce resistance to failure. One of the most important
phenomena recognised in the welding process that reduces the resistance to failure is
residual stress.
A study of various failed structures indicates the importance of residual stress, indeed
the documented observation of failures since the 1990s clearly indicates that the
majority of cracks that lead to failure were found near the welds [2]. The initiation and
growth rate of cracks are determined by the level of stress and the environmental
conditions. A higher level of stress and a corrosive environment increase the
susceptibility of welded structures to Stress Corrosion Cracking (SCC). Fundamentally
stress can be working stress, residual stress, or a combination of both stresses. Vega and
Hallen claimed that the tendency of SCC to occur adjacent to a weld bead is caused by
high residual stress and the corrosion susceptibility of a HAZ microstructure
[3]
. The
tensile residual stress close to a welded joint generally reaches the tensile yield stress of
the base metal. The corrosion rate of the HAZ in a pipeline weld made by submerged
arc welding (SAW) is almost twice that of the parent pipe and is even higher if electric
resistance welding (ERW) is used[2]. Stress in the girth welded joints of pipelines will be
increased further if displacement loads such as soil movement exist [4].The blowout of
three pipelines in Argentina involved residual stress that triggered SCC failure [5], and
an explosion in the pipelines that transport natural gas to Sydney from central Australia,
and the Trans-Canada pipelines that ruptured at Winnipeg
[6]
occurred for similar
reasons.
According to the British R6 or API 579 standards, regarding the significant effects of
residual stress, it is compulsory to evaluate residual stress when assessing the integrity
of welds [7]. Xu et.al. emphasised the importance of reducing the residual stress because
it leads to premature damage from fatigue as well as stress corrosion cracking and
fracture[8].
Figure 1.1. Australias Pipelines [9]
Table 1.1. Next Simdex pipeline project [11]
As materials transported through pipelines can reduce the long-term cost of

transportation, they continue to be extensively installed
[9]
. In most instances, pipelines
are used to transport liquids and gases. On the African continent, pipelines now cover
North Africa, East Africa, Nigeria, South Africa and West Africa. While on the Asian
continent, pipelines spread across Southeast Asia, South Asia and at East Asia. Middle
Eastern countries also use pipelines to transport their oil. Pipelines also cover Europe
and Russia and former Soviet Union. On the American continent pipelines, cover North
America, Central America, South America and Caribbean. The Enbridge Pipeline
3
System is the longest oil pipeline in the world; it exceeds 5,000 km in length and
stretches from Redwater (Canada) to Wisconsin (USA), and it delivers 1.4 million
barrels of crude oil per day
[10]
. Papua New Guinea, New Zealand and Australia also
used pipelines. Pipelines in Australia are mostly used to transport gas (solid red lines)
and oil (green lines), as shown in Figure 1.1. Despite the huge number of existing
pipelines there are plans for even longer pipelines to be installed, as shown in Table 1.1
[11]
. The plans for future pipelines in Australia are shown as dashed red lines in Figure
1.1.
Two types of welded joints are typically used in the construction of pipelines:
longitudinal welds and girth welded joints. A longitudinal welded joint is a shop
fabricated joint where the welding conditions are well-controlled, whereas girth welding
is carried out in situ. As a perfect welding condition in the field is almost impossible to
achieve, it means that the quality of a girth weld is inferior to that of a longitudinal weld
under controlled condition. Furthermore, residual stress in the longitudinal weld joint of
an individual pipe is easily mitigated using post weld heat treatment (PWHT) but it is
quite difficult to perform PWHT on in situ girth welds.
Since conventional welding depends almost entirely on the skill of the operator and is
therefore difficult to reproduce, welding has traditionally been regarded as a craft rather
than a technology. Recent research trends have been devoted to improving the basic
understanding of the welding process, the efficiency of equipment, consumable, control
and automation [12]. This thesis aims to improve the knowledge on girth welding and the
development of residual stress in a girth weld joint.
Due on the nature of pipeline welding, the girth weld is more susceptible to failure from
the effect of residual stress. FEM was used to determine the residual stress in the girth
weld, and the results have been validated experimentally. The assumptions adopted in
the FEM analysis seem to correctly describe the real welding process, particularly of the
girth weld joint. The model can reveal in some detail how the residual stress develops
and the factors controlling it this can in turn will improve our basic understanding and
lead to better control of the process.
1.2 Research objective

The objective of this research is:
to develop a FEM model that can accurately predict residual stress in a girth welded
joint.
This primary objective can be divided into the following sub-tasks:
-
Developing an accurate FEM model based on simple existing data of a welding

process.
FEM residual stress predictions of X-70 (a ferritic steel) girth weld.
Experimental work to validate the FEM model.
1.3 Research approach

The girth welding process was investigated using a combination of FEM and
experimental method because parameters from the experiment are required as input into
FEM, and the experiment can also validate the accuracy of FEM predictions. In Figure
1.2, dashed arrows show that residual stress is obtained from both experiment or FEM
model and some input data in the FEM model are obtained from the experiment.
Residual Stress
Experiment
FEM
Thermal
Metallurgical
Figure 1.2. Research aspects
Mechanical
In the FEM model, three aspects which are interrelated, namely thermal analysis
(temperature field, temperature histories and shape of the weld pool), metallurgical
analysis (CCT, dilation, phase transformation), and mechanical analysis (residual stress)
are investigated. The fully-coupled Thermo-Metallurgical-Mechanical (TMM) analysis
is shown in the Figure 1.2 and how these aspects affect each other will be discussed in
Chapter 2.
The experimental method provides realistic values to validate the FEM model, but the
experiment only provides limited results and is subject to experimental errors. For
example, although an appropriate array of thermocouple based on knowledge of
geometry and material properties may be able to provide sufficient information, but the
temperature history using thermocouples can only be obtained for the point where the
thermocouples are attached, it does not measure the whole temperature field. In another
example, the distribution of residual stress is localized and not distributed over the
whole area. Sometimes the experiments cannot provide the value of a parameter
because of the difficulties of attaching the sensors at desired positions. While FEM
modeling can add more data to the experimental method it is only a prediction, and the
results still depend on assumptions made during modeling.
The FEM analysis was carried out using ANSYS, a general purpose FEM software
package which provides more control for the users. In this research project, the ANSYS
Parametric Design Language (APDL) mode was used because the user can input a
program from the initial stage (geometry and material properties modeling) whilst
another mode (Graphics Users Interface GUI) can utilise the ANSYS tools and menus
which are already available.
Since all of steps and material model must be programmed instead of using available
tools, menus and data, the APDL model is more suited to the objectives of this research
and allows a better understanding of stress development in the girth weld joint. And also
using APDL more flexible logic can be done to accommodate all the programmer ideas.
However, this type of software with more control features poses problems in that the
user can easily make mistakes without realising that the results may not be reliable.
Being aware of this condition, in this thesis the program was continually tested from its
inception. A comparison to the developed analytical method was carried out for a
6
simple moving point-heat-source in a semi-infinite solid. The results of moving the

point-heat-source model were also compared to the uniformly distributed surface heat
source model
[13]
. A verification of the stress developed by a thermal load was also
carried out for a simple elasto-plastic three bars model
[14]
and a thermal analysis for
simple bead-on-plate has also been carried out[15, 16]. Results of FEM prediction were
also compared to the existing data provided by the European Network on Neutron
Techniques Standardization for Structural Integrity (NeT). A prediction of residual
stress for a bead-on-plate weld was obtained through thermo-mechanical FEM analysis,
which was also validated using existing NeTs data[17]. FEM analysis of plate butt joint
welding involving metallurgical analysis was carried out and is presented in Chapter 6.
The FEM prediction was validated against the experiment results by Lee and Chang [18].
Based on the aforementioned experiences, a prediction of residual stress on the girth
weld was carried out and the results were compared to the experimental work to validate
and judge the accuracy of the model.
Furthermore, an application of the available stress mitigation technique, i.e. DC-LSND,
on the girth weld was also studied through FEM simulation. From the FEM results it
can be predicted that the DC-LSND can potentially be applied to advantage on pipeline
girth welds.
1.4 Thesis outlines

This thesis is sub-divided into five main discussions: a general review (Chapter 2), a
modeling exploration based on existing data (Chapters 3 to 6), the pipeline girth weld
(Chapters7 and 8), and a general discussion and conclusions (Chapter 9 and Chapter
10). At the end of each chapter, there is a list of the references that were used. By
following these listed references readers can trace the researchers steps and gain a
deeper insight into the topic discussed.
In Chapter 2 the basic ideas are discussed briefly, and from chapters 3 to 6 the verified
or validated FEM models are discussed. The experience obtained from those chapters
has provided the author confidence and insight to develop the correct FEM ANSYS
model. While the knowledge and experience gained from Chapter 3 to 6 have led to the
7
prediction of the residual stress at the pipe girth welded joint which is presented in
Chapter 7, where the experiment carried out on a real girth weld is described. The
welding parameters were tested to ensure that the conditions were similar to the FEM
simulation. Chapter 8 predicts the stress mitigation when DC-LSND is applied on the
multi pass girt butt weld joint. Chapter 9 presents an overall discussion and finally, in
Chapter 10 conclusions and recommendation for future work are discussed.
References
[1]
P.J. Withers (2007), Residual stress and its role to failure, Reports on Progress in
Physics, vol. 70, pp.2211 2264.
[2]
John Beavers and Narasi Sridhar (2009), Effects of Steel Microstructure and
Ethanol-Gasoline Blend Ratio on SCC of Ethanol Pipeline, Paper no. 09532, Nace
International Conference.
[3]
O.E. Vega and J.M. Hallen (2009), Slow Strain Rate Corrosion of Multiple
Repairs of Girth Weld in Line Pipes, Revista Mexicana De Fisica, 55(1), pp. 14
19.
[4]
Pablo Fazzini, Jose Luis Otegue & Hernan Kunert (2009), Predicting Failure
Conditions of SMAW Girth Welded X70 Pipelines Subjected to Soil Movement,
World Gas Conference.
[5]
C. Manfredi and J.L. Otegui (2002), Failures by SCC in Burried Pipelines,

Engineering Failure Analysis, Vol.9, pp. 495-509.
[6]
J. Wang and A. Atrens (2003), Microstructure and Grain Boundary

Microanalysis of X70 Pipeline Steel, Journal of Material Science, Vol.38, pp.323330.
[7]
Viorel Deaconu (2007), Finite Element Modeling of Residual Stress A Powerful

Tool in The Aid of Structural Integrity Assessment of Welded Structures, 5th Int.
Conference Structural Integrity of Welded Structures, Romania.
[8]
Jinjin Xu, Ligong Chen and Chunzhen Ni (2007), Effect of Vibratory Weld
Conditioning on The Residual Stresses and Distortion in Multipass Girth-butt
Welded Pipes, Pressure Vessels and Piping, no.84, pp.298-303.
[9]
World Pipeline Maps, http://www.theodora.com/pipeline/world_oil_gas_and_

products_ pilines.html.
[10] Enbridge,
Liquid
Pipelines,
http://www.enbridge.com/DeliveringEnergy
/OurPipelines/LiquidsPipelines.aspx.
[11] Simdex, Future Worldwide Pipeline Project, http://www.simdex.com/future_

pipeline_projects/Simdex_Future_Pipeline_Projects_Worldwide_Guide_
statistcs_web
[12] John Norrish (2006), Advance Welding Processes Technologies and Process
Control, Woodhead Publishing in Materials.
[13] Djarot B. Darmadi, John Norrish and Anh Kiet Tieu (2011), Analytic and Finite
Element Solutions for Temperature Profiles in Welding Using Varied Heat Source
Models, World Academy of Science, Engineering and Technology (WASET) 81,
Singapore.
[14] Djarot B. Darmadi (2011), Study of Residual Stress Mechanism Using Three
Elasto-plastic Bars Model, Asian Transactions on Engineering (ATE), Volume 01
Issue 5, pp. 76 80.
[15] Djarot B. Darmadi (2011), Validating The Accuracy of Heat Source Model via
Temperature Histories and Temperature Field in Bead-on-plate Welding,
International Journal of Engineering & Technology, Vol.11, No.5, pp. 12 20.
[16] Djarot B. Darmadi, Anh Kiet Tieu and John Norrish (2012), A validated thermal
model of bead-on-plate welding, Heat and Mass Transfer, Vol.48, pp. 1219
1230.
[17] Djarot B. Darmadi, Muruganant Marimuthu, Kiet Tieu and John Norrish (2012),
Numerical prediction of residual stress in 316L bead on plate weld, 15th
(AMPT), Wollongong Australia, 23-26 September, paper id: welding - 11355.
[18] Chin-Hyung Lee, Kyong-Ho Chang (2009), Finite element simulation of the
residual stresses in high strength carbon steel butt weld incorporating solid-state
phase transformation, Computational Materials Science, vol.46, pp.1014 1022.
10
PART I
BASIC CONCEPTS
Chapter 2
A Review1
Due to its high joint efficiency, lower weight and air (gas) or water tightness, welding is
one of the most effective methods of joining materials, even though it does have several
drawbacks such as distortion defects and residual stresses. Residual stress is regarded as
the key factor to be considered in the structural integrity
[1]
of a weld as it plays an
important role in crack initiation [2], increases the susceptibility to crack propagation [3, 4]
and reduces component fatigue life
[3, 5]
. Residual stress increases the rate of stress-
corrosion-cracking and can cause brittle fracture
[5]
. Moreover, it is also blamed for
causing distortion which is detrimental to assembly tolerances and product quality [6].
Welding is perhaps the most non-linear problem encountered in structural mechanics
because it involves many disciplines such as heat-transfer, mechanics and metallurgical
science [7]. FEM is a versatile tool to gain a good understanding of the welding process
in relation to the above disciplines.
A literature study showed that many important conclusions can be obtained by using an
FEM to model the welding process. Thermal properties especially thermal conductivity,
contribute significantly to the distribution of temperature, and it is this variation in
temperature distribution that causes residual stress in a welded joint
[8]
. A study on
different welding processes showed that the lower heat imbedded by laser beam welding
(LBW) produces a lower peak temperature than gas tungsten arc welding (GTAW)
especially in positions further away from the centre of the heat source. However, the
peak temperature of LBW at the centre of the weld is higher due to a high concentration
of heat
[9]
. By using FEM, the influence of the mechanical properties, especially the
yield strength of the filler metal to the distribution of residual stress, can be observed
[10]
. FEM can also give an insight into how a solid-state phase transformation
contributes to the residual stress of weldments[11]. Thus FEM simulations can
The materials in this chapter has been published in a paper:

Djarot B. Darmadi, Study of residual stress mechanism using three elasto-plastic bars model, Asian
Transaction on Engineering (ATE), Vol.1 Issue 5, pp. 76-80, 2011.
11
19complement welding experiments in providing the necessary data needed for the
welding procedure specification (WPS) [12].
2.1 FEM simulation of welding processes

Since the 1970s, the scope of welding simulation has been to obtain residual stresses
and the corresponding deformation [12]. Not only does temperature contribute to the final
residual stress, so too does the microstructure. How all these aspects affect the residual
stress can be seen in Figure 2.1 and will be discussed in more detailed below.
Figure 2.1 Development of residual stress in welding
The first process to consider in fusion welding is thermal phenomenon i.e. the heat input
from a welding heat source. This heat is concentrated in a small volume of base metal
which causes fusion in this region. In the early stage, this concentrated heat is released
to the cooler region mainly by conduction, although this release of heat from the base
metal to its surroundings also occurs by convection and radiation. The heat is also
absorbed by the base metal when phase transformation occurs (due to latent heat). The
thermal properties of the base metal determine the rate and value of this heat transfer
phenomenon (arrow 1 in Figure 2.1). By considering the law of energy conservation and
boundary conditions, the transient temperature of certain positions can be obtained. If
the distribution of temperature is the final goal, the analysis is called a thermal analysis
of the welding process.
Further analysis should be carried-out to obtain the distribution of residual stress. Based
on the distribution of temperature in the thermal analysis and considering the coefficient
12
of thermal expansion, thermal deformation can be calculated. Since the temperature is

not uniformly distributed the elongation is also not uniform and as a result thermal
stresses will be produced (arrow 2). The deformation that occurs may also change the
thermal boundary condition, such as the contact body, to the welded base metal. Also,
the plastic and elastic-strain produces heat that affects the temperature of the welded
base metal (arrow 3). Analysis up to this stage is called thermo-mechanical (TM)
analysis.
For base metals that experience a phase transformation, the welding process produces
temperatures high enough to initiate a phase transformation. For instance, when the base
metal is ferritic-steel the ferrite is transformed into austenite when heated and since the
cooling rate of welding is high enough, when the base metal has cooled down
martensitic structures may exist in the final weldments. Thermo-metallurgicalmechanical (TMM) analysis considers this solid state phase transformation
phenomenon. The developed microstructure depends on the temperature history (arrow
4). Different microstructures have specific thermal properties which in turn will affect
the heat transferred to the cooler region or influence the transient temperature of certain
positions (arrow 5). The varied microstructure has a certain thermal expansion
coefficient that determines the deformation. Also, significant expansion or shrinkage
occurs due to the solid state phase transformation. This expansion or shrinkage in turn
affect the developed stress (arrow 6), while Lindgren[12] reported that the developed
stress also influences the microstructure (arrow 7).
2.2 Thermal analysis of welding FEM model

Thermal analysis is a fundamental step in modeling welding phenomenon and any
errors in thermal modeling can invalidate subsequent analysis. The welding process has
been modeled as a moving heat source over a solid. The early analysis used the moving
point heat source
model
[14-16]
[13]
. Some paper on FEM modeling also used the point heat source
to represent the heat input from a welding torch. The heat source may be
modeled as a surface heat source q" (J/m2s) [11, 16-18] which can be distributed uniformly
or according to a Gaussian distribution. The heat source can also be represented as a
volumetric heat source q ' ' ' (J/m3s)
[19-21]
which is a body heat load applied for certain

13
volumes. As with the surface heat source, the volumetric heat source can be distributed
uniformly or according a certain pattern.
The heat input to the base metal by the welding heat source is represented by the
aforementioned moving heat source model. The thermal analysis in Figure 2.1 involves
heat transferred to the cooler regions of base metal while the heat is simultaneously
transferred to the surroundings by convection and radiation.
2.2.1 Thermal properties of a material

Conductivity. Incropera and DeWitt[22] defined conductivity as a material property that
describes the transport of thermal energy. If heat is only transferred in the x-direction
. For a certain temperature gradient a
then conductivity can be expressed as:
higher conductivity will increase the heat flux for each unit of time. The other two axes
can be treated similarly. For isotropic material, the conductivity along the three axes is
equal. The actual conductivity for certain materials such as metal depends on the
temperatures. Typical temperature dependent conductivity will be described in Chapters
4, 6 and 7 where FEM material models are discussed.
Specific heat. Specific heat is defined as the energy required to change the temperature
of a unit mass of material by one degree. The higher the specific heat is, the harder it is
to increase its temperature. Cengel and Boles
[23]
described specific heat as the capacity
of materials to store energy. The value of specific heat for certain materials depends on
the temperature and the typical temperatures dependent specific heat are shown in
Chapters 4, 6 and 7.
Enthalpy. Enthalpy (h) was first called the heat content and was proposed by Mollier. It
represents a combination of properties: u + PV, where u is the internal energy stored in
a system and PV is the work done by a system. For phase transformation stages,
enthalpy is also known as latent heat. Enthalpy became important in the welding process
since it involve phase transformation from solid to liquid and from liquid to solid.
Coefficient of convection. The coefficient of convection h, describes the conditions of
the boundary layer on the interface between fluid and solid. A higher h means more heat
is transferred from higher to lower temperatures. Two factors that have a significant
influence on h are the surface geometry and the nature of fluid motion.
14
Emissivity. Emissivity is a radiated property of a material surface. The higher the

emissivity the more is the energy emitted by the radiation. An ideal radiator, that is a
perfect black body, has an emissivity value equal to 1. Typically, a surface has a value
of emissivity below unity.
2.2.2 Heat transfer and heat conservation

As discussed previously, there are three modes of heat transfer: conduction, convection
and radiation; and all three exist simultaneously in the welding process.
Conduction is heat transfer within a material due to differences in temperature. The
energy (heat) is transferred from a higher temperature to a lower temperature. Fouriers
law expressed this heat transfer phenomenon by the following expressions:
(2.1)
(2.2)
(2.3)
= conduction heat transfer rate in the x, y and z directions
where
respectively (W)
= thermal conductivity
A = perpendicular area (m2)
= temperature gradient in the x, y and z directions respectively ( C/m)
Convection is heat transferred by the movement of a fluid. As in conduction, the
difference in temperature between solid and fluid is a prerequisite for convection where
the energy (heat) is transferred from a higher temperature to a lower one. This energy is
transferred by random molecular motion (which is usually called diffusion) and the
relative motion of the bulk fluid. The convection heat transfer can be expressed by the
following expression:
(2.4)
15
where
= heat transfer rate due to convection (W)

h = convection heat transfer coefficient
A = area of contact (m2)
Ts = temperature of solid surface ( C)
T = temperature of fluid far enough from the surface ( C)
Radiation is heat transport by electromagnetic waves. The radiation heat transfer rate
can be expressed by the Stefan Boltzmann law:
(2.5)
where
= heat transfer rate due to radiation (W)

= emissivity ( 0 < < 1)
= Stefan Boltzmann constant
A = surface area (m2)
Ts = surface temperature ( C)
Generally a surface emits and receives radiation; the received radiation is usually called
irradiation, which in turn is reflected, transmitted, and absorbed. The absorbed
irradiation is expressed as:
Gabs = G
where
(2.6)
= absorbtivity of surface ( 0 <
<1)
G = irradiation from others (W)

The value of
depends on the surface properties and the nature of irradiation.
The net radiated heat is the difference between radiation and irradiation which is
expressed as:
(2.7)
where Tsur is the surrounding temperature of the evaluated surface.
16
In welding both convection and radiation occur, so it is convenient to express radiation

in a linear form, similar to convection.
=
=
=
= hr A(Ts - T )
(2.8)
hr is usually called the overall radiation heat transfer coefficient. It should be noted that
hr depends on the thermal properties of the materials and also on the temperature. The
FEM tool (ANSYS) which will be used to analyse residual stress in this thesis has
facility to model non linear radiative heat loss, but this linear model is useful in
analytical stage. Some researchers (see chapter 4) have used this overall heat transfer
coefficient and the above discussion gives some background on this parameter.
In welding, the transient conduction heat transfer problem is connected with a
convection boundary condition on the surfaces of the base metal. If radiation is also
taken into account, the heat transfer by radiation is added to that by convection and it is
preferable to use the overall radiation heat transfer coefficient hr (equation 2.8) since the
heat transfer rate can be expressed as a linear function (instead of power 4) of
temperature difference between the solid surface and the surrounding air.
Whatever the form of heat transfer (conduction, convection, and radiation, or a
combination of these), it always follows the energy conservation law. For a certain
system, the energy conservation law is expressed as:
qin + qgen qout qstr = 0
which for convenience is expressed as:
qin qout + qgen = qstr
where
qin = input heat to the system (Joule)

qout = heat coming out from the system (Joule)
qgen = heat generated in the system (Joule)
qstr = heat stored in the system (Joule)
17
(2.9)
Equation 2.9 is usually called a conservation energy equation. Heat stored in the system
can be written in the following expression:
qstr = cpdTdV
(2.10)
= density (kg/m3)
where
cp = specific heat at constant pressure (J/kg.K)

dV = volume increments (m3)
Applying energy conservation to the incremental volume with sides: dx, dy and dz
results in the following expression:
where q is heat generated for a certain volume. Substitute
in the x-
direction and similarly for the two other axes and expand conduction heat transfer
equation with a constant thermal conductivity can be written as:
(2.11)
where
is known as thermal diffusivity (m2/s).
The temperature in a solid can be determined from equation 2.11 subject to the
boundary conditions of natural or forced convection and radiation.
2.2.3 Parametric study via FEM

In this chapter the simplest moving heat source model (point heat source) is used. How
parameters contribute to the distribution of temperature is demonstrated using an FEM
model. Typical welding parameters and material properties are: heat load rate
4200J/s, welding speed v = 6cm/minute, thermal conductivity
heat cp = 525 (J/kg. C) and density
= 42 W.m/ C, specific
= 8000 kg/m3. These parameters were applied on
thin plate (thickness equal to 1 mm) of 175mm long and 120mm width. At this stage,
adiabatic boundary condition was applied on surfaces to simplify and focus the problem
18
on the effect of parameters. SOLID70 brick elements with 1mm edge length were used
in FEM modeling as shown in Figure 2.2. Isothermal temperatures were evaluated on
top surface of the model and the results are studied and presented in Figures 2.3 2.5.
Figure 2.2. Finite element model for parametric study.
The first parameter considered is the intensity of the heat source at two levels 3150J/s
and 4200 J/s. The origin describes the instantaneous position of the heat source. All the
positions are considered to be relative from the position of the heat source. The heat
source at this stage is modeled as a point heat source and moves from left to the origin.
Isothermal lines for 1200 C, 600 C, 400 C, 150 C and 100 C are plotted after a certain
time, as shown in Figure 2.3. Using a lower heat input results in the isothermal lines
covering a narrower area than for a higher heat input. The longer time means that the
difference between the same isothermal lines is wider. This phenomenon is found in all
time spans up to 200s where the condition reaches a quasi steady state. Once the heat
source had traversed far enough, the shape of the isotherms are steady. This condition is
known as quasi steady state. The quasi steady state will be discussed in more detail in
Chapter 3.
19
(a)
(b)
(c)
(d)
Figure 2.3. Transient isothermal lines for moving heat source with varied heat; (a) 10s,
(b)50s, (c)100s and (d)200s
Figure 2.4 shows the distribution of temperature for different thermal conductivity. The
lines show an isothermal line at 1200 C, 600 C, 400 C, 150 C and 100 C respectively.
The lower thermal conductivity (30 W.m/ C) is used to compare with the previous
value of 42 W.m/ C. It can be seen here that the lower thermal conductivity is
significantly different to the typical parameter in the longitudinal direction, but not in
the transverse direction. A lower thermal conductivity keeps the high temperature close
to the weld line longer than the higher value and as a result Figure 2.4 shows isothermal
lines stretching in the longitudinal direction for certain time.
Figure 2.5 shows the temperature field for different specific heats. For a higher specific
heat 700 (J/kgC), the isothermal lines differ significantly in the transverse direction (y
axis) whilst in the longitudinal direction the isothermal lines cover almost the same
length. With a higher specific heat, a higher heat is needed to increase the temperature
in the transverse direction, but since the heat is kept constant, the isothermal lines for
the higher heat, covers a narrower width than the lower value.
20
42
30
42
30
(a)
(b)
42
30
42
30
(c)
(d)
Figure 2.4. Transient isothermal lines for moving heat source with varied thermal
conductivity; (a) 10s, (b)50s, (c)100s and (d)200s
(a)
(b)
(c)
(d)
Figure 2.5. Transient isothermal lines for moving heat source with varied specific heat;
(a) 10s, (b)50s, (c)100s and (d)200s
21
2.3 Development of stress in welding

The distribution of temperature in the welding process is not uniform, which causes a
non-uniform thermal strain in the welded structure. This strain in turn causes thermal
stress, as a result of misfit which grows until a uniform temperature at the welded
structure is attained. The stress which still exists when the temperature of the welded
structure is already uniform is called residual stress. The simplest explanation of how
residual stress has developed in the welded structure is the three bars model, which will
be presented in the next discussion.
Figure 2.6 Thermal stress in the three bars model (modified from Sindo Kou [24])
Consider the three bars which are clamped together at their ends with a rigid body, as
shown in Figure 2.6. First, as described in Figure 2.6a, the middle bar is heated and
elongates due to thermal expansion. A compressive stress is exerted because its
elongation is constrained by the side bars. Typically in a welding process the
compressive stress exceeds the yield stress and as a result a plastic compressive strain is
developed at the middle bar.
Next, as shown in Figure 2.6b, the middle bar is left to cool by transferring the heat to
the environment. While it is cooling the middle bar shrinks according to its coefficient
22
of thermal expansion but because compressive plastic strain has already developed in
the middle bar, when it cools down to its initial temperature it is now shorter than its
initial length. The final tensile stress is developed in the middle bar as a result of misfit.
In a welded structure, the weld metal and adjacent base metal can be considered as the
middle bar in the three bars model, and areas further away from the weld metal that are
not significantly affected by heat from the welding torch, can be considered as side bars.
The final residual stress in the area close to the weld metal is in tension whilst in the
area further from weld metal is in compression. Typical residual stress of butt joints
obtained from experiments is shown in Figure 2.7.
Figure 2.7. Typical distribution of residual stress in butt joint welding: (a) longitudinal
residual stress, (b) transversal residual stress.
The mechanical properties that contribute to the residual stress are thermal expansion,
yield stress, and the stress-strain diagram for the base metal. The residual stress
determined by both FEM and an analytical calculation for the very simple three bars
model will be discussed in the next section.
2.3.1 A review of the three bars model

Many theories have been proposed to explain the development of thermal stress in the
welding process. The simplest mechanism of the formation of residual stress is
23
proposed by Wells
[25]
and quoted by Lin and Chou
[26]
. The schematic diagram in
Figure 2.8, modified from [26], shows a typical thermal cycle of the region close to the
fusion line in the welding. As the region is heated from room temperature (point O) the
volume expands, but this expansion is restrained by the surrounding bulk of cold
material which first causes elastic compressive stress (point A). Further heating will
develop plastic compressive stress (point B). It should be noted that the compressive
stress at point B is lower than point A as a results of the higher temperature. When this
region begins to cool, this plastic compressive stress still exists and because of misfit an
elastic tensile stress is developed (point C). Further cooling to room temperature will
result in the plastic tensile stress (point D) known as residual stress (also called thermal
residual stress).
Figure 2.8. Thermal residual stress formation
The coefficient of volumetric thermal expansion as a material property is expressed by

equation 2.12.
(2.12)
V is denoting volume and T is temperature.
If only a simple one dimensional case is studied, equation 2.12 can be expressed as the
following expression:
(2.13)
where
material
is coefficient of thermal expansion, l is length of heated material. For isotropic

=
v/3.
If the coefficient of thermal expansion is temperature independence,
equation 2.13 can be simply expressed as the following expression:

24
(2.14)
where lT is final length as a result of temperature rise, l0 is initial length and
T is
temperature rise.
Following the aforementioned theory, the development of residual stresses via
analytical and FEM solutions will be demonstrated here. It is first modeled as a three
bar model, which is shown in Figure 2.9. The three bars are connected at their ends with
two other rigid bars. First, the middle bar is heated to a certain temperature and the
longitudinal thermal stresses that have developed are then analysed. In the next step the
previous temperature load is omitted and the final residual stresses are determined. The
three bars have a 1mm x 1mm cross section and are 10mm long. Their typical properties
are: density
= 8000 (kg/m3), thermal conductivity
= 20 (W/m. C), specific heat cp =
500 (J/kg. C), Youngs modulus E = 15GPa, Poissons ratio
= 0.3, thermal expansion
= 20.10-6 / C. The two bars that clamp the ends of the three bars have the same
properties, except that
= 20.10-6 (W/m. C), = 20.10-20/ C and E = 10.1010 GPa. With
a much lower k it is expected the elevated temperature will only be localised in the
middle bar. The very low means that no thermal expansion exists and a high E is to
model the perfect toughness of the clamps.
Figure 2.9. Three bars model
25
Figure 2.10. Strain stress diagram of the material model by ANSYS
An elasto-plastic material model was used and to avoid the problem of convergence in
the plastic state, a very low E was applied as shown in Figure 2.10 which is represented
in Figure 2.11. Subscript pl refers to the plastic region and subscript y refers to
yield. The first 400 C of temperature applied on the middle bar will give a strain under
the elastic limit. When the temperature load is omitted there should be no stresses left
on the system. The analytical solution which will give an understanding of the thermal
phenomenon follows the flow chart in Figure 2.12. The analytic solution only considers
force and elongation parallel to the three bars and one dimensional analysis was taken
Epl
y
E
y
Figure 2.11. Sketch of the stress-strain diagram.
26
for simplicity. The flow chart follows the equilibrium force law, and since the crosssection of the three bars are the same, the equilibrium force also means equilibrium
stress. Also the flow chart considers the aforementioned residual stress mechanism
theory. Firstly, the tentative stress at the side bars is taken arbitrarily, and by using the
aforementioned mechanism of residual stress, the stress in the middle bar can be
calculated. The stress in the middle bar is checked and should be twice the stress at the
side bars. If this condition is not fulfilled, the tentative stresses in the side bars are
corrected. These steps are repeated iteratively until the condition is met. It should be
noted that the stress in the side bars should be doubled since there are two bars (2
md).
sd
Subscript sd refer to side bar and md refer to middle bar. The analysis for
thermal stress is done when the middle bar is heated and when the temperature load is
omitted. Following the flow chart at Figure 2.12, the stresses when the mid bar is heated
to 400 C are
md
= - 79.58 MPa (compression) and
sd
= 39.79 (tension). The
distribution of stress that resulted from the FEM simulation is shown in Figure 2.13.
The bottom half of the model is shown in Figure 2.13b. The thermal stress in the middle
bar is -79.49 MPa and 39.88 MPa in the side bar. The difference with the analytical
solution may be caused by excluding expansion in the lateral direction as a result of the
temperature load and stress load. Also, the analytical solution does not accommodate
the distribution of stress at the ends of the bars which causes localised high strain in that
region. Although the localised stress may have no effect on the stress at the middle cross
section area (since it is further away) but localised strain will give a different total
length, which is used to determine the theoretical developed stress as shown by Figure
2.12. From the FEM side, the rigid clamping bar is modeled by a high Youngs modulus
which is still not perfectly rigid. However, the difference between the analytical
solutions with FEM simulation is low (0.11%). When the temperature load is omitted,
there are no stresses left on the three bars since no plastic strain existed when the system
was heated initially.
27
Figure 2.12. Flow-chart for evaluating thermal stress in elastic strain state.
(a)
(b)
Figure 2.13. Distribution of stress when the middle bar is heated to 400 C (a) full model
and (b) half model.
28
Figure 2.14. Flow-chart for evaluating thermal stress in plastic strain state.
(a)
(b)
Figure 2.15. Distribution of tress at the three bars model heated to 1100 C. (a) when the
middle bar is heated, (b) when the temperature load is omitted.
29
Figure 2.16. Flow-chart for evaluating the residual stress.

The following analysis demonstrates the three bar model where plastic deformation has
occurred, when the middle bar is heated to 1100 C. The analytical solution for thermal
stress when the middle bar is heated follows the flow chart in Figure 2.14. Following
the flow-chart, when the three bars are heated the stress at the middle bar and side bars
is
md
= -150.36 MPa and
sd
= 75.18 MPa respectively. The FEM simulation produces
a distribution of stress as shown in Figure 2.15a, with a stress at the middle bar of
md
- 150.12 MPa, which shows an insignificant difference with the analytical model
(0.16%).
When the temperature load is removed, there should be a distribution of stress as a
result of previous compressive plastic strain when the system is heated. The FEM
simulation for the remaining stress (residual stress) shows that the stresses are
68.22 MPa and
sd
md
=-34.10 MPa for the middle bar and side bars respectively. The
analytical solution for the residual stress follows the flow chart in Figure 2.16.
Following the flow chart, the residual stress at the middle bar is 67.73 MPa and -33.87
MPa at the side bars, which again shows a small difference with the FEM simulation
(0.72%).
30
Overall, the FEM simulation and analytical solution show good agreement. The residual
tensile stress at the middle bar and side bars can be determined by both methods. The
analytical solution followed assumptions made by Wells
[25]
and the equilibrium stress
theory. The confirmations from the FEM results have demonstrated the proposed
mechanism of residual stress formation.
2.4. Solid state phase transformation and stress

development
As demonstrated in section 2.3 the stress arises as a result of various strains. In section
2.3 only strains resulting from elevated temperature is discussed. When this strain is
constrained it causes stress. It should be noted that since in welding the temperature
distribution is not uniform the thermal strain will also not be uniform and the elongation
in the heated area is constraint by far bulk material which is not significantly elongated
by the thermal stress. Thus it no needs the presence of physical constraint to produce
stress in welding process although the presence of physical constraint alters the
distribution of residual stress. The stress induced by this thermal load can be classified
as elastic and plastic strains. The plastic strain which is irreversible significantly
contributes to the final residual stress since it causes misfit when the welded structure
cooled to the room temperature.
When welding is applied on ferritic steels, it does not only produce thermal, elastic and
plastic strains but also creep, volume change and also transformation plasticity strains as
result of phase transformation. Comprehensive discussion can be found in literatures [27
34]
which is partially discussed below.
Creep strain is not significant in the welding residual stress analysis and usually ignored
[29]
and the total strain can be expressed as in equation 2.15.

Tot
Tot
is total strain,
El
is elastic strain,
=
Pl
El
Pl
Th
is plastic strain,
TrV
Th
is thermal strain,
strain due to volumetric change of phase transformation and
TrP
TrP
(2.15 )
TrV
is
is transformation
plasticity. Concerning strain as result of phase transformation, volumetric strain and

transformation plasticity will be discussed briefly in this section.
31
Since the welding process exposes a welded material to such a high temperature (above
A1), a solid state phase transformation exists for parent materials such as feritic-steels.
When a ferritic-steel is heated to a critical temperature A1, it is transformed into an
austenite, a transformation that is completed at a critical temperature A3. The atomic
structure of feritic-steels is body-centred-cubic (BCC) whilst the atomic-structure of
austenitic-steels is face-centred-cubic (FCC). Since the total amount of atoms is
constant, this atomic arrangement affects the volume of the transformed steels. A non
dimensional parameter called the atomic packaging factor (APF) describes the volume
fraction occupied by atoms.
Mathematically, the APF is expressed as the atomic
volume per total occupied volume. For simplicity, the atoms are described as a solid
ball. The higher the APF the denser the atom which means the volume is less for a
certain number of atoms. The APF of BCC is 0.68 whilst for FCC it is 0.74 which
means when the ferrite transforms to austenite the transformed steel shrinks. It should
be noted that 0.74 is the highest possible APF. This transformation is described with the
solid line in Figure 2.17.
Figure 2.17 Solid state transformations in ferritic steels.

When austenite is cooled to the Ms temperature, it begins to transform to martensite,
and this transformation is finished when the temperature is equal to Mf. Martensite has
a body-centred-tetragonal (BCT) structure with APF equal to 0.67. With this lower
APF, the transformed steel have a higher volume, as described in Figure 2.17. Since the
shrinkage and expansion of the phase transformation affects the thermal stress it should
be considered in the analysis. A more detailed analysis of phase transformation and its
effect on the models will be discussed in Chapter 6.
32
Transformation plasticity is anomalous plastic behavior of steels when a phase

transformation takes place. The plastic strain depends on proportion of phases and
stress. According to Greenwood and Johnson [35] the microscopic plasticity is generated
in the phase with lower yield stress in the presence of stress. The stress permits
macroscopic plastic flow even if the load itself is insufficient to produce plasticity.
Another mechanism is proposed by Magee
[36]
which is only applicable for the
martensitic transformation. According to Magee steels under stress during martensitic

transformations will form martensite plates and this affects the overall shape of the
continuum. The stress causes transformation strains which are at the microscopic scale
and do not average out to zero. No matter which mechanism is followed (GreewoodJohnson or Magee), it can be seen that the presence of residual stress in welding process
while phase transformation takes place causes additional plastic strain which is called
transformation plasticity. This transformation plasticity will be discussed in more
detailed in chapter 7.
33
References
[1]

Tool in The Aid of Structural Integrity Assessment of Welded Structures, 5th Int.
Conference Structural Integrity of Welded Structures, pp.1 - 9, Romania.
[2]
Chiew Sing Ping, Jiang Jin, Lee Chi King & Yu Yi (2011), Effects of Residual
Stress Distributions of High Strength Steel Plate-to-Plate T/Y Joints on Stress
Concentration Factor, 10th International Conference on Steel, Space and
Composite Structures, pp. 1 9, North Cyprus, Turkey.
[3]
G. Ivetic, A. Lanciotti and C. Polese (2009), Electric Strain Gauge Measurement

of Residual Stress in Welded Panels, J. Strain Analysis Vol. 44, pp. 117 - 126.
[4]
Z. Barsoum (2008), Residual Stress Analysis and Fatigue of Multi-pass Welded

Tubular Structures, Engineering Failure Analysis, Vol. 15, pp. 863 874.
[5]
G. Cam, O. Ozdemir and M. Kocak (2010), Progress in Low Transformation

Temperature (LTT) Filler Wires: Review, 63rd Annual Assembly & International
Conference of the IIW, pp. 759 765, Istambul, Turkey.
[6]
Amin Esmaeilzadeh, Mohammad Sadeghi & Farhad Kolahan (2011), Using FEM
for Prediction of Thermal Post-Buckling Behavior of Thin Plates During Welding
Process, World Academy of Science, Engineering and Technology, vol. 74,
pp.780 785.
[7]
P. Marcal (1974), Weld Problems, Structural

Charlottesville, University Press, pp. 191 206.
[8]
E. Armentani, R. Esposito and R. Sepe (2007), The effect of thermal properties

and weld efficiency on residual stresses in welding, Journal of Achievements in
Materials and Manufacturing Engineering, vol. 20, issues 1-2.
[9]
Hwa Teng Lee and Jia Lin Wu (2009), Correlation between corrosion resistance
properties and thermal cycles experienced by gas arc welding and laser beam
welding alloy 690 butt weldments, Corrosion Science, vol. 51, pp.733 743.
Mechanics
Programs,
[10] Dean Deng, Hidekazu Murakawa & Wei Liang (2008), Numerical and
experimental investigations on welding residual stress in multi-pass butt-welded
austenitic stainless steel pipe, Computational Materials Science, Vol.42, pp. 234
244.
[11] Deng Dean & Murakawa Hidekazu (2006), Prediction of welding residual stress
in multi-pass butt-welded modified 9Cr-1Mo steel pipe considering phase
transformation effects, Computational Materials Science, vol.37, pp.209-219.
34
[12] Lars-Erik Lindgren (2001), Finite element modeling and simulation of welding
part 1: increased complexity, Journal of Thermal Stresses, vol.24, pp.141 192.
[13] Daniel Rosenthal (1941), Mathematical theory of heat distribution during welding
and cutting, Welding Journal, 1941, vol.20, pp 220s 234s.
[14] M. Van Elsen, M. Baelmans, P. Mercelis and J.P. Kurth (2007), Solution for
modeling moving heat source in a semi-infinite medium and application to laser
material processing, International Journal of Heat and Mass Transfer, vol. 50,
pp.4872 4882.
[15] Dragi Stamenkovic & Ivana Vasovic (2009), Finite element analysis of residual
stress in butt welding two similar plates, Scientific Technical Review, vol.59,
no.1, pp. 57 60.
[16] Djarot B. Darmadi, John Norrish and Anh Kiet Tieu (2011), Analytic and Finite
Element Solutions for Temperature Profiles in Welding Using Varied Heat Source
Models, World Academy of Science, Engineering and Technology (WASET) 81,
Singapore.
[17] Z. Chai, H. Zhao and A. Lu (2003), Efficient finite element approach for modeling
of actual welded structures, Science and Technology of Welding and Joining,
vol.8, no.3, pp. 195 204.
[18] Fenggui Lu, Shun Yao, Songnian Lou and Yongbing Li (2004), Modeling and
finite element analysis on GTAW arc and weld pool, Computational Materials
Science, Vol.29, pp.371 -278.
[19] Lei Yu-cheng, Yu Wen-xia, Li Chai-hui and Cheng Xiao-nong (2006), Simulation
on temperature field of TIG welding of cooper without preheating, Transaction of
Nonferrous Metals Society of China, Vol.16, pp.838 842.
[20] X. Shan, C.M. Davies, T. Wangsdan, N.P.ODowd and K.M. Nikbin (2009),
Thermo-mechanical modeling of a single-bead-on-plate weld using the finite
element method, International Journal of Pressure Vessels and Piping, vol. 86,
pp.110 121.
[21] Djarot B. Darmadi (2011), Validating The Accuracy of Heat Source Model via
Temperature Histories and Temperature Field in Bead-on-plate Welding,
International Journal of Engineering & Technology, Vol.11, No.5, pp. 12 20.
[22] Frank P. Incropera and David P. DeWitt (2002), Fundamentals of heat and mass
transfer, John Willey & Sons, USA.
35
[23] Yunus A. Cengel and Michael A. Boles (2005), Thermodynamics and engineering
approach, McGraw-Hill, New York.
[24] Sindo Kou (2003), Welding metallurgy, John Willey and Sons, New Jersey.
[25] AA. Wells (1953), The mechanics of brittle fracture, Welding Journal, vol.32,
pp.34 -56.
[26] YC. Lin and CP. Chou (1995), A new technique for reducing the residual stress
induced by welding in type 304 stainless steel, Journal of Materials Processing
Technology, vol.48, pp. 693 698.
[27] A.S. Oddy, J.A. Goldak and J.M.J. McDill (1992), Transformation plasticity and
residual stresses in single-pass repair welds, Journal of Pressure Vessel
Technology, vol. 114, pp. 33 38.
[28] H. Dai, J.A. Francis, H.J. Stone, H.K.D.H. Bhadeshia and P.J. Withers (2008),
Characterizing phase transformation and their effects on ferritic weld residual
stresses with x-rays and Neutrons, Metallurgical and Material Transaction A, vol.
39A, pp. 3070 3078.
[29] H. Dai, J.A. Francis and P.J. Withers (2010), Prediction of residual stress
distributions for single weld beads deposited on to SA508 steel including phase
transformation effects, Materials Science and Technology, vol.26, no.8, pp.940
949.
[30] J.A. Francis, H.K.D.H. Bhadeshia and P.J. Withers (2007), Review Welding
residual stresses in ferritic power plant steels, Materials Science and Technology,
vol.23, no.9, pp. 1009 1020.
[31] A.F. Mark, J.A. Francis, H. Dai, M. Turski, P.R. Hurrell, S.K. Bae, J.R.
Kornmeier and P.J. Withers (2012), On the evolution of local material properties
and residual stress in a three-pass SA508 steel weld, Acta Materialia, vol.60, pp.
3262 3278.
[32] Dean Deng and Hidekazy Murakawa (2008), Prediction of welding distortion and
residual stress in a thin plate butt-welded joint, Computational Materials Science,
vol. 43, pp. 353 365.
[33] Greenwood, G.W. and Johnson, R.H. (1965), The deformation of metals under
small stresses during phase transformation, Proc. Roy. Soc., A283, p.403.
[34] Magee, C.L (1966), Transformation kinetics, microplasticity and aging of
martensite in Fe-31 Ni, Ph.D. Thesis, Carnegie Institute of Technology,
Pittsburgh.
36
PART II
FEM
Vs
ANALYTICAL SOLUTION
Chapter 3
Moving Heat Source Simulations2
Basically, the finite element method (FEM) considers that a structure is constructed
from simple elements which are connected at their nodes, and fulfill equilibrium and
compatibility conditions. Based on this definition, the first step in any FEM analysis is
to divide an observed structure into elements.
Dividing a structure into discrete elements was pioneered by Alexander Hrennikoff
[1]
O. Zienkiewicz (1947) established a mathematical formulation for discrete elements [2].

Since a typical structure consists of a huge number of elements, a large number of
equations must be solved and so involvement with a computer cannot be avoided. The
first software to use the finite element method was NASTRAN, which was established
by NASA in 1965. Today, many software packages are available and in this thesis
ANSYS was used to carry out the finite element analysis.
One of the first applications of the finite element methods to welding can be found in a
1970s in a published paper by Tall [3]. In his paper the thermal analysis was modeled in
a two dimensions whilst the mechanical analysis was essentially a one dimensional
model. Friedman
[4]
calculated the temperature, stress, and distortion, in a two-
dimensional analysis. A constant temperature profile was used as a thermal load which
moves along the weld line. Muraki T. et al.[5] developed a thermo-mechanical finite
element computer program by using an elasto-plastic material model. In the 1970s
other papers were published where a two dimensional model was used
[6 - 9]
. The
possibility of a three dimensional model was restricted by the availability of the

required computing power. Under these limited conditions, plane strain
[10-14]
, plane
stress[15-19] and axis symmetry[20,21] considerations were made in the approach to three
dimensional modelling.
2
Material in this chapter has been published in a paper:

Djarot B. Darmadi, John Norrish and Anh Kiet Tieu, Analytic and finite element solutions for
temperatures profiles in welding using varied heat source models, World Academy of Science,
Engineering and Technology (WASET) 81, Singapore, 2011.
37
The most important aspect of modeling is accuracy which required an integrated

approach which combines the FEM, analytical and experimental method. How the FEM
and the analytical and experiment methods are linked to each other is shown in
Figure3.1.
Ve
rif
ica
tio
n
Numerical Model
(Finite Element)
Analytic Solution
Qualification
Va
lid
at
io
n
Experiment
Figure 3.1 FEM, analytical solution and experimental results.

From Figure 3.1 it can be seen that, FEM is verified by analytical solutions and
validated with experimental measurements. Most published papers that are concerned
with FEM modeling involved validation to judge the correctness of those models. A few
papers, especially on welding simulation used analytical solutions to verify the results
obtained from FEM analysis.
Rahman and Brust evaluated the propagation of cracks in welded pipes. They compared
their proposed solution, which is based on the energy release rate theory, with the FEM
model
[22]
. Anca et al.[23] compared the distribution of temperature obtained from the
analytical method and FEM to predict the liquid region, the mushy region, and the solid
region in the welding processes. Lin et al. evaluated the distribution of stress
surrounding spot weld regions using a mathematical model and FEM
[24]
. Analytical
solutions for temperature profiles using the point heat source and uniform disc surface
heat source model obtained from the mathematical approach were verified with FEM
results [25].
The analytical solution typically solves simple conditions with an exact solution which
is considered to be correct, while FEM gives approximate solutions but can solve
complex conditions (especially geometry). Both may be subject to possible errors in the
assumptions made concerning material properties. Thus, verification can only be done
on simple conditions where analytical solutions can be obtained. Validation compares
the FEM results with experiment measurements. In the FEM model assumptions are
38
made to simplify the real conditions which are complex. A good FEM model can
provide results that are close to both analytical solution and experimental
measurements. This chapter presents verifications that were carried out to ensure a
reliable approach was developed in the early stages of FEM modeling.
3.1 Solutions by Rosenthal

Rosenthal[26] proposed a solution for the moving point heat source which is expressed in
equation (3.1), where T is the rise in temperature, v is the moving speed of the heat
source, and
describes the thermal diffusivity of the solid. This proposed solution is
based on the shape of the weld-pool. Since equation (3.1) is not a time (t) function, it is
a solution for the quasi steady state. A quasi steady state is a steady state if the
parameters are considered from moving coordinates ( ,y,z). The fixed coordinate
system is expressed as (x,y,z), which follows the right hand rule. The coordinates are
described in Figure 3.2. Since the heat source is moving parallel to the x axis, the value
of y is equal to y and the value of z is equal to z.
(3.1)
Equation (3.1) consists of the asymmetric function:
function:
and symmetric
. The asymmetric function is found along the direction parallel to
and the symmetric function is found along lines that are parallel to y and parallel to z. If
the welding speed equals zero, the asymmetric function will be a unity and only the
symmetric function is left. Zero welding speed means a case of heat liberated by a
stationary point which is much easier to obtain a solution for than the moving point heat
source. A final solution proposed by Rosenthal for the moving point heat source on a
semi-infinite solid is expressed in equation (3.2) where
is the conductivity, and R is
the distance from the heat source. T0 is added to account for the initial temperature of
the welded plate. The symmetric function of equation (3.2) is
39
Figure 3.2 Fixed coordinate system and moving coordinate system

(3.2)
Equation (3.2) is a solution for a quasi steady state condition since it is time
independent.
3.2 Moving point heat source model

The moving point heat source is the simplest way to model welding phenomena. Two
analytical solutions are used in this thesis; the first is Rosenthals equation which is
already available, and new analytical solutions based on a solution to the conservation
of energy proposed by H.S. Carslaw and J.C. Jaeger[27]. The solution by Carslaw and
Jaeger in this thesis was developed using a non-dimensional integral techniques to
obtain the solution for a moving point heat source. The non-dimensional integral
follows the method used by R. Komanduri and Z.B. Hou[28].
Analytical solutions to welding heat flow problems were obtained by solving the partial
differential equation of energy conservation (equation 3.3), where T is the temperature,
x,y and z are the three mutually orthogonal directions,
is the diffusivity, and t is the
time. Steady state solutions can be obtained by allowing time t
.
(3.3)
As previously discussed, D. Rosenthal[26] developed quasi steady state solutions for a

40
moving heat source by observing the distribution of temperature around coordinates

which coincided with the moving heat source. Theoretically, after a certain time has
elapsed, the temperature at a given position relative to the moving coordinate is steady
and therefore the condition is called quasi steady state. R. Komanduri and Z.B. Hou[28]
developed a non-dimensional integral to carry out thermal analysis in welding. They
used a Gaussian distributed moving disc heat source to represent the heat load
embedded by the welding torch. A solution for the non-dimensional integration is
obtained through the numerical approach.
3.3 Solutions using non-dimensional integral to

compensate for transient conditions
Carslaw and Jaeger
[27]
have proposed a solution for an instantaneous point heat source
in infinite solid that liberates heat at (xo,yo,zo) as expressed in (3.4).
is a function
which depends on the heat load and material thermal properties.

(3.4)
Komanduri and Hou
[28]
obtained solutions for the distribution of temperature around
the Gaussian distributed disc heat source as it is expressed in (3.5).

(3.5)
where
and I0(p) is a modified Bessel function first kind, order zero. Since
,
is a non-
dimensional term, equation (3.5) is called a non-dimensional integral.

The solution for the moving point heat source is based on (3.4), but if the heat source is
liberated at the origin, equation (3.4) will be simpler as expressed in (3.6).
(3.6)
The total heat for an infinite solid liberated by the heat source can be expressed as in
(3.7), where
represents density and cp is the specific heat.

41
(3.7)
Equation (3.8) is obtained by substituting
considering that for isotropic materials
and
are not a function of x, y and z. Before

is analysed first. Figure 3.3 describes
solving equation (3.8), the solution for

the nature of
T in equation (3.6) to equation (3.7) and
is obtained by calculating the total area under the curve
in Figure 3.3, which is plotted in Figure 3.4. The integral value rapidly converges to
0.886227
. Equation (3.8) is expressing equation (3.7) in a different form.

(3.8)
Figure 3.3 Plotted value for
Figure 3.4 Plotted value for
42
Considering
, an integral term in equation (3.8) can be simplify as
follows:
=
=
The same values are obtained for
and
. The total heat
stored in an infinite solid can then be simplified as in equation (3.9).

=
(3.9)
Considering the total stored heat is equal to the inputted heat q = dt, the value of
now can be determined as
and substituting this value to equation (3.6) yields
equation (3.10).
(3.10)
The rise in temperature resulting from a moving point heat source initially at origin
(0,0,0), and moving parallel to the x axis at a constant speed (Figure 3.5), can be
obtained based on equation (3.10). Equation (3.10) can be interpreted as the rise in
temperature in an infinite solid due to instantaneous heat liberated when t = 0 at the
origin. When t 0 the temperatures rise at all points are zero except at the origin, which
is infinite. (Note: for small values of t at origin x = y = z = 0 the value of
will equal to 1. When t is small the values of
will be very big and
when t 0 it will be infinite and then the combined results will be infinite. For different
place (not at origin) and small t the value of
will be very small and when
t 0 it will be equal to zero. Anything times zero is equal to zero and thus the combined
result is equal to zero. It is one of the shortcomings of point load, even in FEM it is
suggested to transformed point load to surface or volumetric load).
43
Figure 3.5 Point heat source in an infinite media
Equation (3.11) is obtained by adjusting equation (3.10) to the moving coordinates

which coincides with the moving heat source, and where the temperature rise in a small
time increment is expressed as dT.
(3.11)
The total temperature rise at any points can be obtained by integrating equation (3.11).
Considering the initial temperature as T0, the temperature rise is expressed as in (3.12).
(3.12)
This integration can be expressed in a non-dimensional form by substituting the
following expressions:
,
and equation (3.13) is obtained.

(3.13)
For a semi-infinite solid case as in welding, the temperature obtained by equation (3.13)
should be doubled since the heated volume is half that of an infinite solid. (The semi
infinite case applies for fusion welding since the heat source is always applied on a
surface of large material. In case of a large thick, the plate may consider as an infinite
solid. Of course this case study neglect convection + radiation and adiabatic surfaces are
considered as a boundary condition as modeled in section 3.5). For convenience c is
replaced by / and the final temperatures for any points in the welding model can be
represented by equation (3.14).
44
(3.14)
The values for the definite integral in equation (3.14) depend on the value of the upper
limit, which is a time function. Clearly, equation (3.14) is the solution for a transient
state and the results for a quasi steady state can be found by replacing the value of t
, thus the upper integration limit is equal to
. Since a singularity is found when
0, the solution cannot be solved analytically but numerically by setting the lower limit
close to 0 (e.g. 0.001). The integral expression in (14) is plotted and presented
graphically in Figure 3.6. From this figure it can be concluded that when u
3 the
values of the integral are numerically closer to zero which means there is no significant
elevation in temperature. It can also be drawn from Figure 3.6 that at a higher u value,
(which also means a
the definite integration value converges at a higher value of

longer time). For lower u values, it was found that when
5 the value of the integral
converged, thus it is reasonable if the upper limit of integration of 5 is adopted, which

makes the numerical solution much easier to obtain.
Figure 3.6 Integral values for varied

Instead of using
and u
when the steady state condition is achieved, as shown by the
temperature profile in Figures 3.7 and 3.8, the graphics in Figure 3.7 for a longitudinal
path and Figure 3.8 for a transverse path were composed using the analytical solution
expressed in equation (3.14). The legends t10s, t50s, t100s, t200s, t1000s in the legend
box mean temperature profiles when elapsed time equal to 10s, 50s, 100s, 200s, 1000s
respectively that coincide with non-dimensional parameters
45
= 0.25, 1.25, 2.5, 5 and
250 respectively. A longitudinal path with y = 0 and z = 10mm was chosen to describe
the temperature profiles in Figure 3.7.
Figure 3.7 Temperature profiles along a longitudinal path.
Figure 3.8 Temperature profiles along a transversal path.
From Figure 3.7 it can be seen that the temperature profile for 200s coincides with that
for 10000s. A transverse path with
= 0 and z = 10mm was chosen to describe the
temperature profiles in Figure 3.8. The temperature profile at the transversal path, from
Figure 3.8, show that the temperature profiles were already superimposed to each other
when t = 100s. These facts confirm the previous assumption that at
= 5 the
temperature profile at the moving point heat source has already converged, and it is
quite reasonable to adopt 5 as the upper limit of integration of equation (3.14).
46
3.4 Element model (SOLID70)

Basically, the element model describes the relationship between the geometry of a
model with its degree of freedom. The SOLID70 element is suitable for a 3D thermal
analysis model. Typically, the element is comprised of eight nodes with the temperature
as the degree of freedom at each node. However, if a node is formed as a coalescence of
nodes, the element can consist of four, five, or six nodes and a tetrahedral element, a
pyramid element, or a prismatic element are formed as shown in Figure 3.9.
Figure 3.9. SOLID70 3D-thermal element (Release 12.0 Documentation for ANSYS)
3.5 Finite element model for a moving point heat

source
Finite element analysis was carried out using ANSYS software. There are two options
which can be used: ANSYS Parametric Design Language (APDL) and the Graphic User
Interface (GUI). In this thesis the APDL mode was chosen because it provides
flexibility and a greater ease of modification.
The model consists of 23328 SOLID70 thermal elements with 26011 nodes to represent
a 250mm x 180mm x 90mm block. The heat source moves along the x abscissa with a
certain velocity. In APDL it is programmed by changing the position of point heat
source along the line (x,0,0) with changes in the x value according to time (Figure 3.10).
Typical welding parameters were chosen as following: the heat rate
thermal conductivity
= 42 W/(m.C ), thermal diffusivity

47
= 10 mm2/s, and welding
speed v = 1mm/s. In this chapter adiabatic boundary condition was applied on surfaces
means no convection or radiation was involved.
Figure 3.10 FEM model for point heat source
The moving point heat source was modeled by a heat source at a certain point, and heat
is liberated for a certain time, depending on the welding speed and distance between
two consecutive nodes. This distance is determined by the mesh size of the FEM model,
and then after a chosen duration, the heat source is omitted and relocated to the next
position where it liberates heat for the chosen duration. This procedure is repeated to the
end of the weld length. The temperature profiles were observed after the heat source
travels 200mm length (200s). The flowchart of the FEM model is shown in Figure 3.11.
Since
in this position is equal to 5, the temperature profiles can be considered as quasi
steady state for comparison with the analytical solution.
48
Modeling
Material thermal
properties
Geometric model
Element formation
(meshing)
Simulation of Moving Point Heat Source

(Finding solutions)
Heat load at a point for certain time increment
Solve the transient state problem
Defined next point

Previous heat load is omitted
Applied heat load to the next point for certain time
increment.
No
End of
the weld?
Yes
Last heat load is omitted
Post Processing
(Interpreting solutions)
Temperature profiles around heat source center
Figure 3.11 Flow chart of the FEM simulation
3.6 Analytical model vs FE model quasi steady

state
The results from using a point heat source model are presented in Figure 3.12 for the
longitudinal lines and in Figure 3.13 for the transverse lines. The temperature profiles
are evaluated at z = 0mm (at the surface of the plate), z = 10mm, z = 20mm and z =
49
30mm (below the surface). The temperatures were evaluated on these lines by both the
analytical solution and FEM, and the results are shown graphically in Figure 3.12 and
3.13. The analytical solution referred to here, is the solution provided by the nondimensional integral as it was represented in equation (3.14) while the temperatures
were obtained by making the upper limit of the definite integral equal to 5. The solid
lines describe solutions provided by the analytical solution, while the markers describe
the solution provided by FEM. From Figures 3.12 and 3.13 it can be seen that FEM
results are in good agreement with the analytic solutions, especially for those points
which are far enough away from the point heat source.
Figure 3.12 Temperature profiles on longitudinal lines with moving point heat source
model
Figure 3.12 represents the temperature profiles at y = 0 with varied z values (see Figure
3.2 for the axis nomenclature). The analytical solutions and FEM are in better
agreement for temperature data behind the heat source model. Both the analytical
solutions and FEM can describe the peak temperature which lags for higher values of z
(the peak temperature is at negative
values), as it is found in practice, since a longer

50
time is needed to transfer heat from the heat source to higher z values (due to the longer
distance). Meanwhile for a certain negative value of
the heat that is liberated from
previous positions has reached a point on the observed line with the same negative
value. The transverse line shown in Figure 3.13 are lines where
= 0 with varied z
values. Symmetrical graphs are shown in Figure 3.13. This symmetry has emerged
because of equal
by the e-
values. From equation (3.14) an unsymmetrical function is provided
term and it only has an effect when
is varied because the value of V
(V=v/(2 )) depends on the welding parameters: welding speed (v) and thermal
diffusivity,
which are considered constant.
The peak temperatures in Figure 3.13 should have the same values with temperatures at
= 0 in Figure 3.12 since they are in same location. It should however be noted that the
peak temperatures in Figure 3.13 are not the peak temperatures in Figure 3.12 (except
for z = 0) as a result of the lagging effect discussed previously and peak temperatures in
Figure 3.12 are in different depend on distance to the point heat source.
Figure 3.13 Temperature profiles on transversal lines with moving point heat source
model
51
3.7 Analytical models, Rosenthal solution and

FE model quasi steady state
The Rosenthal solution provides a temperature field for the quasi steady state. The quasi
steady state for analytical models (equation 3.14) can be obtained by setting the upper
limit
of the definite integral greater than 5. The solutions obtained from those three
models were compared to each other and are presented in Figure 3.14. The observed
paths are the same as in Section 3.6: a path on the surface, and then 10mm, 20mm, and
30mm below the surface. From Figure 3.14 it can be seen that all those models give
relatively good agreements. However Figure 3.14 for longitudinal paths shows some
deviation, especially for the position furthest behind the centre of the weld torch (further
than -0.12m). Figure 3.14 is enlarged in Figure 3.15 for
further than -0.12m. From
Figure 3.15 it can be said that the solution from the analytical model developed in this
chapter (equation 3.14) gives an even closer solution to the FEM model than the
solution obtained from Rosenthals equation at the far field behind the centre of weld
torch for the quasi steady state.
Figure 3.14 Solutions from analytic, FEM and Rosenthal for temperature profiles at
longitudinal paths (left) and transversal paths (right).
52
Figure 3.15 Enlarged solutions for far field longitudinal paths.
3.8 Analytical models vs FE model

transient conditions
The obvious advantage of a non-dimensional integral (equation 3.14) over the Rosenthal
solution (equation 3.2) is the possibility of providing a solution for the transient state.
The transient state had solutions before a steady state was established (before
5).
The data used are the same as in Section 3.6. The temperature distribution at the surface
of the plate at 10s, 50s, 100s and 200s is presented in Figure 3.16, whilst that through
the cross sectional area is presented in Figure 3.17. The distribution of temperature is
plotted in the isothermal lines of 1200 C, 600 C, 400 C, 150 C and 100 C. Considering
the data used, these elapsed times are equivalent to
= 0.25, 1.25, 2.5 and 5
respectively. The data obtained using FEM (red dashed lines) were compared to the
analytical results (solid blue lines)
53
Figure 3.16 Isothermal profiles at plate surface for transient state

Overall the finite element results show a good agreement with the non-dimensional
integral method. In Figure 3.16, both methods demonstrate that higher elapsed times
will cause higher asymmetric curves for both results. The asymmetric function of
equation (3.14)
can be expressed as
. This asymmetric
which depends on the value of the upper
expression is magnified by
limits of the definite integral. By evaluating the transient condition of Figure 3.16
(before the quasi steady state is achieved), it can be concluded that a longer time t
results in higher values of the definite integral, which in turn produces a more
asymmetric curve.
The temperature profiles on the transverse section (yz plane, refer to Figure 3.2) for
the transient state are shown in Figure 3.17. The results obtained from the finite element
simulation and analytical solution show a symmetrical plot over the origin. From the
perspective of an analytical solution, this result reflects the distribution of temperature
which is expressed by equation 3.14. For = 0 the temperature because of heat liberated
by the heat source is expressed as
which can also be

54
expressed as
. Only y and z at the last expression describe the
position and function as symmetrical for y and z, thus it is a symmetrical function.

These symmetrical results are also confirmed by the FEM results that have shown good
agreement with the analytical solutions.
The temperature profiles at time t = 100 or
t = 200 or
= 2.5 are almost the same as those at time
= 5 which indicates that the temperature have already converged before
= 5. Thus the temperature profiles for the transverse direction converged earlier than
those in a longitudinal direction. This phenomenon is also implicitly described in Figure
3.16 and 3.17.
Figure 3.17 Isothermal profiles at plate cross section for transient state
3.9 Conclusions
In this chapter a simple FEM model is compared with an analytic solution. The
transient state and steady state of the temperature profile in welding can be described
very well by the expression
. This new solution is
developed based on the energy conservation law for a moving point heat source in
infinite solids and follows non-dimensional integration technique of Komanduri [28]. The
55
improvement of this solution over the current Rosenthal equation is that the equation
can describe the profile of the transient state temperature. This transient or steady state
can be determined by the upper limit
of the integral term. When the limit
is close to
5 the quasi steady state is reached, and below this value the condition is considered as
transient. For a steady state evaluation, the solution also shows a better agreement to the
finite element simulation than Rosenthals for longitudinal paths, especially at a
distance further away from the moving point heat source.
The good agreement between FEM and analytic solutions means the developed code
according to the flow-chart in Figure 3.11 is able to represent a moving heat source
phenomenon especially the point heat source. This code can be used as a basis for a
more complex moving heat source in the future.
56
References
[1]
A. Hrennikoff (1941), Solution of Problems of Elasticity by the Frame-Work

Method. ASME J. Appl. Mech. 8, A619A715.
[2]
E. Stein (2009), Olgierd C. Zienkiewicz, a pioneer in the development of the finite

element method in engineering science. Steel Construction, 2 (4), 264-272.
[3]
L. Tall (1964), Residual Stresses in Welded Plates a theoretical study, Welding

Journal, vol. 43, no.1, pp. 10s 23s, 1964.
[4]
E. Friedman (1975), Thermomechanical analysis of the welding process using the

finite element method, Journal of Pressure Vessel Technology, vol. 97, pp. 206213.
[5]
Muraki T, Bryan J and Masubuchi K (1975), Analysis of thermal stresses and

metal movement during welding, ASME Journal of Engineering Materials
Technology, vol. 96, pp. 81 84.
[6]
T. Iwaki and K. Masubuchi (1971), Thermo-elastic analysis of orthotropic plastic

by the finite element method, Japan Society of Naval Architects and Ocean
Engineers, vol. 130, pp. 195-204.
[7]
Y. Ueda and T. Yamakawa (1971), Thermal stress analysis of metals with

temperature dependent mechanical properties, Proceeding International
Conferences Mechanical Behavior of Materials, p.10.
[8]
Y. Ueda and T. Yamakawa (1971), Mechanical cracking characteristics of

cracking of welded joints, Proc. of 1st Int. Symposium on Precaution of Cracking
in Welded Structures Based on Recent Theoretical and Practical Knowledge, The
Japan Welding Socity, Tokyo, p. IC5.1.
[9] Y. Fujita, Y. Takeshi, M. Kitamura and T. Nomoto (1972), Welding stresses with
special reference to cracking, IIW Doc X-655-72.
[10] V.J. Papazoglou and K. Masubuchi (1982), Numerical analysis of thermal stresses
during welding including phase transformation effects, ASME J. Pressure Vessel
Technology, vol. 104, pp. 198-203.
[11] C.K. Leung and R.J. Pick (1986), The Use of Generalized Plane Strain Elements
in the Prediction of Residual Stresses in Welded Flat Plates, Proceeding of the
Symposium on The Computer Modeling of Fabrication Processes and Constitutive
Behavior of Metals, p.563.
[12] L. Wikander, L. Karlsson, L.E. Lindgren and A.S. Oddy (1993), Plane thermomechanical finite element modeling of welding with special reference to the
material behavior, International Conference on Modeling and Control of Joining
Processes.
57
[13] L. Wikander, L. Karlsson, M. Nasstrom and P. Webster (1994), Finite element

simulation and measurements of welding residual stresses, Modeling and
Simulation in Materials Science and Engineering, vol.2, pp. 845-864.
[14] J.B. Roelens (1995), Numerical simulation of multipass submerged arc welding
determination of residual stresses and comparison with experimental
measurements, Welding in the World, vol. 35, no.2, pp 110 117.
[15] Y. Ueda and . Yamakawa (1971), Analysis of thermal elastic-plastic stress and
strain during welding by finite element method, JWRI, vol.2, no.2, pp. 90 100.
[16] B.A.B. Andersson (1978), Thermal Stresses in a Submerged-Arc Welded Joint
Considering Phase Transformations, ASME, J. Engineering Materials and
Technology, vol.100, pp. 356 362.
[17] B. Andersson and L. Karlsson (1981), Thermal stresses in large butt-welded
plates, Journal of Thermal Stresses, vol.4, pp.491 500.
[18] K.W. Mahin, S. MacEwen, W. Winters, W. Mason, M. Kanouff and E.A. Fuchs
(1988), Evaluation of residual stress distribution in a traveling GTA weld using
finite element and experimental techniques, Proce. Modeling of Casting and
Welding Processes IV, The Minerlas, Metals & Materials Society, p.339.
[19] N. Siva Prasad and T.K. Sankaranarayanan (1996), Estimation of residual stresses
in weldments using adative grids, Computer & Structures, vol. 60, no. 6, pp. 1037
1045.
[20] H.D. Hibbit and P.V. Marcal (1973), A Numerical Thermo-Mechanical Model for
the Welding and Subsequent Loading of a Fabricated Structure, Computers &
Structures, vol.3 pp. 1145 1174.
[21] T. Inoue and Z.G.Wang (1983), High temperature behavior of steels with phase
transformation and the simulation of quenching and welding processes, Proc. Of
4th International Conference Mechanical Behaviour of Materials, p. 1015, 1983.
[22] S. Rahman and F.W. Brust (1992), An estimation method for evaluating energy
release rates of circumferential through-wall cracked pipe welds, Engineering
Fracture Mechanics, vol.43, no.3, pp. 417 430.
[23] A. Anca, A. Cardona and J.M. Risso (2004), 3D thermo-mechanical simulation of
welding processes, Mecanica Computational, vol. 23, pp.1 18.
[24] P.C. Lin, S.H. Lin and J. Pan (2006), Modeling of failure near spot weld in lapshear specimens based on a plane stress rigid inclusion analysis, Engineering
Fracture Mechanics, vol. 73, pp. 2229 2249.
[25] Djarot B. Darmadi, John Norrish and Anh Kiet Tieiu (2011), Analytic and finite
element solutions for temperature profiles in welding using varied heat source
58
models, Proc. World Academy of Science, Engineering and Technology, vol. 81,
pp. 154 162.
[27] H.S. Carslaw and J.C. Jaeger (1959), Conduction of Heat in Solid. Clarendon
Press, Oxford.
[28] R. Komanduri and Z.B. Hou (2000), Thermal Analysis of the Arch Welding
Process: Part I. General Solutions. Metallurgical and Material Transaction B,
Vol. 31B. pp. 1353 1370.
59
PART III
MODELING EXPLORATION
BASED ON EXISTING DATA
Chapter 4
FEM Thermal Analysis of NeT
TG1 Specimen3
Welding applications subject materials to cycles of non-uniform temperature that in turn
causes problems which may lead to premature fatigue damage, stress corrosion and
fracture
[1- 4]
. Studies of the distributions of transient temperature in weldments are
important for quality control and they can be used as an aid to the development of
welding procedures [5].
Rosenthal [6] proposed an analytical solution for a moving point heat source that can be
applied to the welding process. Following this, the analytical solutions for moving heat
sources attracted much attention
[7-10]
. More recently the high speed computational
resources needed for numerical analysis have become more readily available and FEM
has become a popular tool for weld modeling. A clear advantage of numerical analysis
over the analytical method is its ability to obtain a solution for complex geometry and
boundary conditions which may be difficult to achieve using analytical solutions.
A major issue in modeling is accuracy and validity
[11-14]
. To this end the European
Network on Neutron Techniques Standardization for Structural Integrity (NeT) has

published
experimental
data
and
procedures
which
can
be
accessed
at
https://odin.jrc.ec.europa.eu[15,16]. Experimental work was carried out using bead-onplate welding where nine thermocouples were attached at different measured points.
Using the NeTs experimental data, a FEM thermal analysis is carried out. At this stage,
the logic and parameters just follow some previous published papers with some
modification when necessary. Experience on how to compose a FEM model in ANSYS
Material in this chapter has been published in papers:

1. Djarot B. Darmadi, Validating the accuracy of heat source model via temperature histories and
temperature field in bead-on-plate welding, International Journal of Engineering & Technology,
Vol.11, No.5, pp.12 20, 2011.
2. Djarot B. Darmadi, Anh Kiet Tieu and John Norrish, A validated thermal model of bead-on-plate
welding, Heat and Mass Transfer, Vol.48, pp. 1219-1230, 2012.
60
APDL code is believed to be valuable contribution to the development of the model of

multi pass girth welds in ferritic steel pipe lines.
A new heat source model using a combination of Goldaks double ellipsoid and uniform
temperature was used. The uniform temperature represents the melting filler metal
which forms the weld bead and the Goldaks heat source model represents the heat that
was transferred to the base metal.
4.1 Welding procedure provided by NeT

A complete welding procedure was obtained at https://odin.jrc.ec.europa.eu[16]. These
welding procedures are summarised on Table 4.1. Four identical plate specimens (called
A11, A12, A21, and A22) were produced. Following the NeT procedures, a sketch
diagram for the welding set up and thermocouple positions are presented in Figures 4.1
and 4.2 respectively. The coordinate system here is following that used by NeT. It
should be noted that the origin of the coordinate system is where the weld stops. The
positions of the thermocouples based on the coordinate system chosen, are tabulated in
Table 4.2.
Table 4.1. Summary of welding procedure.
Parent material:
Size: 120 x 180 x 17mm
AISI Type 316L Austenitic Stainless Steels
Residual stress is relieved using heat treatment
Filler:
0.8mm
Type: 316S96
Spec: A5.9.93 (ASME); ER316H
Welding parameters:
Welding process: Tungsten Inert Gas (TIG).
Type: Bead-on-plate
Weld length: 60mm
Heat Input: 633 J/mm
Welding speed: 2.27 mm/s
Four sets of temperature histories for the four specimens were obtained from nine
thermocouples. As thermocouple Tc8 was not pushed far enough into the hole and there
is no clear reference to where the data was obtained, the Tc8 thermocouple was not
considered. For thermocouple Tc9, the data from specimen A21 were also excluded
61
since they showed a lower temperature and were inconsistent with data from other
specimens [17].
.
Figure 4.1. Sketch of welding process and coordinate system.
Figure 4.2. Thermocouple position on weld on bead welded plate.

Table 4.2. Thermocouple mean-position [16].
Thermocouple
Mean position (mm)

X
T1
8.50
1.24
60
T2
8.00
1.20
30
T3
7.50
1.24
T4
11.50
1.21
60
T5
11.50
1.23
30
T6
12.00
1.18
T7
10.50
15
T8
4.67
46
T9
17.00
30
62
The temperature dependent properties of the base metal and welding filler were
obtained from NeT
[16]
. The material properties for the base metal and weld metal are
shown in Table 4.3. The density and Poissons ratio for both materials are 7966 kg/m3
and 0.294 respectively. In Figure 4.3 the thermal properties for both metals are shown
graphically.
Table 4.3. Properties of materials as a function of temperature [16].
Specific Heat, cp
Temperature
kJ/(kg C)
Conductivity,
Thermal expansion,
Young's Modulus
(Gpa)
W/(m C)
-1
x 10 ( C )
Weld
Parent
Weld & Parent
Parent
Weld
20
0.488
0.492
14.12
14.56
Weld & Parent
195.6
171.0
100
0.502
0.502
15.26
15.39
191.2
165.0
200
0.520
0.514
16.69
16.21
185.7
157.5
300
0.537
0.526
18.11
16.86
179.6
150.0
400
0.555
0.538
19.54
17.37
172.6
142.5
500
0.572
0.550
20.96
17.78
164.5
135.0
600
0.589
0.562
22.38
18.12
155.0
127.5
700
0.589
0.575
23.81
18.43
144.1
120.0
800
0.589
0.587
25.23
18.72
131.4
109.0
900
0.589
0.599
26.66
18.99
116.8
96.9
1000
0.589
0.611
28.08
19.27
100.0
83.0
1100
0.589
0.623
29.50
19.53
80.0
66.4
1200
0.589
0.635
30.93
19.79
57.0
47.3
1300
0.589
0.647
32.35
20.02
30.0
25.0
1400
0.589
0.659
33.78
20.21
2.0
1.7
Figure 4.3. Temperature dependence of thermal properties [16].

63
4.2 Previous work in modeling NeT welding

procedures
There are two main techniques used in weld modeling, the standard technique and the
birth-and-death technique. The standard technique models basically deal with the weld
bead in situ, whilst the element birth-and-death technique considers the growing weld
bead. Further details can be found in Sections 4.3.3 and 4.3.4
There are at least eight participants involved in the NeT weld modeling exercise: The
British Energy-Frazer-Nash Consultancy, UK
[18]
modeled the welding process using
ABAQUS. The moving heat source is modeled as Goldaks ellipsoidal distributed heat
source. Two different sets of parameters for the heat source were used: rx = 0.7mm, ry =
2.8mm, rz = 2.8mm and rx = 0.4mm, ry = 1.6mm, rz = 1.6mm with the heat efficiency set
to 75% and 60% respectively. Different dwell times were used for each model at the
starting point of the weld; with1.5s dwell for the first model and 0.7s dwell for the
second model. Convection was represented by hconv = 4.2W/m2K at 100 C and hconv =
11.8 W/m2K at 1000 C. Radiation was modeled using an emissivity of 0.4. The
element birth and death technique were used for both models.
Imperial College, UK also used the ABAQUS code for the welding simulation
[19]
. The
standard and the element birth and death techniques were used in conjunction with a
moving Goldaks ellipsoidal heat source model. For the standard technique the heat
source parameters were: rx = 2.8mm, ry = 3.2mm, rz = 2.9mm while for the element
birth and death techniques rx = 1.4mm, ry = 3.4mm, rz = 2.9mm. The dwell at the
starting point of the weld start point was set to 1s. The convective heat transfer (CHT)
coefficient is 5.7 W/m2K and radiation emissivity was set to 0.75. The heat efficiency
is 100%.
INSA-Framatone, France used the SYSWELD code with three different moving heat
source models: Gaussian surface heat source, conical heat source and prismatic heat
source
[20]
. The heat efficiency was 75%, the CHT coefficient 15W/m2C and the
radiation emissivity 0.7.

The Joint Research Centre, Institute for Energy, Petten, Netherlands
[21]
used the
ANSYS code. The moving heat source was modeled with two different models: the
melting temperature field and uniform heat generated field. The melting temperature
64
was loaded at a prescribed melted zone while the heat rate for the uniform heat
generated field was applied at a prescribed melting zone. The heat efficiency was set at
70%. The CHT coefficient was set to 10 W/m2C and no radiation was considered.
Serco, a Technical and Assurance Service, UK [18] used the SYSWELD code with
Goldaks double ellipsoidal moving heat source model rx1 = 1.6mm and rx2 = 3.2mm, ry
= 1.6mm, rz = 3.2mm. (Serco made 2 models with 2 different heat source model; the
first is Goldaks heat source with rx = 1.6mm, ry = 1.6mm , rz = 3.2mm and the second is
Goldaks heat source with rx = 3.2mm, ry = 1.6mm, rz = 3.2mm).The heat efficiency was
set to 80% while the overall convective coefficient was 13.57 W/m2K at 100 C.
Three other participants also made contributions: The Institute for Nuclear Research
Romania, Korea Power Engineering Company and University of the West of England,
UK
[17]
. The Institute for Nuclear Research used ANSYS code. Uniform heat was
generated at nodes inside the melted zone. Convection was modeled with a convective
coefficient of 10W/m2K . The Korea Power Engineering Company used the ABAQUS
code with the heat source modeled as a combination of heat flux and body heat load.
The heat flux was set to 60% and the body heat load set to 40% of the total heat. The
element birth and death technique was used. The CHT coefficient was 10 W/m2K and
the heat efficiency 75%. The University of the West of England also used the ABAQUS
code with a uniform moving heat source in the melted zone. The overall convective
coefficient at 1000 C was 89.2 W/m2K while the CHT coefficient was 10 W/m2K and
the radiation emissivity 0.4. The heat efficiency was 80%.
From the above, it can be seen that the assumed welding efficiency varied from 60% to
100%. The participants used various welding efficiencies to describe the heat absorbed
by the plate. Three participants used Goldaks heat source model, but the parameters (rx,
ry and rz) differed, depending on the researchers assumptions. At the starting point of
the weld, the participants used dwell times that varied, from 0 s (no dwell) to 1.5 s. The
cooling mode due to convection and radiation were modeled using a combination of the
following: CHT coefficient, radiation emissivity, or the overall CHT coefficient;
however the values chosen varied widely. In summary it can be seen that the welding
models gave different results depending on the basic assumptions made.
65
4.3 Comparison of FE model to NeT data

From Figure 3.1 it can be seen how FEM is related to the other methods (Analytical
method and experimental method). Whilst verification compares the FEM results with
the analytical solutions, validation compares the FEM results with experimental
measurements.
Modeling
Geometric model
Material thermal
properties
Simulation of Moving Point Heat Source

(Finding solutions)
Heat Source Model
Initial Condition
Iteration
Element formation
(meshing)
Boundary Condition:
- Convection at surfaces
- Radiation at surfaces
Simulate moving heat source model
Solving the problem
Post Processing
(Interpreting solutions)
- Temperature history of nodes
- Temperature distribution for certain time
-
Figure 4.4 Flowchart for moving heat source model

In this chapter secondary data are used to validate the FEM model, and some important
theoretical considerations that were used in the FEM program are discussed. These
considerations are non-linearity, birth&death technique and the heat source model.
Welding is typically modeled by a moving heat source over the solid. The flowchart in
Figure 4.4 describes the general idea of how welding is simulated in FEM. First comes
the modeling, where a geometric model and thermal properties of material, such as
66
conductivity, density, and specific heat, are defined. Generally, a non-linear material
model is involved since the thermal properties of the material depend on the
temperature which is unknown. The last step in modeling is element formation (building
mesh) based on the type of element chosen that describes the relationship between the
geometry and degree of freedom of the element.
The second step is finding a solution which may also be called simulation. In this stage
the initial condition, such as the initial temperature of the plate, and boundary
conditions such as convection and radiation at the surfaces, as well as the load, are
defined. For the welding model, the load is a moving heat source. There is a varied heat
source model that can be used, and it is usually divided into three classes: the point heat
source, the surface heat source and the volumetric heat source. For a transient analysis
the solution is obtained at each time step based on the initial condition, boundary
condition and load. For a non-linear material model, an iteration is needed to obtain the
final solution.
The last step is interpreting the solution. Generally the default outputs for thermal
analysis are the distribution of temperature and temperature history at the nodes.
The three main configurations of moving heat source used are: the point heat source, the
surface heat source and the volumetric heat source. In this thesis, the volumetric heat
source is based on Goldaks ellipsoidal distributed heat source since it is the most
general form of heat source models. Also from the point of view of FEM simulation, the
volumetric heat source as a body load to model welding processes has at least two
advantages over a surface load: a better accuracy
[20]
and flexibility. The model in this
thesis is based on a moving heat source over an infinite solid. Following the coordinate
system set by NeT, welding process moves parallel to the z axis and if
represents a
moving abscissa parallel to the z axis, the heat load at a certain small increment volume
can be expressed by equation (4.1).
(4.1)
A, B and C are constants. Considering the energy conservation and that welding is
applied onto a plate, i.e. a semi-infinite solid, equation (4.2) is obtained.
(4.2)
67
It is mathematically known that
, and at the limits
As a result, equation (4.2) can be written as in (4.3).

(4.3)
For the Gaussian distributed surface disc heat source, the heat varies according to
[22]
distance from the centre which is expressed as
. ri is the distance
from the centre point, and ro is the outside radius of the disc. For ri = 0 the heat flux will
whilst for ri = ro the minimum heat flux will be
have the maximum value of
. The ratio between the minimum heat flux at the edge of the disc and
maximum heat flux at the centre of the disc is about 5%. Similarly, the heat rate for a
Goldak heat source model that is generated at the edge elements was considered to be
equal to 5% of the heat rate applied at the centre element. For elements at (rx,0,0),
(0,ry,0), and (0,0,rz),
= 5%
; and using equation (4.1) and (4.3), equation
(4.4) can be obtained for the Goldak heat source model.

(4.4)
The maximum value of equation (4.4) is
at position (0,0,0). By
considering this condition and substituting
, equation (4.4) can be
simplified to (4.5).
(4.5)
To obtain a good model of the heat source, a very fine mesh should be used in the area
where the heat source will pass through. Equation (4.5) can be used to judge how fine
the meshes need to be, but they depend on the rx, ry and rz values; the higher these
values are the coarser the mesh needs to be. Brick elements of 0.3mm brick mesh have
been used. The adequacy of the mesh size will be analysed after the rx, ry and rz values
have been obtained. Further details can be found in Section 4.3.3.
68
Figure 4.5. FEM mesh design.
In this work the finite element model and simulation have been carried out using
ANSYS Parametric Design Language (APDL) mode because it is more flexible than the
Graphics User Interface (GUI) mode. A one-half model was used due to symmetry. The
model consists of 43,639 nodes and 71,128 elements of SOLID70. Denser meshes close
to the weld line are needed due to the high temperature gradient at this position. Very
fine meshes in the weld area are needed to match the moving heat source closely. Dense
meshes are also applied at positions surrounding a thermocouple to locate its position
accurately. The final FEM mesh design is shown in Figure 4.5.
The general material properties obtained from https://odin.jrc.ec.europa.eu are given
in Table 4.3 and the thermal properties are presented graphically in Figure 4.3. Since a
non-linear material model is used, iteration is performed using the Newton-Raphson
iteration method.
This study used two approaches; a standard approach and an element birth-and-death
technique approach. In the standard approach the deposited weld bead is modeled insitu with the welded plate, while in the birth and death techniques the weld deposit
metal is developed in accordance with the position of the moving heat source model.
The standard approach can rapidly provide temperature history data, and although the
69
element birth and death technique is closer to the real conditions, a longer computing
time is needed.
4.3.1 Non-linear material model

Many physical situations present nonlinearities too large to be ignored. The sources of
nonlinearities can be divided into three classes: a changing status (contact element),
geometry and nonlinear material properties. The last is the one most frequently involved
in the practical world.
In thermal analysis, the temperature is mostly a unit which should be obtained. For one
dimensional conduction problems, the relationship between temperature and heat rate,
material thermal properties and geometric condition can be expressed by Fourier law:
. In a linear material model k is considered independent of T, whilst in a
non-linear material model k depends on T. Since T is not known a-priori the problem
becomes more complicated and an iterative method is normally used. Three important
methods will be presented in this chapter: the direct method, the modified method, and
the Newton-Raphson method.
To present nonlinearity in general terms, it can be expressed mathematically as in
equation 4.6 [23].
(4.6)
Where x is the unknown which should be obtained and k is a constant. For a certain
value of y the degree of freedom x, should be found.
Figure 4.6 represent the direct method graphically. For a certain value of yA the value of
x should be xA. In the direct method, the first f (x) is considered equal to zero. The
tentative first x values, which is symbolised as x1 can be found easily. Applying this x1
value to f (x), the tentative second x value, x2, can be obtained. This method is applied
iteratively until the tentative x values are converged. Red lines describe each iteration
slope. It can be seen from Figure 4.6, that the lines start from the same point and its
values depend on previous x values.
70
The second is a modified method that alters equation (4.6) in the form of
First, f (x) is considered as zero and an x1 value can be obtained. By applying this x1
value to the term xf(x), the tentative x2 value can be found. This method is applied
iteratively until the x values are converged. Figure 4.7 represents the direct method
graphically. It can be seen from Figure 4.7 that the same slope is used at each iteration,
and although values of x1, x2, .. obtained by the modified method may differ from those
obtained by the direct method, when it has converged the value will be same at xA.
Figure 4.6 Direct method
Figure 4.7 Modified method
The Newton-Raphson iteration method is depicted in Figure 4.8. At an initiation point

A, the values of y and x are xA and yA respectively. The x value should be obtained
when y is yB.
71
When developing equation (4.6), the slope at any point (usually called the tangent
stiffness) can be obtained using equation (4.7).
(4.7)
Using the initial point (xA, yA),
. The coincidence y1 can be obtained
by substituting x1 to the equation (4.7). Now (x1, y1) as an initial point and x2 can be
found iteratively as
. The procedure can be repeated until x is
converged or until y is converged to yB.
Figure 4.8 Newton Raphson iteration method
4.3.2 Input parameters

Basically the inputted parameters in this thesis follow previous published papers. It
should be noted that the goal of this chapter is to develop an ANSYS code that can
provide good results. The obtained experiences proved valuable when composing the
more complex girth-weld model.
The first parameter which should be determined is welding efficiency. It was
recommended by NeT to use 75% efficiency which fit with normal welding efficiency
of TIG ranges between 65% and 88% [24].
Other parameters (rx, ry, rz) were determined by fitting mid length thermocouples with a
peak temperature (Tc2, Tc5 and Tc7). It should be noted that, the APDL code can
accommodate adjustable rx, ry and rz. Mid-length thermocouples are insensitive to arc
72
dwell and initiation mode at the starting point of the weld, and are also insensitive to the
arc extinction model at the end point of the weld.
The next step is modeling the arc dwell and initiation at the starting point of the weld.
The arc dwell time just follows previous published paper while considering NeT
recommendations and 1 second dwell time was used. Two choices of arc initiation mode
can be used, either a stepped load mode or a ramped load mode and ramped load mode
was chosen. The arc extinction was also modeled by a ramp load which is removed after
0.001s. The 0.001s extinction time is practically a step load and the ramp load mode is
easier to be converge.
The parameters for convection (h) and radiation ( ) were determined from the mean
values of previous published paper. Since convection heat transfer coefficient from
previous papers varied from 4.2 15 (W/m2. C), a value of CHT 10 (W/m2. C) was
adopted while emissivity was assumed to be 0.4 considering variation of 0 (no
radiation) 0.75.
4.3.3 Standard technique

In the standard technique the weld bead is modeled before the moving heat source load
is simulated. The moving heat source is modeled by applying the Goldaks ellipsoidal
model at a certain point and heat is liberated for a given time duration depending on the
welding speed and distance between one point and the next consecutive point. This
distance is determined by the mesh size of the FEM model. Since the mesh size on the
heat source path is 0.3 mm and the welding speed is 2.27 mm/s, the time duration is
0.13s, after which the heat source is removed and relocated to the next position, which
then liberates the heat for the same time duration. This procedure is repeated up to the
end of the weld length. The welding is simulated using the standard technique with the
flow-chart shown in Figure 4.4, with a detail block for simulating the moving heat
source shown in Figure 4.9.
Following the NeT procedures, it should be noted that the data were obtained from four
tests (A11, A12, A21 and A22) and each test comprised nine initial temperature data
(Tc1 Tc9 Figure 4.2). The initial temperatures varied from 21 C to 29 C. From the
NeT temperature histories, the initial temperature assigned to the base metal was 25 C.
73
Simulate moving heat source using standard technique

Determine first heat source center
Defined elements relative position to the heat source

center
Inside
ellipse?
No
Yes
Apply volumetric heat according to the
relative position for certain time duration
Iteration
Determine next heat source center

center
Inside
ellipse?
No
Yes

Applied heat load to the next point for certain duration
No
Iteration
End of
the weld?
Yes
Iteration
Figure 4.9 Standard technique
The parameters rx = 3mm, ry = 5mm, rz = 2mm to describe heat source
(equation
4.4) were obtained after evaluating the peak temperatures at the mid-length position.
Considering the arc dwell in some published paper varied from 0.7s to 1.5s
[17]
, the
dwell time at the start of the weld is taken as 1s. The mid-length macrograph was also
helpful in giving tentative values of the heat source parameters especially rx and ry
whilst the bi-lobed transverse boundaries described tentative values of rz.
The brick mesh sizes along the heat source trajectory is 0.3mm. The values of the body
heat source depend on the distance from the centroid of the elements to the centre of the
74
heat source. Since the brick mesh sizes is 0.3mm, the closest
when applied to equation (4.5) results in
is 0.03645mm2 which
. The maximum body heat
will be represented by 96.42% of its value, and it can be assumed that the mesh size
along the weld path is sufficiently fine. (It would be good to study what mesh size is
optimum for convergence but this is likely to involve a large number of models with
varying mesh size and needs careful and time consuming study, however the most
important factor in the welding is probably how the heat source is modeled. The values
of the volumetric heat rate is really dependant on the mesh size since it determine the
).
4.3.4 Birth and Death technique

The birth and death technique is very useful when modeling welding phenomenon, and
ANSYS uses the EKILL and EALIVE commands. When an element is killed it will be
deactivated although it remains in the model. Deactivated elements have a conductivity
close to zero (for mechanical analysis, stiffness is set close to zero). Stress, plastic
strain, creep strain, and other dependent variables are set to zero. The next discussion
will explore the birth and death technique in thermal analysis, from a very simple
model.
How a structure with deactivated elements behaves in thermal analysis is simply shown
at Figure 4.10. Figure 4.10 shows temperature distribution over a plate. In Figure 4.10a,
a solid square is modeled with its left side maintained at 200 C and its right side at 0 C.
Steady state analysis was carried out and the distribution of temperature is shown. In
Figure 4.10b a solid square with a circular hole in the middle is modeled. The left side
of the square was loaded with a temperature of 200 C and its right side was maintained
at 0 C. The temperature distribution was plotted after a steady state condition had been
achieved.
In Figure 4.10c a solid square is modeled. Elements inside a circle are deactivated. The
circle is at the centre of the square with a diameter that is exactly the same as the hole in
Figure 4.10b model. It can be seen by observing the active element, that the distribution
of temperature shown is exactly the same as the model of a square with a hole (Figure
75
4.10b). If the birth and death technique is applied in single pass weld it behaves like
there is isolation for the weld bead ahead of torch.
(a)
(b)
(c)
Figure 4.10 Temperature distribution for (a)solid rectangle, (b)rectangle with a hole and
(c) rectangle with deactivated elements inside the same hole as (b)
In Figure 4.11, instead of temperature distributions, the thermal flux is plotted for a
solid rectangular model, a solid rectangle with a hole in the centre, and a solid rectangle
with deactivated elements inside the circle. By observing the active elements of the
rectangle with deactivated elements inside the hole, the conclusion is that the thermal
flux of the active elements shows exactly the same vectors as the rectangle with a hole.
Overall it can be concluded that, when the elements have been deactivated, it looks like
they have been omitted, even though they still remain in the model. This condition can
be used to model a growing weld bead, a real geometrical model for which in FEM is
not easy.
76
(a)
(b)
(c)
Figure 4.11 Temperature flux for (a)solid rectangle, (b)rectangle with a hole and (c)
rectangle with deactivated elements inside the same hole as (b)
In an FEM model of bead on plate welding using the element birth and death technique,
the metal bead grows with the moving heat source. How the welding is simulated using
the birth-and-death technique follows the flow-chart shown in Figure 4.4 with a detail
block for simulating the moving heat source shown in Figure 4.12. First, all of the weld
bead elements were omitted using the EKILL command in ANSYS and then the
growth was modeled using the EALIVE command. The position of weld bead elements
was retrieved using *GET command and compared to the position of instantaneous heat
source centre. Born elements are elements of weld bead which lay behind the moving
heat source.
77
Moving heat source using birth and death technique

All weld bulge elements are deactivated
Determine first heat source center
Weld bulge elements behind the heat source are activated and temperature
load T=2400 C is applied for the new activated elements

center
Inside
ellipse?
No
Yes
Iteration
Determine next heat source center


center
Inside
ellipse?
No
Yes

Apply heat load to the next point for certain duration
No
Iteration
End of
the weld?
Yes
All weld bulge elements are activated
Iteration
Figure 4.12 Birth and death technique

When the elements are born, their temperature should be at the superheated temperature
2400 C [25] of the melting embedded filler metal. Instead of applying a temperature node
load, the body heat load rate ( ) at the elements was applied. The body heat load
value is such that it can produce a temperature of 2400 C at the growing weld bead. The
heat needed to elevate the weld pool from the initial temperature of 25 C to 2400 C was
78
evaluated using q = mc T. It should be noted that the specific heat cp, is temperaturedependent as expressed in Table 4.3 and in Figure 4.3. The specific heat at a certain
temperature range was taken as the mean value between these ranges. The heat rate can
be obtain using
= mc T
expression. Finally, the body heat load which should be
applied at the born weld bead can be obtained using equation (4.8).
(4.8)
4.3.5 Results and discussion

The temperature histories for the standard technique are shown in Figure 4.13 to 4.17.
From these figures it can be claimed that this simulation has shown a good agreement
with the experimental results, but apart from showing the general trend, the transient
temperatures also matched the experimental data. For the close field thermocouples,
especially at mid-length and the weld end (Tc2 and Tc3), the FEM model predicted a
lower temperature than the measured peak temperature. The lower prediction may be
caused by radiation from the heat source which was not modeled. The below bead
thermocouple (Tc7), showed a higher value in the model than the measured value. The
below bead thermocouple was inserted 6.5mm deep in a hole with a diameter of
1.2mm. The possibility that the thermocouple was not fully located at the tip of the hole
is high. Incomplete insertion could result in a larger distance from the heat source and a
lower peak temperature.
The predicted temperature histories at the thermocouple positions resulting from the
application of both techniques are shown in Figure 4.18 to 4.21. It can be seen that the
temperature profiles using both techniques do not significantly differ. The conclusion is
that welding thermal solution of structures like the bead on plate, with a minimal weld
bead, are not sensitive to either the presence of the bead ahead of the torch (standard
technique), or any forward conduction suppression made to account for its absence.
Certainly this is not the case for the multi-pass girth weld joint where the weld beads are
too large to be ignored, and the using of birth and death technique become compulsory.
79
Figure 4.13. Temperature histories using standard technique for location of

thermocouple Tc1 & Tc4 (refer to figure 4.2).
Figure 4.14. Temperature histories using the standard technique for the location of
thermocouples Tc2 & Tc5 (refer to figure 4.2).
80
Figure 4.15. Temperature histories using standard technique for the location of
thermocouples Tc3 & Tc6 (refer to figure 4.2).
Figure 4.16. Temperature history using standard technique for the location of
thermocouple Tc7 (refer to figure 4.2).
81
Figure 4.17. Temperature history using standard technique for the location of
thermocouple Tc9 (refer to figure 4.2).
Figure 4.18. Temperature histories using standard and birth & death techniques for the
location of thermocouples Tc1 & Tc4 (refer to Figure 4.2).
82
83
(a)
(b)
Figure 4.22 (a) actual weld pool[16] and (b) temperature contour from standard technique
84
(a)
(b)
Figure 4.23 (a) actual weld pool and (b) temperature contour from birth and death
technique
Figure 4.24. Transverse cross section of weld pool at weld-start and weld-end points
The shape of the weld pool at the mid-length cross section was compared with actual
shape of the weld pool in Figure 4.22 and 4.23, for both standard and birth&death
techniques. The actual cross section of the weld pool was obtained by cutting the
welded plate[17]. The shape of the weld pool predicted by the FEM model is the region
where the temperature was above 1400 C (melting point of AISI 316 stainless steel).
The weld pool here can be modeled quite well using both techniques, which follow the
procedure shown in Figures 4.9 and 4.12.
85
Figure 4.24 shows the transverse cross section of weld start and weld end and the actual
cross section obtained from NeT [16, 17]. The differences are caused by the mass transfer
mode that was not modeled in the developed FEM model. This mass transfer mode is
not sensitive at the mid-length cross section that makes the actual weld-pool fit with the
FEM model.
(a)
(b)
Figure 4.25. Comparison of temperature contour. (a)standard technique, (b)birth and
death technique.
In Figure 4.25 the sections of weld pool predicted by both techniques was compared.
Although the distribution of temperature, especially for high temperatures above
1900 C shows a different distribution for both techniques, the shape of the weld pool is
the same.
The heat source in this model was differentiated into two types, uniform heat load in the
weld bulge and the distributed heat load following Goldaks double ellipsoid function
with the sketch shown in Figure 4.26. Due to the high concentrated heat load of
Goldaks along the weld line the isothermal temperature develops from the weld line
which is seen as a dot in the cross section area. Since in the standard technique there is a
cold weld bead in front of the heat source part of the heat should be transferred to the
cold weld bead. On the other hand, in the birth and death technique this cold weld bead
is deactivated. Based on this description it is possibly why in standard technique the
high temperature (1900C) occupied a narrower area and was deeper to the centre as a
result of a lower heat since a part of the heat was transferred to the cold weld bead
which was not deactivated.
86
Figure 4.26. Heat source model.
4.4 Conclusion
A comprehensive model of the transient thermal behaviour of a weld was established
and validated using the experimental results from the NeT consortium.
The shape of the weld pool predicted by the standard technique and birth-and-death
technique show well matched results, and an evaluation of the predicted shape of the
weld pool reinforces the validity of the welding model.
The validated APDL code and obtained experiences while solving the thermal analysis
in this chapter are a good start for the next more complicated analysis.
87
References
[1]
Jinjin Xu, Ligong Chen and Chunzhen Ni (2007), Effect of Vibratory Weld
Conditioning on The Residual Stresses and Distortion in Multi-pass Girth-butt
Welded Pipes, Pressure Vessels and Piping, no.84: pp. 298-303.
[2]
A.H. Yaghi, T.H. Hyde, A.A. Becker and W. Sun (2008), Finite element
simulation of welding and residual stresses in a P91 steel pipe incorporating
solid-state phase transformation and post-weld heat treatment, Journal Strain
Analysis, vol.43, pp. 275 293.
[3]
S.F. Estefen, T. Gurova, X. Castello and A. Leontiev (2010), Surface residual

stress evaluation in double-electrode butt welde steel plates, Materials and
Design, pp.1622 1627.
[4]
Mochammad Noer Ilman and Triyono (2010), Fatigue Crack Growth Behaviour
of Friction Stir Welded Aluminium Alloy 2024-T3 Under Local Preheating,
International Journal of Material Science, Vol.5 No.6, pp. 791 -800.
[5]

Tool in The Aid of Structural Integrity Assessment of Welded Structures. 5th Int.
Conference Structural Integrity of Welded Structures, Romania.
[7]
H.S. Carslaw and J.C. Jaeger (1959), Conduction of Heat in Solid. Clarendon
Press, Oxford.
[8]
Paley, Zvi; Lynch, J.N. and Adam C.M. JR. (1964), Heat Flow in Welding Heavy
Steel Plate, Welding Research Supplement, pp.71-79.
[9]
N. Christensen, V. Davies and K. Gjermundsen (1965), Distribution of

Temperature in Arc Welding. British Welding Journal 12(2), pp. 54-75.
[10] C.L. Tsai (1982), Heat Flow in Fusion Welding. Proceeding of The Conference
on Trends in Welding Research in the United States, ASM International, New
Orleans, pp.91 108.
[11] Timothy Hasselman, Keng Yap, John Cafeo and James Cavendish (2002),
Predictive accuracy of resistance spot welding simulation, Proc. 20th IMAC, LA,
pp.904 909.
88
[12] C. Mares, J.E. Mottershread and M.I. Friswell (2004), Stochastic model updating
of a spot welded structure, Proc. International Conference on Noise and Vibration
Engineering - ISMA, Belgium, pp. 1885 1898.
[13] M.J. Bayarri, J.O. Berger, R. Paulo, J. Sacks, J.A. Cafeo, J. Cavendish, C.H. Lin
and J. Tu (2005), A framework for validation of computer models, Technical
Report no. 162, NISS.
[14] Koishi Kumagai, Masakazu Shirooka, Jirou Ohachi and Toshirou Ogawa (2007),
Rupture modeling of spot welds suitable for crash FE analysis in vehicle
development process, 6th European LS-DYNA Users Conference, Sweden, pp.
3.1 3.13.
[15] C.E. Truman & M.C. Smith (2009) The NeT residual stress measurement and
modeling round robin on a single weld bead-on-plate specimen. International
Journal of Pressure Vessels and Piping Vol. 84, pp.1-2.
[16] Protocol for Phase 2 finite element simulation on the NET single-bead-on-plate
test specimen; https://odin.jrc.ec.europa.eu
[17] M.C. Smith and A.C. Smith (2009) NeT bead-on-plate round robin: Comparison
of transient thermal prediction and measurements; International Journal of
Pressure Vessels and Piping Vol. 86, pp.96-109.
[18] X. Ficquet, D.J. Smith, C.E. Truman, E.J. Kingston and R.J. Dennis (2009)
Measurement and prediction of residual stress in a bead-on-plate weld
benchmark specimen. International Journal of Pressure Vessels and Piping, Vol.
86, pp. 20 30.
[19] X, Shan; Davies,CM; Wangsdan,T; ODowd,NP & Nikbin,KM (2009) Thermomechanical modeling of a single-bead-on-plate weld using the finite element
method, International Journal of Pressure Vessels and Piping, vol. 86, pp.110
121.
[20] Philippe Gilles, Walid El-Ahmar and Jean-Francois Jullien (2009) Robustness
analyses of numerical simulation of fusion welding NeT-TG1 application:
Single weld-bead-on-plate. International Journal of Pressure Vessels and
Piping, vol.86, pp.3 12.
[21] C. Ohms, R.C. Wimpory, D.E. Katsareas and A.G. Youtsos (2009) NeT TG1:
Residual stress assessment by neutron diffraction and finite element modeling on
89
a single bead weld on a steel plate; International Journal of Pressure Vessels and
Piping, vol.86, pp.63 72.
[22] S.K. Bate, R. Charles and A. Warren (2009) Finite element analysis of a single
bead-on-plate specimen using SYSWELD. International Journal of Pressure
Vessels and Piping, vol.86, pp.73 78.
[23] Robert D. Cook, David S. Malkus & Michael E. Plesha (1989), Concepts and
Applications of Finite Element Analysis. John Wiley & Sons.
[24] Kou, Sindo, Welding Metallurgy, John Willey & Sons Inc. Publication; 2003.
[25] M.J. Painter and M.H. Davies, Numerical Modeling of The Gas Metal Arc
Welding Process, Annual Report CRC Project: 93.04; 1994.
90
Chapter 5
Mechanical Analysis of NeT TG1
Specimen4
The temperature in the welding process is not distributed uniformly and causes a nonuniform thermal strain in a welded structure, which in turn, causes thermal stress. This
thermal stress continues to grow until the welded structure reaches a uniform
temperature. Any stress remaining while the temperature of the welded structure is
already uniform is called residual stress and is known as a major cause of failure in
welded structures especially in pipelines [1- 3].
Figure 5.1 Pattern of longitudinal distribution of residual stress.
Longitudinal residual stress close to the weld line (Figure 5.1) can be roughly predicted
using Masubuchis equation (equation 5.1)
x/b 0.6.
z(x)
[4]
. This equation generally is only valid for
to describes the distribution of longitudinal residual stress which
depends on the transverse distance from the centre of the weld (x);
4
max
is the maximum
Material in this chapter has been published in a paper:

Djarot B. Darmadi, Muruganant Marimuthu, Anh Kiet Tieu and John Norrish, Numerical prediction of
th
residual stress in 316L bead-on-plate welds, 15 International Conference on Advances in Material &
Processing Technology (AMPT), Wollongong Australia, 23-26 September 2012, paper id: 11355.
91
tensile residual stress, and b is the width of the region of tensile stress. In many cases
the maximum tensile residual stress can be represented by the
yield
of the base metal
whilst the value of b remains unknown, based on the thermal analysis and welding
parameters. Sindo Kou modified Masubuchis equation using an exponential function,
as shown in equation 5.2 [5].
(5.1)
(5.2)
Actually, the development of residual stress in the welding phenomenon is very
complex because the pattern of residual stress depends on the type of welding process
(heat input model), and the geometry of the welded joint and the base metal [6]. As there
is no single mathematical equation available to predict the residual stress pattern, a
numerical prediction such as FEM is more suited to deal with this complexity.
The basic mechanism of the development of residual stress was described by Wells
[7]
and is presented in Chapter 2. To understand the development of residual stress, Sindo

Kou
[5]
used the three bars model. The analytical solution regarding the elasto-platic
material model compared to the FEM provides a clearer idea on the development of
residual stress
[8]
. First, heating and cooling during welding always causes
elongation/shrinkage in the base metal and the weld metal, but this elongation/shrinkage
depends on the elevation of temperature and coefficient of thermal expansion, and this
elongation/shrinkage causes distortion and stresses. Some FEM model papers focused
on this distortion
[9 13]
. If the distortion is constrained it will cause residual stress.
These constraints can be the rigid surrounding materials, or the cool region away from
the weld line. Some FEM model papers focused only on this residual stress
[14 - 17]
and
others on both phenomenon (distortion and residual stress) [18 - 23] , since they co-exist in
a welded structure.
This chapter discusses the FEM model of the development of residual stress in a beadon-plate welding of 316L Austenitic Stainless Steel. The thermal analysis of the model
has been presented in Chapter 4. The welding procedure follows the experimental data
provided by the NeT consortium [24] and the result (residual stress) is validated using the
92
NeT consortium member data which is published in international journals
[25-29]
. Details
of the welding procedure have also been described in chapter 4.
5.1. Thermo mechanical analysis

Thermo-mechanical analysis involves thermal analysis and mechanical analysis. From
thermal analysis the distribution of temperature at certain time can be predicted, where
as in mechanical analysis, the distribution of temperature from thermal analysis is
applied as a thermal load. This thermal load in turn causes the elements to expand and
since the temperature is not distributed uniformly, a non uniform elongation exerted to
the welded structures will cause thermal stresses. When all of the temperature in a
welded structure is uniform (achieved after certain time duration), no further stress is
added and the remaining stress is called residual stress.
In a thermo-mechanical analysis a mechanical analysis is applied sequentially after
thermal analysis has been completed. This uncoupled model is reasonable since the
distribution of temperature does affect the development of residual stress, even though
the residual stress has an insignificant effect on the distribution of temperature in a
welded structure as it has been discussed and shown in Figure 2.1.
This chapter discusses the residual stress exerted on the bead-on-plate welding
discussed in Chapter 4 where the thermal analysis was carried out.
5.2. Element model (SOLID45)

The element model used in the thermo-mechanical analysis is SOLID45. These
elements accommodate structural analysis. They consist of eight nodes and three
degrees of freedoms (dof) at each node. The degrees of freedom are translation in three
perpendicular directions: x, y and z respectively. The nodes at the element can coincide
with each other and the element can be in a prism or tetrahedral forms, as shown in
Figure 5.2. Since SOLID70 and SOLID45 have basically the same geometry, they are
often used as a pair in a thermo-mechanical model, i.e. SOLID70 is used in thermal
analysis while SOLID45 is used in mechanical analysis. It should be noted that the
93
SOLID45 element cannot act in the pyramid form like the SOLID70 element, and since
the elements that compose the geometric model in thermal analysis should be exactly
the same as in mechanical analysis, the pyramid form of SOLID70 should be avoided
[8,17]
Figure 5.2. SOLID45 3D-structural element (Release 12.0 Documentation for ANSYS)
5.3. Mechanical properties

The mechanical properties depend on temperature and the mechanical properties
experienced in the welding process vary significantly with the wide range of
temperatures, from ambient to the material melting point. Stress-strain diagrams of the
weld metal and base metal material model for various temperatures are shown in Figure
5.3a and 5.3b respectively. In FEM modeling, a multi-linear kinematic hardening
material model was used to accommodate those temperature dependent stress-strain
relationships. The multi-linear model consists of linear lines which represent a nonlinear plasticity. In APDL mode for a database input of certain properties at certain
variables and between consecutive defined variables the defined property is considered
linear. In case of Figure 5.3a, as an example to describe non linear stress-strain curve at
20C the properties
input were
, i.e. 210MPa, 238MPa, 292MPa, for the
variables , i.e. 0.107%, 0.3%, 1.1%, 2.1% . And between those defined points
ANSYS applies linear interpolation. If a nonlinear model is defined by enough points in
a linear model it can be considered the multi-linear model represent the non-linear
model. The multi-linear kinematic hardening model in ANSYS is used to represent nonlinear kinematic hardening in the real model.
94
The other mechanical properties have been discussed in Chapter 4. Since in the ANSYS
APDL mode the material properties should be defined in a consistent range of
temperatures, the material properties for the defined temperatures are obtained using
linear interpolation. Those material properties, after being interpolated, are presented in
Table 5.1a and 5.1b, for base metal and weld metal respectively.
Figure 5.3a. Temperature dependent stress-strain diagram for base metal [9]
Figure 5.3b. Temperature dependence stress-strain diagram for weld metal [9]
95
It should be noted that, the thermal expansions ( ) provided by NeT in Tables 5.1 are
means between zero and the relevant temperature. The instantaneous thermal expansion
which was input into ANSYS APDL is shown in Table 5.2.
Table 5.1a. Base metal properties [9].
T( C)
20
275
550
750
800
900
1100
1400
(W/mC ) cp(J/kgC )
14.12
492
17.76
523
21.67
556
24.52
581
25.23
587
26.66
599
29.50
623
33.78
659
x 10-6/C )
14.56
16.70
17.95
18.58
18.72
18.99
19.53
20.21
E (GPa)
195.60
181.13
159.75
137.75
131.40
116.80
80
2
x 10-6/C )
14.56
16.70
17.87
18.43
18.86
19.27
20.21
E (GPa)
171
151.88
133.13
120
102.95
83
1.70
Table 5.1b. Weld metal properties [9].

T( C)
20
275
525
700
850
1000
1400
(W/mC ) cp(J/kgC )
14.12
488
17.76
532.75
21.32
576.25
23.81
589
25.95
589
28.08
589
33.78
589
Table 5.2. Instantaneous thermal expansion

T( C)
20
275
550
750
800
900
1100
1400
Base Metal
14.56
19.18
21.53
21.43
21.33
21.48
22.77
23.45
T( C)
20
275
525
700
850
1000
1400
x 10-6/C )
Weld Metal
14.56
19.18
21.45
21.06
21.62
22.32
23.53
It is reported that varied hardening models (elasto-plastic, kinematic and isotropic) were
used as material model by some investigators of weld simulation
[27]
. Regarding the
stress-strain diagram in Figures 5.3 the elasto-plastic mode was dropped as the
hardening model in this chapter. In this Chapter kinematic hardening is simulated first.
The results using isotropic hardening are then compared with the results of the
kinematic hardening model and discussed.
96
Others properties such as density and Poissons ratio for base metal and weld metal are
not considered to be affected by temperature. The values for density and Poissons ratio
are 7966kg/m3 and 0.294 respectively.
5.4. Exploring the birth and death technique

How the birth and death technique behaves in thermal analysis was discussed in Section
4.3.4. The element model in thermal analysis (SOLID70) considers temperature as a
degree of freedom whilst displacement is considered in mechanical analysis using
SOLID45 element model. In mechanical analysis, when the element model is dead
(deactivated) the EKILL command has been applied, the stiffness of the element is set
close to zero. How the birth and death technique behaves in mechanical analysis is
shown simply in Figure 5.4 that describes displacement distribution.
(a)
(b)
(c)
Figure 5.4. Displacement distribution for (a) solid rectangle, (b) rectangle with a hole
and (c) rectangle with deactivated elements inside the same hole as (b)
97
In Figure 5.4a, a solid rectangle is constrained in all directions on the left surface. On
the right surface the rectangle is loaded by a small displacement to the positive xdirection, but constrained in the other two directions (y and z). The x-displacement
distribution of steady state in the rectangle is then plotted. In Figure 5.4b the rectangle
has a hole at the centre and the same displacement load and constraint as in Figure 5.4a
are applied in Figure 5.4b. The x-displacement distribution of steady state in the
rectangle with a hole is then plotted. In Figure 5.4c the solid rectangle has dead
(deactivated) elements inside a hole which has a position and size equal to the hole in
Figure 5.4b. The same displacement load and constraint are applied and the xdisplacement distribution of steady state in the rectangle with the dead elements is
plotted. It can be seen that the x-displacement distribution in the active elements of
Figure 5.4c has exactly the same pattern as in Figure 5.4b.
(a)
(b)
(c)
Figure 5.5. Vector plot of displacement for (a) solid rectangle, (b) rectangle with a hole
and (c) rectangle with deactivated elements inside the same hole as (b)
In Figure 5.5, instead of the distributions of displacement, the displacement vectors are
plotted for a solid rectangular model as a solid rectangle with a hole in its centre, and a
98
solid rectangle with deactivated elements inside the circle. By observing the active
elements on the rectangle with deactivated elements inside the hole, a conclusion was
reached that the displacement vector of active elements showed the same vectors as the
rectangle with a hole.
Overall it can be concluded that deactivated elements appear to have been omitted,
although they remain in the model. In mechanical analysis, this condition can be used to
model a growing weld bead because a real growing geometric model is not easily
created in ANSYS-APDL. First, the base metal and weld bead were modeled, then all
the weld bead elements were deactivated, and then parts of the weld bead elements were
activated after the relative position of those elements to the heat source centre had been
evaluated.
5.5. FEM for the thermo-mechanical model

The flow-chart for the comprehensive procedure to obtain the residual stress in the
thermo-mechanical analysis is shown in Figure 5.6. The following will discuss each step
in more details.
Considering the equality of element geometry and the degree of freedom as discussed in
Section 5.2, the SOLID45 element was used in the mechanical analysis. Since the nodal
temperature in thermal analysis will be used as a thermal load in the mechanical
analysis, the meshes in the mechanical analysis should be exactly same as thermal
analysis to accommodate temperature load mapping. This can be achieved by resuming
the geometric model in the thermal analysis and replacing the element (SOLID70) with
the structural elements model (SOLID45).
At this stage, a proper geometric model has been built, so the next steps are to define the
mechanical properties of the element and define the boundary condition and initial
condition of the welded structure. The boundary condition is determined by how the
welded element is constrained in real conditions. The initial condition is the initial
temperature that is considered when no thermal stress has been exerted. Since the
geometry of the welding process was symmetric, to save computing time only a half
99
model was built. For the simulated bead-on-plate welding which follows the NeT
welding procedure (described in Chapter 4), the boundary conditions were defined for
the clamping vice and plane of symmetry. The vice was considered as a light clamping,
thus no constraint was applied, while the symmetry plane was modeled as an area with
no displacement in a normal direction (parallel to the x axis).
Thermal Analysis Results
Next Time Step
Geometric model
First Time Step
Resuming the geometry
Check the position of
weld bead elements
Behind weld
torch?
Modify elements to structural

elements (SOLID45)
Temperature distribution
for certain time
Deactivated all of weld bead

elements
Apply temperature
as thermal load
Y
The elements are
activated
Define boundary and initial

conditions
Thermal strain and thermal

stress for defined time
The Data-base
Elements temperature
for certain time
> Melting point?
Stiffness 0
Elements temperature
at certain time
N
> Melting point?
Y
Plastic
strain = 0
End of thermal
load?
Residual Stress
Figure 5.6. FEM procedures to predict the distribution of residual stress.

100
The temperature history of each node in thermal analysis is saved in the *.rth file. For a
certain time the temperature at each load can be retrieved using the LDREAD command
and applied as a thermal load. Applying this temperature as a thermal load produces
strain and thermal stress that can be calculated by considering the stress-strain
relationship. This application of thermal load is repeated until no more significant
thermal stress has developed. The remaining thermal stress is considered to be residual
stress.
5.5.1. Accommodating a growing weld bead

As in thermal analysis, a growing weld bead is accommodated using a combination of
the EKILL command and EALIVE command. This technique can also be applied in
mechanical analysis. First, all the weld bead elements are deactivated using the EKILL
command and then the position of the weld bead elements is checked relative to the
position of the welding torch. When the elements lie behind the welding torch, the
EALIVE command is applied and the element activated. The temperature distributions
for a certain time from the thermal analysis are applied to this geometric model with the
growing weld bead.
5.5.2. Accommodating the melting phenomenon

While a material is melting, its stiffness and plastic strain are significantly decreased.
Shan et.al
[25]
put zero values on the stiffness and plastic strain. To identify which
elements are melting, the temperature of the elements should be checked in the thermomechanical analysis.
Temperature data in the thermal analysis are saved in the file *.rth. It is time consuming
to retrieve those files to be used as parameters, so a data-base which can provide
parameters that give the temperature of each element at a certain time has been
constructed. It was constructed during the thermal analysis stage, and it should be fast
and easy to open in the thermo-mechanical analysis stage. In this FEM procedure, a
*.txt database file was made. Based on the parameters retrieved from the data-base, the
elements which melted in the thermo-mechanical analysis can be determined. To save
101
computing time, only those elements close enough to the weld-line were checked. With
the melted elements, the stiffness was set close to zero before a temperature load was
applied, and the plastic strain was set to zero after the temperature load was applied.
Deactivating an element using the birth and death technique causes near zero stiffness
(10-6N by ANSYS default), and the element is deactivated before a thermal load is
applied. Plastic strain can be set to zero by applying the DESOL post processor
command. This command is used to manipulate the ANSYS output file, i.e. plastic
strain of the elements.
5.6. Results and discussion

As discussed in Chapter 4 and shown in Figure 4.4, ANSYS solved the FEM problem in
three steps: modeling, simulation, and post processing. NeT provided experimental data
for residual stress measurement. Those residual stresses are for longitudinal and
transverse directions and no normal stress is available
[27]
. Following the available
residual stresses data, Figures 5.7 and 5.8 show pictures of the distribution of
longitudinal and transverse stress respectively, where it can be seen that the high tensile
longitudinal or transverse residual stress in the region close to the weld bead are
balanced by compressive stress in the field away from the weld bead. It should be noted
especially in Figure 5.8 that those stresses are varied along the weld bead. When the
welding starts, the entire region in the plate is set equal to its initial condition. For the
next position the plate temperature changes due to the previous welding and this
condition develops further with ongoing welding. Moreover the geometry also changes
due to the embedded weld bulge on the plate surface.
In Figure 5.9 the distribution of longitudinal residual stress for the cross sectional area
in the middle of the weld line (at z = 30mm refer to Figure 4.1 and 4.2) is shown. In the
a similar manner the distribution of transverse residual stress is shown in Figure 5.10.
Experiment data are obtained from the white dashed B2 line which is superimposed in
the Figures 5.9 and 5.10. The line is lied 2mm below the surface of base metal, and
along the line the numerical values of stresses (longitudinal and transversal) were
retrieved from the ANSYS *.rst file and will be compared to the measured distribution
of residual stress.
102
Figure 5.7 FEM longitudinal residual stress (Sz) prediction of the bead on plate welding
Figure 5.8 FEM transverse residual stress (Sx) prediction of the bead on plate welding
103
Figure 5.9 FEM longitudinal residual stress (Sz) prediction of the bead on plate welding
at the mid cross section area
Figure 5.10 FEM transverse residual stress (Sx) prediction of the bead on plate welding
at the mid cross section area
Experimental results for residual stress were provided by four participants in the NeT
consortium: 1) Joint Research Centre, Institute for Energy, The Netherlands JRC
2) Nuclear Physics Institute, Czech Republic NPI
Open University, England OU
[28]
[27]
[26]
; 3) Materials Engineering, The
and 4) Heinz Maier-Leibnitz Institute HMI
[29]
All these participants used neutron diffraction to measure the residual stress and the
results were used to validate the FEM prediction discussed in this paper.
104
In Figure 5.11 the data from all four participants was compared to the FEM data that
predicted the distribution of longitudinal and transverse residual stresses along the 2mm
path shown in Figures 5.9 and 5.10. The trend of the predicted residual stress can be
seen to match those of the measured data. Predicting the longitudinal residual stress in
the area close to the weld line is slightly lower compared to the general measured data,
but it is close to the data provided by NPI.
Figure 5.11 Residual stresses along the 2mm path in the mid cross section area.
The predicted longitudinal residual stress close to the weld line is higher than transverse
residual stress unlike the area further away from the weld. This prediction is confirmed
by the experiment results. The higher longitudinal residual stress is qualitatively
understandable considering the discussion on the three bar elasto-plastic model in
Chapter 2.31 which suggests that the residual stress is mainly formed in the longitudinal
direction. Considering the forces balance, the higher residual stress in the area close to
the weld line result in lower residual stress for the further area.
105
Figure 5.12 shows a normal view of the distribution of longitudinal residual stress in the
area of symmetry. It also shows two selected paths 2mm and 3mm below the top
surface. The 2mm path is shown as a red dashed line and the 3mm path is shown as a
white dash-dot line (D2 and D3 lines). The blue dot line is also shown which is mid
length through thickness of the base metal (BD line). The numerical data for those paths
can be retrieved from the ANSYS *.rst file and compared to the NeT consortium
experimental measurement. In the same manner Figure 5.13 shows predicted transverse
residual stress.
Figure 5.12 FEM longitudinal residual stresses (Sz) prediction at symmetry area in a
normal view.
Figure 5.13 FEM transverse residual stresses (Sx) prediction at symmetry area in a
normal view.
A comparison of the predicted longitudinal and transverse residual stresses with the
experimental result at the 2mm path (D2 line) is shown in Figures 5.14. The HMI did
not report any result for the 2mm path. It should be noted that the end of the weld is laid
on the origin of the coordinate system and the weld line spans to z = 60mm (Figure 4.2)
where the weld started. By comparing the predicted and measured transverse residual
106
stress at the 2mm path, as shown in Figure 5.14, it can be concluded that the FEM
prediction agrees with the measured results quite well.
The other evaluation was done on the residual stresses at the 3mm path (D3 line) as
shown in Figure 5.15. Only the HMI and JRC reported the distribution of residual stress
at the 3mm path. When the predicted distribution of residual stress and the measured
data were evaluated, the FEM prediction again gave well matched results.
The last evaluation is a residual stress validation on BD line. The longitudinal residual
stress validation is shown in Figure 5.16 while Figure 5.17 is for transverse residual
stress. All participants provide the experiment data. Zero y shows top surface position
of the base metal and negative y shows the weld metal position of bop. The prediction
on this line is also validated with the experiment.
Figure 5.14 Residual stresses on the D2 line
107
Figure 5.15 Residual stresses on the D3 line
Figure 5.16 Longitudinal residual stress (Sz) on BD line
108
Figure 5.17 Transverse residual stress (Sx) on BD line
5.6.1 Using isotropic hardening model
Figure 5.18 Kinematic hardening vs Isotropic haredening model

109
In the previous NeT simulation discussed in this chapter the kinematic hardening model
was used. Figure 5.18 compares the results obtained from kinematic hardening and
isotropic hardening model. Instead of using MKIN command for kinematic hardening,
ANSYS APDL uses MISO command to represent isotropic hardening. From Figure
5.18 no significant differences were found when the isotropic hardening model was
used. However using isotropic hardening gave a slightly higher result compared to the
kinematic hardening models.
5.6.2 Comparing to Masubuchis

Figure 5.18 compares the results predicted by FEM and Masubuchis equation. Based
on the FEM results the maximum residual stress of 250 MPa and 20mm width of the
tensile area are used and applied to Masubuchis equation. Results from both
Masubuchis initial (equation 5.1) and modified (equation 5.2) equation are plotted
together with FEM prediction.
Figure 5.18 FEM vs Masubuchis

Figure 5.18 shows that modified Masubuchis equation can describe residual stress
levels close to those derived by FEM. However it should be noted that, the maximum
residual stress and width of tensile area were obtained from FEM prediction; which
means the modified Masubuchis technique alone cannot provide the residual stress
results.
110
5.7. Conclusion
A validated residual stress FEM prediction for bead on plate can be provided through
the developed FEM model. This validated result confirmed the correctness of
assumptions made and also the appropriateness of the developed code and techniques in
ANSYS APDL. The melting phenomenon was approached by low stiffness and the
plastic stress on the melted materials was omitted. A certain database contains an array
of parameters which describe the temperature history created in the developed FEM
model, and which can accelerate computer time. This database was proved useful in the
Thermo-Mechanical-Metallurgical analysis which will be discussed in the next chapters.
111
References
[1] John Beavers and Narasi Sridhar (2009), Effects of Steel Microstructure and
Ethanol-Gasoline Blend Ratio on SCC of Ethanol Pipeline, Paper no. 09532, Nace
International Conference.
[2] C. Manfredi and J.L. Otegui (2002), Failures by SCC in Burried Pipelines,
Engineering Failure Analysis, Vol.9, pp. 495-509.
[3]
J. Wang and A. Atrens (2003), Microstructure and Grain Boundary

Microanalysis of X70 Pipeline Steel, Journal of Material Science, Vol.38, pp.323330.
[4]
Koichi Masubuchi & D.C. Martin (1961), Investigation of residual stresses by use
of hydrogen cracking, Welding Journal, vol. 40, pp. 553s 563s.
[5]
Sindo Kou (2003), Welding metallurgy, John Wiley & Sons, Canada.
[6]
Radaj (2003), Welding residual stresses and distortion, DVS Verlag, Germany.
[7]
A.A. Weels (1953), The mechanism of brittle fracture, Welding Journal, vol.32,
pp. 34 56.
[8]
Djarot B. Darmadi (2011), Study of residual stress mechanism using three elastoplastic bars model, Asian Transaction on Engineering (ATE), volume 01 issue 5,
pp. 76 80, 2011.
[9]
Jijin Xu, Ligong Chen, Jianhua Wang and Chunzhen Ni (2008), Prediction of
welding distortion in multipass girth-but welded pipes of different wall thickness,
Int. J. Advance Manufacturing Technology, vol. 35, 987 993.
[10] Wang Rui, Rashed Sherif, Serizawa Hisashi, Murakawa Hidekazu and Jianxun
Zhang (2008), Study on welding inherent deformations in welded structural
materials, Trans. Of JWRI, vol. 37, no. 1.
[11] Deng Dean, Murakawa Hidekazu, MA Ninshu and Serizawa Hisashi (2010),
Application of inherent deformation and interface element to prediction of
welding distortion during assembly process, Transaction of JWRI, vol.39 no. 2.
112
[12] Wang Jiangchao, MA Ninshu, Murakwa Hidekazu, Teng Bugang and Yuan
Shijian (2010), Prediction and measurement of welding distortion of thin shell
structure, Transaction of JWRI, vol. 39, no.2, pp.88 90.
[13] Farhad Kolahan, Amin Esmaeilzadeh, Mohammad Sadeghi and Mortaza Mosleh
(2011), Finite element simulation of post-buckling distortion for thin plate
welding positioned by fixture, World Academy of Science, Engineering and
Technology (WASET), vol. 74, pp. 77 81.
[14] Li Yajiang, Wang Juan, Chen Maoai and Shen Xiaoqin (2004), Finite element
analysis of residual stress in the welded zone of a high strength steel, Bull. Mater.
Sci., Vol. 27, No. 2, pp.127 132.
[15] Lixing Huo, Furong Chen, Yufeng Zhang, Li Zhang, Fangjun Liu and Gang Chen
(2004), 3D finite element numerical simulation of residual stresses on electron
beam welded BT20 plates, J. Mater Sci. Technolo, vol. 20, no.1, pp.117 120.
[16] Deng Dean, Kiyoshima Shoichi, Serizawa Hisashi, Murakawa Hidekazu and Horii
Yukihiko (2007), Numerical investigation on welding residual stress in 2.25Cr1Mo Steel Pipes, Transactions of JWRI, vol.36, no.1, pp. 73 90.
[17] Djarot B. Darmadi, Murugananth Marimuthu, Anh Kiet Tieu and John Norrish
(2012), Numerical prediction of residual stress in 316L bead on plate welds, 15th
(AMPT), Wollongong Australia, 23-26 September 2012, paper id:11355.
[18] M Abid, M Siddique and R.A Mufti (2005), Prediction of welding distortions and
residual stresses in a pipe-flange joint using the finite element technique,
Modeling and Simulation in Materials Science and Engineering, vol. 13, pp. 455
470.
[19] Dean Deng, Hidekazu Murakawa (2008), Prediction of welding distortion and
residual stress in a thin plate butt-welded joint, Computational Materials Science,
vol. 43, pp. 253 365.
[20] Dean Deng (2009), FEM prediction of welding residual stress and distortion in
carbon steel considering phase transformation effects, Materials and Design, vol.
30, pp. 359 366.
113
[21] M. Seyyedian Choobi, M. Haghpanahi and M. Sedighi (2010), Investigation of the

effect of clamping on residual stresses and distortion in butt-welded plates,
Scientia Iranica, Vol.17, no.5, pp. 387 394.
[22] The Thao Doan, Thien Phuc Tran, Tuong Quan Vo and Ill Soo Kim (2010), Finite
element prediction of residual stresses and distortion in T joint fillet welds , Proc.
APIEMS, pp. 1 8.
[23] Primos Mrvar, Jozef Medved and Sebastjan Kastelic (2011), Welding sequence
definition using numerical calculation, Welding Journal, vol. 90, pp. 148s 151s.
[24] Protocol for Phase 2 finite element simulation on the NET single-bead-on-plate
test specimen; https://odin.jrc.ec.europa.eu.
[25] X. Shan, C.M. Davies, T. Wangsdan, N.P. ODowd and K.M. Nikbin (2009),
Thermo mechanical modeling of a single-bead-on-plate weld using the finite
element method, International Journal of Pressure Vessels and Piping, vol. 86,
pp.110 121.
[26] C. Ohms, R.C. Wimpory, D.E. Katsareas and A.G. Youtsos (2009), NeT TG1:
Residual stress assessment by neutron diffraction and finite element modeling on
a single bead weld on a steel plate, International Journal of Pressure Vessels and
Piping, vol. 86, pp. 63- 72.
[27] M.C. Smith and A.C. Smith (2009), NeT bead-on-plate round robin: Comparison
of residual stress predictions and measurements, International Journal of Pressure
Vessels and Piping, vol. 86, pp.79 -95.
[28] S. Pratihar, M. Turski, L. Edwards, P.J. Bouchard (2009), Neutron diffraction
residual stress measurements in a 316L stainless steel bead-on-plate weld
specimen, International Journal of Pressure Vessels and Piping, vol. 86, pp. 13
19.
[29] M. Hofmann, R.C. Wimpory (2009), NeT TG1: Residual stress analysis on a
single bead weld on a steel plate using neutron diffraction at the new engineering
instrument STRESS-SPEC, International Journal of Pressure Vessels and Piping,
vol. 86, pp. 122 125.
114
Chapter 6
Finite Element Model of
Residual Stress Involving
Metallurgical Considerations
There is an increasing need for light steel structures involving the use of high strength
ferritic-steels. During welding, these ferritic steels undergo a solid state phase
transformation (SSPT) which affects the development of residual stress in the welded
structures. Taljat et.al. [1] defined residual stress as the stress that remains in a weld as a
result of the transformation from a liquid to solid phase associated with the
solidification and non-uniform cooling on the welded plate altered by SSPT. From that
definition, it can be seen that for ferritic-steels welding SSPT plays an important role in
the development of residual stress. From previously published papers
[2 14]
it can be
concluded that phase transformation alters the residual stress not only due to changes in
the material properties, but also by volumetric change.
SSPT alters the properties of materials and induces volumetric changes as a result of
different atomic arrangements (see Figure 2.17). When a ferritic steel is heated above its
transformation temperature (A1) its ferrite/pearlite structure (BCC) is starting to
transform to Austenite (FCC). This transformation finishes at a certain temperature
(A3). Due to the high Atomic Packaging Factor (APF) of austenite, when ferrite/pearlite
is transformed to austenite a volumetric shrinkage is induced while the structure is
heated. These existing austenitic structures, when cooled down to room temperature
may be transformed into martensite structures (BCT) or bainite, depend on the cooling
rate [15, 16]. For very low cooling rates, the austenite may transform to ferrite/pearlite (see
Figure 6.1). A CCT diagram for X70 pipeline steel is represented in Figure 6.1. Unlike
the other, microstructural components bainite has an acicular microstructure (not a
phase) found by E.S. Davenport and Edgar Bain
[17]
. Since austenitic structures (FCC)
have the highest possible Atomic Packaging Factor (APF) the phase transformation will
expand the volume of the structures as it cools down to room temperature. Since this
115
transformation significantly affects the development of residual stress, it should be

included in the FEM model.
C
Si
Mn
P
S
Cr
Ni
B
Ca
Cu
Mo
N
Nb
Ti
V
Al
0.053
0.19
1.6
0.012
0.0008
0.01
0.17
0.0001
0.0023
0.01
0.24
0.0047
0.043
0.023
0.044
0.042
Figure 6.1. CCT diagram for high strength X-70 steel [16]
This chapter is a preliminary study which intended to develop ANSYS code that was
able to model SSPT phenomenon especially the volumetric and material property
changes. Transformation plasticity was ignored at this stage and but is accommodated in
the next girth weld model (chapter 7). The developed logic and experimental procedure
and results mainly follow a paper by Lee and Chang[18] with some modification
elsewhere. From this paper; full martensite SSPT is implied for elements which are
heated above A1. Also the cooling rate was considered high that only martensite was
formed. In this thesis, martensite which is reheated above A1 will experience aged
martensite SSPT. A kinematic hardening model was used following Lee and Chang.
6.1. Welding procedures

This section describes welding procedures carried out by Lee and Chang
[18]
. The base
material was a 30mm thick, high strength carbon steel (POSTEN80) plate. The
electrode for the Flux-Cored Arc Welding (FCAW) process was a MGS-80 type, 1.2
mm in diameter. The joint was a butt-joint type with a double V surface preparation, as
shown in Figure 6.2. Five passes of FCAW were carried out to join the plate. First, three
welding passes were laid on the top V groove, and then the plate was turned over and
116
the next two passes were laid. The welding parameters for each pass are listed in Table
6.1. There is no information about the cross sectional shape of the weld bead for each
pass, so the shape of the weld bead was approached using the melting rate equation
[19]
for FCAW (equation 6.1).
Figure 6.2. Geometry and pass sequences of the welding process.

Table 6.1. Welding parameters for each pass
Pass
Current (A)
Voltage (E)
1
2
3
4
5
250
260
260
250
250
30
32
35
30
30
Welding Speed
(mm/s)
2.7
2
1.9
4.1
3.3
(6.1)
The melting rate MR (gr/min) is the mass of filler materials melted into the welding
joint. K, , and
are the constants; l is the stick-out length and A is the cross section
area of the filler materials. The relative area of the cross section of the weld-bead for
each pass can be evaluated by dividing MR by the welding speed.
A preheating temperature of 110 C for the first pass was applied and the inter-pass
temperature was 200 C 10 C.
6.2. Material properties

The chemical composition of both metals, plate and electrode, is shown in Table 6.2,
and the thermal properties and mechanical properties for both metals, which are
temperature dependent, are shown in Figures 6.3 and 6.4 respectively. There is no
difference in the thermal properties of both metals and Figure 6.3 describes the thermal
properties for both the base metal and weld metal. The temperature dependent Youngs
117
modulus of both metals is also superimposed over each other. The differences are found
in the temperature dependence of the yield stress and ultimate stress, as shown in Figure
6.4. The linear expansion coefficient and Poissons ratio are temperature independent
with values equal to 1.2 x 10-5 (1/K) and 0.3 respectively.
Table 6.2. Materials content of base metal and filler metal
Material
C
Si
Mn
P
S
Cr
Ni
Cu
V
Mo
B
POSTEN80 (mass,%)
0.07
0.3
0.91
0.015
0.004
0.45
0.97
0.02
0.038
0.45
0.0016
MGS-80 (mass,%)
0.05
0.44
1.35
0.006
0.001
0.6
2.3
0.25
Figure 6.3. Temperature dependent thermal properties of materials
Figure 6.4. Temperature dependent mechanical properties of materials

118
6.3. Considerations for phase transformation

Lee & Chang
[18]
observed the microstructure of the weld region and HAZ, and found
martensite in both regions, which confirmed that the weld region and HAZ experienced
a martensitic transformation. The values needed to determine the volumetric change
during martensitic transformation are the martensite start temperature, the martensite
finish temperature, and fraction volume of the martensite.
The martensite start temperature can be approached using a formula proposed by
Andrew[20], as shown in equation (6.2), which depends on the chemical content of the
materials.
Ms = 539 423C 30.4Mn 12.1Cr 17.7Ni 7.5Mo
(6.2)
Using the chemical content at Table 6.2, the martensite start for HAZ is 456 C and
427 C for the weld metal.
The martensite finish for both metals is 200 C. The martensite fraction is obtained using
Koistinen Marburgers [21] equation for carbon steels, as shown in equation (6.3).
(6.3)
For a small temperature increment, the martensite fraction differential can be obtained
from equation (6.4).
(6.4)
The volumetric change due to martensite phase transformation can be obtained using
equation (6.5).
sspt
Where
22]
sspt
= dfm x
expresses the volumetric change increment and
mar
mar
(6.5)
equal to 1.5 x 10-3 [18,
is the volumetric change of full martensite due to a solid state phase transformation.
The total volumetric change increment is a summation of volumetric change due to the
martensite phase transformation and to the difference in temperature as follows.
T
SSPT
119
(6.6)
By evaluating equation (6.6), it is clear that the total volumetric change depends on the
Ms values which are different for the base metal (456 C) and the weld metal (427 C).
Using these values, Figure 6.5 shows the effect of the martensite solid state phase
transformation on the coefficient of thermal expansion for the base metal and weld
metal. The coefficient of thermal expansion, without considering SSPT, is also plotted
on the figure for a comparison.
Figure 6.5. Effect of martensite on the coefficient of thermal expansion

SSPT is also found when ferrite/pearlite transforms to austenite. The transformation
started at A1 (760 C) and finished at A3 (920 C). The total austenitic transformation
strain is 2.288 x 10-3
[18, 20]
. For simplification, the volumetric change is considered
linear, and the volumetric change for any temperature at an austenitic phase
transformation range is approached using linear interpolation, as shown in equation
(6.7).
SSPT
(6.7)
The total volumetric change is the sum of solid state phase transformation and
elongation due to the temperature load, as shown below.
120
SSPT
(6.8)
The effect of austenitic solid state phase transformation to the coefficient of thermal
expansion is shown in Figure 6.6.
A1
A3
Figure 6.6. Effect of austenite on the coefficient of thermal expansion
The solid liquid phase transformation was also found and based on previous FEM
simulation (chapter 5), the effects should be considered. The solid liquid phase
transformation in thermal analysis is typically modeled by accommodating the latent
heat. A latent heat of 270 kJ/kg at a temperature transformation range of 1450 C 1500 C was applied
[22]
. When a metal becomes a liquid, the thermal conductivity
linearly increases three times, between 1450 C and 1727 C
[1]
. This increased thermal
conductivity is due to the effects of convective stirring.

Latent heat is also found when SSPT takes place. The latent heat of martensite SSPT
can be obtained by evaluating the enthalpy difference between Ms and Mf. The enthalpy
of martensite SSPT depends on the Ms and Mf temperatures which are determined by
the composition of the materials as shown in equation (6.2). Enthalpy at a certain
121
temperature can be evaluated using an equation proposed by Lee
[23]
, as shown in
equation (6.9).
(6.9)
Using equations (6.9) and (6.2), and considering the material compositions for weld
metal and base metal as shown in Table 6.2, the latent heat for martensitic SSTP of the
base metal is 75.07 kJ/kg and for the weld metal 66.51 kJ/kg.
The latent heat for austenite SSPT is 4200 J/mol
[24]
. Considering the materials
composition, the austenite SSPT latent heat for base metal and weld metal is 25.35
kJ/kg and 26.27 kJ/kg, respectively. Table 6.3 shows the values of this latent heat and
the temperature range of the suitable SSPT.
Table 6.3. Latent heat of the base metal and weld metal.
Latent Heat (kJ/kg)
Martensite SSPT
Austenite SSPT
Base Metal
Weld Metal
75.07
25.35
66.51
26.27
Melting
270
Temperature range ( C)
Base Metal
Weld Metal
456 - 200
427 200
760 920
1450 1500

As discussed in Chapter 5 the thermo-mechanical uncoupled analysis is acceptable
because the mechanical effects had an insignificant effect on the thermal analysis.
Uncoupled analysis cannot be applied to thermo-metallurgical effects because the phase
transformation alters the thermal properties of the materials whilst the temperature
history of certain elements determines the developed phase at the element. In
mechanical-metallurgical analysis phase transformation does affect mechanical analysis
(in this case it is residual stress) due to an alteration of the mechanical properties and
volumetric change, but the mechanical results have an insignificant effect on
metallurgical analysis. Deaconu
[25]
described how those analyses affect each other in
the Thermo-Mechanical-Metallurgical analysis shown in Figure 6.7.
122
Figure 6.7. Thermo-Mechanical-Metallurgical analysis
From the above considerations thermo-mechanical and mechanical-metallurgical

uncoupled analysis were applied to the FEM model whilst coupled analysis was applied
to the thermo-metallurgical analysis.
6.4.1. Coupled thermo-metallurgical model

As shown in Figure 6.7, the temperature histories of a certain element determined the
phase transformation of an observed element. When a ferrite/pearlite element is heated
above A1 the element transforms to austenite. This transformation is completed at
temperature A3. When cooled down to room temperature, this austenite transforms to
prime martensite, it started at an Ms temperature and finished at an Mf temperature
following Koistinen Marburgers law.
In multi-pass welding in this study, a prime martensite heated above A1 transforms to
austenite and when cooled down to room temperature transforms to aged martensite in
the Ms-Mf temperature range.
When a phase transformation takes place, a certain element alters its thermal properties,
which in turn affects the heat flow in the elements. The thermal property which is
altered significantly when phase transformation occurs is latent heat, which is needed to
transform one phase to another phase. The latent heats applied in this developed model
are shown in Table 6.3.
123
The ANSYS procedure for thermo-mechanical analysis, is described as a flow-chart in

Figure 6.8. First, the results from the thermal analysis without involving phase
transformation were obtained for the initial analysis. From the thermal analysis, the
temperature history for each pass can be retrieved from the ANSYS *.rth file and the
peak temperature for each element can be determined. If the peak temperature exceeds
A1 (760 C for both materials: weld metal and base metal), a solid state transformation
occurs on the observed elements. The element will experience pearlite/ferrite austenite
prime martensite in the first welding pass, and for the next passes the elements with
peak temperature exceeding A1 will experience pealite/ferrite austenite prime
martensite, prime martensite austenite aged martensite or aged martensite
austenite aged martensite depending on its initial phase.
The peak temperature can be obtained by comparing three consecutive temperature
histories for a certain element, namely Tn-1, Tn and Tn+1. Tpeak which equals Tn, is
obtained when the condition Tn-1 Tn Tn+1 is fulfilled. In the austenitic transformation
range, the transformation takes place when the observed element is heated, whilst at the
martensitic transformation range, the transformation takes place when the element is
cooling down. By comparing Tn to the previous temperature (Tn-1) it can be determined
if an element is heated or cooled, i.e., when Tn Tn-1 the element is heated whilst if Tn
Tn-1 the observed element is cooled.
Since the initial peak temperature was observed from thermal analysis without
considering phase transformation, it should be reformulated when phase transformation
takes place. The peak temperature of elements from previous analysis was replaced
using the last analysis with the phase transformation considered. This process was
iterated until a convergent solution was obtained. The convergence was evaluated by
evaluating the microstructure configured at a certain cross sectional area. From the
coupled thermo-metallurgy analysis, the initial phase, developed phase, and final phase
for certain passes can be determined. Those phases will be used in the thermometallurgy-mechanical analysis to determine the residual stress.
124
Figure 6.8. Coupled thermo metallurgical analysis
6.4.2. Development of residual stress while

considering the phase transformation
Thermal stress is basically the result of strain in an observed element. This strain,
regarding the stress-strain diagram, determines the thermal stress. When a temperature
load is applied to an element, due to the coefficient of thermal expansion and elevation
125
in temperature, the element will either expand or contract. When phase transformation
takes place, expansion/contraction due to volumetric change as a result of APF
difference occurs. The simple technique to incorporate the expansion/contraction of the
phase transformation is use of a modified coefficient of thermal expansion, as described
in equations (6.6) and (6.8).
Figure 6.9. Effects of prime and aged martensite on the yield stress of base metal
The phase transformation also alters the material properties of the observed element.
The most important material properties that affect thermal stress is yield stress. When
martensite is formed, the yield stress of the element increases, but an escalation of the
yield stress depends on the fraction of martensite which is calculated using equation
(6.3). The prime martensitic yield stress is 1050 MPa whilst the aged martensitic yield
stress is 1120 MPa[26]. The combined yield stress is obtained by weighting the initial
yield stress and martensitic yield stress according the existing martensitic fraction.
Figures 6.9 and 6.10 show the alteration in the yield stresses due to the existence of
martensite, for the base metal and weld metal respectively.
126
Figure 6.10. Effects of prime and aged martensite to yield stress of weld metal
6.4.3. Solutions and strategies

Some important phenomena obtained from the above discussion are summarised as
follows:
a) From coupled thermo metallurgical analysis the phases of each elements for a
certain welding pass can be predicted.
b) The existing phase is determined by three important factors: the initial phase, the
temperature gradient (heated or cooled), and the peak temperature of an element.
c) The initial phase for the first pass is ferrite/pearlite for all elements, and for the
next pass depends on the previous thermo-metallurgical analysis.
d) The predicted phase transformation in the coupled thermo metallurgical
analysis determined material properties of the materials which will be used in
uncoupled thermo mechanical and metallurgical mechanical analysis.
e) Three most important material properties that are affected by solid state phase
transformation are: the coefficient of thermal expansion ( ), the yield stress ( y),
and the existence of latent heat when phase transformation takes place.
127
There are three possible initial phases, depending on which pass the analysis is carried
out on: ferrite-pearlite, prime martensite, or aged martensite. The temperature gradient
has only two possibilities, heated or cooled, whilst the peak temperature may be divided
into two classes: below A1 and equal to or above A1, in which a determination will be
made as to whether the state transformation exists in a certain pass or not. Thus, there
will be 3 x 2 x 2 = 12 possible conditions where certain elements at certain times should
be modeled.
Table 6.4 shows all of the possible material models which should be accommodated in
the TMM analysis. As has been noted in item e) above, in mechanical analysis the
change due to SSPT is coefficient of thermal expansion
and yield stress y. Latent
heat is also altered because of SSPT but this is in the thermal analysis stage. The
subscripts in Table 6.4 express respective phases: - , M and M for ferrite/pearlite,
prime martensite and aged martensite. Since the yield stress of martensite (both prime
and aged martensite) as a function of temperature differs when it is cooled or heated
subscripts c and h are added, i.e.,
y(Mc) is yield stress temperature dependence
which follows Prime Cooled curve in Figures 6.9 or 6.10 whilst
y(Mh) follows
Prime Heated. Aged martensite follows the same manner. These materials models
interchange with each other depending on the previous three factors already mentioned
in item b): the initial condition, the temperature gradient and the peak temperature. The
initial condition depends on the predicted phase of previous passes, except for the first
pass where the initial condition is always pearlite-ferrite.
When an element is heated, the material model is equal to initial model, despite the peak
temperature experienced by certain observed elements. When an element is cooled and
its peak temperature is below A1, the material behaves in its initial phase, but if its peak
temperature is above A1, the behaviour of the element is altered due to martensitic
SSPT. Thus, for certain initial phases, only two element models are needed: the first is
for heated elements and cooled elements with peak temperature below A1, and the
128
Table 6.4. Change in material properties due to SSPT

Initial
Phase
Heated
Ferrite/Pearlite
Below A1, without austenite SSPT
Cooled
y( -
No SSPT
y( -
( - )
( - )
Above A1, austenite SSPT takes places
y( -
Prime Martensite SSPT
y(Mc)
(M)
( )
No SSPT
y(Mh)
Prime martensite
y(Mh)
( - )
( - )
y(Mh)
Aged Martensite SSPT
y(Mc)
( )
c(M)
No SSPT
y(Mh)
Aged martensite
y(Mh)
( - )
( - )
Aged Martensite SSPT
y(Mc)
y(Mh)
( )
(M)
second is for cooled elements that experienced a peak temperature above A1. Since
there will be three possible initial phases, the total number of element models required
129
for certain materials are six instead of twelve which will simplify the ANSYS
simulation. The material model for the base metal and weld metal is different so the
ANSYS program should model 6 x 2 = 12 models.
In the mechanical stage, the growing weld bead should also be accommodated using the
birth and death techniques, as discussed in Chapter 5. For a certain thermal load, a
certain grown weld bead is modeled and the rest are deactivated. Another modeling
phenomenon is melting. When an element has melted its plastic strain is removed. The
melting condition can be obtained from the temperature data of each element for a
certain time step in thermal analysis. To remove the plastic strain, the DESOL
command is used, as discussed in Chapter 5.
Figure 6.11. Procedure in the TMM analysis
The thermal load can be obtained directly from the thermal.rth file which is produced
by the thermal analysis in ANSYS. Unfortunately extracting data from this file is time
consuming when accessed for parameters, so a shadow database was created to save
130
computing time (Chapter 5). This database is in txt format and a certain parameter can
be quickly and easily retrieved from it.
The flow chart in Figure 6.11 describes how all those aspects are carried out in the
developed ANSYS procedure.
FIRST PASS
COUPLED THERMO - METALLURGICAL

Initial Elm.
Configuration
Elm. Config.
After heat load
Initial Elm.
Configuration
Iteration
Thist 1
Data Base 1
SECOND PASS
MECHANICAL
Initial Elm.
Configuration
Elm. Config.
After heat load
Elm. Config.
After heat load
Initial Elm.
Configuration
Iteration
Thist 2
Data Base 2
Elm. Config.
After heat load
NEXT PASS
Figure 6.12 Parameters transferred in multi-pass TMM
In multi-pass welding, the exchange of data for each step should be tightly controlled.
This data includes element configuration, the temperature histories of each element, the
time step and other miscellaneous data. The other important parameters are: the initial
phase while the element is heated, the predicted phase for each element when the
welded plate is cooling down to room temperature, and the peak temperature of each
131
element which is in the form of an array. How those data are interchanged between
passes is described in Figure 6.12. It should be noted that only the thermal load is
retrieved directly from the ANSYS result files, whilst the other data is obtained from the
created database.
Element meshes and configurations have been resumed from the first data base used in
first pass of the coupled thermo-metallurgical analysis. Since a mechanical analysis was
carried out sequentially after the coupled thermo-metallurgy analysis was completely
finished, the data and mesh configuration cannot be retrieved directly after each pass
has finished. After each pass has been completed the data are duplicated and used in
appropriate time step in the mechanical analysis. It should be noted that each pass
consists of many time steps, and in each time step a certain temperature for a certain
element is recorded in the ANSYS *.rth files. Since the data base was created outside
the normal ANSYS procedure, the data should be correctly flagged to guarantee the
appropriate temperature load data mapping for certain time steps and certain elements.
The element configuration and parameter transfer for the next pass (3, 4, and 5) follow
the same pattern as in pass 2.
6.4.4. The ANSYS APDL modeling

All of the above considerations need to be written in ANSYS program, but it the
standard ANSYS system needs some adjustment to accommodate them. ANSYS APDL
basically follows the pattern shown in Figure 6.13. Four types of consecutive blocks
should be defined: 1) ANSYS setting, 2) Pre-Processing, 3) Solution/Processing and 4)
Post-Processing. ANSYS setting defines what kind of analysis will be carried out such
as: transient or steady state analysis, thermal or mechanical analysis, etc. The preprocessing block consists of the material model, the geometrical model and the load
definition. In the material model the type of element, material properties, etc., are
defined. The geometrical model defines the geometry of the specimen, meshing
configuration, the relationship between geometries, and initial configuration of the
microstructure etc. Load definition block describes loads, constraints, initial and
boundary conditions. In the thermal analysis, the load was defined as a moving heat
source. The heat source was represented by a combination of Goldaks distributed
volumetric heat source model and the uniform heat source model, as discussed in
132
Chapter 4. Goldaks distributed volumetric heat source represents the heat embedded by
a welding torch. The heat source parameters rx, ry and rz are equal to those used
previously in Chapter 4, i.e. 5mm, 2mm and 3mm respectively. The value of heat for
certain relative position depends on the voltage, current and welding efficiency for each
pass. The voltage and current are tabulated in Table 6.1 and the welding efficiency was
considered equal to 0.85 [9,18]. The efficiency used here is different from that used in the
previous chapter since the welding process is different, in this case FCAW is used
whereas previously TIG was considered. Even in the same welding process the
efficiency may differ depending on various parameters. In the current analysis the data
used in references [9, 18] is employed since it aims to conduct a modeling exploration
based on existing data. The heat intensity load for a each element was calculated using
equation 4.4. Following data by Lee
[9]
and Lee & Chang
[18]
, convection and radiation
were represented by convection coefficient of 15W.m-2.K and 0.2 emissivity

respectively. It should be noted that this value differs from values in chapters 4 and 5
which may be caused by the different environment. The initial temperature was equal to
the preheat temperature of 110 C. The uniform temperature load of 2400 C represents
the heat from the melted filler metal that formed the weld bead for each pass. The heat
of Goldaks heat source should be reduced by the heat to melt the filler metal.
ANSYS setting
PRE-Process
Material model
Geometry model
Load definition
Solution/Process
Post-Process
Figure 6.13. ANSYS APDL basic pattern
133
The solution/process block typically executes the calculation based on FEM analysis
solely by the ANSYS program. The results are retrieved and interpreted in the postprocess block. To obtain an effective process and results, the program should be built by
following the basic structure shown in Figure 6.13.
Constants or values (called parameters) can be defined in any process (Pre-processing,
processing or post processing). Another benefit of using parameters is that they can be
implemented more quickly than using the results files obtained from ANSYS postprocessing approach discussed previously.
From Figure 6.7 which described the residual stress prediction involving the
metallurgical consideration, three different analyses were carried out: thermal analysis,
metallurgical analysis and mechanical analysis. Microstructures are defined in the
material model and as mentioned in Section 6.4.3, twelve different material models are
defined to represent all possible microstructures.
Microstructures are defined as a material which follows the properties described in
Section 6.3, especially those shown in Figures 6.5 and 6.6, and Figures 6.9 and 6.10 in
Section 6.4.2. The microstructures were represented by material models and this
requires 12 models (see section 6.4.3). When transformation takes places in an element
in a certain time the element is switched to the proper element model as explained
previously (see Table 6.4). The existing microstructure depends on the initial
microstructure and the peak temperature reached in a certain pass and its temperature
gradient in a certain time step. The flow chart of the thermo-metallurgy-mechanical
analysis in ANSYS APDL is shown in Figure 6.14. To ensure a consistent geometrical
model in each analysis, the RESUME command was used to copy the preprocessing
stage from each analysis.
In the metallurgical analysis, there is no finite element calculation performed by
ANSYS. The geometry model especially the configuration of the microstructure was
duplicated from the thermal analysis. This configuration of the microstructure was
recalculated when welding was ongoing. All the predicted microstructures configuration
were saved in PRED1(i), PRED2(i), PRED3(i), PRED4(i) and PRED5(i) for element i
at pass 1st, pass 2nd, pass 3th, pass 4th and pass 5th respectively whilst for initial
134
microstructure configurations are saved in INIT1(i), INIT2(i), INIT3(i), INIT4(i) and

INIT5(i). The initial microstructure for all elements in the first pass was austenite.
Figure 6.14. TMM ANSYS program.

Some of the parameters were obtained from the post processing stage in the thermal
analysis using the *GET command. Parameters are in an array form. Typical parameter
configurations are shown in Figure 6.15. By default the dimension of parameters array
135
is 3, but the real dimension is defined in *DIM command. The TIME(i) parameters are
in one dimensional array and are described in real time (in seconds) for certain ith time
steps. The second and third values in definition for the one dimension array are kept as
1. TEMP(i, j) defines the temperature (in C) at ith time step of node j. TPEAK(i) shows
the peak temperature of elements i at a certain pass. TMAX1(i) defines real time
(second) when the peak temperature of element i was reached for the 1st pass whilst
TMAX2(i), TMAX3(i), TMAX4(i) and TMAX5(i) are for the 2nd pass, 3th pass, 4th pass
and 5th pass respectively. This database was formed from ANSYS post processing files
in the thermal analysis stage, however when it is coupled with metallurgical analysis,
parameters in the database can be revised. The data is shown in Figure 6.15 in its raw
form.
*DIM,TEMP
.
.
.
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*SET,TEMP
*DIM,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
*SET,TMAX1
.
.
.
,ARRAY,
147,
(
126,
65402,
(
127,
65402,
(
128,
65402,
(
129,
65402,
(
130,
65402,
(
131,
65402,
(
132,
65402,
(
133,
65402,
(
134,
65402,
(
135,
65402,
(
136,
65402,
(
137,
65402,
(
138,
65402,
(
139,
65402,
(
140,
65402,
(
141,
65402,
(
142,
65402,
(
143,
65402,
(
144,
65402,
(
145,
65402,
(
146,
65402,
(
147,
65402,
,ARRAY,
59016,
(
1,
1,
(
2,
1,
(
3,
1,
(
4,
1,
(
5,
1,
(
6,
1,
(
7,
1,
(
8,
1,
(
9,
1,
(
10,
1,
(
11,
1,
(
12,
1,
65402,
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1,
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1),
1,
200.5419023837
199.8736012935
199.2146155278
198.4845373333
197.7660651966
196.0865018539
194.4666987213
192.9016331259
191.3881535053
189.9233473592
187.3004339905
184.8254453043
182.4814782146
180.2567707373
177.4590127784
174.8348604693
172.3638655197
170.0298484041
165.7207786258
160.9040351313
156.5949137780
152.6955366859
1,
11.85184800000
11.85184800000
7.407405000000
10.37036700000
8.888886000000
11.85184800000
10.37036700000
10.37036700000
11.85184800000
11.85184800000
13.33332900000
13.33332900000
Figure 6.15. Saved in parameters.
Beside any previous array parameters which were derived from the post-processing of
the thermal analysis, others were generated based on considered conditions. The
136
generated parameters are the initial and predicted microstructure configurations. The
initial microstructure for element i are labeled in INIT1(i), INIT2(i), INIT3(i), INIT4(i)
and INIT5(i) for the 1st pass, 2nd pass, 3th pass, 4th pass and 5th pass respectively, whilst
PRED1(i), PRED2(i), PRED3(i), PRED4(i) and PRED5(i) described the predicted
microstructure. For the next pass the initial microstructure was set directly from
microstructure predicted from the previous pass as shown graphically in Table 6.4.
The microstructure configuration and temperature at the ends of each pass were
evaluated. If the microstructure and temperature did not converge, the thermal analysis
was repeated. In this thermal analysis the microstructure configuration was redefined
based on parameters modified in the metallurgical analysis. The initial microstructure
configuration was defined using INIT1(i), INIT2(i), INIT3(i), INIT4(i) and INIT5(i)
parameters. When the time step was higher than TMAX1(i), TMAX2(i), TMAX3(i),
TMAX4(i) and TMAX5(i) for pass 1, 2, 3, 4 and 5 respectively this meant that the
element(i) had cooled, the element(i) had swapped to PRED1(i), PRED2(i), PRED3(i),
PRED4(i) and PRED5(i). This coupled thermo-metallurgical analysis was iterated until
the microstructure and temperature converged.
After the coupled thermo-metallurgical analysis was finished the last step was
Mechanical Analysis (Figure 6.14). As in the metallurgical analysis, the pre-process
step from thermal analysis was resumed to ensure that equal geometry and data
mapping can be carried out correctly. The microstructure for each pass for a certain time
step was generated again using parameters in the database. For a certain pass the first
microstructure configuration was determined by INIT1(i), INIT2(i), INIT3(i), INIT4(i)
and INIT5(i). As in thermal analysis which considered microstructure, when a certain
element had cooled down (time step was equal or greater than TMAX1, TMAX2,
TMAX3, TMAX4 and TMAX5 for the 1st, 2nd, 3th, 4th and 5th passes, respectively) the
element microstructure was swapped to PRED1(i), PRED2(i), PRED3(i), PRED4(i) and
PRED5(i).
In mechanical analysis the distribution of temperature in thermal analysis was applied as
a thermal load using the LDREAD command. The FEM method which was solely
calculated by ANSYS was performed in the solution/process block and the solution can
be evaluated in the post-processing stage.
137
6.5. Results and discussion

It should be noted that some simplification in this TMM model have been made. The
main goals of this chapter were to check and evaluate ANSYS APDL program to
determine whether it can manifest the logic discussed in this chapter. The correctness of
the logic will be evaluated later after the results have been obtained.
As shown in Figure 6.1, at a slower cooling rate, not only martensite will be developed
in the weldment, bainite and ferrite may also be developed and mixed with martensite.
Also, not all austenite when cooled down to the room temperature at a fast cooling rate
transformed to martensite, some austenite may have also retained.
The first simplification considered that only martensite would be developed when a
certain element was heated above A1 and the cooling rate was neglected. This
assumption is acceptable since phase transformation has a significant effect on the
residual stress at areas close to the weld line (this will be discussed in more detail later
in Chapter 9). In this region the cooling rate is typically high and as a consequence
martensite will be formed. Martensitic volumetric change has high value of SSPT and
thus affects significantly the residual stress. The other reason for the simplification is
that a different X-70 material with a different CCT diagram will be obtained. Since this
CCT diagram in Figure 6.1 is not a CCT diagram for POSTEN80 and MGS-80, sticking
with the CCT diagram will not be useful.
The second simplification considered that all of austenite was transformed to martensite.
In previous discussed TMM theory, the residual stress can alter due to phase
transformation and the corresponding volumetric change in the material properties, but
since martensite has the lowest possible APF and the highest volume expansion,
neglecting any retained austenite is an extreme condition. A consideration of any
retained austenite in the residual stress prediction will provide values between the TMM
analysis with 100% martensite and TMM without any martensite formation, i.e. no
SSPT. Another reason is that no data is available on what percentage of retained
austenite is available for certain cooling rate for POSTEN80 and MGS-80.
138
Figure 6.16 Predicted microstructure configuration.
The microstructure after convergence for each pass is presented in Figure 6.16. In
Figure 6.16a, i.e. with the microstructure at the end of the 1st pass, the elements close to
the weld bead which have a peak temperature above A1 were transformed into
martensite (shown in yellow). The same observation can be found in Figure 6.16b with
the microstructure at the end of the 2nd pass, except for the elements which were already
in a martensitic phase, the peak temperature above A1 caused the element to transform
139
to aged martensite (shown in red). The microstructure configured at the end of the 3rd,
4th, and 5th pass follow the same logic as at the end of the 2nd pass. Figure 6.16
demonstrates that the ANSYS code can follow the developed logic. However, this
microstructure configuration is not validated with experiment data which is beyond the
scope of this thesis.
The distribution of residual stress by considering SSPT is presented in Figure 6.17.
Longitudinal residual stress was measured on the surface in the mid section of the plate.
It should be noted that only experimental longitudinal residual stress is provided by Lee
& Chang [18] which is considered enough for the objective of this chapter; development
of ANSYS code that can model the SSPT phenomenon. The FEM prediction of the
distribution of longitudinal residual stress in the mid cross section is presented in Figure
6.18. An inspection of a path on the top surface of Figure 6.18 indicates a longitudinal
residual stress which is comparable to the experimental results.
Figure 6.17 Longitudinal residual stress distributed on the whole model
140
Figure 6.18 Longitudinal residual stress distributed on mid cross section area
Figure 6.19 Longitudinal residual stress distributed on the top path of the mid cross
section area.
Figure 6.19 shows a comparison between FEM predictions with and without SSPT and
the experimental results to obtain a general insight into the effects of SSPT on residual
stress. It can be argued from Figure 6.19 that the martensitic SSPT decreased the tensile
residual stress in the area close to the weld beads due to an expansion of martensitic
SSPT that occurred while the elements were cooled to room temperature. Since the
expansion was constrained by the surrounding area away from the weld line, it caused
141
compressive stress which reduced the predicted residual stress without SSPT. Figure
6.19 shows that the SSPT led to a prediction that was closer to the experimental results,
thus indicating that SSPT should be used to predict residual stress of high strength
steels.
Figure 6.20 Longitudinal residual stresses without SSPT consideration.

The prediction of longitudinal residual stress on the same cross sectional area without
SSPT is presented in Figure 6.20. When Figures 6.18 and 6.20 are compared it shows
that the residual stress using SSPT caused residual stress that was not distributed
smoothly. The distribution of residual stress from the FEM prediction without SSPT
was redistributed similar to the predicted phase shown in Figure 6.16.
The peak temperature in the top surface of both models (with and without SSPT) was
evaluated and as shown in Figure 6.21, there are no significant differences. As discussed
with the developed model, SSPT does not change the thermal properties, apart from the
latent heat, when austenitic SSPT and martensitic SSPT are exhibited. Moreover, this
latent heat is much lower than the latent heat of melting (270 kJ/kg). As a result of all
the assumptions mentioned above, the peak temperature obtained from a thermal
analysis of the evaluated path has almost the same for both cases.
On the other hand the mechanical properties that do change due to SSPT are the
coefficient of thermal expansion and the yield stress which created residual stress in the
142
model with SSPT and was different to the case without SSPT. As a result, the residual
tensile stress in the model with SSPT was reduced near the weld line and increased
away from the weld line. It is interesting to relate the distribution of peak temperature
(Figure 6.21) with the prediction of residual stress (Figure 6.19) and these figures are
represented again in Figures 6.22 and 6.23 to evaluate the area where the peak
temperature was above A1. Figure 6.22 shows that the peak temperature is equal to A1
when the distance from centre of the weld was between 15.63mm and 18.75mm.
Residual stress in the model involving SSPT decreased close to the weld line and
increased away from it, and coincidently this change from decreased residual stress and
increased residual stress also lie between 15.63mm and 18.75mm (Figure 6.23). If the
peak temperature of an element is above A1, the element experience SSPT and as a
consequences it will expand when cooling from Ms to Mf temperatures. This expansion
that is constrained by the surrounding materials causes compressive stress in the area
where the peak temperature is above A1. That is why, in the area where the peak
temperatures were above A1 (close to the weld line) the high tensile residual stress was
compensated by the compressive stress developed due to martensitic SSPT.
Furthermore, the reduction in residual stress close to the weld line caused an increase in
residual stress away from the weld line. This phenomenon was caused by the force
equilibrium principal.
Figure 6.21 Peak temperatures on the top path at mid cross section area.
143
Figure 6.22 Distributions of peak temperature on the path close to the weld line.
Figure 6.23 Distributions of residual stress on the path close to the weld line.
6.6. Conclusion
The prediction of residual stress using the logic described in this chapter can provide
results that match with previous experiment measurements. Using ANSYS TMM
analysis was undertaken following the logic shown in Figure 6.14. For the APDL mode,
144
this logic must be written in ANSYS code which is understood by the ANSYS program.
Much effort was spent to develop this code. It can be claimed that the developed
ANSYS code can follow very well the logic shown in Figure 6.14.
Principally, SSPT affects residual stress through three phenomenons: volumetric change
due to atomic arrangement, material properties alterations and transformation plasticity.
The first two have been discussed in this chapter and the transformation plasticity will
be discussed in the model of pipeline girth weld joints (Chapters 7 and 9). The
ommission of transformation plasticity may have caused the residual stress overestimation by the FEM model.
Involving martensitic SSPT in the prediction of residual stress reduced the residual
tensile stress near the centre of the weld. The martensite SSPT causes expansion while
the transformed elements are cooling to room temperature, and this expansion is
constrained by the surrounding area that causes compressive stress. The elements near
the weld line experienced SSPT because their peak temperature was higher than A1,
which is why the compressive stress that developed near the weld line compensated for
the residual tensile stress as a result of thermal load.
A new feature in this thesis is the consideration of both prime martensite and aged
martensite in a multi-pass welding model. Aged martensite was also considered for
elements which were in previous phase had prime martensite SSPT.
The use of the parametric database is also a new feature. As shown in Figure 6.14, these
parameters play an important role in the development of the ANSYS code. There are at
least three advantages in using a parametric database: 1) Faster simulation time because
it is easy to use and fast to retrieve compared to the ANSYS post processing files (*.rth,
*.rts files), and 2) the results can be accessed easily, and 3) it makes available the
required parameters that are not provided by ANSYS default. An evaluation of peak
temperature on the top path, which was presented in Figure 6.21, was obtained directly
from this parametric database and it may be difficult to obtain it directly from the
ANSYS post processing files. Furthermore the database can be accessed by another
program, since the database is in *.txt format. This fact makes it possible to combine the
ANSYS results with another program.
145
References
[1]
B. Taljat, B. Radhakrishnan, T. Zacharia (1998), Numerical analysis of GTA

welding process with emphasis on post-solidification phase transformation effects
on residual stresses, Materials Science and Engineering A, vol.246, pp. 45 54.
[2]
J.B. Leblond and J. Devaux (1984), A new kinetic model for anisothermal
metallurgical transformations in steels including effect of austenite grain size,
Acta Metall, vol. 32, no.1, pp. 137 146.
[3]
J.B. Leblond, G. Mottet and J.C. Devaux (1986), A theoretical and numerical
approach to the plastic behavior of steels during phase transformation II. Study
of classical plasticity for ideal-plastic phases, J. Mech. Phys. Solid, Vol. 34, no. 4,
pp. 411 432.
[4]
J.B. Leblond, G. Mottet and J.C. Devaux (1986), A theoretical and numerical
approach to the plastic behavior of steels during phase transformation I.
Derivation of general relation, J. Mech. Phys. Solid, Vol. 34, no. 4, pp. 395 409.
[5]
P. Ferro, H. Porzner, A. Tiziani and F. Bonollo (2006), The influence of phase

transformations on residual stresses induced by the welding process 3D and 2D
numerical models, Modeling and Simulation in Materials Science and
Engineering, Vol. 14, pp. 117 136.
[6]
Deng Dean, Murakawa Hidekazu (2006), Prediction of welding residual stress in

multi-pass butt-welde modified 9Cr-1Mo steel pipe considering phase
transformation effects, Computational Materials Science, vol. 37, pp. 209 219.
[7]
A.H. Yaghi, T.H. Hyde, A.A. Becker and W. Sun (2008), Finite element
simulation of welding and residual stresses in a P91 steel pipe incorporating
solid-state phase transformation and post-weld heat treatment, J. Strain Analysis,
vol. 43, pp. 275 293.
[8]
Dean Deng (2009), FEM prediction of welding residual stress and distortion in
carbon steel considering phase transformation effects, Materials and Design, vol.
30, pp. 359 366.
146
[9]
Chin-Hyung Lee (2008), Computational modeling of the residual stress evolution

due to solid-state phase transformation during welding, Modeling and Simulation
in Materials Science and Engineering, Vol. 16, pp. 1 16.
[10] D Camilleri and TGF Gray (2005), Computationally efficient welding distortion
simulation techniques, Modeling and Simulation in Materials Science and
Engineering, vol. 13, pp. 1365 1382.
[11] Murakawa Hidekazu, Beres Miloslav, Vega Adan, Rashed Sherif, Davies Catrin
M., Dye David and Nikbin Kamran M. (2008), Effect of Phase Transformation
onset Temperature on Residual Stress in Welded Thin Steel Plates, Transactions
on JWRI, Vol. 37, No. 2, pp. 75 80.
[12] J.A. Francis, H.K.D.H. Bhadeshia and P.J. Withers (2007), Welding residual
stresses in ferritic power plant steels, Materials Science and Technology, vol.23,
no.9, pp.1009 1020.
[13] Masao Toyoda and Masahito Mochizuki (2004), Control of mechanical properties
in structureal steel welds by numerical simulation of coupling among
temperature, microstructure and macro-mechanics, Science and Technology of
Advanced Materials, vol. 5, pp. 255 266.
[14] M. Becker, C. Jordan, SK Lachhander, A Mengel, M. Renauld (2005), Prediction
and measurement of phase transformation, phase-dependent properties and
residual stresses in steels, Research Report LM-05K049.
[15] M.I. Onsoien, M. Mhamdi and A. Mo (2009), A CCT diagram for an offshore
pipeline steel of X70 type, Welding Journal, vol. 88, pp. 1s 6s.
[16] M.R. Forouzan, A. Heidari and J. Golestaneh (2009), FE simulation of submerged
arc welding of API 5L-X70 straight seam oil and gas pipes, Istiqlaal, vol.28, no.1,
pp. 93 110.
[17] Smith, Cyril Stanley (1960), A history of metallography, University of Chicago
press, p. 225.
147
phase transformation, Computational Materials Science, vol. 46, pp.1014 1022.
[19] John Norrish (2006), Advanced welding processes, Woodhead Publishing in
Materials, Cambridge, England.
[20] K.W. Andrews (1960), Empirical formulae for the calculation of some
transformation temperature, J. Iron Steel Res. Int. 203, pp. 721 727.
[21] D.P. Koistinen and R.E. Marburger (1959), A general equation prescribing the
extent of the austenite-martensite transformation in pure iron-carbon alloys and
plain carbon steels, Acta Metallurgica, vol. 7, pp. 59 60.
[22] Ching-Hyung Lee (2008), Computational modeling of the residual stress
evolution due to solid-state phase transformation during welding, Modeling and
Simulation in Materials Science and Engineering, vol. 16, pp. 1 16.
[23] Seok-Jae Lee and Young-Kook Lee (2009), Latent heat of martensitic
transformation in a medium-carbon low-alloy steel, Scripta Materialia, vol. 60,
pp. 1016 1019.
[24] R.A. Jaramillo, G.M. Ludtka, R.A. Kisner, D.M. Micholson, J.B. Wilgen, G.
Mackiewicz-Ludka, N. Bembridge and P.N. Kalu (2005), Investigation of phase
transformation kinetics and microstructural evaluation in 1045 and 52100 steel
under large magnetic field, Solid-to-solid phase transformations in inorganic
materials: Diffusional transformations, edited by J.M. Howe, D.E. Laughlin, J.K.
Lee, U. Dahmen, W.A. Soffa, The Minerals, Metals and Materials Society, pp.
893-898.
[25] Viorel Deaconu (2007), Finite element modeling of residual stress A powerfull
tool in the aid of structural integrity assessment of welded structure, 5th
International Conference Structural Integrity of Welded Structures, Romania.
[26] RWK Honeycombe and HKDH Bhadeshia (1995), Steels microstructure and
properties, 2nd edition, Gray Publishing, UK.
148
PART IV
THE PIPELINE
GIRTH WELD
Chapter 7
Residual Stress Analysis of
The Girth Weld Joint
This chapter discusses the modeling of the girth weld joint and the experimental work
used as a validation. The first thing to be discussed is the welding procedure and
parameters that were used in the girth welding. Weld modeling is then being carried out
using approach previously described. The third part of the investigation is the
measurement of residual stress and the last part of this chapter is a comparison between
the FEM model and experimental results. Comparison of the actual temperature
histories from certain nodes which represent the position of the thermocouples are made
with those predicted using the coupled thermo-metallurgical analysis. A second
comparison is made between the measured and the FEM prediction of residual stress.
7.1. Welding procedure and parameters

Two X70 pipes 42cm diameter, 21cm length and 8mm thickness were girth welded. The
girth weld was carried out using an automated rig (Figure 7.1) which provided constant
parameters. Three control motors with attached welding torches provided motions
relative to the welded pipe. The first motor moves the welding torch parallel to the axial
centerline of the pipe, and the second motor moves the welding torch in a radial
direction. Both motors were used to place the welding torch in the desired weld line and
to adjust the CTWD (Contact Tip to Work Distance) respectively. The third motor
moves the welding torch around the pipe at a speed that determines the welding
velocity.
149
Figure 7.1.
The rig for girth welding
It may have been easier to let the pipe revolve around its own axis while keeping the
torch stationary. However moving the welding torch is closer to the real situation since
it is impossible to turn the pipe when girth welding of pipelines in the field. As in actual
girth weld line of pipelines, no heat treatment was applied (such as stress relieving or
annealing) to the specimen.
Figure 7.2 shows the main interface seen on the computer screen when the girth weld
controller is operating. The welding speed, CTWD, and the length of the stringer bead
are shown on the screen. Zero point positioning in the circumferential direction can be
applied by clicking the JOG tab bar and the rotational motor can be moved manually
using the pendant controller shown in Figure 7.3. Two buttons are available, i.e. the red
and the green button, which move the welding torch clockwise or anticlockwise
respectively.
150
Figure 7.2 Main menu of mechanised girth weld program.
Figure 7.3 The pendant controller.
Choosing the setup menu of the main menu in Figure 7.2 will display the System Setup
window shown in Figure 7.4, which is basically used to define the zero point. After
entry of input data (Pipe Thickness, Root Face, Gas Shield Base to Contact Tip, Gas
Shield Base to Pipe measured, Pendant Sensitivity and Welding/Motion Delay), the
Calibrate tab bar is chosen to set the CTWD real distance using the red and green
151
buttons on the pendant which control the movement of the torch away from and towards
the pipe surface respectively. The Centre Torch tab bar is used to manually control the
initial transverse position of the welding torch, again using the red and green buttons on
the pendant. Pushing the OK tab bar will again display the main menu in Figure 7.2.
Once the zero point has been set, the system is ready to perform girth weld by pushing
the START task bar of the main menu.
Figure 7.4 System Setup menu

Before welding can commence, the pipe should be setup (centred) on the rig (Figure
7.5). There are two goals that must be achieved when centring the pipe, the first is to
provide constant CTWD and the second is to provide an equal transverse distance to the
weld bug. The transverse movement by pendant control is limited by the construction of
the rig ( 3cm).
152
Figure 7.5. Centering the pipe: a) radial (left) and b) transverse position (right)
Figure 7.6. The pipe is tilted from it ideal position: a) ideal position, b) tilted
horizontally, c) tilted vertically and d) tilted arbitrarily.
153
Figure 7.7. The pipe is shifted from it ideal position: a) ideal position, b) shifted
horizontally, c) shifted vertically and d) shifted arbitrarily.
Figure 7.8. Presenting dial indicator measurements in a diagram: a) radial position (left
side) and transversal position (right)
154
Positioning the pipe in a desired position was not easy because the pipe can be tilted or
shifted horizontally, vertically, or arbitrarily as shown in Figures 7.6 and 7.7. The
typical results from the dial indicator measurement are presented in graphics and shown
in Figure 7.8. Using both graphics it can be determined what should be done to obtain
the ideal pipe position.
Furthermore, the pipe was not always perfectly round and thus a surface on one side
was not matched with its opposite side. This is known as a hi-lo problem, so to
provide good joint, a hydraulic jack was used to push the pipe externally or internally to
align the edges of the pipes (as shown in Figure 7.9).
Figure 7.9. Using hydraulic jack to minimise the hi-lo problem
Figure 7.10 End pipe preparation (gap) and design of the welding passes
End preparations on the 8mm thick pipe wall followed AS 2885.2 as shown in Figure
7.10. First a machined surface preparation provided the trimmed 60 edge was made.
This edge was refined using a manual grinder and a 1.60.8 mm land is formed with a
hand file.
155
Figure 7.11 Camera and its box for welding torch transversal positioning.
Figure 7.12Welding arc and gap as seen on camera: a) before welding and b) when
welding is in process.
A perfectly consistent gap parallel to the welding jig is almost impossible to be achieved
due to the sum of small errors in surface preparation and positioning the pipe in the
welding rig. To accommodate the inconsistency a pendant and a camera to detect the
relative position of the welding torch to the gap were used. Figure 7.11 shows the
camera with a cover to protect the camera from spatter and also to attach the light filter.
The pendant initiates one of the motors mentioned above which control the transverse
movement of welding torch using the dial (Figure 7.3). The sensitivity of the dial can be
adjusted in the System Setup menu (Figure 7.4). By monitor the welding arc and the gap
ahead (as shown in Figure 7.12) where the welding torch should be positioned
transversally by the pendant can be predicted.
156
While welding was ongoing, the welding parameters (current and voltage) were
recorded using an AMV4000 data logger (Figure 7.13). A sensor probe 1001 (Figure
7.14) was used to detect the current whilst the difference in voltage was measured
directly at the pipe and filler wire.
Figure 7.13 AMV4000
Figure 7.14 Current probe
The welding speed of 300 mm/min (Figure 7.2) is controlled by the motor on the rig.
The speed of wire feed is controlled by the GMAW welding machine and different
values are applied for the root pass and filling passes. A semi automatic STT (Surface
Tension Transfer) Lincoln GMAW machine as shown in Figure 7.15 was used. The
STT is a controlled short-circuit transfer mode patented by Lincoln Electric that makes
the GMAW process more applicable to the welding of open root weld preparations. STT
uses a closed loop control where the arc current and voltage were automatically adjusted
157
to provide the desired weld bead. The arc current and the speed of the wire feed were
controlled independently while the heat input was kept constant regardless of any
alteration to the electrode extension [1-4].
Figure 7.15 STT Lincoln GMAW machine

Table 7.1. Suggested parameters for high strength carbon steels [2].
Included angle
60
Root opening
1.6 2.8 mm
Root face (land)
1.6mm
Welding position
5G vertical down
Shielding gas
100%CO2
Backing gas
NA
Electrode
L-56 ER70S-6 ( 1.2mm)
Peak current
360A
Background current
55-65A
Tail-out current
0
3m/min
Wire feed speed
The parameters controlled by this machine are the wire feed, peak current, background
current and tail-out current. The wire feed controls the rate of deposition and the peak
current controls the arc length. The background current and the tail-out current control
the heat-input. Tail-out current serves as a coarse heat control and background current is
a fine heat control. The peak current affects the shape of the root face whilst the
background current and tail-out current affect the back bead. Suggested parameters for
welding the root pass of high-strength steels are shown in Table 7.1. Following the
suggested parameters especially welding position, it should be noted that for each pass,
158
two different runs were applied from the 12 oclock position to the 6 oclock position:
either clock wise or anti clock wise. The filling pass, especially passes 2,3 and 4, used a
higher wire feed speed (4 mm/min) and as consequence other parameters were changed
to provide a good weld bead, i.e., the peak current was 415 Amp and the background
current 90 amp.
FIGURE 7.16 Data logger and thermocouples attached to the surface of the pipes
Figure 7.17 Experiment set up

A National Instruments USB 6211 with a seven port data logger was used to record the
temperature histories at different position on the welded pipe (Figure 7.16). These
temperature histories were used as validation of the thermal analysis. A sketch of the
overall experimental set up is shown in Figure 7.17.

The mesh design of the pipe is shown in Figure 7.18. A very fine mesh was applied
close to the weld line because it was necessary to model the moving heat source and
159
also an abrupt thermal profile close to the weld. A symmetrical model to save
computation could not be applied because the individual weld bead was not always
symmetric to the centre of V groove especially for the 3rd and 4th passes (Figure 7.10).
The model consists of 78380 elements (SOLID70 and SOLID45 for coupled thermometallurgical analysis and mechanical analysis respectively) and 94976 nodes.
Figure 7.18 Mesh configurations for the girth weld joint

A cross section of the weld bead was obtained following the pass design shown in
Figure 7.10 and thus the cross sectional area for each pass can be calculated by
considering the diameter of the wire (dwire), wire feed speed (wfs) and welding speed (v)
using equation 7.1. Using Figure 7.10 and equation 7.1, the configuration of the weld
beads was predicted as shown by the ANSYS graphs in Figure 7.19.
Figure 7.19 Predicted cross sectional area (left) and mesh design on the cross section
(right)
160
(7.1)
SSPT was taken into account when the FEM for the girth weld was carried out. The
finite element model followed the methodology discussed in Chapter 6. First, a coupled
thermo-metallurgical analysis was performed iteratively until the phase configuration
and temperature at a chosen position converged. Using the results from the coupled
thermo-metallurgical analysis, a sequential (uncoupled) mechanical analysis was carried
out to obtain the residual stress.
Using the suggested parameter
[2]
, two runs were carried out for each pass: clock wise
and anti clock wise as shown in Figure 7.20. The weld started at 12 oclock and ended
at 6 oclock. The residual stress was measured at 3 and 9 oclock paths which were far
enough from the welding start and end, and the FEM prediction on these paths were
compared to the measurement to evaluate the correctness of FEM model.
Figure 7.20 Sketch of the girth welding process.

Some important aspects which are different with plate welding are discussed in the next
sections.
7.2.1. Coordinates transformation

The obvious difference between a girth weld and welding on a plate is the coordinate
transformation required to define the moving heat source model. As was explained in
161
Section 4.3, Goldaks distributed volumetric heat source model was expressed in a
position relative to the centre of a moving heat source where the intensity of the
volumetric heat source in a certain element was determined by its relative position to the
centre of the heat source. Besides the relative position, in pipe welding the relative
rotation is also needed. The right hand rule as in Figure 3.2 was also used with an
additional rotation rule shown in Figure 7.21. Rotation around axis x, y, and z axes is
known as Rx, Ry and Rz respectively. The Rx, Ry and Rz shown in Figure 7.21 are the
positive values.
In Figure 7.22 the coordinate of a certain point is tabulated, evaluated from the initial
coordinate and from a new coordinate which was rotated with an angle equal to . The
new position can be expressed as follows; in terms of its initial values and rotation
angle by a simple trigonometry function, as shown in Figure 7.22
Figure 7.21 Cartesian axes with rotations.
Figue 7.22a. A point at (y,z) observed from a coordinate Rx =
162
Figure 7.22b A point at (x,z) observed from a coordinate Ry =
Figure 7.22c A point at (x,y) observed from a coordinate Rz =

Rx :
Ry :
Rz :
The square matrix on the right side is called the transformation matrix and Goldaks
heat source model can be defined for the rotated coordinate system shown in Figure
7.23. For example, Figure 7.23 shows a coordinate system rotated around the z axis
(Rz ). All other relative positions are evaluated from the new coordinate system (xr, yr)
with the origin at the centre of the heat source. This initial coordinate system is usually
called a global coordinate system (GCS) whilst the new local coordinate system is
163
termed users coordinate system (UCS). The angle of rotation can be obtained using
trigonometric expression shown in equation 7.2.
(7.2)
Figure 7.23. Goldaks heat source model in rotated coordinate system

After the global coordinate of elements has been transformed to the local coordinate
system with the origin at the centre of the heat source, the combined heat source models
(Goldaks volumetric heat source and uniform temperature load) as previously used in
Chapter 6 can be applied. The relative angle was used in the girth weld model, instead
of the relative longitudinal position (evaluating the x coordinate position in a logic
shown by Figures 4.14 and 5.6) to apply the birth and death technique and uniform
temperature load.
Figure 7.24 Elements configuration of the girth weld joint.

164
The element configuration close to the weld line with killed (deactivated) elements and
temperature distribution at a certain time step after the first pass are shown in Figures
7.24 and 7.25 respectively. Figure 7.25 shows that the coordinate transformation and the
evaluation of relative angle developed are able to describe the birth and death technique
in the girth welding process.
Figure 7.25 Distributions of temperature when the heat source is moving.
7.2.2. Filtering peak temperature

As discussed in Chapter 6, the defined peak temperature of an element for a certain pass
has an important role in determining the real time phase of an element. The peak
temperature T2 is obtained by evaluating temperatures of three consecutive time steps,
i.e. T1, T2 and T3 to satisfy the condition T1<T2 T3.
Figure 7.26 shows the temperature profile for an element close to the 6 oclock position.
On the left side is the temperature history of the element for the first pass since welding
started until it cooled down to room temperature, and on the right side the second peak
temperature of the element has been observed.
165
Figure 7.26 Temperature history of an element at 6 oclock.
It can be concluded from Figure 7.26 that there are two temperatures which fulfill the
condition required to identify the peak temperature (T1<T2 T3): 2400 C and 21.52 C.
The database for the temperature histories and prediction of peak temperature as shown
in Figure 6.14 were developed after thermal analysis was completed which means that
the phase transformation was applied based on the predicted peak temperature and
predicted time when the peak temperatures were reached. All this information is saved
in a database of parameters. Unfortunately the 21.52 C occurred after reaching 2400 C,
and after a completed iteration the peak temperature was identified as 21.52 C. Since
21.52 C is far below 760 C the observed element does not experience SSPT whilst for
2400 C the peak temperature of the element should experience SSPT.
The second peak temperature is caused by numerical oscillation and existed especially
in elements close to 6 oclock position. To avoid wrongly identifying the peak
temperature, peak temperature filtering was programmed to follow the logic shown in
Figure 7.27. First, all the initial peak temperatures were set to equal zero as a tentative
peak temperature. Once the certain peak temperature has been found, it is saved in a
dummy parameter which will be compared to the previous tentative peak temperature. If
the dummy parameter is higher than the tentative peak temperature, the tentative peak
temperature is swapped with the dummy parameter. This step is iterated until the end of
time step for certain pass. The last tentative peak temperature that is the highest one (for
a certain path) is assigned as the peak temperature for the evaluated element. In Figure
7.28 the microstructure predicted close to the 6 oclock position was compared for the
166
first pass with and without the peak temperature filtering and element configuration is
displayed for a half model. From Figure 7.28, it can clearly be seen that the logic in
Figure 7.27 has handled the peak temperature quite well, and so a sound predicted phase
could be provided.
Figure 7.27 Filtering identification of peak temperature
Figure 7.28 Microstructure configurations at area close to 6 oclock position for first
pass with (right side) and without (left side) filtering peak temperature.
167
7.2.3. Material properties for X-70 base metal

and ER70S-6 filler metal.
The coefficient of thermal expansion has a significant effect on the residual stress as
discussed in Section 2.3. In this section the properties of both yield stress and thermal
expansion will be discussed below, although their roles will be discussed further in
Chapter 9.
The first material property to consider is the coefficient of thermal expansion for base
and filler metal when SSPT is considered. Based on previous literature study
[5-7]
, the
coefficient of thermal expansion at room temperature was considered equal to 14

m/m C and 13.24
m/m C for base metal and filler metal respectively. When an
austenitic transformation takes place between 760 C (A1) and 920 C (A3), the total
austenitic transformation strain for filler and base metal (2.29 x 10-3) [8-10] is assumed to
be linear and can be represented by a coefficient of thermal expansion equal to -2.3
m/m C (equation 6.8). The negative thermal expansion means when temperature
increases the heated material shrinks. In equation 2.14 for a coefficient of thermal
expansion the volume is expanded linearly while due to the austenite transformation the
volume is shrunk and the net result is that the material is shrunk. This result is
confirmed by typical temperature-dilatation diagram of steels when austenite SSPT take
place (Figure 2.17).
The coefficient of thermal expansion at atomic level depends on the distance between
atoms, which is described as the APF in SSPT. When austenite is formed, the
coefficient of thermal expansion was considered to have shifted in proportion to the
APF (0.68 for the initial phase and 0.74 for austenitic phase).
The chemical composition for ER70S-6 filler metal and X70 base metal is shown in
Table 7.2. Using Andrews equation (equation 6.2) and the chemical composition from
the table, the martensite start (Ms) for base metal is 454 C and 455 C for the weld
metal. The coefficient of thermal expansion can be obtained using equation 6.6 which is
a differential form of Koistinen-Marburgers law (equation 6.3). When a fully
martensitic phase is achieved, the coefficient of thermal expansion is again shifted
proportional to the APF of martensite (0.67). The coefficients of thermal expansion for
base metal and filler metal are presented graphically in Figure 7.29.
168
Table 7.2 Chemical composition of weld metal (wire) and base metal [4,5]:
%C
%Mn
%Si
%S
%P
%Cr
wire
0.08-0.09
1.42-1.65
0.81-0.87
0.006-0.010
0.004-0.010
0.01-0.05
BM
0.053
1.9
0.19
0.0008
0.012
0.01
0.17
0.24
%V
%Cu
%B
Ca
Nb
Ti
Al
0.17-0.22
0.0033
0.0047
0.043
0.023
0.042
wire
0.01
BM
0.044
0.01
0.0001
%Ni
%Mo
0.04
0.01
Figure 7.29 Coefficient of thermal expansion for base metal (left) and filler metal
(right).
The next discussion will describe how the yield stress at elevated temperature was
modeled for filler metal. Determination of the mechanical properties, especially the
yield stress at an elevated temperature, is challenging and little published experimental
data is available. Some researchers predicted the mechanical properties at elevated
temperature using an artificial neuron network
[11-14]
. In this thesis the yield stress is
expressed by an empirical equation based on some previous experimental results.

Theoretically, most of the strength of metal alloy at high temperature is divided into two
regions which are called normal low temperature and the creep control regions. A
typical graphic of yield strength at an elevated temperature is shown in Figure 7.30. The
material switches between each region at a switching temperature which is strain rate
dependent [15, 16].
To evaluate the effect of the high temperature on the yield strength, instead of using its
absolute value, it is better to use the reduction factor Kd that denotes the ratio between
the value of yield strength at an elevated temperature to its value at room temperature.
The Kd values obtained experimentally from previous published papers
[6, 11]
are shown
in Figure 7.31. The normalised yield stress at an elevated temperature from the weld and
169
base metal model developed in Chapter 6 is included in the figure. In the figure there is
also the assumed Kd as an input parameters in the ANSYS APDL model. The equation
for the Kd model in the normal low temperature region is shown in equation 7.3, whilst
in the creep control region it is shown in equation 7.4. The switching temperature is
assumed to be 650 C.
y
normal low temperature
creep control region
T switching
Figure 7.30 Yield strength at elevated temperature.
Figure 7.31. Normalised yield stress at elevated temperature.

Kd = -2x10-6T2 7x10-5T + 1
(7.3)
Kd = 2x10-7T2 5.7x10-4T + 0.414
(7.4)
170
Figure 7.32 shows the yield stress for base metal and filler metal. Yield stress at an
elevated temperature for base metal (X70) was obtained from previously published
paper[5] and data by Zhixiong Zhu[17] whilst for filler metal it was derived using the Kd
predictions shown in Figure 7.31, equations 7.3 and 7.4. The yield stress of ER70S-6 at
room temperature depends on the inert gas used when welding is carried out as shown
in Table 7.3[18] and since 90% Ar and 10% CO2 were used the yield stress was 470MPa.
Figure 7.32 Yield stresses of base metal and filler metal.

Table 7.3. Mechanical properties of AWS ER706S-6[17]
Inert Gas
100% CO2
75%Ar, 25%CO2
90%Ar, 10%CO2
98%Ar, 2%O2
Yield Strength
(Mpa)
440
460
470
455
Tensile Strength
(MPa)
560
565
580
565
Elongation
(%)
29
27
28
27
When martensitic SSPT takes place the yield stress was altered due to the contribution
of yield stress of prime martensite and aged martensite. Considering the carbon content
and the initial yield stress, the prime martensite yield stresses were 1027 MPa and 937
MPa for base metal and filler metal respectively [19], whilst the aged martensite yield
stresses were 1082 MPa and 1003 MPa. As discussed in Chapter 6 (Figures 6.9 and
6.10), the combined yield stress was obtained by weighting the initial and martensite
yield stress according the existing martensitic fraction from the Koistinen Marburgers
equation (6.3). The corrected yield stresses when SSPT occurs are shown in Figures
7.33 and 7.34 for base metal and filler metal respectively.
171
Figure 7.33. Effects of prime and aged martensite on the yield stress of base metal
Figure 7.34. Effects of prime and aged martensite on the yield stress of filler metal
172
Figure 7.35 Youngs modulus (E) models for base metal and weld metal.
Figure 7.36 Thermal conductivity models for base metal and weld metal.
The other properties required followed previously published papers
[5-7]
and they are
shown graphically in Figures 7.35 7.37. The density of both filler and parent metal are
considered to be 7850 kg/m3.
173
Figure 7.37 Specific heat models for base metal and weld metal.
The latent heat when martensite SSPT takes place follows equation (6.9) whilst for
austenite SSPT was considered to be 4200 J/mol[20]. By considering the composition,
and equations 6.9 and 6.2 the martensitic SSPT latent heat for each metal can be
obtained. Austenitic SSPT in kJ/kg unit can be calculated based on its composition and
assumption that the austenite SSPT latent heat is 4200 J/mol. The latent heat during
solid-liquid phase transformation was assumed to be 270 kJ/kg in the temperature range
1450 C - 1500 C[19]. The latent heats and temperatures when phase transformations
occur are summarised in Table 7.4. In this chapter isotropic hardening model was used
instead of kinematic hardening since in some papers it has been shown that kinematic
hardening under predicts residual stress [21, 22].
Table 7.4. Latent heat of the base metal and weld metal.
Latent Heat (kJ/kg)
Base Metal
Weld Metal
Martensite SSPT
74.360
76.502
Austenite SSPT
25.031
25.630
Melting
270
Temperature range ( C)
Base Metal
Weld Metal
454 - 200
455 - 200
760 - 920
1450 - 1500
174
7.2.4. Considering transformation plasticity

When SSPT takes place, anomalous plastic flow is reported
[23-25]
: while the stress or
temperature is kept constant a plastic deformation is observed. This plastic deformation

is called as transformation plasticity; that is plastic flow arising from SSPT. A
comprehensive discussion of transformation plasticity can be obtained from the
publications of J.B. Leblond [26-29], and this information was used as a material model in
this thesis.
In terms of welding FEM model, the transformation plasticity occurs in element (grains)
which develops SSPT under stresses resulted by an instantaneous thermal load. When
transformation plasticity is included, it alters the residual stress distribution
[30-31]
and in
this chapter,unlike the previous model, transformation plasticity is incorporated to refine

the FEM estimates.To underlined the role of transformation plasticity, Oddy et. al.[30]
even stated that there are two ways that SSPT alters residual stress which both are
irreversible: first is volume change and the second is the transformation plasticity.
Irreversible transformation plasticity means when the transformed phase back to the
initial phase does not cancel the plastic deformation. Transformation plasticity does not
depend on transformation rate, therefore is not a creep phenomenon and simply ceases
when SSPT is completed.
Mathematically, the transformation plasticity can be expressed as a general solution
shown in equation 7.5.
(7.5)
tp
is transformation plastic strain, K is a constant and (fm) is a normalized function of
martensite fraction which when no martensite formed equal to zero and when
transformed fully to martensite equal to unity. There are some normalized function have
been proposed and in this thesis a function proposed by Desalos
[32]
as expressed in
equation (7.6) is used. It should be noted that using equation 7.6 (0) = 0 and (1) = 1
which meet a prerequisite of a normalized functions.
(7.6)
K based on experiment data for several steels varied between 4.5e-5 to 10e-5 (MPa-1)
[25]
and the mid value of 7.25e-5 (MPa-1) is taken in this thesis. Indeed, there are some
175
function proposed for K as a function of relative difference between the two phases
29]
[26-
but because of the lack of data, a moderate constant value as mentioned above was
taken.
For instantaneous temperature in the range of martensite SSPT, fm can be obtained
following Koistenen Marburger (equation 6.3) and considering equation 7.6 (fm) can
be determined. For a certain temperature then
tp
is linearly proportional to , which
simplify the material model.
7.2.5. Considering cooling rate

Based on the CCT diagram for X-70 in Figure 6.1 page 113, bainite and ferrite SSPT
can be developed depending on the cooling rate. Ferrite may not formed as indicated by
the nature of temperature history in the girth weld. In the welding model, only elements
close to the weld line will experience peak temperature above A1. The cooling rate in
these elements is high which restricts the possibility of ferrite formation.
Figure 7.38. Temperature histories of nodes at varied transverse distances
The composition of base and weld metals, as shown in Table 7.2 in page 167, is
different from that indicated in Figure 6.1 but it was considered that this will not
significantly alter the CCT diagram. The cooling rate of elements which have peak
temperature higher than Ac1 can be observed by evaluating temperature histories of
elements closed to the ongoing weld bead. For simplicity, evaluation was done when the
176
first pass was applied. The temperature histories which are shown in Figure 7.38 are
obtained from nodes at the inner surface of the pipe at 3 oclock position. In Table 7.5
temperature histories of nodes with varied transverse distances are numerically
tabulated. The cooling rate is expressed as time needed to cool from 800 C to 500 C
(t8/5).
Table 7.5. Cooling rate of nodes.
Nodes
Distance (mm)
Tpeak ( C)
n3
0
2400.00
n4
1
2400.00
n47055
2.8
2070.28
n47056
4.6
1287.52
n47057
6.4
907.25
n47058
8.2
646.276
t8/5 (s)
9.596
9.596
10.396
11.195
11.995
n/a
Considering table 7.6 and Figures 6.1 and 7.38, the t8/5 of elements with peak
temperature exceeding Ac1 will not be longer than 13s and only bainite or martensite
transformations will occur as it was expected.
The bainite transition was modeled using JMAK[33] (Johnson-Mehl-AvramiKolmogrov) equation. The JMAK equation shown in (7.7) describes the bainite fraction
formed when SSPT takes place. k and n are material properties, and t is cooling time.
fb = 1 exp(-ktn)
(7.7)
Following equation (7.7), the values of fb varied between zero and unity. It should be
noted from (7.7) that fb never reach zero or unity that is why 5% fb was considered to
represent full martensite transformation and 95% fb represents full bainite
transformation. From the CCT diagram in Figure 6.1 full martensite transformation is
achieved when t8/5 is less than or equal to 7s whilst full bainite transformation takes
place when the t8/5 is higher or equal to 100s. Fitting these two conditions into JMAK
equation, k is equal to 0.003 and n = 1.5 (more or less). Substituting those values into
equation (7.7), the bainite fraction can be expressed as cooling time function as shown
by equation (7.8). A correction factor 1.053 is applied to obtain full bainite
transformation for cooling rate equal or higher than 100s.
fb = 1.053(1 exp(-0.003t1.5))
177
(7.8)
When the bainite transformation finished (temperature of an evaluated element equal to

martensite start (Ms) then martensite transformation is started and will be ended at
martensite finish (Mf) temperature. The volume fraction of martensite was determined
by Koistinen Marburger law
[34]
. Since the bainite is formed prior to martensite
transformation, the Koistinen Marburger law should be modified as shown in equation

(7.9) with T as instantaneous temperature. It should be noted that equation (7.9)
expressed martensite fraction at temperature between Ms and Mf.
fm = (1 1.053(1-exp(-0.003t1.5)) (1-exp(0.011(T Ms)))
(7.9)
The final portion of martensite can be obtained by equation (7.10). Using equations
(7.8) and (7.10) the final fraction of bainite and martensite for varied t8/5 can be obtained
and shown graphically in Figure 7.39. The shaded area is the possible phase in the girth
welding process. Based on Figure 7.39, it is believed to be acceptable if only full
martensite transformation is considered which makes the coupled TMM analysis much
simpler.
fm = (1 1.053(1-exp(-0.003t1.5))
(7.10)
Figure 7.39. Final phase as t8/5 function
7.3. Residual stress measurements

The residual stress measurement was carried out at ANSTO (Australian Nuclear
Science and Technology Organization) using the X-ray diffraction technique (XRD).
178
Basically XRD uses Braggs law which describes the constructive interference of an Xray beam [35-37].
One disadvantage of X-ray diffraction is that it is only suitable for measuring residual
stress on surfaces. Measurement depths of only 10-20 m are standard, but when
coupled with electro polishing surface removal, depths of up to 1-1.5mm are achievable.
The present study was prompted by a need to assess the effect of residual stress on
stress corrosion cracking and these surface stress measurements were considered
valuable. Nominal accuracy depends on the observed material, and for steel the nominal
accuracy is 30MPa
[38]
, an accuracy that is seriously affected by the grain size and
texture.
7.4. Analysis and discussion

The important parameters for moving the heat source model are the heat input, which is
obtained from the product of voltage and current divided by the welding speed, and heat
source parameters rx, ry, and r which were equal to 5mm, 3mm and 2mm respectively.
Real voltage and current were measured in the experiment using an AMV4000 welding
monitor where the welding speed in every pass was set to constantly equal 300mm/min.
Real time heat input can then be calculated based on real voltage and real current.
Figure 7.40 shows the voltages, current and power. In this figure the mean power for
each pass were 1950, 2722, 2500, 2500 and 1950 W for pass 1, pass 2, pass 3, pass 4,
and pass 5 respectively. As in previous chapters the mixed mode heat source model was
used and the heat transferred by the welding torch was divided into the uniform heat
load needed to increase the temperature of the filler metal to the melting temperature,
and Goldaks distributed heat source that represents the heat applied to the base metal.
The heat needed to melt the filler metal was equal to c T which was equal to 10.872
GJ/m3 which can be transformed into W by multiplying it with the cross sectional area
of each bead times the welding speed (300 mm/min = 0.005 m/s), or it can be
transformed to
which is equal to 82.40 MW/m3. The cross sectional area can be
obtained by evaluating the FEM geometry model shown in Figure 7.19. It should be
noted that there are small differences between the theoretical cross sections using
equation 7.1 and the cross section of the ANSYS model, although an attempt was made
to make the cross section as close to the theoretical approach as possible. The body heat
179
load in the centre of the heat source model can be obtained using equation 4.4. The
input data into the ANSYS APDL program for each pass are tabulated in Table 7.6.
All the parameters and material properties discussed in Section 7.2.3 were used in the
FEM simulation. The logic for FEM was discussed in Chapter 6 and the complete
methodology is represented on one page, as shown in Figure 7.41, where again the
important role of the parameter database developed in this thesis can be seen.
Figure 7.40 Welding parameters for the girth weld for (a) first pass, (b) second pass,
(c) third pass, (d) fourth pass and (e) fifth pass.
180
Table 7.6. Parameters for moving heat source model.
pass1
pass2
pass3
pass4
pass5
Power
(Volt. Amp)
1950
2722
2500
2500
1950
Ac (mm3)
11.3
15.0
14.9
14.9
11.0
Huniform
(W)
612.51
813.62
810.02
807.41
597.17
Hgoldak
qmax
(W)
(GW/m3)
1337.49
83.21
1908.38
118.72
1689.98
105.13
1292.59
105.30
1352.83
84.16
The first run was thermal analysis without SSPT, followed by the first metallurgical
analysis. The important output from this metallurgical analysis was the real time
(second) for each time step, the temperature of each element in each time step, the peak
temperature of each element, the time when the peak temperature was reached, and the
predicted SSPT configuration. All these parameters were saved in a data base of
parameters from where they could be retrieved quickly and easily.
The SSPT configuration predicted by metallurgical analysis was compared to the SSPT
configuration from a previous thermal analysis and when a different configuration was
found the thermal analysis considering SSPT was run again. The thermal analysis
considering SSPT executes light blue blocks whilst the thermal analysis without SSPT
does not. SSPT was carried out after each time step by evaluating the current time
compared to the time when peak temperature for each element was reached. When the
current time was higher than the time when peak temperature was reached this means
that the observed element has cooled and a martensitic SSPT can be executed at a
temperature between Ms and Mf based on the predicted phase for each element. After
the second thermal analysis considering SSPT had finished a metallurgical analysis was
run and the resulting SSPT configuration was again compared with the SSPT
configuration from the previous thermal analysis. This sequence was repeated until the
SSPT configuration used in the thermal analysis was equal to the SSPT configuration
predicted by the metallurgical analysis. After the coupled thermo-metallurgical analysis
was completed a sequential mechanical analysis (uncoupled analysis) considering SSPT
was then conducted to obtain the residual stress. A thermal load was applied using the
thermal analysis post processing file (thermal.rth) and SSPT was conducted using
information obtained from parameters saved in the txt data base.
181
THERMAL ANALYSIS
METALLURGY ANALYSIS
MODELING
MODELING
Geometry model, Material properties,

Initial condition, Boundary condition.

Initial phase configuration
Initial phase configuration
SIMULATION (Finding Solutions)
Resume SSPT from Metallurgical

analysis
Retrieve temperature histories for all

nodes from thermal.rth
Identify peak temperature (Tpeak) for each

element and time (tmax) when the peak
temperature is reached for certain pass
Determine predicted phase (pred) of each

element based on initial phase (init) and peak
temperature
SIMULATION (Finding Solutions)
All weld bulge elements are deactivated

Post Processing
Determine current heat source center
Predicted microstructure configuration
Previous Predicted microstructure
Coordinate transformation
Define relative position of nodes/elements to

instantaneous heat source center
Data Base of
PARAMETERS
Yes
Defined elements relative position to the heat source center
Inside
ellipse?
No
No
Thermal analysis considering SSPT
Thermo-Metallurgy
Finish
Yes
Yes
Solve the non-linear transient state problem using Newton-Raphson iteration
End of
Moving HS?
Execute Martensitic SSPT

based on initial phase
Equal micro structure

configuration?
No
time
tmax?
No
Identify does the

element experience
SSPT and check the
slope of elements
temperature history
MECHANICAL ANALYSIS
MODELING
Yes
All weld bulge elements are activated
Ongoing phase configuration
Last heat and temperature load is omitted

SIMULATION (Finding Solution)
Cooling due to convection/radiation
Next time step
Solve the non-linear transient state problem using Newton-Raphson iteration
Coordinate transformation
Execute Martensitic SSPT

based on initial phase
First time step
Yes
time
End of
PASS?
No
tmax?
Check position of weld bead

elements
Modify elements to
structural elements
(SOLID45)
No
Yes
Identify does the

element experience
SSPT and check the
slope of elements
temperature history
Temperature distribution for

certain time
Behind weld
torch?
All weld bulge elements are

deactivated
No
The elements are
activated
Apply temperatures as
thermal load
Yes
Post Processing
Temperature histories of each

element
Element temperature for
certain time
Thermal strain and thermal stress for

defined time
Element temperature for

certain time
Yes
Stiffness
> Melting
point?
0
No
>melting
point?
Yes
Plastic strain
No
Identify does the element experience SSPT and check

the slope of elements temperature history
time
tmax?
Execute Martensitic SSPT based on

initial phase
End of thermal
load?
Yes
Post Processing
RESIDUAL STRESS
TMM Finish
Figure 7.41 TMM analysis of the girth weld joint
182
No
7.4.1. Validation of the thermal results

Using all the material models and the logic discussed previously, the ANSYS APDL
was programmed. The results of the welding simulation for thermal analysis, especially
the temperature history for a certain node, can be retrieved from the *.rth file. The FEM
predicted temperature histories are validated with temperature histories recorded by the
thermocouples attached inside and outside the pipe, and laid on a path close to the 3
oclock position in the transverse position shown in Figure 7.42. For practical reasons
the transverse positions of the thermocouple were measured relative to the edges of the
gap, and then the values were transferred into the global coordinate system, as shown in
Table 7.7. Figure 7.42 shows the mesh with flagged node numbers which represent the
thermocouples. The transverse positions of the FEM nodes in the global coordinate
system and the thermocouples in close proximity are included in Table 7.7. Table 7.8
shows the thermocouples and corresponding FEM nodes. Since Tc4 was detached while
welding was on-going (pass 4) and the temperature history of the thermocouple was
incomplete, the temperature history of Tc4 was not used as a validation tool. The
temperature histories of the FEM selected nodes and the measured results of the
thermocouple close to the nodes are shown in Figures 7.43 7.46.
Figure 7.42. Sketch of measuring transversal position of thermocouples (left) and nodal
numbers with transversal position close to the thermocouples (right)
From Figures 7.43 7.46 it can be argued that the model provided quite good
temperature histories at the evaluated nodes. The minor differences between the
183
predictions and the experiment were due to some simplification in the FEM model and
experimental inaccuracy. Those simplifications in FEM that may have caused the
discrepancies are:
Table 7.7. Thermocouples and nodes transversal position
Thermocouples
measured (mm)
GCS (mm)
Nodes
GCS (mm)
Tc1
Tc2
Tc3
1.86
3.30
3.70
2.86
4.30
4.70
Tc4
Tc5
Tc6
Tc7
1.50
2.20
3.20
3.60
6.54
7.24
8.24
8.64
4
47055
47056
47057
27908
46397
46398
46068
1.000
2.800
4.600
6.400
6.328
7.552
8.776
10.000
Table 7.8. Thermocouples and corresponding FEM nodes

Experiment
Tc1
Tc2
Tc3
Tc4
Tc5
Tc6
Tc7
ANSYS
N 47055
N 47056
N 27908
N 46397
N 46398
Contact between the jig and pipe was not modeled. In fact part of the heat was
transferred through this contact with the jig which produces a higher cooling rate
than if the pipe is held with an adiabatic contact.
Thermal properties were not directly measured on the welded pipe. As mentioned
previously in Section 7.2.3, the thermal properties were obtained from a previous
literature study. Although the materials were the same as that used in the literature
(X70 and ER70S-6), since X70 has wide variations in composition, a small
difference in the thermal properties may have existed.
184
Figure 7.43 Validation of temperature history at node 47055 for all passes.
185
186
The coefficient of convective heat transfer and emissivity for radiation were also
based on literature and previous simulations (chapters 4 6).
The heat source power in the FEM model was represented by its mean value and as
shown in Figure 7.40, the actual heat source power varied over the length of the
weldment.
From the experimental side some shortcomings that might have contributed to the
deviation between the experimental results and ANSYS prediction are listed below:
Some errors in the prediction of the weld bead cross section. As described
previously in Section 7.2, the weld beads were predicted using a theoretical
approach (equation 7.1) so it may differ from the actual formed weld beads. On the
other hand the weld beads varied at different positions in the weld line due to the
pendant control used to accommodate the inconsistent gap.
The small error in identification of thermocouple position and since the evaluated
position is close to the weld line this small error might cause discrepancies in peak
temperature prediction.
Actual relative transverse distance to the centre of the weld bead may have shifted
due to the pendant control used to accommodate the inconsistent gap. If the relative
actual transverse distance was closer to the ideal position the temperature of the
thermocouple would be higher, but it would be lower if the actual distance was
further away.
Figure 7.47 Mistakes due to thermocouple attachments.

Errors on aimed measured temperatures due to thermocouple attachments as
described graphically in Figure 7.47. The temperature measured by the
thermocouple detects temperature at a slightly different position than the assumed
target position. This may also impose a different environment with different
coefficients of convection heat transfer. The larger the diameter of the
187
thermocouple wire the greater the distance between the target aimed for and the
measured temperature. The diameter of the thermocouple was 0.5mm and the
combined junction size was varied between 1 1.5mm.
Despite all these shortcomings the FEM predicted the temperature histories quite well,
as shown in Figures 7.43 7.46. These temperature histories indicate that the thermal
model developed can be considered to be the correct model. When the developed phase
configurations are incorporated with the correct thermal results, it is expected that
residual stress prediction will be close to the measured results.
7.4.2. Residual stress validation

Residual stress measurement was carried out at ANSTO using XRD and a review was
described in Section 7.3. The most important result from this measurement is the
longitudinal residual stress, or in terms of the girth weld may known as tangential stress.
The longitudinal stress was measured at the 3 oclock and 9 oclock transverse paths on
the outer surface of the pipes. Normal and transverse stress are also obtained from the
XRD measurement. It should be noted that the nature of XRD only provides residual
stress on the surfaces. The two samples of girth weld were called sample A and sample
B. The welding parameters for the samples were kept constant. (Note: A and B are
different samples. Although it was attempted to maintain the welding parameters
constant for both samples but welding itself is a noisy process and the result will
always be slightly different as it is shown). Stress on an unwelded sample was measured
to include initial stress as a correcting factor to the residual stress of the welded
samples. All those measurement results are shown in Figures 7.48 7.52. Figures 7.48
and 7.49 show measured residual stress at 3 oclock path for samples A and B
respectively whilst Figures 7.50 and 7.51 are for 9 oclock path. The data of sample A
for 3 oclock transverse residual stress at distance 45mm from weld centre line then was
dropped since it is inconsistence with the others and considered as an outlayer data.
Figure 7.52 shows initial stress on a path in a transverse direction to the weld line. To
obtain residual stress as a result of the welding process alone, the measured residual
stress should be decrease by the initial stress. Initial stress will vanish in the area close
to the weld line since it will be melted when the welding is applied. On the other hand,
188
initial stress is close to zero when it is close the weld centre line. For this reason the
measured residual stress is decreased by initial residual stress in area beyond 20mm
from the weld centre line.
Figure 7.48. X70A 3 oclock residual stress chart
Figure 7.49. X70B 3 oclock residual stress chart
189
Figure 7.50. X70A 9 oclock residual stress chart
Figure 7.51. X70B 9 oclock residual stress chart
190
Figure 7.52. X70 residual stress unwelded chart
Initial residual stress varied from one sample to another and considering Figure 7.52 and
residual stress for far field area in Figures 7.48 - 7.51, it is quite fair if the initial
longitudinal stress is considered equal to 100Mpa while initial normal and transverse
stress are assumed to be 150Mpa. Dropping the outlier data in Figure 7.48 and
excluding initial residual stress, Figures 7.53 7.56 shows the residual stress
distribution as a result of the girth welding process. Compared with the ANSYS
prediction, the mean data of samples A and B as shown in Figures 7.57 and 7.58 are
produced. Using the mean data, as representative, will decrease the number of scattered
data on a graph while keep the general trends. The 30 Mpa error bars are added to
show the nominal accuracy of the XRD measurement.
191
Figure 7.53. Welding residual stress at 3 oclock of sample A
Figure 7.54. Welding residual stress at 3 oclock of sample B
192
Figure 7.55. Welding residual stress at 9 oclock of sample A
Figure 7.56. Welding residual stress at 9 oclock of sample B
193
Figure 7.57. Mean value of welding residual stress at 3 oclock
Figure 7.58. Mean value of welding residual stress at 9 oclock
194
ANSYS provided results in the global coordinates that in term of stresses are those
stresses in x, y, and z directions and which are denoted in the ANSYS post process
results as Sx, Sy and Sz respectively. To obtain the longitudinal and normal stresses, all
stress components in a longitudinal (tangential) direction and normal direction should
be retrieved for a correct interpretation of the ANSYS results.
The tangential stress was obtained as follows: First, the position of element was
identified to obtain its local coordinate system with regards to the ANSYS geometric
model (Figure 7.18) and right hand rule for translation and rotation (Figure 7.21). It can
be concluded from those figures that there is no component of Sx to the longitudinal and
normal stresses because it is perpendicular to the longitudinal and normal directions.
Basically an element can be in tensile stress or compressive stress in all direction or it
may be tensile in one direction and compressive in another. It should be noted that this
stress is not really a vector unit and to add one stress to another is not possible using the
vector additional rule. The components of tensile stress will always be a tensile stress
despite their direction and the component of a compressive stress has the same rule.
This means the components of a positive value of stress (tensile) will always be positive
and a negative value of stress (compressive) will always be negative. Based on this
argument, the transformation matrix in equation 7.11, is used absolute values instead of
the real values, i.e. |sin x| not sin x.
In Figure 7.59 a component is studied which is in tensile stress with the yz Cartesian
coordinate system. The yz coordinate system is a plane that if viewed from a positive x
axis and using the right hand rule, the positive angle is anti-clock wise. The local
coordinate consist of a tangential axis and a radial axis and the component of a stress on
this direction can be obtained using simple trigonometric relations. The components of
those stresses (Sy and Sz) on the tangential (longitudinal stress) and radial directions
(normal stress) are St = Sy sin
+ Sz cos
and Sr = Sy cos
when expressed in a matrix, as shown in equation 7.11. The angle

equation 7.12.
195
+ Sz sin
or
is obtained from
Figure 7.59 Component of stresses in local coordinate system
(7.11)
(7.12)
Using equations 7.11 and 7.12 the distribution of longitudinal stress on the girth weld is
shown in Figures 7.60 7.61. In Figure 7.60 the longitudinal residual stress was shown
in a full specimen and in a half specimen by passing through a cutting plane. On this
cutting plane lay the evaluated paths (3 and 9 oclock paths) where the FEM predictions
will be verified with the previous discussed experimental results. Figure 7.61 enlarges
the 3 and 9 oclock parts for the half model in Figure 7.60. With the equal manner of the
longitudinal stress, Figures 7.62 and 7.63 describe transversal stress while Figures 7.64
and 7.65 present normal stress. Numerical residual stress, which has been shown
graphically on Figures 7.60 7.65, on defined paths can be retrieved from ANSYS post
process using a series of PATH commands. Those numerical results are compared with
the experiment results as shown in Figure 7.66 for 3 oclock path and Figure 7.67 for 9
oclock path.
196
Figure 7.60 Longitudinal residual stress profiles on full and half model
Figure 7.61 Longitudinal residual stress on enlarge portion of the half model
Figure 7.62 Transversal residual stress profiles on full and half model
197
Figure 7.63 Transversal residual stress on enlarge portion of the half model
Figure 7.64 Normal residual stress profiles on full and half model
Figure 7.65 Normal residual stress on enlarge portion of the half model
198
Figure 7.66. Residual stress on 3 oclock path
Figure 7.67. Residual stress on 9 oclock path

199
Some asymmetry in the stress distribution is expected due to in process variations in the
welding process. From Figures 7.66 and 7.67 it can be concluded the FEM prediction
can provide results that have a good agreement with the experiment measurements. It
should be noted that, although the final bead model look symmetric (Figure 7.19) the
residual stress distribution by FEM was not symmetric over the weld centre line since
the actual joint geometry varied. It is obvious the 3rd and 4th passes were not symmetric
(see Figures 7.10 and 7.19 for the passes). When the 3rd pass was applied the constraint
conditions were added by previous 1st and 2nd passes, and similarly with the 4th pass the
previous 3 passes all contribute to the constraint. Beside the difference in constraints,
the geometry and initial conditions were not symmetric when pass 3 and pass 4 were
applied. Thus, it is understandable that the final residual stress prediction by FEM was
not symmetric over the weld centre and this was also consistent with the experimental
results.
Higher longitudinal residual stress (compared to the other two component) can be
demonstrated by FEM prediction that validated by the experiment results. Principally,
as it has been theoretically discussed and presented with three elasto-plastic bars model
the residual stress is a result of plastic strain in the mid bar. The value of the plastic
strain depends on the magnitude of compressive stress when the mid bar is heated. This
compressive stress is determined by temperature difference between the mid bar and the
side bars or in welding case by temperature difference between adjacent areas. From the
temperature profile for a certain time, the temperature difference is higher in the
transverse direction than in the longitudinal direction which would explain why the
longitudinal residual stress is higher than transverse or normal stress. Certainly the other
phenomenon previously discussed as volumetric and material properties change and
also transformation plasticity due to SSPT, stiffness and plastic strain mitigation when
melting takes place also contribute to the final residual stress as additional aspects.
All the residual stress (longitudinal, transversal and normal) on 9 oclock path as shown
by FEM prediction are generally higher than 3 oclock path validated with the
experiment results. The residual stress difference between 3 oclock and 9 oclock paths
is caused by the constraint which is different when clock wise and counter clock wise
welding, where 3 oclock and 9 oclock paths laid respectively, are applied. When
counter clock wise welding is applied the previous clock wise welding added constraint
200
to the welded specimen and this additional constraint increases the final residual stress.
The FEM prediction as shown in Figure 7.67 overestimates the longitudinal residual
stress. In actual condition due to temperature elongation when the first clock wise run,
where 3 oclock path laid, was applied the gap in anti clock part was changed slightly.
The gap change may cause lower measured longitudinal residual stress which was not
modeled in FEM simulation.
The other minor discrepancies were found due to imperfections from both the FEM
method and experimental work. From the FEM side the errors are likely to be caused by
the phenomenon listed below:
The additional mechanical load as a result of pipe setting on the rig was not
modeled.
In FEM when the elements had cooled down from temperatures above A1 only the
martensitic transformation was modeled, which is the extreme condition.
The properties of materials, especially at elevated temperatures, were based on
theoretical approaches which only represented the general or mean value,
regardless of the unique behavior of the X70 pipes.
The errors due to experimental work are listed below:
Basically the welding itself is a noisy process as shown by Figures 7.48 and 7.49
and also Figures 7.50 and 7.51. From samples A and B with constant welding
parameters, some differences in the distribution of residual stress were still
exhibited.
A perfectly uniform surface preparation was not provided and since a consistent
gap was not provided a manual pendant control was needed and the weld beads are
not perfectly straight. This condition differs from situation which was modeled in
FEM where an ideal condition was assumed.
The available interface program does not provide changeable welding parameters
whilst the weld bead for constant welding parameters varied with different position
due to force of gravity.
201
The accuracy of X-ray diffraction methods is quantitatively around 30 MPa for

steel materials.
Regardless all the above mentioned imperfections, quite a good prediction of residual
stress has been obtained, as shown by Figures 7.66 and 7.67. From those validated
results it can be said that the assumptions taken to develop the FEM model can be
considered to be reasonable.
7.5. Conclusions
Based on the validation of temperature histories and residual stress of the girth weld, it
can be argued that there is good agreement between the FEM predictions and
experiments, or in other words the theoretical approaches developed in plate welding
(Chapter 6) can be applied to girth welding. However, from the simulation point of
view, several factors should be noted:
-
A coordinate transformation is needed because the ANSYS default is the

Cartesian coordinate system. This transformation is needed to express Goldaks
moving heat source in Cartesian coordinate relative to the centre of the heat
source.
To obtain the distribution of longitudinal and normal stresses, manipulations

based on the available perpendicular Cartesian stresses of elements namely Sx,
Sy, Sz and the position of the elements, were made. This manipulation was made
using a data base of parameters.
Peak temperature filtering should be done for the half pipe welding method (two
runs for each pass) to avoid mistakes in identifying peak temperature.
Some deviation between the FEM analysis and experimental results were caused by
imperfections from both simulation and experiment. Despite all the shortcomings, a
quite close prediction was obtained which means the FEM procedure for the girth weld
joint developed in this thesis can be considered to represent the actual phenomenon,
especially for the thermal analysis and residual stress (mechanical). Further
improvements to the girth weld modeling and validation could be achieved in both the
202
experimental work and FEM modeling. For experimental work it is suggested that the
activities listed below should be performed:
Preheat treatment may be applied to the pipes to omit the previously imposed
residual stress due to fabrication and surface preparation.
Improved automatic tracking of the weld seam could be utilized eg. using laser seam
tracking.
To provide a more consistent heat input a closed loop controller that gives feedback
to the welding machine could be developed. The inconsistent heat input may be
accommodated by welding parameters (current or voltage) or CTWD adjustment.
A program that can adjust the welding parameters to provide a consistent weld bead
in all positions of the girth weld may be very useful. Certainly, a study of the shape
of the weld bead with respect to its different positions is needed.
A flexible and portable jig with all the above improvement could be produced to
make it applicable to real girth welding process in pipelines.
From the FEM point of view the ideas listed below may be improve the accuracy of the
available program:
It is better to obtain all of the modeled properties by directly measuring the observed
materials, especially their properties at elevated temperatures, instead of the
theoretical approach used in this thesis. However a huge of amount of effort and the
availability of extensive elevated temperature measuring instruments would be
needed.
Considering the effect of the cooling rate on the phase development would increase
the accuracy of the program instead of considering its extreme condition by only
regarding the martensitic phase transformation. Developing this kind of program
needs a primary data of experimental results suggested in the first point.
In real conditions not all the austenite was transformed to martensite, and certainly
consideration of retained austenite will increase the accuracy of the developed
program, but again it needs a primary data based on experimental observation of the
materials under study.
203
Considering a full martensite transformation (neglecting retained austenite) is an

extreme condition, so the results involving retained austenite should be between the
model without SSPT and that with full martensite transformation. The model without
SSPT will be compared with model with full martensite transformation in the
Chapter 9.
204
References
[1]
Lincoln Electric, Process waveform control technology, surface tension transfer

(STT),
http://www.lincolnelectric.com/assets/US/EN/literature
/NX220.pdf,
accessed 14 October 2012.
[2]
Lincoln Electric, Application waveform control technology, surface tension

transfer (STT), STT open root, http://www.lincolnelectric.com/assets/US/EN/
literature/NX220.pdf, accessed 14 October 2012.
[3]
Bruce D. DeRuntz (2003), Assessing the benefits of surface tension transfer

welding to industry, Jounal of Industrial Technology, vol. 19, no.4, pp.2 8.
[4]
Lincoln Electric, SuperArc L-56 Mild Steel, Copper Coated AWS ER70S-6 &
EH11K,
http://www.lincolnelectric.com/assets/en_us/Products/Consumable_
MIGGMAWWires-SuperArc-SuperArcL-56/c4105.pdf, accessed 8 November
2012.
[5]
M.R. Forouzan, A. Heidari and J. Golestaneh (2009), FE simulation of submerged

arc welding of API 5L-X70 straight seam oil and gas pipes, Istiqlaal, vol.28, no.1,
pp. 93 110.
[6]
R.I. Karlsson and B.L. Josefson (1990), Three-dimensional finite element analysis
of temperature and stresses in a single-pass butt-welded pipe, Transactions of the
ASME, vol.112, pp. 76-84.
[7]
Riswan Alim Mufti (2008), Mechanical and microstructural investigation of weld

based rapid prototyping, PhD Thesis, Faculty of Mechanical Engineering, Ghulam
Ishaq Khan Institute of Engineering Sciences and Technology.
[8]
D. Deng, Y.Lou, H. Serizawa, M. Shibahara & H. Murakawa, Trans. JWRI 32

(2003) 325-333.
[9]
Chin-Hyung Lee (2008), Computational modeling of the residual stress evolution

due to solid-state phase transformation during welding, Modeling and Simulation
in Materials Science and Engineering, vol. 16, pp. 1 16.
phase transformation, Computational Materials Science, vol. 46, pp. 1014-1022.
[11] Moudar H. Zgoul (2010), Predicting the mechanical behavior of steel frames
exposed to elevated temperatures using artificial neural networks, WSEAS
Transactions on System, Vol.9, Issue8, pp. 895 904.
205
[12] Thanuja Ranawaka and Mahen Mahendran (2009), Experimental study of the
mechanical properties of light gauge cold-formed steels at elevated temperatures,
Fire Safety Journal, vol.44, pp. 219-229.
[13] T. Hozjan, G. Turk and S. Srpcic (2007), Fire analysis of steel frames with the use
of artificial neural networks, Journal of Constructional Steel Research vol.63, pp.
1396-1403.
[14] Z. Sterjovski, D. Nolan, K.R. Carpenter, D.P. Dunne, J. Norrish (2005), Artificial
neural networks for modeling the mechanical of steels is various applications,
Journal of Materials Processing Technology, vol. 170, pp. 536-544.
[15] N. Saunders, Z. Guo, A.P. Miodownik and J.P. Schille (2004), Modeling the
material properties and behavior of multicomponent alloys, 22nd CAD-FEM
users meeting 2004, International Congress on FEM Technology.
[16] Zhanli Guo, Nigel Saunders, Peter Miodownik and Jean-Philippe Schille (2007),
Modeling the strain-life relationship of commercial alloys, Proceedings of Eight
International Conference on Creep and Fatigue at Elevated Temperature, San
Antonio, Texas.
[17] Zhixiong Zhu (2013), Structure property correlation in the weld HAZ of high
strength line pipe steels, PhD Theses, University of Wollongong.
[18] The Lincoln Electric Company, MIG & TIG GMAW/GTAW consumeable,
http://www.iwsco.co.kr/data/goodsImages/520110530155236.pdf, accessed 18
November 2012.
[19] RWK Honeycombe and HKDH Bhadeshia (1995), Steels microstructure and
properties, 2nd edition, Gray Publishing, UK.
[20] R.A. Jaramillo, G.M. Ludtka, R.A. Kisner, D.M. Micholson, J.B. Wilgen, G.
Mackiewicz-Ludka, N. Bembridge and P.N. Kalu (2005), Investigation of phase
transformation kinetics and microstructural evaluation in 1045 and 52100 steel
under large magnetic field, Solid-to-solid phase transformations in inorganic
materials: Diffusional transformations, edited by J.M. Howe, D.E. Laughlin, J.K.
Lee, U. Dahmen, W.A. Soffa, The Minerals, Metals and Materials Society, pp.
893-898.
[21] O. Muransky, C.J. Hamelin, M.C. Smith, P.J. Bendeich & L. Edwards (2012), The
effect of plasticity theory on predicted residual stress fields in numerical weld
analysis, Computational Materials Science, Vol. 54, pp. 125 134.
206
[22] O. Muransky, M.C. Smith, P.J. Bendeich, P.J. Bendeich, T.M. Holden, V. Luzin,
R.V. Martins and L. Edwards (2012), Comprehensive numerical analysis of a
three-pass bead-in-slot weld and its critical validation using neutron and
synchrotron diffraction residual stress measurements, International Journal of
Solids and Structures, Vol. 49, Issue 9, pp. 1045-1062.
[23] G.W. Greenwood and R.H. Johnson (1965), The Deformation of Metals Under
Small Stresses During Phase Transformation, Pro. Roy. Soc., A283, 403.
[24] C.L. Magee (1966), Transformation Kinetics, Microplasticity and Aging of
Martensite in Fe-31 Ni, PhD Thesis, Carnegie Institute of Technology, Pittsburgh,
PA.
[25] S. Denis, A. Simon and G. Beck (1982), Estimation of the Effect of Stress/Phase
Transformation Interaction when Calculating Internal Stress During Martensite
Quenching of Steel, Trans. Iron Steel Inst. Japan, 22, 505.
[26] J.B. Leblond, G. Mottet and J.C. Devaux (1986), A Theoretical and Numerical
Approach to The Plastic Behaviour of Steels During Phase Transformations 1:
Derivation of General Relations, J. Mechj. Phys. Solids, Vol. 34, p.395.
[27] J.B. Leblond, G. Mottet and J.C. Devaux (1986), A Theoretical and Numerical
Approach to The Plastic Behaviour of Steels During Phase Transformations 2
Study of Classical Plasticity for Ideal Plastic Phases, J. Mechj. Phys. Solids, Vol.
34, p.411.
[28] J.B. Leblond, J. Devaux and J.C. Devaux (1989), Mathematical Modelling of
Transformation Plasticity in Steels 1: Case of Ideal-Plastic Phases, International
Journal of Plasticity, Vol.5, pp. 551-572.
[29] J.B. Leblond, J. Devaux and J.C. Devaux (1989), Mathematical Modelling of
Transformation Plasticity in Steels 2: Coupling with Strain Hardening
Phenomena, International Journal of Plasticity, Vol.5, pp. 573-591.
[30] A.S. Oddy, J.A. Goldak and J.M.J. McDill (1992), Transformation Plasticity and
Residual Stresses in Single-Pass Reparir Welds, Journal of Pressure Vessel
Technology, Vol.114, pp. 33-38.
[31] H. Dai, J.A. Francis and P.J. Withers (2010), Prediction of Residual Stress
Distribution for Single Weld Beads Deposited on to SA508 Steel Including Phase
Transformation Effect, Materials Science and Technology, Vol.26, no.8, pp. 940
949.
[32] Y. Desalos, Comportement Dilatometrique et Mecanique de lAustenite
Metastable dun Acier A 533, IRSID Report #95349401 in J.B. Leblond, J.
Devaux and J.C. Devaux (1989), Mathematical Modelling of Transformation
207
Plasticity in Steels 1: Case of Ideal-Plastic Phases, International Journal of

Plasticity, Vol.5, pp. 551-572.
[33] Dean Deng, Yangang Tong, Ninshu Ma and Hidekazu Murakawa (2013),
Prediction of the residual welding stress in 2.25Cr-1Mo steel by by taking into
account the effect of solid-state phase transformations, Acta Metallurgica Sinica
(English Letters) vol.26, no.3 pp. 333-339.
[34] D.P. Koistinen and R.E. Marburger (1959), A general equation prescribing the
extent of the austenite-martensite transformation in pure iron-carbon alloys and
plain carbon steels, Acta Metallurgica, vol. 7, pp. 59 60.
[35] James W. Dally and William F. Riley (1991), Experimental Stress Analysis,
McGraw Hill International Editions.
[36] M. Birkholz (2006), Thin film analysis by X-ray scattering, Wiley-VCH Verlag
GmbH&Co. KGaA, Weinheim.
[37] Paul S. Prevey, Current application of X-ray diffraction residual stress
measurement, Lambda Techologies, http://www.lambdatechs.com/documents/
214.pdf, accessed December 2012.
[38] Veqter,
X-Ray
Diffraction,
http://www.veqter.co.uk/residual-stress-
measurement/x-ray-diffraction, accessed December 2012.
208
Chapter 8
Simulating DC-LSND on Girth
Butt Welding
A great deal of efforts has been spent to mitigate residual stress with the most prominent
technique being the application of extra heating or cooling. This can be divided into
three different classes: applying extra heating/cooling before welding, while welding is
ongoing, and after welding. Applying heating or cooling at the same time as welding
saves production time and has economic advantages.
One technique for applying heating or cooling while welding is ongoing is called as
DC-LSND (Dynamically Controlled Low Stress No Distortion) one version of which
was patented by Q.Guan, C.X. Zhang and D.L. Guo[1, 2]. This technique applies a
cooling source behind the welding torch.
Many subsequent researchers have published details of their own experiment and
analyses of DC-LSND [3-12]. Typically DC-LSND is applied on single bead butt joint of
thin plate with the main goal of reducing distortion. No published papers which discuss
the effect of DC-LSND on multi pass girth welds and the resultant residual stress could
be found. Experimental investigation of the technique was beyond the scope of the
current work but based on the development of the FEM residual stress prediction work
reported above the possible effect of DC-LSND on multi pass girth butt weld joint was
studied using the FEM approach.
Applying a trailing heat sink increases the complexity of the welding phenomenon
which is already complex. An analytical solution to this complex phenomenon is almost
impossible to obtain and the only suitable approach may be a numerical analysis using
FEM. Although one author has mentioned that an ideal trailing heat sink is not her main
goal, some excellent data applicable to the current analysis was found in a series of
writings by E.M. Van der Aa
[10-12]
. This data was used as a basis for the model of a
trailing heat sink in this chapter. Van der Aa used CO2 snow as a cooling media for the
209
trailing heat sink and applied it to a single pass of thin plate welding. A trailing heat
sink was applied onto the flat surface of a 100mm wide by 200mm long, and either
1.5mm or 2mm thick AISI 316L stainless steel plate with a welded joint in the middle
and parallel to the 200mm long side. The sketch of the DC-LSND system used by E.M.
Van der Aa is shown in Figure 8.1.
Figure 8.1. DC-LSND of a thin plate welding [10-12]
The cooling heat source model in this chapter utilises the CO2 snow trailing heat sink
that is used in the E.M. Van der Aa papers and it will be studied the effects of applying
DC-LSND on the residual stress in the butt girth weld joint discussed in Chapter 7.
The heat which is transferred from a surface area where DC-LSND was applied was
assumed to be by a convection mechanism. The area where convection takes place is a
double ellipsoid that can be expressed in terms of a local coordinate with an origin at the
artificial centre of the trailing heat sink, as shown in equation 8.1. AHsink is an area
where convection takes place,
and
are the relative transverse and longitudinal
distances from artificial centre of the heat sink respectively, w is the transverse
parameter of the area considered equal to 0.025m and gi is longitudinal parameter of the
area where it equal to 0.10m when
has positive values and 0.02m when
has
negative values. 0.10m and 0.02m are denotes as g1 and g2 respectively.

(8.1)
210
The cooling strength on the cooling area depends on the convection coefficient and
temperature of the cooling media. The Gaussian distributed coefficient of convection [1012]
after having been modified to suit the APDL code, can be expressed as a relative
coordinate to the heat sink centre, as shown in equation 8.2. It should be noted that
equation 8.2 is a general form of double ellipsoid of Gaussian distribution model and
constants are used to match the model with the experimental measurements.
(8.2)
hmax expresses the maximum trailing heat sinks coefficient of convection which is
equal to 2000 W.m-2.K-1; p, m and n are constants equal to 1.2, 1, and 8 respectively,
based on experimental results
[10-12]
. Two important notes should be considered when
equation 8.2 is used, the first is the equation is only applied inside area which fulfills
equation 8.1 and the second is that the maximum value of the equation is constrained to
2000 W.m-2.K-1 whilst the minimum value is 25 W.m-2.K-1. Observing hsink along the
longitudinal axis (assigning
equal to zero), the hsink can be represented as shown in
Figure 8.2.
From Figure 8.2 it can be seen that the value of hsink for
equal to 25(W.m-2.K-1) and equal to 2000(W.m-2.K-1) for
higher is the
L
L
of equation 8.2. The
is 100mm but evaluating hsink at this position has
no meaning with regards to equation 8.1 where the value of

full value of
greater than 84.90mm. The
value the higher the term
maximum value is equal to 1 when
from -20 to 24.47mm is
(equal to w = 25mm) is found at
will be equal to zero. The
= 0 but at this position, from Figure
8.2 the value of hsink achieves its allowed minimum value, so at this position hsink will be
uniformly distributed equal to 20 W.m-2.K-1. It may be beneficial to evaluate hsink at a
path with
= 84.904mm which lay inside the area defined by equation 8.1 and the
is
spanned from -13.208 to 13.208mm. The value of hsink along this path is shown in
Figure 8.3.
211
Figure 8.2 hsink at
Figure 8.3 hsink at
L=
=0
84.904mm
Figure 8.4 Temperature and hsink distribution at

212
=0
The temperature of the cooling media (CO2 snow) which is symbolised as Tmedia also
varies in the cooling area depending on the longitudinal position relative to the artificial
centre of the heat sink, as shown in equation 8.3. As in equation 8.2, equation 8.3 has
been modified to fit the ANSYS APDL code. kmedia is a constant which is equal to 38
based on experimental results[10-12] and
state, which is -78 C. At a path where
is the temperature of CO2 snow in a solid

T
= 0, the distribution of temperature is plotted
as shown in Figure 8.4 with hsink also superimposed on the graph. Figure 8.4 is the same
as that of Van der Aa which means the modifications made to suit ANSYS APDL code
are consistent with previous work. The cooling strength (heat flux) of the heat sink
which follows the convection phenomenon depends on the coefficient of convection
hsink and temperature difference between the surface and the cooling media, as shown in
equation 8.4. Tplate is the temperature of the plate which has increased from room
temperature due to welding.
(8.3)
(8.4)
A high hsink and a low Tmedia at high
will cause a very high cooling effect, which also
depends on the temperature of the plate from the results of FEM thermal analysis. Since
the higher
means it is closer to the welding torch which means a higher Tplate and as a
result the temperature difference has increased, and the cooling strength close behind
the welding torch, is very high. The cooling effect expressed by equation 8.4 indicates
that a higher
but higher hsink (see Figure 8.4) means stronger cooling strength. Of
course this is only defined at AHsink. In the AHsink higher
lower Tmedia (see Figure 8.4)
results in stronger cooling strength. The initial temperature of plate as a result of

welding Tplate will be higher if closer to the heat source centre (higher
inside
AHsink), and in AHsink the small ellipse of Figure 8.16 is when all the discussed values are
in the conditions that caused the cooling strength maximum: high hsink; low Tmedia and
high Tplate.
8.1.
Heat sink modeling in ANSYS
There are published papers which attempt to model the heat sink in an FEM model:
using a negative heat flux (distributed or uniform) and applying convection onto a
213
surface. As described implicitly, this research considers that the heat sink relies on
convection phenomenon. There are two reason for using this model: the first is that
applying a negative heat flux resulted in a very low temperature, even below 0 K,
which is unrealistic [4] and the second is that the comprehensive model used by Van der
Aa also used the convection phenomenon.
Evaluates the position of an elements

Apply coordinate transformation
T
(mm)
0?
0?
AHsink
1?
dummy>2000?
dummy<20?
hsink
Figure 8.5 Defines hsink in ANSYS APDL
214
Figure 8.6 hsink at the defined area
Evaluates the position of an elements

Apply coordinate transformation
T
(mm)
0?
0?
AHsink
1?
Tmedia
Figure 8.7 Defines Tmedia in ANSYS APDL
How the heat sink was modeled in ANSYS APDL is described in the flow-chart shown
in Figure 8.5. Here the position of an element represented by its relative position to a
centre point was first evaluated using the *GET command. This initial position in
215
ANSYS was defined in Cartesian coordinates and a coordinate transformation is needed

as discussed in Section 7.2.1. The transformed coordinate can be obtained from the
elements relative local coordinate ( L,
T).
The area where the convection caused by the
trailing heat sink was applied by a double-ellipsoid evaluated by equation 8.1 depends
on the value of
L.
On the elements inside the ellipsoid the values of hsink were evaluated
based on equation 8.2 with the following constraints: the maximum value is 2000 (W.m2
.K-1) and the minimum value is 20 (W.m-2.K-1). By following the flow-chart in Figure
8.5, the distributed hsink inside the double ellipsoid is determined and shown in Figure
8.6.
Figure 8.7 describes how ANSYS APDL obtains the distribution of temperature of the
cooling media on the double ellipsoid area where the convection heat transfer is applied.
Similar to the case of hsink, the relative local position of an observed element is
evaluated first. Based on the relative local position ( T,
L)
it can be determined whether
the element is laid inside the double ellipsoid and when the element is inside the double
ellipsoid, the temperature of the cooling media follows equation 8.3. Applying the flowchart shown in Figure 8.7 using ANSYS APDL, the temperature distribution of the
cooling media is shown in Figure 8.8. The artificial centre of the trailing heat sink has
lagged 133 mm (see Figure 8.1) behind the heat source centre which modeled the
welding torch. Considering Rz equal to 3mm the DC-LSND have 30mm offset with the
heat source model.
Figure 8.8 Tmedia at the defined area
216
The next thing to be studied through FEM simulation is the application of the DCLSND model to the girth weld joint. As previously described, it was assumed that the
cooling strength and temperature can be expressed solely in terms of the relative
transverse position ( T) and relative longitudinal position ( L) to the artificial centre of
the heat sink. Indeed equations 8.1 to 8.4 are based on flat surfaces, however this
approach can be accepted as long as the distance from the surface to the nozzle does not
exceed the core length where the drop in the pressure head is not less than 95%
[13]
. For
the case of impinging jet applied to a curved surface it has been shown that on a convex
surface Nusselts number will increase on a convex surface and decrease on a concave
surface, but the difference in Nusselts number is not really significant[14]. Since the heat
transfer coefficient of convection depends solely on the Reynolds and Nusselts
number, while for almost equal speed of impinging media the Reynolds number will
not change significantly so it is hoped that the convection heat transfer coefficient will
not alter much due to the radius of curvature. However, a more precise approach needs
an experimental investigation of an actual heat sink applied onto the surface of the pipe
which is beyond the discussion of this thesis.
8.2. Applying DC-LSND on the girth weld

The most suitable available command that can be used for distributed convection used
in DC-LSND simulation are the following two consecutive SFE commands:
SFE, noELM, noSURF, CONV, 1, hsink
SFE, noELM, noSURF, CONV, 2, Tmedia
noELM is the number of an evaluated element, noSURF is the number of relative
surface of an observed element, hsink is the coefficient of convection obtained using the
flow-chart in Figure 8.5 and Tmedia is the temperature of the cooling media obtained by
following a logic described as a flow-chart in Figure 8.7. For SOLID70 and SOLID45
elements there are six possible numbers of relative surfaces that should be carefully
evaluated to obtain the correct surface where convection will be applied. Elements
where convection due to DC-LSND are applied will also be altered when the birth and
death technique is applied that should be carefully modeled.
217
As discussed in Section 7.2.1, it is easier in the girth weld model if the global Cartesian
coordinate was transferred to the local cylindrical coordinate. To fit with the
experimental work, it was assumed that 12 oclock position where the welding was
started
= 0 and considering the right hand rule (Figure 7.21) the counter clock wise
directions, i.e. 12, 11, 10, 9 oclock are positive values that coincide with 0, /6, /3,
/2 radian respectively. The angle of certain elements or nodes can be obtained using
equation 8.5.
(8.5)
It should be noted that welding procedures suggested for the STT Lincoln GMAW
machine meant that the welding was done with 5G vertical down position; that is there
were two welding stages for every pass, i.e., anti-clock wise and clock wise. The local
transverse coordinate ( T) did not change with the Rx
(Figure 7.21) and the
mathematical expression for the local longitudinal coordinate ( L) depended on the

direction of welding, i.e. clock wise or anti clock-wise, the position of the cooling heat
source centre and the observed element position where z > 0 were equal to quadrants 1
and 2, and z < 0 was equal to quadrants 3 and 4. The mathematical formulae for
regarding all of those prerequisites are tabulated in Table 8.1.
Table 8.1. Mathematical formulas for
Welding direction : Anti clock wise

z0_DC 0
for zj 0
for zj 0
=
(
)R
L
j
0
j
L = ( j - 0 - 2 )Rj
z0_DC 0
for zj 0
for zj 0
L = ( j - 0 + 2 )Rj
L = ( j - 0)Rj
Welding direction : Clock wise
z0_DC 0
for zj 0
for zj 0
L = ( 0 - j)Rj
L = ( 0 - j + 2 )Rj
z0_DC 0
for zj 0
for zj 0
=
(
2
)R
L
0
j
j
L = ( 0 - j)Rj
z0_DC = z global coordinate for DC-LSND artificial centre.
zj = z global coordinate for evaluated elements.
0 = angle position of the DC-LSND artificial centre.
j = angle position of evaluated elements.
218
Using the formulae in Table 8.1 and previously discussed logic, a typical convection
coefficient and temperature of the cooling media are shown in Figures 8.9 and 8.10
respectively. These figures describe the trailing heat sink when applied at the first pass
with clock wise welding direction. Note that the heat sink model can follow the contour
of V-grove and the active elements, i.e. the bead top surface of the first pass. It can be
concluded from the figures that ANSYS can defined correctly the DC-LSND model of
Van der Aa, in the girth weld model. However unlike Van der Aa the current analysis is
applied to a thicker material on a curved surface.
Figure 8.11 show the temperature distribution of a certain time step of the first pass
when the heat source of girth weld with trailing heat sink model is applied. It can be
seen that the trailing heat sink alters the temperature profiles (compared to Figure 7.25).
This temperature profile will be applied as thermal load in TMM analysis and since the
thermal load has changed, the final residual stress should be altered. How this new
thermal load affects the residual stress will be discussed in more detail in section 8.3.
Figure 8.9. Ditributed hsink (W.m-2,K-1) of DC-LSND model applied on the girth weld
219
Figure 8.10 Distributed Tmedia ( C) of DC-LSND model applied on the girth weld
Figure 8.11 Typical temperature profile of girth welding with DC-LSND
8. 3.
Results and discussions
Figure 8.11 showed temperature contours which were altered due to the application of
DC-LSND in which can be clearly seen at the isothermal lines especially for 200 C line.
220
How the DC-LSND affects the residual stress, can be studied by evaluating the final
residual stress when DC-LSND is applied in a similar manner to that in Section 7 of the
girth-weld model. Figure 8.12 shows residual stress profiles for a model with DCLSND in the areas of 3 and 9 oclock which are comparable to Figures 7.59, 7.61 and
7.63 for girth-weld model without DC-LSND. Figure 8.13 compares residual stress for
the girth-weld model (Chapter 7) evaluated at 3 and 9 oclock paths and the residual
stress resultingfrom a girth-weld with DC-LSND application for the equal paths. From
these figures especially Figure 8.13 it can clearly be seen that the application of DCLSND embeds compression residual stresses for the area close to the weld line. When a
heat source is applied such as the welding torch, the final residual stress is tensile
residual stress in the area close to the weld centre as a result of plastic strain and misfit
that has been discussed thoroughly in section 2.3. Using an equal analogy, when a
cooling source is applied, i.e. the DC-LSND, simply the final residual stress is the
opposite of tensile that is compressive residual stress. This compressive residual stress
resulting fromthe DC-LSND compensates for the tensile residual stress due to the
leading welding torch as shown in Figure 8.13.
Practically, the main concern for structure integrity from residual stress is the tensile
residual stress which in turn decreases the lifetime of the welded structure through its
effect on fatigue load or stress corrosion cracking . The analysis shows that the
application of DC-LSND would be expected to decrease the residual tensile stress near
the centre of weld line on both paths. The maximum values of longitudinal residual
stress decrease by 39.16% and 31.16% at 3 and 9 oclock paths respectively. Whilst for
transversal residual stresses are 50.59% and 38.25% and for normal residual stress are
34.81% and 19.35% at 3 and 9 oclock paths respectively. Taking the mean values, the
residual stress decrease 35.55% by application of DC-LSND.
221
8.12 Residual stress profile at 3 and 9 oclock areas when DC-LSND is applied
222
Figure 8.13 Residual stress at 3 (right) and 9 (left) oclock paths for girth weld with and
without DC-LSND
As discussed in Section 2.3, thermal stresses exist as a result of differences in
temperature of the weld line with its adjacent transverse position which had been
described as side bars. Based on this theory it may useful to compare temperature
histories of nodes which have different transverse distance. Nodes 4, 47056, 47057,
45903, 74728, and 74729 on the pipes inner surface have transverse distance from weld
centre equal to 1 mm, 4.6 mm, 6.4 mm, 10 mm, 13.82 mm, and 17.24 mm respectively
were chosen. The temperature histories for these nodes when the first pass was applied
223
are presented in Figures 8.14 and 8.15 for girth weld with and without DC-LSND. It can
be seen from Figure 8.14 that the temperature history of nodes 4, 47056 and 47057 are
Equilibrium temperature
Figure 8.14 Temperature histories of nodes with varied transversal position of girth
weld without DC-LSND
altered by DC-LSND application but not for the rest. This result can be evaluated using
convection coefficient of DC-LSND model which is mathematically expressed by
equation 8.2 and shown graphically in Figures 8.2 and 8.4. Considering equation 8.2
and the permitted maximum hsink, the area inside the small ellipse in Figure 8.16 has
hsink equal to 2000 W.m2.K-1 and will decrease exponentially to the surrounding area.
Only the area inside AHsink of the small ellipse actually experiences DC-LSND and this
area is shown by the grey area in Figure 8.16. This area may called as AHmax where the
coefficient of convection equal to 2000 W.m2.K-1.
224
Equilibrium Temperature
Figure 8.15 Temperature histories of nodes with varied transversal position of girth
weld with DC-LSND
The centre of the small ellipse can be expressed by
and
longitudinal term of the equation 8.2 :
is obtained when the
is equal to unity and
to zero since the equation is a symmetric function to

= 100mm and
T.
T;
equal
so the centre of ellipse is at
= 0. The major axis of the ellipse is in the longitudinal direction and
equal to
direction is equal to
= 0.01510m = 15.10mm whilst the minor axis in the transverse

= 0.01068m = 10.68mm. But it should be noted
that the hsink is only applied in the area which satisfies equation 8.1 which graphically is
shown in Figure 8.16 as part of the large ellipse behind the small one. The intersection
between the big and the small ellipses are at
= 92.73mm and
= 9.36mm. As only
nodes 4, 47056 and 47057 have transverse distance below 9.36mm, their temperature
histories are altered by the application of DC-LSND model behind the welding torch
while the rest (nodes 45903, 74728 and 74729) are not as shown by temperature
histories of those nodes in Figure 8.15.
225
Figure 8.16. Evaluating hsink of equation 8.2.
It should be emphasised that the thermal stress is caused by temperature distribution

which is not uniform, so if a material is uniformly heated and cooled no thermal stress is
formed as long as no constraint is applied. Without DC-LSND the temperature of those
selected nodes is uniform at around 130s and 48 C whilst with DC-LSND at around 78s
and 180 C which means the thermal stress will not developed further. The 48 C and
180 C may be considered as the equilibrium temperature where temperature for varied
transversal position especially the closest node (node 4) with its adjacent (node 47056)
can be considered same.
Basically the magnitude of developed thermal stress depends on the temperature
distribution and yield stress. When an element near the weld line reaches its peak
temperature the temperature differs significantly with its adjacent transverse elements.
However at this high temperature the stiffness of the element is very low and as a result
the thermal stress may not high. Furthermore, the resetting plastic strain is mitigating
the thermal stress in the near melted element (this will be discussed further in Section
9.3), which is why it is sometime said that basically the residual stress is formed when
the weld line is cooled down to the room temperature. This thermal stresses is
developed further until the element near the weld line is cooled down to 48 C if DCLSND is not applied. Put simply the thermal stress is developed from 855 C to 48 C for
the elements near the weld line which is represented by node 4. At 48 C the difference
of temperatures may not be high but since the yield stress is high enough the thermal
226
stress may be quite high. When DC-LSND is applied, elements near to the weld line
develop thermal stress from 855 C to 180 C which has a narrower temperature span
from peak temperature to equilibrium temperature than the girth weld without DCLSND and also the stiffness at 180 C is lower than that at 48 C. This discussion may
explain qualitatively why the final residual stress for girth weld especially close to the
weld line with DC-LSND is lower.
8.4. Conclusions
Based on FEM predictions, the application of DC-LSND on multi-pass girth butt weld
joint of thick pipe has been demonstrated to reduce the residual stress significantly. The
stress reduction is a result of increasing equilibrium temperature especially for area
close to the weld centre. The change of temperature was caused by the high coefficient
of convection in AHmax ellipse (shown by grey area in Figure 8.16). Coincidently the
ellipse is in front part of AHsink where the temperature caused by the welding heat source
(Tplate) is higher than the rear part. Furthermore Tmedia in the AHmax is lower and as a
consequence the cooling strength of the AHmax ellipse area is very high.
227
References
[1]
Q. Guan and D.L. Guo (1987), Method and apparatus for low stress no distortion
welding of thin walled structural elements, Chinese patent no. 87100959.5
[2]
Q. Guan, C.X. Zhang and D.L. Guo (1993), Dynamically controlled low stress no
distortion welding method and its facility, Chinese patent no.93101690.8.
[3]
Q.Guan, D.L. Guo, C.Q.Li and R.H. Leggatt (1994), Low stress non distortion
(LSND) welding a new technique for thin materials, Welding in the World (UK),
Vol.33, no.3, pp. 160-167.
[4]
D. Dye, S.M. Roberts, A.M. Korsunsky, K.E. James, B. Benn and R.C. Read
(2002), Application of low stress no distortion welding to the gas tungsten arc
welding of wrought nickel-base alloy C263, Mathematical Modeling of Welding
Phenomena 6, Graz, Austria.
[5]
Li Ju, Guan Qiao, Shi You Wu, Guo De Lun, Sun Yong Chun and Du Yu Xiao
(2004), Stress field of GTAW with a spot heat sink on titanium alloy thin plate,
Journal of Mechanical Strength, Vol.26, no.3, pp. 278 283.
[6]
J. Li, Q. Guan, Y. Shi, D. Guo, Y. Du and Y. Sun (2004), Studies on

characteristics of temperature field during GTAW with a trailing heat sink for
titanium sheet, Journal of Materials and Processing Technology, Vol. 147, no.3,
pp. 328 335.
[7]
J. Li, Q. Guan, Y.W. Shi and D.L. Guo (2004), Stress and distortion mitigation
technique for welding titanium alloy thin sheet, Science and Technology of
Welding and Joining, Vol.9 no.5, pp. 451 458.
[8]
Q. Guan (2005), Control of buckling distortion in plates and shells, in Zhili Feng,
Processes and Mechanism of Welding Residual Stress and Distortion, Woodhead
Publishing Limited, Cambridge, England, pp. 295 343.
[9]
Guan Qiao (2006), Thermal tensioning effects to prevent welding buckling

distortion in manufacturing of thin-walled aerospace shells and panels,
Engineering Sciences, Vol.4, No.3, pp. 1 10.
[10] E.M. Van der Aa, M.J.M. Hermans, I.M. Richardson, N.M. Van der Pers and R.
Delhez (2006), Experimental study of the influence of a trailing heat sink on the
welding residual stress distribution, Materials Science Forum, Vol. 524-525, pp.
479-484.
[11] E.M. Van der Aa, M.J.M. Hermans and I.M. Richardson (2006), Conceptual
model for stress and strain development during welding with trailing heat sink,
Science and Technology of Welding and Joining, Vol.11, No.4, pp. 488-495.
228
[12] Elisabeth Maria Van der Aa (2007), Prediction and control of residual stresses
and buckling distortion, PhD Thesis, TU Delf.
[13] Holger Martin (1977), Heat and mass transfer between impinging gas jet and
solid surfaces, in James P. Hartnett and Thomas F. Irvine, Jr., Advances in heat
transfer, Vol.13, Academic Press, New York, San Francisco, London.
[14] A.M. Tahsini and S. Tadayon Mousavi (2012), Laminar impinging jet heat
transfer for curved plates, World Academy of Science, Engineering and
Technology vol. 72, pp. 1280 1285.
229
PART V
GENERAL DISCUSSIONS
Chapter 9
General Discussions
9.1. Overview
Specific discussions concerning sources of error and the approach adopted have been
included at the end of each chapter but the following summarises the overall outcomes
of the work.
The study reported here indicates that FEM is a powerful tool to model a technically
complex phenomenon. However, the results depended on assumptions made by users.
When using a FEM software package such as ANSYS it is important that the
programmer sets out the problem in logic and syntax which are understood by the
chosen package.
There are two ways to ensure the reliability of an FEM model, shown in Figure 3.1, the
first is; comparing the results obtained from FEM with the analytic solution. Since the
analytic solution provides a solution for a simple condition, only a simple FEM model
can be verified. Comparing the FEM model with Rosenthals solution for a welding heat
source indicated the validity of the logic and syntaxes that were made and discussed in
Chapter 3. The FEM model in Chapter 3 represented the thermal analysis of a moving
point heat source in an infinite solid and Rosenthals solution represents an analytic
solution of the temperature field of a moving point heat source in an infinite solid.
A general analytical solution that covers the transient and steady state was developed
based on Carslaw and Jagers[1] general solution for energy conservation in an infinite
solid. This solution was obtained using a non dimensional integral technique proposed
by Komanduri and Hou[2]. The transient solution obtained from FEM showed a good
agreement with the analytical solution. For a steady state, the solution obtained from
Rosenthal, the non dimensional integral and FEM also showed a good agreement.
230
Another approach to validating a FEM model was undertaken by comparing the FEM
prediction with the experimental results. Under practical conditions welding is not as
simple as most analytical models suggest. The FEM model should accommodate many
factors to make it comparable to the experiment. However, some reasonable
simplifications still need to be taken to make sure the problem can be solved using
FEM.
Figure 9.1 indicates the difference between real welding conditions and simplifications
which need to be adopted taken in the FEM thermal analysis [3] even for the very simple
TIG process. In a thermal analysis using the FEM model, and the energy conservation
law, the welding phenomenon was modeled as a moving heat source, with heat transfer
due to conduction to the surrounding cooler metal, heat loss to environment due to
convection and radiation, and latent heat as a result of phase transformation.
The values of transient temperatures are determined by the response of the materials
thermal properties which are temperature dependent subject to boundary condition of
the heat source, conduction, convection, radiation and latent heat.
Figure 9.1. Welding phenomenon, an FEM model and an actual condition [3].
The heat source can be classified into three different ways: point heat source, surface
heat source and volumetric heat source. The most prominent moving heat source model
was Goldaks volumetric heat source where the intensity of heat at certain position is
231
determined by its position relative to the centre of the heat source. Goldaks heat source
model was discussed thoroughly in Chapter 4.
In this thesis, a new heat source model was developed. The heat source was modeled as
combination of Goldaks heat source model and a uniform temperature volumetric heat
source. Goldaks heat source was used to represent heat embedded into the welded
metal by the welding torch and the uniform temperature volumetric heat source model
was used to represent the melted filler metal that forms the weld bead. The validity of
the heat source model was cross-checked against the experimental data as discussed in
Chapter 4.
By using the validated temperature distribution as the thermal load, the transient thermal
stresses in the welded plate can be obtained. The residual stress is a steady-state
thermal stress, i.e. it is a thermal stress when the plate has cooled down close to ambient
temperature. Relating the temperature distribution to thermal stress is usually known as
thermo-mechanical analysis, but since the distribution of temperature is insignificantly
influenced by the stress state, sequential thermo-mechanical analysis was used in this
thesis and has been discussed in Chapter 5.
Since in this thesis the girth weld joint of ferritic steels pipes was discussed, SSPT
needed to be considered. The Thermo-Metallurgy-Mechanical (TMM) analysis
considers SSPT in the development of residual stress. The method used to carry out
TMM has been shown in Figure 6.7 and in more detail in Figures 6.14 and 7.39.
9.2. TMM in plate welding

The FEM model that considers the SSPT for plate welding has been discussed in
Chapter 6. From Figure 6.11 it can be concluded that SSPT affects the residual
development of stress in welding because of changes in the properties of the material
and volumetric change. The most important material property affected by SSPT was
yield stress. The yield stress at a certain condition depends on each phase fraction which
is accounted for according to the KoistenenMarburger equation (equation 6.3). The
effect of SSPT on yield stresses is presented graphically in Figures 6.9 and 6.10 for base
metal and weld metal respectively.
232
As mentioned above, another aspect of SSPT that influences the residual stress is the
volumetric change that occurs as a result of an atomic arrangement of each phase
(Sections 2.4 and 6.3). Following the Koistenen-Marburger equation when martensitic
SSPT takes place the volumetric change was presented graphically, as shown in Figure
6.5 (equation 6.6), but when austenitic SSPT takes places the volumetric change was
considered linear (equation 6.8) and was presented graphically in Figure 6.6.
How this SSPT influenced the development of residual stress was shown in Figures
6.19 and 6.23, by comparing to the residual stress prediction without involving SSPT.
As discussed above, the SSPT influences residual stress due to alteration of material
properties and volumetric change. How each aspect influences the residual stress and to
what extent it alters the development of residual stress is an interesting question.
9.2.1. Change of material properties in SSPT

In this chapter the volumetric change due to SSPT will be excluded as was considered
in the material model stage. To exclude volumetric change, the FEM model follows the
logic shown in Figure 9.2, which is a modification of the flow-chart in Figure 6.11
where volumetric change due to SSPT was excluded. (Note: mechanical property which
was altered is yield stress).
It should be noted that altering the thermal properties due to SSPT, especially latent
heat, was still considered in the coupled thermo-metallurgical analysis to determine the
predicted phase. Using a flow-chart as in Figure 9.2, the predicted residual stress was
obtained, as shown in Figure 9.3. The predicted residual stress was compared with the
predicted residual stress without considering SSPT.
233
Figure 9.2. Excluding volumetric change in SSPT
It can be concluded from Figure 9.3, that if only the mechanical properties, i.e.,the yield
stress, was included in the residual stress prediction, the tensile residual stress close to
the weld joint would increase. Since the initial mechanical response due to thermal load
was elongation, simply increasing yield stress will raise the residual stress. When yield
stress changes due to SSPT is considered without a volumetric change, the residual
stress in the weld line increases by 27.30% and the maximum residual stress increase by
23.84% compare to residual stress without SSPT. (Note: SSPT results in a volumetric
change, but not only that; the material properties especially yield stress are also altered.
In the SSPT model both contributes to the final residual stress).
234
Figure 9.3. Residual stress as a result of SSPT by neglecting volumetric change
9.2.2. Volumetric change in SSPT

Whilst in Section 9.2.1 the volumetric change was neglected in subsequent mechanical
analysis, in this chapter the change in mechanical properties was neglected. The goal
was to separate the effect of volumetric change from the change in the mechanical
properties. The FEM model followed the flowchart in Figure 9.4 and as in Section 9.2.1
in coupled thermo-metallurgical analysis the change in thermal properties due to SSPT
should be included.
The prediction of residual stress by excluding mechanical properties of the material is
presented in Figure 9.5. Residual stress prediction without SSPT was also plotted in the
same figure to study the effect of the volumetric change due to SSPT.
Two volumetric changes occurred due to SSPT, when austenite and martensite
transformation took place. Volumetric shrinkage occurred while the elements were
heated due to Austenite transformation with increased the tensile stress whilst
volumetric expansion while the elements were cooled due to martensite transformation
produced a compressive stress close to the weld line. The Austenite transformation
existed at high temperatures (760 C - 920 C) when the yield strength of the weld metal
and base metal were low. On the other hand, Martensite transformation took place in the
lower temperature range (456 C - 200 C for base metal and 427 C - 200 C for weld
235
metal) when the yield strength of the weld metal and base metal were higher than
austenitic SSPT. Using the analogy used in Section 9.2.1: the lower yield strength
altered the residual stress less; so it is understandable that the net effect of volumetric
change due to SSPT reduced the tensile residual stress close to the weld line as shown in
Figure 9.5. The volumetric change due to SSPT, excluding yield stress changes, reduce
the residual stress in the weld line by 67.04% and the maximum residual stress reduce
by 62.68% compare to residual stress without SSPT.
Figure 9.4. Excluding mechanical properties change in SSPT
236
Figure 9.5. Residual stress as a result of SSPT by neglecting material properties change
9.2.3. Aspects involved in SSPT

Results obtained by FEM prediction, as described in Section 9.2.1 and Section 9.2.2 are
compared to evaluate each parameter that was changed due to SSPT in residual stress
development. Those results were plotted together with the residual stress prediction with
SSPT and without SSPT, as shown in Figure 9.6.
From Figure 9.6 it can be seen that the prediction of residual stress with SSPT is a
combined result of a model with a change in material properties only and a model that
only accommodates volumetric change. Neglecting volumetric change due to SSPT will
overestimate the residual stress close to the weld line whilst neglecting the change in
material properties will give a lower prediction. Accommodating both phenomena gives
results intermediate between those two cases. As discussed in Chapter 6, the
consideration of residual stress with SSPT (by accommodating alterations in material
properties and volumetric change) gave a residual stress prediction close to the
experimental results compared to the prediction without SSPT.
237
Figure 9.6. Evaluating aspects that affect residual stress
9.3. Resetting plastic strain in the girth butt

welding.
How SSPT affects residual-stress has been discussed thoroughly in Section 9.2 using
the plate butt welding studied in Chapter 6. Another important aspect that needs to be
discussed is how the melting phenomenon is considered in the APDL program
developed in this thesis. The girth butt welding procedures discussed in Chapter 7 will
be taken as a case study in this section.
As shown in Figure 7.39, when melting exists in a certain element the element loses it
stiffness and plastic strain. Element stiffness at a high temperature is already low due to
its low yield point. From the ANSYS manuals, when an element is deactivated (using
the EKILL command) by default the stiffness is set close to zero. Deactivating the
melting element was used to develop the APDL program in Chapters 6 and 7. But as
mentioned above, since the material model has exhibited low stiffness at melting
temperature the deactivating technique will not have any significant effect on the
residual stress. Secondly, the melting phenomenon reduces plastic strain in the melting
elements. Basically this can be accommodated by resetting the plastic strain of the
melting element using the DESOL command. DESOL is an ANSYS post processing
238
command and since melting can exist at any time step on certain elements, the program
should toggle from the simulation block to the post processing block at each time step.
Results with plastic strain relieving when melting takes places have been presented in
Chapter 7. To study the effect of this plastic strain relieving, in this section the FEM
model without plastic strain relieving blocks is simulated and the result is compared to
the previous model with plastic strain relieving. Figures 9.7 9.9 describe longitudinal,
transversal and normal stress respectively for FEM model without plastic strain
relieving in areas of 3 and 9 oclock. The numerical residual stress then retrieved from
ANSYS post process results at 3 and 9 oclock paths and presented graphically in
Figures 9.10 9.12. Figure 9.10 shows longitudinal stress distribution on 3 and 9
oclock paths for a model with and without plastic strain resetting. The measurement
result from both A and B samples are also plotted. Note that the mean values as in
Chapter 7 (Figures 7.55, 7.56, 7.64 and 7.65) are not plotted but the longitudinal
residual of both A and B specimens. Plotting both experiment data gives insights of the
position of ANSYS results compared to the results obtained from the real noisy
welding process.
It can be seen that without resetting the plastic strain the FEM prediction overestimates
the residual stress as a result of the welding process. Indeed, the FEM longitudinal
residual stress prediction on 9 oclock path also overestimated the residual stress. The
reason for the overestimation has been discussed in chapter 7 that is; a result of gap
geometry which changed in the real welding process. This change is caused by
distributed temperature when a previous pass in clock wise direction was applied.
Moreover, the difference between the FEM model and measured longitudinal residual
stress is not as high as predicted by model without accommodating the plastic strain
reset and it may still acceptable considering the varied parameters in the real welding
process.
Figures 9.11 and 9.12 describe transverse and normal residual stress on the 3 and 9
oclock paths which the residual stress profiles on the pipe are shown in Figures 9.8 and
9.9 respectively. From those figures it can be concluded that as in longitudinal stress,
the FEM model without plastic strain reset always overestimates the all component of
residual stress (longitudinal, transversal and normal).
239
Figure 9.7. Longitudinal residual stress at 3 oclock (right) and 9 oclock (left) areas
Figure 9.8. Transversal residual stress at 3 oclock (right) and 9 oclock (left) areas
Figure 9.9. Normal residual stress at 3 oclock (right) and 9 oclock (left) areas
240
Figure 9.10. Longitudinal residual stress at 3 oclock (right) and 9 oclock (left) paths
Figure 9.11. Transversal residual stress at 3 oclock (right) and 9 oclock (left) paths
Figure 9.12. Normal residual stress at 3 oclock (right) and 9 oclock (left) paths
241
An explanation of the overestimated prediction can be obtained by following the

mechanism of residual stress formation described in Section 2.3. When an element is
heated, the melting elements which are laid very close to the weld line are typically in
compression because its temperature is higher than the surrounding elements. The
higher temperature means higher thermal expansion and because expansion is
constrained by the bulk of cooler far field materials, those melting elements are in state
of compression that caused elastic and plastic shrinkage. The elastic shrinkage will
recover while the load (thermal load) is omitted whilst the plastic shrinkage is kept in
the elements. The plastic shrinkage that exists, as a result of misfit, finally produces
tensile residual stress when the welded materials are cooled to room temperature. By
resetting the plastic strain in the melting element the final net tensile residual stress in
the melted element will decrease, which is why if resetting the plastic strain is
considered to be correct, without resetting the plastic strain the FEM predicts a higher
tensile residual stress close to the centre of the weld joint where elements experience
peak temperature equal to or exceeding the melting temperature. The correctness of
mitigating the plastic strain when an element has melted was confirmed with predictions
closer to the measurement of residual stresses.
9.4. Considering lower annealing temperature.

In previous models when an element is at its melting temperature, it will lose its
stiffness and plastic strain. As has been discussed in Chapter 5, deactivating an element
by application of EKILL command caused the element to lose it stiffness and removing
plastic strain can be performed by application of DESOL command. It was suggested
that this phenomenon happens even at temperature below melting which is called as
annealing temperature. Theoretically annealing temperature depends on the carbon
content which is divided into two regimes, hypoeutectoid and hypereutectoid. In
hypoeutectoid the annealing temperature is around 50C above AC3 temperature whilst
50C above austenite cementite line for hypereutectoid steels. Since the steel under
this study is hypoeutectoid then the annealing temperature is 50C above AC3 which is
equal to 970C. The Girth weld FEM model in Chapter 7 used this annealing
temperature. To observe the effect of this lower annealing temperature, in this section
annealing is considered to take place at melting temperature (1400 C). The result from
242
the model of this section then compared to the previous outcomes as shown in Figures
9.16 up to 9.18. By comparing both results the effect of a lower annealing temperature
can be observed. Figures 9.13 9.15 show residual stress profile for areas close to the 3
and 9 oclockpaths which are comparable to Figures 7.59, 7.61 and 7.63.
Figure 9.13 Longitudinal residual stress at 3 oclock (right) and 9 oclock (left) areas for
a model with annealing temperature equal to 1400 C.
Figure 9.14. Transversal residual stress at 3 oclock (right) and 9 oclock (left) areas for
a model a model with annealing temperature equal to 1400 C.
Figure 9.15. Normal residual stress at 3 oclock (right) and 9 oclock (left) areas for a
model a model with annealing temperature equal to 1400 C.
243
for a model with varied annealing temperature
Figure 9.17. Transversal residual stress at 3 oclock (right) and 9 oclock (left) paths for
a model with varied annealing temperature
Figure 9.18. Normal residual stress at 3 oclock (right) and 9 oclock (left) paths for a
model with varied annealing temperature
244
From Figures 9.16 9.18 it can be concluded that the different residual stress prediction
following those two assumptions of annealing temperature is clearly seen in the area
close to the weld line. It is not surprising since the annealing takes place in this area, no
matter what the annealing temperature is assumed (970 C or 1400 C). It should be
noted that temperature slope is very high in the weld centre line and the covered area in
transverse distance for element with temperature above 1400 C and 900 C will not
significantly change. Indeed, a small difference can be seen in far field residual stress
for varied annealing temperature but is just a consequence of maintaining forces balance
in the observed area.
Transverse residual stress for 1400 C is almost equal to predicted residual stress with
970 C annealing temperature. The geometric constraint for transverse direction is
different to other direction; in this direction the bulk solid low temperature which never
experience annealing is laid whilst in the two other direction only material which
experienced annealing is laid. While annealing takes place strain is revised by omitting
plastic strain but in the transverse direction this strain correction is constrained by far
field solid with low temperature. As has been mentioned before, the distance from the
weld centre line in the transverse direction for element with peak temperature above
1400 C and 970 C will not be altered significantly due to the high temperature slope in
this direction. This different condition of constraint finally may cause a different
contribution of altered annealing temperature those results in a significant residual stress
change in the longitudinal and normal direction but not in the transverse direction.
With the higher annealing temperature, residual stress should be higher since the
residual stress is basically formed when a material is cooled down from this annealing
to the room temperature. Higher annealing temperature means residual stress will be
formed in the wider temperature span which in turn results in higher residual stress. But
longitudinal residual stress distributions shown in Figure 9.16 exhibits reversed trend. It
must be underlined that in the selected paths (3 and 9 oclock surface paths) is laid
layers 3 and 4 (pass 6 and 8 respectively), above this layer is layer 5 (pass 9 and 10
respectively for clock wise and anti clock wise down ward welding). To understand the
lower longitudinal residual stress for 1400 C annealing temperature the temperature
history of a certain node in those nodes must be observed (nodes 14541 and 27905)
layers (layers 3 and 4) close to the selected paths. The typical temperature history is
245
shown in Figure 9.19.Node 14541 is laid at layer 3 whilst node 27905 at layer 4. When
layer 3 is applied, node 14541 has a peak temperature equal to 2400 C which equal to
applied uniform temperature load in the mixed mode of the heat source model. Node
27905 shows the same manner that when layer 4 is applied the peak temperature reaches
2400 C and below it in the other layers.
Figure 9.19. Temperature history of nodes at layer 3 and layer 4 close to the evaluated
path.
From Figures 9.19 it can be seen that, when layer 5 is applied the node under
observation experienced temperature history with peak temperature equal to 1222 C
and 1035 C. If annealing temperature is considered equal to 1400 then when the 5th
layer is applied no future annealing temperature is found instead when the 3th or 4th
layers are applied for node 14541 and node 27905 respectively. But if annealing
temperature is considered equal to 970 C the second annealing temperature is exhibited
that when the 5th is performed. From previous discussion if annealing phenomenon is
not considered in the residual stress formation the residual stress is formed as a result of
compressive plastic strain which is developed when the material in the centre line is
heated while the expansion is constrained by cool far field bulk material. The developed
plastic strain caused misfit which in turn is considered as a reason for developed tensile
stress in the area close to the weld centre. All this mechanism has been discussed in
chapter 2.3 when discussing the three bar elasto-plastic material model. When annealing
takes place, residual stress is formed due to contraction of elements in the weld centre
line when cooled down to room temperature which is constrained by far bulk cooler
material. Since all nodes under observation when 5th layer isapplied have peak
246
temperature above 970 C and below 1400 C which mechanism will be followed
depends on the initial assumptions. When annealing temperature is considered equal to
970 C all those observed nodes develop residual stress when 5th layer is applied due to
contraction which is constrained when those nodes cooled down to room temperature
which in turn caused higher tensile residual stress result. On the other hand when
annealing temperature is considered equal to 1400 C the developed residual stress
produced by misfit as a result of plastic compressive stress when those nodes are heated
which in turn caused lower tensile residual stress result.
9.5. Transformation plasticity.

When SSPT takes place not only volumetric and mechanical properties changes exist
but also transformation plasticity that describes additional strain when SSPT takes place
under instantaneous stress. A theoretical approach and description of how this
transformation plasticity was included in FEM model has been discussed in section
7.2.4. In this section the contribution of the transformation plasticity to the developed
residual stress by comparing the model to the model without transformation plasticity
considerations will be discussed.
Figure 9.20. Longitudinal residual stress at 3 oclock (right) and 9 oclock (left) areas
247
Figure 9.21. Transversal residual stress at 3 oclock (right) and 9 oclock (left) areas
Figure 9.22. Normal residual stress at 3 oclock (right) and 9 oclock (left) areas
The residual stress profiles in 3 and 9 oclock areas ; ignoring transformation plasticity
are presented in Figures 9.20 up to 9.22 for longitudinal, transversal and normal residual
stress respectively which are comparable to Figures 7.59, 7.61 and 7.63 respectively. In
Figure 9.23 the longitudinal residual stress without incorporating transformation
plasticity is compared to the Chapter 7 model which included transformation plasticity
phenomenon. Experimental results for both specimens (namely specimens A and B) are
also plotted in Figure 9.23 to validate the FEM predictions. In the same way, Figures
9.24 and 9.25 show residual stress predictions as well as experimental results for
transversal and normal residual stress respectively.
From these figures it can be understood that ignoring transformation plasticity cause the
residual stress prediction to be more tensile especially in the area close to the weld
line. It should be noted that basically the initial excitation in the mechanical analysis
248
stage of the welding FEM model is strain ( ). The strain is obtained from the elongation
due to temperature load which is retrieved from FEM thermal analysis results. When the
strain is in a proportional region the embedded stress equal to the strain times the slope
of the stress-strain curve which equal to modulus Young E. While if the strain exceeds
the yield strain the stress produced is equal to the yield stress plus the strain times the
slope of the plastic lines. Since the stress-strain curves are modeled by multi linear
isotropic lines there will be always slopes.
Figure 9.24. Transversal residual stress at 3 oclock (right) and 9 oclock (left) paths
When transformation plasticity is modeled, a part of the strain is considered to be a

transformation plasticity (or it can be said that apart of strain is due to the
transformation plasticity strain) and the strain that produces stress is decreased.
Although the FEM model of the girth weld is not as simple as discussed, but it can be
qualitatively explain that the residual stress by involving transformation plasticity will
have lower tensile stress results compare to the model that ignores the transformation
249
plasticity. If the model incorporating the transformation plasticity is considered to be

correct, ignoring the transformation plasticity cause over prediction of the residual
stress.
Figure 9.25. Normal residual stress at 3 oclock (right) and 9 oclock (left) paths
9.6. Aged martensite considerations.

As has been discussed in Chapter 7, aged martensite SSPT was considered in the girth
weld model. Aged martensite is a phase which is formed at the cooling stage of a
martensite when reheated above A1. Figure 9.26 shows comparison of a model
involving aged martensite transformation with a model considers prime martensite only.
Figure 9.26. Residual stress at 9 oclock (left) and 3 oclock (right) for model with and
without aged martensite SSPT.
From the figure it can be concluded that involving aged martensite lowers all final
residual stresses. Considering the SSPT model of prime martensite and aged martensite
presented in Figures 7.33 and 7.34, basically aged martensite represents a higher yield
250
stress material compared to prime martensite. How this higher yield stress produces
lower final residual stress will be demonstrated below.
In the previous three bars model it has been demonstrated how residual stress developed
as a result of thermal load. Figure 2.15 page 29 shows the stress when the middle bar is
heated to 1100 C. It should be noted that in the model the yield stress of all bars are 150
MPa. It can be demonstrated using the three bars model that increasing the yield stress
of middle bar will lower the final tensile stress in the mid bar. The yield stress of the
middle bar in the next model is increased to 160 MPa. The stress-strain diagram for both
model are shown in Figure 9.27. Figure 9.28 shows residual stress development for this
increased yield stress of the middle bar.
Figure 9.27. Stress strain diagram for material model used in Figure 2.15 (left) and
Figure 9.27 (right).
Figure 9.28. Three bars model with higher yield stress of middle bar; when the middle
bar heated (left) and temperature load omitted (right)
251
From Figure 2.15 it can be seen that when the middle bar is heated the stress in the side
bars is 75 Mpa and -150 for middle bar and when the temperature load is omitted the
stresses in side bars and middle bar are -34.1 MPa and 68.2 MPa respectively. Figure
9.28 shows that when the middle bar is heated the stresses in side bars and middle bar
are 80 MPa and -160 MPa respectively. When the temperature load is omitted (right
side), the stress in side bars and middle bar are -33.9 MPa and 67.8 MPa respectively.
This result is consistent with FEM model involving higher yield stress (Aged
martensite).
How the lower residual stress is produced can be understood by evaluating not only
stress in the 3 bars model but also the strain. Stress and strain for the model with yield
stress of middle bar equal to 150 MPa are shown in Figure 9.29 whilst Figure 9.30
shows the result for 160 MPa. The first column of each graph which is green shaded
describes the stress and the remaining three columns describe strain. All strains in the
figures use mm/m units.
Figure 9.29. Stress and strain for yield stress of middle bar equal to 150 MPa when
middle bar is heated (left) and when temperature load is omitted (right).
Figure 9.30. Stress and strain for yield stress of middle bar equal to 160 MPa when
middle bar is heated (left) and when temperature load is omitted (right).
252
Indeed the higher yield stress in the middle bar (160 MPa) results in higher stress for
those bars when the middle bar is heated (80.05 Mpa and -160.1 MPa) whilst for 150
MPa the stresses are lower (75.06 MPa and -150.12 MPa) but it does not mean the
strains especially plastic strain are higher. The total strains to keep the force balance are
5.004 mm/m for side bars and -12.269 mm/m for middle bar when 150 MPa model is
used. Whilst 5.336 mm/m and -11.667 mm/m for 160 MPa model. Since the elastic
strain limited to 10 mm/m (or -10 mm/m) the plastic strains are -2.257 mm/m and 1.661 mm/m for 150 MPa and 160 MPa models respectively. The higher observed stress
seems to result from the higher Youngs modulus of 160 MPa material. This lower
plastic strain finally produced lower residual stress which is caused by misfit. This
discussion may explain qualitatively why the model involving aged martensite in the
multi-pass girth weld FEM model caused lower residual stress in the area close to the
weld line. Since the results consider aged martensite SSPT close to the experiment
results, this argument is considered to be valid.
253
References
[1]
H.S. Carslaw and J.C. Jaeger (1959), Conduction of Heat in Solid.Clarendon

Press, Oxford.
[2]
R. Komanduri and Z.B. Hou (2000), Thermal Analysis of the Arch Welding
Process: Part I. General Solutions. Metallurgical and Material Transaction B,
Vol. 31B. pp. 1353 1370.
[3]
Philippe Gilles, Walid El-Ahmar and Jean-Fracois Jullien (2009), Robustness

analyses of numerical simulation of fusion welding NeT-TG1application: Single
weld-bead-on-plate. International Journal of Pressure Vessels and Piping, Vol.
86, pp. 3 12.
254
PART VI
GENERAL
CONCLUSSIONS
Chapter 10
General Conclusions
Following the discussions presented in Chapters 1 to 9 some conclusions and future
work can be recommended. All the conclusions were obtained from the FEM results
described in Chapters 2 - 9. All the FEM results were verified with the analytical results
(Section 2.3 and Chapter 3) or validated with either secondary experiment results
(Chapters 4, 5, and 6) or primary experimental results (Chapter 7). Chapter 8 evaluated
the possibility to apply DC-LSND in multi pass butt girth weld joint of thick pipelines
and Chapter 9 discussed effects of SSPT and melting phenomenon in more details.
10.1. Conclusions
1. Welding modeling for high strength steel must involve three coupled analyses:
thermal, metallurgical, and mechanical analysis. Thermal and metallurgical
analysis need to be fully coupled because they have a significant effect on each
other whilst mechanical analysis can be done sequentially due to the weak effect
of mechanical results on the thermal and metallurgical results.
2. The new heat source model that combined Goldaks heat source model and
uniform temperature load provided a good prediction of thermal and mechanical
results.
3. The developed general database for parameters has accelerated computing time,
and provided flexibility and inter-changeability between different analyses.
4. The birth and death technique enables realistic modeling of the growing weld
bead. The main idea of the birth and death technique is to deactivate all the weld
bead elements and then activate the proper weld bead elements by evaluating the
relative position of those elements to the centre of the weld.
5. SSPT changes the prediction of residual stress especially in the area close to the
weld line through thermal properties, mechanical properties, and volumetric
change. The change in thermal properties had an insignificant effect whereas for
255
mechanical properties the elevated yield stress due to the formation of martensite
and volumetric change from phase transformation had clear effects on the
residual stress. The elevated yield stress developed a higher tensile residual
stress whilst the SSPT volumetric alteration developed a compressive residual
stress near the centre of the weld. When austenitic transformation takes place,
the shrinkage due to SSPT caused increased tensile residual stress in the centre
of the weld, however austenite SSPT was exhibited at high temperature where
the yield stress of the material was low, and as a result, a relatively small tensile
residual stress was formed. On the other hand, expansion due to a martensitic
transformation which formed a residual compression stress near the weld line
took place at a lower temperature and caused a relatively high residual
compressive stresses to form.
6. Another important aspect concerning SSPT that should be incorporated in the
FEM model is transformation plasticity. Ignoring transformation plasticity
causes overprediction of the final residual stress since a part of strain which
should be a portion of transformation plasticity is converted to stresses.
7. The maximum residual longitudinal stress for the girth weld obtained from the
ANSYS prediction was 452 MPa for the 3 oclock position and 552 MPa for the
9 oclock position, whilst the measurements were (428
30) MPa and (379
30) MPa for the 3 oclock and 9 oclock paths respectively. These predicted and
measured values were close to the yield stress of materials at room temperature
(600 MPa and 470 MPa for base metal and filler metal respectively) this
underlined the importance of considering residual stress when assessing the
integrity of welded structures.
8. When SSPT takes place yield stress and artificial coefficient of thermal
expansion are significantly changed. The artificial coefficient of thermal
expansion was used to accommodate the volumetric change due to SSPT
combined with the initial thermal expansion. From a case study in chapter 9 it
can be concluded the effect of the volumetric change (64.86%) is more
significant than yield stress change (25.57%).
9. When melting occurred in an element of the FEM model, it is important to
remove its plastic strain otherwise FEM will overestimate the residual stress,
especially close to the weld line. By resetting the plastic strain, the maximum
256
predicted residual stress at 3 oclock was 452 MPa, 1.82% higher than the
measured value, which is acceptable regarding some of the limitations
mentioned in Section 7.5. Ignoring the resetting plastic strain when an element
was melting gave a maximum predicted residual stress that was 194% higher
than the measured value. At the 9 oclock position, and considering plastic strain
resetting, the maximum predicted residual stress of 552 MPa was 34.96% higher
than the measured value (379.03 30 MPa). Note that the 34.96% is when
compared to the mean value of the measured results, but if compared to the
maximum results (sample A) it will be 20.26%. This is acceptable considering
the imperfections in both sides (modeling and experiment work). However
ignoring the resetting of plastic strain means that the maximum residual stress
was 900 MPa, i.e. 137% higher than the measured value. The higher deviation at
9 oclock (compared to the prediction at 3 oclock) is understandable since the
initial gap was changed further after welding at the opposite side had been
carried-out due to thermal expansion. From those facts it can be considered that
FEM without resetting plastic strain when melting takes place overestimated the
residual stress.
10. The plastic strain and stiffness removal mentioned in point 9 above,exists even
below melting temperature which called as annealing temperature.
11. A validated FEM model for girth weld was developed and the parametric study
based on the validated FEM can be carried out. In this thesis the application of
DC-LSND was modeled to study the possibilities of decreasing residual stress
which in turn may increase the lifetime of the girth weld of pipelines.
12. From the girth weld with DC-LSND using the FEM parametric study, the
residual stresses decreased when a CO2 trailing heat sink was applied. The DCLSND reduces the maximum tensile stress by 35.55%.
13. The DC-LSND reduces the residual stress by increasing the equilibrium
temperature and decreasing the duration from peak temperature to the
equilibrium temperature.
257
The contributions of this research can be summarised as follows:

1. A new analytical solution for the moving point heat source. This new solution
for the transient state complements Rosenthals existing solution for the
distribution of temperature of a moving point heat source at a steady state.
2. A new heat source model that is a combination of existing Goldaks heat source
model and uniform temperature load.
3. An FEM model that considers both prime martensite and aged martensite in
multi-pass welding using a strength reduction factor (Kd) in the FEM model to
represent the filler metal properties at elevated temperature.
4. A new database of parameters developed to interchange the parameters required
between analysis (thermal, metallurgical and mechanical) and between blocks in
certain analyses.
5. A validated FEM model for a girth-welded joint of X-70 pipes.
6. The predictions of residual stresses when DC-LSND is applied on the girth
welding.
10.2. Future work

1. As shown in Chapters 6 and 7, this thesis only considered martensite SSPT
which is the extreme condition. However, the discussion in this thesis pointed
out that future development of FEM modeling should consider some further
aspects that should be included in the modeling assumptions for specific
materials. In real high strength steel, not only is martensite formed, so too is
bainite depending on cooling rate. There are two kinds of bainite: upper bainite
and lower bainite and the best way to include this bainite in the model is a
mathematical model that describes the volumetric change as a result of SSPT
such as Koistenen-Marburger equation for martensitic SSPT. The other aspect is
the cooling rate that may be obtained by evaluating the temperature history of a
certain element.
2. Combining the parameter data base with the other program that can optimise
welding parameters. Since the data base in the general format (*.txt) the data
should be able to be opened by another program.
258
3. Fully coupled TMM may also be a worthy of future work, but it needs a strong
material science basis, especially in the coupled mechanical-metallurgical
aspect.
4. The experiment work proved that improving infrastructure of the mechanized
girth weld has practical benefits. These improvements cover two aspects: surface
preparation and controlling the welding parameter. Controlling the welding
parameters means improving control of automatic path and automatic welding
parameters. Improving the automatic path can be achieved using surface sensors
that can guide the welding torch on a correct path, and improving control of the
welding parameter can be done with automatically adjustable welding parameter
control. The control must depend on the position of the welding torch because
different positions need different parameters. An expert learning system that can
give optimised parameter for each position may be able to be developed. Overall
it can be said that the application of single-sided girth weld is a challenging topic
with a wide field of operating practice where many problems have not yet been
solved.
5. Experimental work with DC-LSND on the girth weld should be carried out
because from FEM study, the application of DC-LSND proved that it can
decrease the residual stress in the girth weld.
259

Residual Stress Analysis of Pipeline Girth Weld Joints

Uploaded by

Copyright:

Available Formats

Residual Stress Analysis of Pipeline Girth Weld Joints

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Residual Stress Analysis of Pipeline Girth Weld Joints

Uploaded by

Copyright:

Available Formats

University of Wollongong

University of Wollongong Thesis Collections

Residual stress analysis of pipeline girth weld joints

Research Online is the open access institutional repository for the

RESIDUAL STRESS ANALYSIS

The representation of a physical system usually by a computer or physical

Authors own words: Modeling and simulation are a pair of powerful

PART I BASIC CONCEPTS

PART II FEM vs ANALYTIC SOLUTION

PART III MODELING EXPLORATION BASED ON EXISTING DATA

5. Mechanical Analysis of NeT TG1 Specimen ....

6. Finite Element Model of Residual Stress Involving Metallurgical

PART IV Pipeline Girth Welding

7.2.4. Considering transformation plasticity

8. Simulating DC-LSND on the girth butt welding ...

Development of Residual stress in welding

FEM, analytic solution and experiment results .

Plotted value for

Point heat source in infinite media

Sketch of welding process and coordinate system .

Pattern of longitudinal distribution of residual stress .

Predicted microstructure configurations .

The rig for girth welding .

DC-LSNDI of a thin plate welding ..

Defines hsink in ANSYS APDL

Typical temperature profile of girth welding with DC-LSND .

Next Simdex pipeline project 3

Djarot B. Darmadi, Study of residual stress mechanism using three elasto-plastic

Djarot B. Darmadi, Muruganant Marimuthu, Kiet Tieu and John Norrish,

List of symbols and abbreviations

discussed the role of residual stress on failure at various loads. Residual

occurred for similar

Figure 1.1. Australias Pipelines [9]

Table 1.1. Next Simdex pipeline project [11]

As materials transported through pipelines can reduce the long-term cost of

. In most instances, pipelines

. Papua New Guinea, New Zealand and Australia also

1.2 Research objective

Developing an accurate FEM model based on simple existing data of a welding

FEM residual stress predictions of X-70 (a ferritic steel) girth weld.

Experimental work to validate the FEM model.

1.3 Research approach

Figure 1.2. Research aspects

simple moving point-heat-source in a semi-infinite solid. The results of moving the

. A verification of the stress developed by a thermal load was also

carried out for a simple elasto-plastic three bars model

and a thermal analysis for

1.4 Thesis outlines

C. Manfredi and J.L. Otegui (2002), Failures by SCC in Burried Pipelines,

J. Wang and A. Atrens (2003), Microstructure and Grain Boundary

Viorel Deaconu (2007), Finite Element Modeling of Residual Stress A Powerful

World Pipeline Maps, http://www.theodora.com/pipeline/world_oil_gas_and_

[11] Simdex, Future Worldwide Pipeline Project, http://www.simdex.com/future_

. Residual stress increases the rate of stress-

corrosion-cracking and can cause brittle fracture

. Moreover, it is also blamed for

. By using FEM, the influence of the mechanical properties, especially the

contributes to the residual stress of weldments[11]. Thus FEM simulations can

The materials in this chapter has been published in a paper: