Anda di halaman 1dari 149

CRANFIELD UNIVERSITY

NICHOLAS STONE







MODELLING NARROW GROOVE PIPE WELDING OF X100
PIPELINE STEEL







SCHOOL OF APPLIED SCIENCES







MSc Thesis








2

CRANFIELD UNIVERSITY




SCHOOL OF APPLIED SCIENCES




MSc Thesis




Academic Year 2006 2007




Nicholas Stone




Supervisors: D Yapp, P Colgrove

Sept 2007



This thesis is submitted in partial fulfilment of the requirements for the
degree of Master of Science


Cranfield University 2007. All rights reserved. No part of this
publication may be reproduced without permission of the copyright owner.

3

Abstract


Knowledge of temperature distribution patterns is useful in any welding process to
predict microstructure and distortion. I n the current work a model has been
developed to predict the thermal cycles during welding of a narrow groove joint of
X100 pipeline steel. The model was developed in the COMSOL Finite Element
package and considered a tandem arc welding power source, including temperature
dependent material properties. I n addition the work looked to evaluate the process
efficiency using liquid nitrogen calorimetry. Results show a good comparison
between model and experimental results, although further work is needed to more
closely match temperature profiles close to the heat source.
























4
List of Contents


Abstract...................................................................................................................... 3
List of Contents......................................................................................................... 4
List of Figures........................................................................................................... 6
List of Tables........................................................................................................... 10
1. I ntroduction ....................................................................................................... 11
Project Aims and Objectives............................................................................. 11
2. Literature Review............................................................................................. 12
I ntroduction......................................................................................................... 12
X100 Steels used for Pipelines.......................................................................... 13
Tandem Pulsed Gas Metal Arc Welding of X100 Steel................................. 17
Modelling the Welding Process ........................................................................ 19
Numerical Methods............................................................................................ 21
Heat Source Models............................................................................................ 22
Microstructure Modelling.................................................................................. 26
Thermal Material Properties for Welding Simulation................................. 28
Heat Source Efficiency Measurement ............................................................. 31
3. Modelling the Welding Process........................................................................ 36
Modelling the Geometry.................................................................................... 37
Heat Source Modelling....................................................................................... 40
Model Subdomain & Boundary Conditions.................................................... 46
Mesh Size............................................................................................................. 50
Material Properties............................................................................................ 51
4. Model Validation through Experiment........................................................... 79
I ntroduction......................................................................................................... 79
Materials.............................................................................................................. 80
Welding Equipment............................................................................................ 80
Recording Equipment ........................................................................................ 83
Metallographic Examination............................................................................ 84
Calorimeter Measurements .............................................................................. 85
Experimental Procedure.................................................................................... 86
Welding Tests...................................................................................................... 86
Set up of Welding Rig......................................................................................... 86
Welding Test 1: Flat Plate Test 3 Thermocouples........................................ 87
Welding Test 2: Flat Plate Test 4 Thermocouples........................................ 89
Welding Test 3: Bead on Plate Welds for Weld Wool Measurement.......... 90
Welding Test 4: Multipass Groove Weld with Multiple Thermocouples... 91
Thermal Efficiency Measurements.................................................................. 94
5. Experimental Results..................................................................................... 100
Welding Tests 1&2: Bead on Plate Tests...................................................... 100
Weld Test 3: Weld Pool Measurement .......................................................... 103
Weld Test 4: Multipass Groove Weld............................................................ 104
5
Process Efficiency Tests................................................................................... 111
6. Discussion of experimental results................................................................ 115
Weld Test 1........................................................................................................ 115
Weld Test 2........................................................................................................ 115
Weld Pool Measurement.................................................................................. 116
Weld Test 4: Multipass Groove Weld............................................................ 117
Welding Efficiency Calorimeter Tests........................................................... 119
7. I ncorporation and comparison of Experimental Results........................... 122
Weld Test 1........................................................................................................ 122
Weld Test 2........................................................................................................ 126
Weld Test 4: Multipass Groove Weld............................................................ 128
8. Discussion of Model results............................................................................ 141
9. Conclusions and Further Work...................................................................... 145
10. References........................................................................................................ 147


6
List of Figures

Figure 1: Comparison of weld and base metal areas with FeC phase diagram
[6]....................................................................................................................... 16
Figure 2: Variation of hardness profiles in HAZ of TCMP steel welded with
varied heat input [6]....................................................................................... 17
Figure 3: Diagram showing Rosenthal model description [7]......................... 20
Figure 4: Goldak Double Ellipsoidal heat source.............................................. 23
Figure 5: Comparison of normal Gaussian heat source and spread heat
source [25]........................................................................................................ 25
Figure 6: Thermal cycle comparision between heat treatment and welding
[6]....................................................................................................................... 26
Figure 7: Effect of peak temperature on CCT diagram [6].............................. 27
Figure 8: Typical temperature dependent properties for mild steel [8] ........ 29
Figure 9: More common form of temperature-dependent data for welding
simulation [31] ................................................................................................ 30
Figure 10: Heat source efficiencies for various processes [7].......................... 32
Figure 11: Typical measurement of arc efficiency by Kou [7]......................... 32
Figure 12: Typical groove geometry.................................................................... 37
Figure 13: Model geometry dimensions.............................................................. 39
Figure 14: Heat source dimensions..................................................................... 41
Figure 15: Distributed heat flux of Goldak heat source................................... 43
Figure 16: Distributed heat flux of modified heat source................................ 43
Figure 17: Heat source shape comparison.......................................................... 44
Figure 18: Heat source located in the groove..................................................... 45
Figure 19: Heat source applied to top of deposited weld.................................. 46
Figure 20: Prescribed temperature boundary condition
Figure 21: Top surface heat loss boundary condition 48
Figure 22: Bottom surface heat loss boundary condition
Figure 23: Thermal insulation & symmetry boundary condition........... 48
Figure 24: Convective flux boundary condition................................................. 49
Figure 25: Example of meshed geometry ........................................................... 50
Figure 26: Refined mesh around heat source.................................................... 52
Figure 27: Temperature dependent thermal conductivity models................. 53
Figure 28: Thermal conductivity variation 0mm.............................................. 54
Figure 29: Thermal conductivity variation 4mm.............................................. 54
Figure 30: Thermal conductivity variation 8mm.............................................. 55
Figure 31: Thermal conductivity variation 11.51mm....................................... 55
Figure 32: Thermal conductivity variation 21.32mm....................................... 56
Figure 33: Thermal conductivity variation 31.8mm......................................... 56
Figure 34: Specific heat capacity models............................................................ 58
Figure 35: Specific Heat variation 0mm............................................................. 59
Figure 36: Specific Heat variation 4mm............................................................. 59
7
Figure 37: Specific Heat variation 8mm............................................................. 60
Figure 38: Specific Heat variation 11.51mm..................................................... 60
Figure 39: Specific Heat variation 21.32mm..................................................... 61
Figure 40: Specific Heat variation 31.8mm........................................................ 61
Figure 41: Density and Emissivity models......................................................... 62
Figure 42: Density variation 0mm....................................................................... 63
Figure 43: Density variation 4mm....................................................................... 63
Figure 44: Density variation 8mm....................................................................... 64
Figure 45: Density variation 11.51mm............................................................... 64
Figure 46: Density variation 21.32mm............................................................... 65
Figure 47: Density variation 31.8mm................................................................. 65
Figure 48: Emissivity variation 0mm................................................................. 66
Figure 49: Emissivity variation 4mm................................................................. 66
Figure 50: Emissivity variation 8mm................................................................. 67
Figure 51: Emissivity variation 11.51mm.......................................................... 67
Figure 52: Emissivity variation 21.32mm.......................................................... 68
Figure 53: Emissivity variation 31.8mm............................................................ 68
Figure 54: Heat-transfer coefficient variation 0mm......................................... 69
Figure 55: Heat-transfer coefficient variation 4mm......................................... 70
Figure 56: Heat-transfer coefficient variation 8mm......................................... 70
Figure 57: Heat-transfer coefficient variation 11.51mm................................. 71
Figure 58: Heat-transfer coefficient variation 21.32mm................................. 71
Figure 59: Heat-transfer coefficient variation 31.8mm.................................... 72
Figure 60: Thermal efficiency variation 0mm................................................... 73
Figure 61: Thermal efficiency variation 4mm................................................... 73
Figure 62: Thermal efficiency variation 8mm................................................... 74
Figure 63: Thermal efficiency variation 11.51mm............................................ 74
Figure 64: Thermal efficiency variation 21.32mm............................................ 75
Figure 65: Thermal efficiency variation 31.8mm.............................................. 75
Figure 66: Travel speed variation 0mm.............................................................. 76
Figure 67: Travel speed variation 4mm.............................................................. 76
Figure 68: Travel speed variation 8mm.............................................................. 77
Figure 69: Travel speed variation 11.51mm...................................................... 77
Figure 70: Travel speed variation 21.32mm...................................................... 78
Figure 71: Travel speed variation 31.8mm........................................................ 78
Figure 72: Twin contact tip tandem torch.......................................................... 82
Figure 73: Groove plate sample secured onto the welding rig........................ 82
Figure 74: Welding rig with table motion control box...................................... 83
Figure 75: Calorimeter experimental setup....................................................... 85
Figure 76: Welding table travel speed calibration............................................ 87
Figure 77: Test 1 thermocouple locations........................................................... 88
Figure 78: Test 1 thermocouple distances from the weld line......................... 89
Figure 79: Test 2 thermocouple locations........................................................... 90
Figure 80: Bottom and top drill hole positions for thermocouples................. 92
Figure 81: Test 4 multipass weld thermocouples locations............................. 92
Figure 82: Groove geometry dimensions ............................................................ 93
8
Figure 83: Welded 1000g sample......................................................................... 95
Figure 84: Sample with added clamping feature.............................................. 96
Figure 85: Groove sample with added clamping feature................................. 96
Figure 86: Welding energy composition.............................................................. 98
Figure 87: Oscilloscope data for Test 1............................................................. 100
Figure 88: Oscilloscope data for Test 2............................................................. 101
Figure 89: Oscilloscope data for Test 3............................................................. 101
Figure 90: Weld Test 1 thermocouple data...................................................... 102
Figure 91: Weld Test 2 Thermocouple data..................................................... 102
Figure 92: Weld pool size dimensions, 0.617 m/min above, 0.793 m/min
below............................................................................................................... 103
Figure 93: Bead on plate weld dimensions....................................................... 103
Figure 94: Bead on plate with torch oscillation weld dimensions................ 104
Figure 95: Pass 1 thermocouple......................................................................... 105
Figure 96: Pass 2 thermocouple data................................................................ 105
Figure 97: Pass 3 thermocouple data................................................................ 106
Figure 98: Pass 4 thermocouple data................................................................ 106
Figure 99: Pass 5 thermocouple data................................................................ 107
Figure 100: Pass 6 thermocouple data.............................................................. 107
Figure 101: Cap pass thermocouple data......................................................... 108
Figure 102: Macro of multipass groove weld.................................................... 108
Figure 103: Multipass weld location of thermocouple 4................................. 109
Figure 104: Multipass weld location of thermocouple 5................................. 109
Figure 105: Multipass weld location of thermocouples 1,2,3 & 6................. 110
Figure 106: Multipass weld location of thermocouple 8................................. 110
Figure 107: Liquid nitrogen normal evapouration rate................................. 112
Figure 108: Calorimeter tests example weight loss profile........................... 113
Figure 109: Calorimeter test data showing variation with temperature.... 113
Figure 110: Calorimeter test data showing variation with weight .............. 114
Figure 111: Weld test 1 model data comparison dataset 1............................ 123
Figure 112: Weld test 1 model data comparison dataset 2............................ 124
Figure 113: Weld test 1 model data comparison dataset 3............................ 125
Figure 114: Heat Affected Zone model size for 5 mm heat model depth..... 125
Figure 115: Heat Affected Zone model size for 2 mm heat model depth..... 126
Figure 116: Weld test 2 model data comparison dataset 1............................ 127
Figure 117: Weld test 2 model data comparison dataset 2............................ 128
Figure 118: Multipass weld test 4 model data comparison dataset 1, pass 1
......................................................................................................................... 129
Figure 119: Multipass weld test 4 model data comparison dataset 1, pass 1
......................................................................................................................... 130
Figure 120: Multipass weld test 4 model data comparison dataset 1, pass 2
......................................................................................................................... 130
Figure 121: Multipass weld test 4 model data comparison dataset 1, pass 2
......................................................................................................................... 131
Figure 122: Multipass weld test 4 model data comparison dataset 1, pass 3
......................................................................................................................... 131
9
Figure 123: Multipass weld test 4 model data comparison dataset 1, pass 3
......................................................................................................................... 132
Figure 124: Multipass weld test 4 model data comparison dataset 1, pass 4
......................................................................................................................... 132
Figure 125: Multipass weld test 4 model data comparison dataset 1, pass 4
......................................................................................................................... 133
Figure 126: Multipass weld test 4 model data comparison dataset 1, pass 5
......................................................................................................................... 133
Figure 127: Multipass weld test 4 model data comparison dataset 1, pass 5
......................................................................................................................... 134
Figure 128: Multipass weld test 4 model data comparison dataset 2, pass 1
......................................................................................................................... 135
Figure 129: Multipass weld test 4 model data comparison dataset 2, pass 1
......................................................................................................................... 135
Figure 130: Multipass weld test 4 model data comparison dataset 2, pass 2
......................................................................................................................... 136
Figure 131: Multipass weld test 4 model data comparison dataset 2, pass 2
......................................................................................................................... 136
Figure 132: Multipass weld test 4 model data comparison dataset 2, pass 3
......................................................................................................................... 137
Figure 133: Multipass weld test 4 model data comparison dataset 2, pass 3
......................................................................................................................... 137
Figure 134: Multipass weld test 4 model data comparison dataset 2, pass 4
......................................................................................................................... 138
Figure 135: Multipass weld test 4 model data comparison dataset 2, pass 4
......................................................................................................................... 138
Figure 136: Multipass weld test 4 model data comparison dataset 2, pass 5
......................................................................................................................... 139
Figure 137: Multipass weld test 4 model data comparison dataset 2, pass 5
......................................................................................................................... 139
Figure 138: Heat Affected Zone size with width of 7.9 mm........................... 140
Figure 139: Heat Affected Zone size with width of 10 mm............................ 140










10


List of Tables


Table 1: Typical composition and properties of X100 steel [3]........................ 14
Table 2: Equation based material properties [21]............................................. 31
Table 3: Welding parameters for bead on plate tests..................................... 100
Table 4: Multipass groove weld input parameters and weld deposition..... 104
Table 5: Mutipass weld power input results.................................................... 104
Table 6: Multipass weld thermocouple location summary............................ 111
Table 7: Calorimeter test calibration summary.............................................. 111
Table 8: Welding efficiency results.................................................................... 112
Table 9: Model input data for Weld Test 4....................................................... 128





11
1. Introduction
Project Aims and Objectives

The main aim of this project has been to develop an accurate model that predicts
the thermal cycles present in a narrow groove pipe weld, utilising the Tandem
Pulsed Gas Metal Arc Welding (Tandem GMAW-P) process. The prediction of these
thermal cycles, due to each successive welding pass, allows an understanding of the
microstructure that will develop upon cooling of the weld, which in turn provides
estimates of mechanical properties of such welds.

The complete task of modelling the problem was broken down into smaller areas,
both theoretical and experimental. Therefore the overall project aims became:

Understanding of the fundamentals of the Tandem GMAW-P process and
how narrow groove welds are typically welded.
Knowledge of welding X100 pipeline steel, detailing welding parameters for
the process that are suitable to create defect-free welds.
Understanding how to model a welding process, involving characterisation
of welding heat sources and typical model properties and boundary
conditions.
Validation of the model using conventional narrow groove Tandem GMAW-
P and subsequent measurement of thermal cycles and weld properties.
An in depth study into the efficiency of the Tandem GMAW-P process in
order to utilise a sensible and accurate value in the developed model.

Another objective of this project has been to develop typical project working and
report writing skills, known as soft skills. Therefore the main soft skills that have
been tested and improved during the project have been:
Time management, developing skills for scheduling experiment and working
to deadlines.
Refinement of report writing skills, presenting concisely and eloquently the
details of our project.
12
2. Literature Review

Introduction

Demand for high strength steel pipelines is increasing, with recent studies showing
there will be a doubling in demand by 2030. This is due to continuing increased
demand for energy across the globe. Products such as oil and gas need to be
transported across large distances which can be done in a very cost effective way
with pipelines. Oil products can also be transported by tanker across the sea and
stored locally. This is a more economical alternative and one that is receiving a
large investment by oil and gas majors. Until recently sea tanker was not a viable
option for the transportation of Natural Gas, and so pipelines have been the main
method employed. The Liquified Natural Gas product has allowed a similar
technique of transportation to that of oil products, by way of tanker, but as
Aristotelle [1] states this is not yet economically feasible. There is some debate as
to whether this is the case, owing to a large number of LNG terminals that are
being built around Europe to store the natural gas produced from Russia.
Whatever the final outcome the near future holds a requirement for continued
pipeline construction both on and offshore.

I t has been stated [1] that the speed of laying a pipeline, and the quality of the in-
field welded joints during construction, is fundamental to the feasibility of such a
project. I f too many repairs must be made during construction, the cost of the
project increases. To avoid this problem robust welding technology and procedures
are required. I n essence the outcome of welding such steels in the correct
configurations must be known before the project is started. I n order to have
confidence in a particular welding procedure or method the usual technique is to
carry out a set of trial and error tests, to quantify the welding variables associated
with the method. I n fact this is a requirement of all of the welding standards
currently employed. This is due mainly to the fact that welding is a complex science
and it is difficult to predict the integrity of a joint by specifying merely input
13
variables. Such procedure qualification tests are time consuming and expensive to
perform as they expend manpower and welding consumables.

Mathematical modelling has obvious economic advantages in that, given a
workable model, much of the procedure development could be computationally
based. Computational work is far cheaper than experimental welding trials and so
would offer considerable cost savings. Mathematical models have been employed
for all aspects of welding, with varying degrees of success. The fact that welding
procedure specification and qualification is still experimentally based shows that
welding model performance is limited. The main reason mathematical models are
limited in performance is due to complexity of the input variables that must be
adequately accounted for to characterise a specific welding process.

The project undertaken will develop a model to characterise and predict the
thermal cycle of a typical narrow groove geometry used for welding pipeline steel in
service. The model has to take into account an appropriate heat source model to
represent the weld heat input, in addition to the correct geometry of joint. The aim
of the project is to predict with accuracy the thermal cycle in the Heat Affected
Zone (HAZ) and therefore allow prediction of microstructural changes that will
occur due to the imposed thermal cycle. This literature review will first discuss the
X100 thermomechanically produced steel, that is increasingly used for modern
pipeline installations. The manufacturing procedure and original microstructure of
such steels is important to understand before it is possible to consider the added
effect of welding thermal cycles. Typical welding procedures will be discussed in
relation to the heat input used and the effect on the base metal, especially the
HAZ. The review will then concentrate on the modelling of welding in general,
discussing the various models that have been historically used.

X100 Steels used for Pipelines

The relatively new X100 material has a specified minimum yield strength of
690MPa. Use of increased strength steels for pipelines allow either the operating
pressure to be increased, allowing more carrying capacity, or the thickness of the
14
pipe to be decreased. Either way there are distinct economic advantages. X100 is
part of the wider S690 family of steels that have this minimum specified yield
strength. These steels can show a large difference in their overall properties due to
their diverse manufacturing procedures and chemical composition. Various
standards exist for these steels such as: ASTM A543 Class2, API 5XL X100 or
MI L-S-12616 HY100 [2].
The previous generation of pipeline steels were produced by quenching and
tempering. Quenching of the steel is done after the material reaches the
austenising temperature of about 900 C. Quenching of the steel involves rapid
cooling that prevents formation of soft microstructural components such as ferrite.
I nstead the hard micro-constituent of martensite is formed. Pure martensite is far
too brittle for structural purposes and so it must be tempered. The tempering
process reduces the super-saturation of the matrix by forming carbides, and causes
some annealing which also reduces the dislocation. Quench and tempered steels
display a combination of good tensile and toughness properties. I n order that high
strength steels can be produced in such a manner, the ideal carbon level is between
0.12% - 0.18% to facilitate the martensite formation for such steels [2]. The
thickness of such materials poses a problem as it is difficult to achieve the required
cooling rate for martensite formation in the centre of thick section steels. This
problem may be rectified by adding alloying elements that increase the materials
hardenability, or the ability to form martensite. The carbon equivalent for such a
material is often used as a measure for the weldability of such steels. High carbon
equivalent steels are difficult to weld due to the high hardenability that may cause
cracking during cooling. I t is not possible to produce very thick quench and
tempered steels with low carbon equivalent, and hence they are difficult to weld.

Table 1: Typical composition and properties of X100 steel [3]

15
To overcome these problems, thermomechanically produced steels have been
developed which take advantage of thermomechanical rolling followed by
accelerated cooling. The rolling improves the grain refinement and dislocation
density. Most strengthening mechanisms produce an increase in strength at the
expense of a fall in material toughness. This is not so with grain refinement and it
is the principle reason why steels such as X100 are sought after. I t is a feature of
the very fine ferrite and second phase structure present [6]. A fine ferrite structure
is achieved before the austenite-ferrite transformation with microalloying additions
such as niobium, vanadium and titanium. These combine with carbon, oxygen or
nitrogen to pin the grain boundary movement and prevent growth [6, 2]. Work
hardening of the material is done after the material has been transformed to
acicular ferrite, which further increases the strength of the material. The major
advantage of this material is the much lower carbon content, which makes it more
a weldable steel due to a higher resistance to hydrogen cracking.
X100 is a type of thermomechanically produced steel, whose thickness is limited to
about 20 mm due to production methods. There are a number of different
approaches to producing X100 steel. The manufacturer can choose which
composition and manufacturing process is used provided the material meets the
required strength and impact toughness. Experience has shown the an optimised
two-stage rolling process allows the use of a low carbon content but quite a high
carbon equivalent [3].

Welding of such material means the strength due to work hardening and grain
refinement is lost in the HAZ. Figure 2 shows the coarse grains of heat affected
zone material produced by grain growth which reduces the strength of the welded
steel. I n addition high peak temperatures dissolve the precipitates that pin the
grains and prevent growth, leading to coarser grains [6]. Welding of X100 steel
poses similar problems to those of thick gauge X80 steel [3]. The use of precipitates
that will not easily dissolve, such as titanium nitride, TiN, is a solution to the
problem [6]. The representation of welding thermal cycles using an equilibrium
phase diagram, as in figure 2, is not however appropriate as welding produces far
from equilibrium conditions. The remedy to this is discussed further in the
literature review regarding microstructural models in welding.

16

Figure 1: Comparison of weld and base metal areas with FeC phase diagram [6]

A number of studies have been carried out to evaluate the effect of various welding
procedures on the final properties of welded pipeline girth welds [4, 5]. Welds are
designed to act as crack arresters that essentially prevent a crack from propagating
the length of a pipe, causing catastrophic failure. For this reason the weld metal is
usually designed to have a higher strength than the base metal. This is known as
overmatching. The selection of welding electrode is where this is put into practise
and there are a number of choices with regards to electrode composition. Laratzis
[5] and Hudson [4] have both performed extensive tests regarding suitable welding
electrodes. The advanced nature of welding procedure specification is not relevant
to this project. The important aspect to focus on with respect to modelling such
procedures is the heat input of the process and welding geometry. Finally an
appreciation of the final weld strength or hardness is important.
Hudson showed that 100 C preheat was required to produce good welds with Gas
Metal Arc Welding (GMAW) and Shielded Metal Arc Welding (SMAW) welding
procedures on X100 [4]. Typical hardness values were 281 HV10 for the HAZ of the
base metal and 299 HV10 for the weld metal using SMAW electrodes. Hudson
17
limited the amount of heat input into the joint to 1.5kJ /mm to reduce the likelihood
of mechanical strength reduction. Laratzis also found that high heat input was not
acceptable. Submerged Arc Welding (SAW) of joints failed due to a high heat input
of 2.5KJ /mm. Dual tandem GMAW trials proved more successful employing a lower
heat input of 0.5kJ /mm.


Figure 2: Variation of hardness profiles in HAZ of TCMP steel welded with varied
heat input [6]

Figure 3 shows how the heat input of the welding process affects the hardness in
the HAZ of a typical Thermo-Mechanical-Controlled-Process (TCMP) steel. High
heat inputs reduce the hardness and, in consequence, the strength of the HAZ.
Hudson states that the pipe wall thickness, bevel angle and welding process have a
major impact on the HAZ dimensions and hardness levels. I t is therefore essential
that these aspects are modelled correctly to allow accurate prediction of the
thermal cycle and hence final weld composition.
Tandem Pulsed Gas Metal Arc Welding of X100 Steel

Welding of gas pipelines has become an automated process as automation provides
more stability during welding and hence a higher quality weld. Arc stability, for
18
example, is far more constant when the welding torch is mounted on to a purpose-
built pipe jig than when used by hand for manual operation. The Tandem Pulsed
Gas Metal Arc Welding (Tandem GMAW-P) is a welding process that has been
developed for welding pipeline steel and is an area for which much research has
been done at Cranfield University. The major advantage of the process is the high
production rate, due to use of two welding electrodes to feed one weld pool.
I mproved positional welding around the pipe is also possible due to a fast freezing
weld pool with pulsed power supply used.
Gas Metal Arc Welding (GMAW) is probably the most widely used process to join
ferrous materials. The process creates a welding arc between a continuously fed
wire electrode and the work surface with additional gas shielding present to
displace the natural atmosphere. The arc melts the wire electrode creating a weld
bead. The GMAW process is mostly used with the Direct Current Electrode
Positive (DCEP) power configuration, meaning that DC voltage is used and the
electrode represents the positive terminal of the supply, with the work piece the
negative terminal. GMAW is mostly a constant voltage process, whereby the
current is regulated by the wire-feed speed of the electrode wire [6]. This allows a
constant arc length to be maintained even if the welding torch is moved away from
the work piece. The transfer of metal droplets from the electrode tip to the work
piece is achieved by globular transfer below a certain welding current. Globular
transfer is characterised by large metal droplets that fall from the electrode tip due
to their own weight. Above a certain current, known as the Transition Current, the
transfer of metal becomes more ordered with smaller droplets and is known as
Spray Transfer. The advantage of spray transfer is that, as the droplets are
accelerated across the arc with electromagnetic force, there is greater possibility to
weld in positions other than down-hand or directly on top of a plate. This has
obvious advantage for pipe welding where the entire pipe must be welded at all
angles. The spray transfer mode is also less susceptible to spatter.

The high current required to produce spray transfer is associated with a high heat
input for the process. I t has already been shown that materials can be sensitive to
high heat input processes with associated loss of mechanical strength. A
compromise to this is the Pulsed Current process. This has an overall low heat
input but can be used for greater positional work. The GMAW-P process utilises
19
finite periods of high current, above the transition current, with an overall low base
level of current to maintain the welding arc. To define a pulsed current waveform
requires selecting the peak current and time in addition to the base current and a
frequency of pulsing. The magnitude of the pulse waveform affects the molten
metal detachment from the electrode. Trial and error tests are often performed to
determine the correct set of parameters to produce satisfactory welds [30].
The Narrow gap groove geometry is used in conjunction with the GMAW-P or the
tandem version of the process. The advantage of the narrow gap configuration is
that a smaller amount of material is required to weld the joint, thereby increasing
productivity. I n addition the heat input into the material is less due to reduced
welding time. The narrow groove profile does require accurate electrode positioning
and for this reason the automated process is always used.
Modelling the Welding Process

Mathematical modelling of welding phenomena has become extremely popular in
academia as researchers look to predict processes performance. The proposed
advantages of modelling are mostly due to cost. Welding simulation is, however,
not readily used in industry where a pragmatic approach to welding still
dominates. The applications that have found use have been in the aerospace and
nuclear industries, where safety plays and important role in the production of
components [10].
Looking only at temperature field prediction due to welding, there are a number of
general approaches that can be used. Experimental methods, both full and small
scale allow evaluation of the permeating welding temperature fields. Often, due to
component complexity, this is the only realistic method of obtaining useful,
workable data. To obtain this experimental data, however, is time consuming and
expensive. Extrapolation of data obtained from experimental methods may not
characterise the welding process effectively and may not be useful for purposes
outside of which the original test was designed. Analytical and numerical
modelling offer the advantage of characterising a wider area of welding
phenomenon, although the complexity of the science has severely limited the
application of reliable models.
20
Analytical models quantify a phenomenon using mathematical equations or
relationships. These relationships look to characterise behaviour by simplifying the
problem greatly. Rosenthal was the first to create a useful mathematical model for
welding, based on Fouriers heat flow equations [7]. He used an infinitely small
point heat source to represent the arc heating. I n addition the line heat source is
often used for analytical models. Rosenthal postulated that, given a heat source
moving with a constant speed, on a constant plate thickness the analysis could be
done as quasi-stationary. Quasi-stationary means that the temperature locally
around the heat source remains constant. This is one major simplification which is
very useful for all welding models. I n addition to the point heat source analytical
models use the assumptions that no convection occurs in the weld pool, there are
no convection or radiation heat losses and there is negligible heat of fusion. The
models are designed for heavily simplified geometries to further aid analysis.
Another major simplification, used with all analytical methods, is to presume that
the thermal properties of the material are not temperature dependent. I n fact, due
to the complexity of inclusion, no analytical model accounts for temperature
dependent material properties or latent heat effects. The Rosenthal model is
highlighted in figure 3.

Figure 3: Diagram showing Rosenthal model description [7]

The diagram shows the temperature evaluated at a point from the heat source,
characterised by the equation:
( )
0
0
2
exp
2 2
s r
T T kg V V
K
Q

_ _


, ,
(1) [7]
21
Where T is temperature, T0 is workpiece temperature, k is thermal conductivity, g
material thickness, Q is heat input, V is travel speed, is thermal diffusivity, r is
radial distance from the heat source and K0 is a modified Bessel function of second
kind.

Basic geometries are characterised by the semi-infinitely extended solid, infinitely
extended plate and the indefinitely extended rod [26]. Choice of geometry is based
on the heat flow analysis that is of interest.
Even though the original Rosenthal model has great simplifications and was
developed in the 1940s it is still a very quick and accurate way of determining the
thermal cycles due to welding on simple geometries such as flat plates. One well
known problem with the Rosenthal model is that, by use of the point heat source,
the accuracy of temperature prediction is not good near the welding arc. I n fact the
predicted temperature tends to infinity. Adams [6] accounted for the peak
temperature with another model to predict the temperature at any distance from
the fusion line. The Adams model is:
0 0
4.13
1 1
g
p m
VY C
T T Q T T

+

(2) [6]
Analytical models give quick solutions, where simplifications to the problem are
sufficient, but complex structures using these models are difficult to analyse.
Ramirez [29] states that Rosenthals analytical heat flow models used for bead on
plate welding are not adequate for representing multipass welding of medium thick
plates. The major field of work is now focussed on numerical analyses.

Numerical Methods

Since the 1970s the work of computational analysis in welding engineering has
increased [8]. The advantage of using computers for analysis is calculation speed.
Numerical models are often used where a closed form of analytical solution is not
available. I n welding, as previously discussed, the closed form solution of thermal
cycle models grossly simplifies the actual welding setup. Therefore numerical
models allow better prediction of the real weld properties. Numerical models can be
based on different methodologies such as Finite Difference, Finite Element and
22
Computational Fluid Dynamics. All of these methods involve solution of partial
differential equations which describe the problem over a mesh created from the
geometry. The method of solving these equations, however, is different for each
method. Given the nature of heat flow and stress analysis the primary methods
used for computational welding simulations are either Finite Difference or Finite
Element techniques. The finite difference technique was more popular before
computational power allowed Finite Element methods to improve.
There have been a number of reviews already done of the vast amount of numerical
models used for welding simulation [8,9]. These models have different objectives in
that some wish to predict weld thermal cycles, others are trying to quantify the
amount of distortion using complex thermo-mechanical coupling relationships. As
this project is focused on evaluating thermal cycles, alluding to prediction of effects
on the base metal, the review has not looked in depth into distortion and stress
analysis models. Although it must be pointed out that all welding simulation
models have to ascertain the thermal cycle due to welding and so they all become
relevant to the current work.
Lindgren [9], Komanduri [9] and Goldak [10] have all done extensive reviews of
many welding models from both the analytical and numerical perspective. Common
goals of simulation have been to quantify the effects of weld thermal cycle on
structure or thermal expansion and volume change due to phase transformations.
This is known as thermal dilatation. Multipass welding and diverse geometries has
also been a constant area of model development. I t has been shown by Wahab [12]
that the heat source model has a very influential effect on the validity of the entire
welding model. For this reason the main analysis of the literature review is
focussed on the various heat source models and their applicability to different
welding applications.

Heat Source Models

The original point heat source model was, as previous described, the Rosenthal
solution. Lindgren points out that the original method of Rosenthal was enhanced
by Christensen [9], in order to make the analysis dimensionless. This allowed
many different welding processes to be compared easily. Eager and Tsai [11] stated
23
that the heat source due to welding should be defined as a distributed source over
the surface of the material. They developed the 2D Gaussian distributed heat
source. Similar to Rosenthal the model did not include convective or radiative heat
flow. The model also used constant thermal properties due to the analytical nature
and was applicable only to quasi-steady-state analysis. The model did provide a
comparison between actual welding inputs, such as current, voltage and travel
speed, and the dimensions of the weld pool. By far the most widely used heat
source model for numerical models of welding is that of Goldak [10]. Goldak states
that the analyst requires a heat source model that accurately predicts the
temperature field in the weldment. This argument is that the heat source is not
purely surface based but has an associated volume due to the weld pool
dimensions. The Double Ellipsoidal heat source model, proposed by Goldak, allows
the volumetric nature of actual welding arcs to be taken into account. The model
can be shown to be of general form, with the other distributed models of Pavel and
Paley [10] to be special cases. The model is highlighted in figure 4.


Figure 4: Goldak Double Ellipsoidal heat source

24
The coefficients of a, b, c1 and c2 shown are used to specify the boundary of the
fusion zone. The following equation is used to characterise the heat source in front
of the arc:
( )
2
2 2
1 1
6 3
, , exp 3 3 3
f
R Q
x y z
q x y z
a b c abc
1
_
_ _
1

, , 1
,
]
(3) [10]
and the following for behind the arc:
( )
2
2 2
2 2
6 3
, , exp 3 3 3
b
R Q x y z
q x y z
a b c abc
1
_
_ _
1

, , 1
,
]
(4) [10]
where x, y, z are spatial coordinates, Q is the total heat input in the process, and R
is a balancing factor between the front and back equations whereby both the sum
of the integral of both ellipsoids should equal the total heat input Q.

The advantage of the model is that a variety of different welding processes, such as
arc or laser for example, can be specified by using different multiplication factors
for the above equations. The problem exists as to how to determine these model
coefficients and it has been shown by Gery et al. [13] that correct heat source model
parameters are vital in producing an accurate prediction of the fusion zone and
heat affected zone size. Moore, Bibby and Goldak [14] state that setting the values
to 10% smaller than those of the weld pool gives good results. Alternatively the
width of the weld pool can be used by way of standard multiplication factors to
characterise the heat source. With this technique the front of the heat source
parameter, c1, is set to half the bead width, a. The distance behind the heat source,
c2, is set to twice the bead width, four times a. The problem still exists as to how
the weld pool dimensions can be measured. The typical method is to produce bead
on plate samples with a known set of welding parameters. The width and depth
can then be used for the model input values. The weld pool length still poses a
problem using this technique if the bead profile is very smooth. A novel approach
was proposed by Wahab et al. [12]. An apparatus was designed to eject molten
metal, by workpeice acceleration, from the welded material during welding. The
crater that is left can be measured and gives an accurate measure of the weld pool
geometry for a given set of welding parameters. The authors found that welding
speed and current had the greatest influence on weld pool length. I ncreased heat
input values produced an increase in all of the weld pool dimensions [12]. Gery et
25
al point out that this is not a linear relationship [13]. The crater method is feasible
but using it would require some time and experience to develop. Thermal boundary
measurement is another method of characterising the heat source. This requires
expensive thermal imaging equipment and only gives surface weld pool
measurements.
Without doubt the largest difficulty in using a mathematical model for a
distributed heat source is ascertaining the correct input parameters. Gery showed
that the peak temperatures and distributions are extremely sensitive to small
changes in the heat source model [13].
An added complication to the heat source model is due to the weaving or oscillating
nature of the arc when performing narrow groove welding of pipelines. The
movement of the heat source could be programmed as oscillating from side to side
but Sabapathy [25] used a widened version of the Goldak heat source model. The
advantage of this method is that the path of the arc can remain linear along the
length of the weld. Sabapathy modified the Goldak model, quoted previously, to:
( )
1 2 3
, , exp 3 3 3
n n n
f
x y z
q x y z Q
a b c
1
_ _ _
1

1
, , ,
]
(5) [25]
The value of n2 allows a broadening of the heat source as shown in figure 5 below.
Although this solves the problem of an oscillating source it does complicate model
input parameters.

Figure 5: Comparison of normal Gaussian heat source and spread heat source [25]

26
Microstructure Modelling

The high thermal gradients caused during welding produce far from equilibrium
cooling conditions, as previously discussed. I n order to ascertain the effect of a weld
thermal cycle on a particular ferritic steel it is not possible to use methods that
may be applicable to other heat treatments. The figure below shows the difference
between a typical heat treatment, on the left, and a welding thermal cycle, on the
right, at various points in the weldment. Heat treatments involve a soak time at a
specified temperature, usually around the Ac3 temperature, followed by a constant
cooling period. The graph on the right of the figure shows that the peak
temperature during welding varies with distance from the weld centreline, and the
accompanying cooling rate is not constant. Continuous Cooling Curves or CCT
diagrams are often used to predict final cooled microstructures of metals. These are
produced with a single peak temperature and so cannot easily be applied to
welding. The rate of heating and peak austenising temperature highly affect the
phase transformation upon cooling.

Figure 6: Thermal cycle comparision between heat treatment and welding [6]

The figure below shows that the effect of increased peak temperature is to shift the
CCT curve to the right. The larger austenite grain size increases the hardenability
of the steel, allowing more time for the harder martensite structure to form [6].
27

Figure 7: Effect of peak temperature on CCT diagram [6]

For the prediction of final microstructure the important cooling time from 800 C to
500 C is used. This time interval directly determines the hardenability of the
weldment as it is during this time that austenite decomposes to bainite or
martensite [15][14]. Various microstructure prediction models are based on the
T8/5 time. Yurioka, for example, has developed a number of equations to
characterise the HAZ hardness, tensile strength and weld metal toughness, based
on a modified Rosenthal model by predicting peak temperature and thermal
profiles [17,18]. I n addition the T15/1 time, that is the time for cooling from 1500 C
to 100 C, is also considered as this dictates the time for hydrogen to escape the
HAZ and prevents cold cracking [14].
The problems associated with using a CCT diagram for welding purposes has not
prevented most work on microstructure prediction from utilising them. Often a
database of CCT diagrams for different peak temperatures and various material
chemical compositions is employed. Okada et al [16] used an estimated thermal
cycle in combination with the CCT diagram database. This is obviously difficult to
achieve without the necessary data to support it.
Odanovic simulated the microstructure in the HAZ of quench and tempered HY-
100 steel. The study used only two CCT diagrams, one for the Course-Grained HAZ
and another for the Grain-Refined HAZ. The prediction of HAZ dimensions was
within 13% of the experimental values. The hardness prediction was not very
28
accurate using this method which was thought to be primary due to the inaccurate
thermal cycle analysis.

Thermal Material Properties for Welding Simulation

Welding produces a large temperature gradient in the base material being welded.
Materials do not exhibit constant thermal properties, but instead vary with
temperature and the relationships are not necessary simple. For analytical
solutions there exists the problem of whether to use room temperature, maximum
temperature or average temperature properties. Komanduri [8] suggests that
average temperature values should offer the most logical choice for analytical
models. Moore [14] shows that if a very low value of 25 W/mC for thermal
conductivity is used, as a constant value for analytical methods, the final thermal
analysis and HAZ structure can be well compared. I t is thought that the low value
of thermal conductivity compensates for neglecting the effect of latent heat or the
effect of keeping other quantities constant that are actually temperature
dependent. I t was pointed out that these results may have been specific to the case
at hand.
The choice of whether to use constant or temperature dependent material
properties has been solved by the introduction of Finite Element techniques for
analysis with added computational functionality. I n fact Lindgren states that there
is no longer a need for analytical methods for thermal analysis due to the ease of
FE analyses [9]. The new computational methods can accommodate quantities that
vary with temperature and the argument for their inclusion is too strong to ignore.
The problem of a suitable source of data does, however, still exist. I deally material
data would be available for all material types to allow good input data for
simulations. I nstead most studies use a set of density, thermal conductivity,
specific heat and latent heat of fusion for mild steel, with the presumption that the
data does not vary significantly with chemical composition. Figures 8 and 9 show
the variance of data gathered from two different sources. I t should be noted that
the relationships are far from linear. Zhu [24] points out that the thermal
conductivity has the most effect of the all material properties on the simulation
results, due mainly to the fact that conduction is the primary mode of heat
29
transfer. The plot in figure 9 shows that there is a jump in the specific heat plots
due to the latent heat of transformation which may be difficult to incorporate into a
model. Goldak [10] suggests that the simplest way to include latent heat is to
compute specific heat from the enthalpy, with the specific heat being the derivative
of enthalpy.

Figure 8: Typical temperature dependent properties for mild steel [8]
30
Temperature Variable Material Properties
0
10
20
30
40
50
60
0 500 1000 1500 2000 2500 3000
Temperature DegC
T
h
e
r
m
a
l

C
o
n
c
d
u
c
t
i
v
i
t
y

(
W
/
m
K
)
0
200
400
600
800
1000
1200
1400
1600
S
p
e
c
i
f
i
c

H
e
a
t

(
J
/
k
g
m
3
)
Conductivity
Specific Heat

Temperature Variable Material Properties
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
0 500 1000 1500 2000 2500 3000
Temperature DegC
D
e
n
s
i
t
y

(
k
g
/
m
3
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
E
m
i
s
s
i
v
i
t
y
Density
Emissivity

Figure 9: More common form of temperature-dependent data for welding
simulation [31]

31

Table 2: Equation based material properties [21]

Heat Source Efficiency Measurement

Given the primary input into the distributed heat source is the overall heat from
the welding arc, it is clear that an accurate evaluation of arc heating is required.
Not all of the electrical input power of the welding arc is converted to useable heat
in the process. Heat losses due to convection and radiation to the environment are
also present. The ratio of actual heat input to power input for the process gives the
overall process efficiency. The process efficiency for GMAW is quoted at between
62% - 85% [21]. Such a spread is difficult to incorporate into an accurate heat
source model. The typical efficiencies of various welding processes are shown in
figure 10, where LBW is Laser Beam Welding, showing the lowest efficiency,
although this is dependent on the surface reflectivity. PAW in the diagram is
Plasma Arc Welding, GTAW is Gas Tungsten Arc Welding, SMAW is Shielded
Metal Arc Welding, GMAW is Gas Metal Arc Welding, SAW is Submerged Arc
Welding and EBW is Electron Beam Welding, with the higher heat source
efficiency. These thermal efficiencies quoted are a combination of three calorimeter
measurements estimating heat transfer from the arc, filler metal drops and
cathode heating [7] and therefore represent quite a detailed interpretation of
efficiency.

32

Figure 10: Heat source efficiencies for various processes [7]

The process used in the project will be Tandem GMAW-P which will also have its
own associated process efficiency and must be investigated experimentally to
determine the value.
A number of approaches can be used to estimate the process efficiency. Goncalves
et al [20] point to two main methods, the cooled anode technique and
measurements from actual weldments. The cooled anode approach is also
highlighted by Kou [7].

Figure 11: Typical measurement of arc efficiency by Kou [7]

The figure above highlights how the temperature change in cooling water passing
through a welded section can account for the energy input into the system. The
integral of the thermal cycle, given by equation 7, can be used to calculate heat
33
input. The power input will be known based on the current and voltage measured
during the trial and equation 6 can then be used to obtain a value for efficiency.

weld
weld
Qt Q
EIt EI
(6) [7]
( ) ( )
0 0
weld out in out in
Qt WC T T dt WC WC T T dt



(7) [7]
The technique of Tandem Pulsed GMAW (GMAW-P), to be modelled in this project,
has a complex power waveform where the average current of the process remains
below the spray transition temperature for fast cooling, but pulses of current above
the transition temperature create spray transfer for improved positional welding.
The method of assessing process efficiency of GMAW-P is complicated by the fact
that, due to the very rapid variations of the pulsed voltage and currents, it is not
possible to use the simple relationship given in equation 6 as the input process
power. I n addition it has been stated that efficiency is function of voltage and
current of the arc [10].
J oseph et al [21] gave a review of the various methods of assessing power input
into the GMAW-P process. The studies focussed on the three ways of accounting for
power input:

Root Mean Squared measurement of pulsed voltage and current waveforms
o
2
1
2
1
.
RMS RMS RMS
n
i
RMS
i
n
i
RMS
i
P I V
I
I
n
V
V
n

(8) [21]
o where I is welding current, V is welding voltage and n is the number
of points in the waveform
Average values of voltage and current
34
o
1
1
.
AV AV AV
n
i
i
AV
n
i
i
AV
P I V
I
I
n
V
V
n

(9) [21]
I nstantaneous Averages of voltage and current
o
1
.
n
i i
inst
i
I E
P
n


(10) [21]

The study concludes that Root Mean Squared or RMS values should only be taken
for non-pulsed waveforms where there may be a slight ripple present on the power
supply. The Average value method takes the average of the voltage or current over
a complete cycle, while the I nstantaneous Average method calculates the power by
multiplication of the voltage and current at each point in the cycle. The advantage
of the instantaneous method is that it can account for any spikes in the power
supply that may occur during welding.
The studies point out the process efficiency measured by RMS values give 10%
higher than instantaneous measurement, while the average method gives
efficiencies of 12% lower [21]. Therefore the study concludes that the instantaneous
power should be used when calculating power input for process efficiency
measurements of pulsed power supplies.

I n addition to the power input review the study of J oseph et al also highlights the
use of liquid nitrogen to measure the heat input during GMAW-P. I n effect a small
welded coupon is placed into a dewar containing liquid nitrogen, the initial and
final weight of which can be used to calculate the heat input into the coupon. I f the
process is done quickly the losses to the environment are minimal. A more detailed
description of the process is given in the Experimental Procedure section of this
report.

The literature study showed that some researchers have investigated the variation
of welding process efficiency when different geometries are welded. Rykalin [29]
proposes a correction factor for the heat input of the process based on the joint
35
geometry of a V bevel groove weld. Ramirez states that Rykalins model implies
that arc energy losses to the surrounding environment are greater than the losses
when welds are made within a groove geometry. The experimental study of process
efficiency in this project should prove whether this is the case.

36
3. Modelling the Welding Process

The Finite Element Method was chosen to model the process of multipass pipe
welding. The literature review has highlighted the fact that computational
methods are far more adaptable to complex welding cases and it is for this reason
the FEM solution was chosen. FEM obtains approximate solutions of engineering
problems by cutting the structure in elements and finds solutions for a number of
connected elements.
The software package chosen for modelling during the project was COMSOL
Multiphysics, formerly FEMLAB. COMSOL is an FEM analysis and solver package
that has many modules for different engineering applications [27]. Very often,
when using FEM packages, a very in-depth knowledge of cell structures and types
must be known before correct characterisation of the process can be achieved.
COMSOL allows modelling of physical problems without great knowledge of
required solution structures. The package allows standard construction of the
domain geometry, independent of the problem type, from heat transfer to fluid
flow. COMSOL also has the ability to combine various physical phenomena, for
example coupled problems such as thermal cycles and associated residual stress.
This would be an ideal application for welding analysis.
The Heat Transfer Module was used in COMSOL to create a model that combined
conduction, convection and radiation.
A thermal problem is setup by specification of the domain, material parameters,
initial conditions and boundary conditions. Solution to the problem is temperature
at all points in the domain.
The field equation for heat conduction in a 3-dimensional solid of dimensions x, y
and z is given as:
2 2 2
2 2 2
1
v
Q T T T T
t c x y z c t


_
+ + +


,
(11) [32]
where the parameter T is temperature, t is time, is the thermal conductivity, c
the mass-specific heat capacity and the density. Q is the heat energy released per
unit volume.
By adding a convective term and simplifying the 3-dimensional notation this
becomes:
37

( ) . .
T
c T Q c u T
t

(12) [27]
where u is the velocity vector.

The aim of the project was to correctly model the typical geometry of a narrow
groove pipeweld. Additional suitable boundary conditions were then set to allow
evaluation of thermal cycles due to an applied heat source. Figure 12 shows the
dimensions of the joint that was modelled. The bevel angle was 5.


Figure 12: Typical groove geometry

The actual weld is completed in a number of separate weld passes. Therefore it was
decided that a separated model would be setup for each weld pass. Each model
would have different groove height and width dimensions, due to the amount of
weld deposited in the groove, in addition to different heat input and travel speeds.
The boundary conditions would be identical for all of the models. The model would
simplify the welded components by not considering weld metal and base metal
separately. I nstead the entire volume would be considered as one homogeneous
volume with the same material properties.
Modelling the Geometry

I t was decided from the outset that the best way to model the geometry in
COMSOL was to use a scripting language. The COMSOL script is a language that
5.00
9.00
22.8 All measurements in mm
38
allows the user to write code to be implemented in COMSOL. Using a COMSOL
script the user can actually specify the complete model, from the geometry to the
visualisation parameters of the solution.
The advantage of using a script to build the geometry is that, should the user wish
to change the dimensions of a model, it can be done by merely changing a pre-
defined variable in the script. Without a script the complete model must be rebuilt
from scratch for each change in geometry. The time to write the script was,
therefore, easily less than the time to make any changes to model geometries
further down the line.
The model geometry is defined in terms of plate size (length, width and thickness)
in addition to the groove dimensions. The dimensions of the groove are shown in
figure 13. The input variables for the model geometry are:

plate_t - Plate thickness in metres
plate_l - Plate length in metres (actually width in model)
plate_w - Plate width in metres (actually length in model)
back_t - From groove base to bottom of the model in
metres
bevel_deg - Bevel angle
grve_botdim - Width of the groove at the lowest point

The script uses the above input parameters to calculate the other geometry
dimensions as follows:

grve_ht = plate_t - back_t; - Groove height
bevel_rad = bevel_deg*(pi/180); - Bevel angle in radians
grve_topdim = (grve_ht*tan(bevel_rad))+grve_botdim; - Groove top
dimension

The calculated variables are used as coordinates to specify the geometry
automatically once the COMSOL script is executed. Due to symmetry only half of
the groove geometry needs to be created, as the other half would be redundant.
This also means less computation is required to solve the model.
I t should be noted the zero reference height, or zero position on the Z axis, is
always the bottom of the groove. The zero position therefore moves as the height of
the groove is varied. This was done to aid positioning of the heat source model to a
simple reference plane.
39





Figure 13: Model geometry dimensions

40
For characterisation of the multipass weld 5 models were developed, each to model
a single pass. Accurate dimensions of the groove geometry could not be specified
until experimental results had been collected from actual weld tests.
Model length was varied depending on the temperature profile time required. To
obtain a time from the point of heating, the distance from the heat source was used
in conjunction with the travel speed.
The COMSOL script file that creates the mesh not only specifies the grooved
sample dimensions but also defines a region in the groove where the heat source is
situated and is used to simulate insulation of the weld pool from heat losses to the
atmosphere. The setting of the boundary conditions is described in more detail in a
subsequent section. To create this geometry two semi-ellipse shapes were modelled
in COMSOL and applied to the groove geometry. The COMSOL model file can be
used vary the size of this insulation shape by varying the size of the input
variables.


Heat Source Modelling

The literature review showed that correct characterisation of the heat source
influences greatly the final accuracy of the welding model. For arc welding the
most common form of heat source model is the Goldak Double-Ellipsoidal model.
The model is made up from two semi-ellipsoids, one at the front and one at the
rear. Figure 14 shows the configuration of the standard heat source model.
Dimensions of the model in figure 14 correspond to the geometric coefficients in the
following equations:

( )
2
2 2
6 3
, , exp 3 3 3
f
f f
R Q
x y z
q x y z
c a b abc
1
_
_ _
1



1 , ,
,
]
(13) for the front
( )
2
2 2
6 3
, , exp 3 3 3
b
b b
R Q x y z
q x y z
c a b abc
1
_
_ _
1

, , 1
,
]
(14) for the rear

41
where Q is the total heat input from the welding arc. The two equations must be
balanced by the quantity R so that Rf + Rb =2
The project aim is to model Tandem Pulsed Gas Metal Arc Welding, which had
some big implications on the choice of heat source model. I t has previously been
shown that there is not a great deal of published literature available with regards
to modelling the tandem arc process. Even though there are two arcs present in the
process, the choice was made to create a model with only one Goldak type heat
source. The reasoning was that the main input variables for such a model are
based on weld pool dimensions, and the tandem process has one weld pool even
though two electrodes feed the same weld pool.
The pulsed nature of the process was not seen to influence the heat source model in
terms of the concept used, but merely affect the heat input in the model and
therefore possibly the weld pool dimensions.



Figure 14: Heat source dimensions

Narrow groove welding is an automated process. Part of the automation of the
process is utilisation of torch oscillation. The oscillation of the torch allows the weld
to be completed in less pass,s as more weld metal can be deposited for each pass.
A transient model could be setup in COMSOL whereby the heat source would
42
follow a designated path. This path could be a straight line, or it could be a sine
wave along the length of the groove geometry. This type of analysis would certainly
be interesting to run, but due to the complex nature would take a long time to
produce a solution. The chosen steady-state model could not produce a heat source
that moves in such a way. To simplify the model, therefore the approach used was
to adopt a modified heat source model that takes into account the torch oscillation.
The model proposed by Sabapathy (25), documented in the literature review, was
investigated. The model is a modification of the Goldak heat source model, with the
power of the width component varied to spread the heat source in the direction of
oscillation. The general form of the model:
( )
1 2 3
, , exp 3 3 3
n n n
f
x y z
q x y z Q
a b c
1
_ _ _
1

1
, , ,
]
(15) [25]
Which was modified to:
( )
2
10 2
, , exp 3 3 3
f
f
EIr x y z
q x y z
R c a b

1
_
_ _
1



1
, ,
,
]
(16)
for the front of the heat source model and
( )
2 10 2
, , exp 3 3 3
b
b
x y z EIr
q x y z
R c a b

1
_ _ _
1

1
, , ,
]
(17)
for the rear of the model.

The value of 10 for the power of y dimension was chosen as it produced a
satisfactory shape that approximated an ellipse with a degree of oscillation.
Figures 15-16 show the comparison of heat flux distribution between the original
Goldak heat source model and the new proposed model.
43

Figure 15: Distributed heat flux of Goldak heat source


Figure 16: Distributed heat flux of modified heat source

Figure 17 shows how the shape of the model is affected as the y dimension power is
changed from its original value of n=2 to n=10. The width and height of the model
44
is also increased from 4 to 6 in the diagram to further highlight the difference
between the two models. I t can be seen that the proposed model could be used to
characterise a heat source that is being oscillated in the y axis direction.

Figure 17: Heat source shape comparison

Equations 16-17 contain the term
EIr
R

(18)
where is the thermal efficiency of the process, E is the welding voltage, I is the
welding current, r is the balance variable between front and back sources and R is
the reduction factor.
The total heat input due to welding is equal to EI , whereby the power input is
reduced due to thermal inefficiencies of the welding arc. These inefficiencies
manifest themselves in heat losses to the surroundings during welding. The factor
of R must be determined depending on the size of the heat source model and
effectively modifies the expression in the model so that the heat input is no more
than the available power input from the welding power supply. I n order to set this
correction factor the integral of both expressions 16 & 17 must be taken over the
volume of the heat source that is contained within the plate geometry. The value of
R must be changed until the integral of the expression, which represents the total
heat input, matches the value of EI .
45
The placement of the heat source is on the zero line of the model. This puts the
heat source at the centre of the groove geometry. Figure 18 shows the groove
geometry with the heat source applied to the edge of the groove face.


Figure 18: Heat source located in the groove

The COMSOL script file defines the size of the heat source using the following
variables:
gold_a = grve_botdim+0.001; - Heat source width
gold_b = - Heat source depth
gold_cf = - Heat source front length
gold_cb = - Heat source rear length
gold_bt = - Heat source insulation height

The width of the heat source was defined as the width of the groove +1mm. This
means that the heat source protrudes into the side of the groove, thought to better
characterise the effect of arc oscillation during welding.
46

Figure 19: Heat source applied to top of deposited weld

Figure 19 shows that for each weld pass the heat source model was defined as
having maximum heat flux at the upper edge of the groove geometry. This
corresponds to top of the weld metal deposited during the pass being modelled and
may be unrealistic in terms of actual heat input and metal deposition during
welding. The heat source model also has heat flux above the reference in addition
to below the reference plane (the top surface of the aforementioned weld). This
allows a heat source boundary to be produced as it is shown in the diagram. I t is
thought this is quite satisfactory for characterising the actual flow during narrow
gap welding.
Model Subdomain & Boundary Conditions

The subdomain settings are those that describe the properties of the created
geometry. The heat equation that describes the subdomain used is:
( ) . .
p
k T Q C u T (19) [27]
where k is thermal conductivity, is density, Cp is specific heat capacity, Q is heat
input and u is speed.
These properties are mainly material properties, which will be discussed in
greater detail in a separate section. The Q term is where the heat input can be
expressed and is in the form of an expression of volumetric heat flux distributed
47
through the volume of x, y and z. The exact equation used for the heat source is
discussed in a previous section.
The travel speed, u, was defined depending on the welding test or pass number in
the mutlipass weld. The initial temperature for the subdomain was set to ambient
temperature of 293 degrees Kelvin.

Boundary conditions are a compromise between an accurate definition of the
problem and many simplifications to create a workable model. The weld pool
dynamics, for example, are a very complex set of interactions. Modelling such
behaviour for the purposes of thermal profiles would be extremely difficult. I t is
obviously extremely important, however, to use correct boundary conditions where
possible.
Given that the model is essentially a homogenous solid the boundary conditions are
external boundary conditions only. The model uses 5 different boundary conditions
applied to the 11 boundary faces of the geometry.

Prescribed temperature was set for the front face of the model, shown in figure 20.
This boundary condition sets the boundary to a known temperature, which in this
case is the ambient temperature of the sample 293K.

Insulation or symmetry specifies where the domain is well insulated, or can reduce
model size due to symmetry.
( )
. 0
p
n k T C uT +
(20) [27]
where n is the normal vector to the direction of heat flow.
The above equation shows that there is no heat transfer across the specified
boundary. This condition was set for the surfaces shown in figure 23. The surface
at the origin of the axes is only there due to a reduction in size of the model due to
symmetry. For that reason symmetry or insulation boundary condition was
applied. The far surface in figure 23, furthest from the origin, has the insulation
boundary condition also. Strictly speaking this may not be correct, but was set due
to the large size of sample to be modelled. As the welding direction is parallel to
this face, and the heat flux does not reach that face, the effect of this boundary
condition is highly limited.
48

Figure 20: Prescribed temperature boundary condition Figure 21: Top surface heat loss boundary condition

Figure 22: Bottom surface heat loss boundary condition Figure 23: Thermal insulation & symmetry boundary condition
49


Figure 24: Convective flux boundary condition

Thermal insulation was also set for the two faces covering the location of the heat
source. This was done so that no heat losses could exist at the position of the heat
source. I n reality the intense nature of the welding arc would prevent heat losses
from this area.
Heat flux with surface to ambient radiation is a boundary condition that defines
heat losses from the boundary, for both convective heat losses and radiation heat
losses.
The equation describing the heat across the boundary is:
( ) ( ) ( )
4 4
0 inf
.
p amb
n k T C uT q h T T T T + + + (21) [27]
where h is the heat transfer coefficient, Tinf is the external temperature, is surface
emissivity, is Stefan-Boltmann constant and Tamb is the ambient radiation
temperature. The heat transfer coefficient and emissivity were varied, as will be
later discussed, with the Tamb temperature set to 293K. This boundary condition
was set for the bottom and top surface of the plate, as well as the groove faces of
the model, shown in figures 21-22. I n reality these are the faces that will lose heat
to the surrounding atmosphere.
Convective flux assumes passing energy through the boundary must pass by the
convection mode and so assumes heat conduction does not occur. The equation for
the boundary is:
( )
. .
p
q n C uT n (22) [27]
50
I t was applied to the outer face shown in figure 24 as the outlet temperature was
unknown.

Mesh Size

The mesh size of the model has a direct influence on the refinement of the solution
as it affects the number of nodes that must be solved for any given model. Figure
25 shows the typical mesh that was applied to the groove geometry. The heat was
applied at the location of the volumetric heat source. For this reason the mesh was
limited to a minimum size of 0.0015 m or 1.5 mm in size in this region. The top
surface was limited to a minimum size of 0.01m or 10mm per cell. These settings
were the best that could be achieved using the linear solvers. Cell sizes smaller
than this resulted in mesh matrix sizes that could not be solved given the
hardware configuration.

Figure 25: Example of meshed geometry


51
Material Properties

Choice of material properties obviously has a major effect on the accuracy of any
finite element model. As detailed in the previous section the material properties
required to model a heat conduction problem are:
Thermal conductivity
Specific heat capacity
Density
Heat transfer coefficient
Emissivity
The main advantage of the Finite Element method is that these quantities can be
defined as temperature dependent, and therefore reflect a more realistic
characterisation of the material being modelled, rather than using constant values.
Unfortunately, the main aim of this project was to model a rare material type of
X100 steel. This material type posed a large problem in that there is no published
information available detailing the required material properties. I n fact detailed
temperature dependent material data for thermal conductivity and specific heat
are not available for all but the most common structural materials. The data
chosen for the model was therefore temperature dependent data for any standard
low carbon steel. The data for thermal conductivity, specific heat capacity, density
and emissivity was originally taken from Frewin [31], although it was adapted for
the final model. The suitability of the chosen properties will surely be judged by
how well the final model temperature profiles match the actual welding
temperature profiles.

During modelling it was necessary to understand the influence of individual
material properties on the thermal profile produced from the model. To achieve this
a systematic variation of material properties was devised, to allow separate
evaluation of each material property. I f a material property does not affect the
model results to any great extent it would be prudent to not waste time further
refining that particular property value in the model. A further advantage of
studying the effect of material property variation is that it allows a more educated
choice of parameters that can be changed, should the model not accurately match
the experimental thermal profiles acquired during welding.
52
For the purpose of material property evaluation a flat plate model was created,
with thickness of 23mm, matching that of the actual plate. Figure 26 shows that
the mesh was refined near the heat source and was also limited in size around this
area. This was found to be necessary to produce smooth temperature profiles, as
some constant material property values made it difficult to find convergent finite
element solutions.

Figure 26: Refined mesh around heat source

The model was solved using the material properties under consideration with a
constant set of input parameters for heat input and heat source dimensions for
each test. A set of temperature profiles were then taken from the model at locations
of 0, 4, 8, 11.51, 21.32, and 31.8 mm from the weld source centreline, or the bottom
edge of the model shown in figure 26. These chosen values would then show how
the model temperature profiles vary at the heat source, close to the heat source and
at some distance from the heat source.

The effect of varied thermal conductivity was evaluated by using 5 models. Figure
27 shows the two temperature dependent models considered. Model 1 was directly
taken from Frewin [31]. Model 2 includes a high thermal conductivity once the
53
temperature is above the materials melting point. This approach has been
employed by other modellers [10] to take account of the heat transfer by convection
due to fluidity of the weld pool. The other values of thermal conductivity were
constant values of 26, 52 and 120. The values were chosen as they represented
extreme and average values of the temperature dependent models.

Figure 27: Temperature dependent thermal conductivity models


Figures 28-33 show the temperature profiles taken from the model for varied
values of thermal conductivity. Figure 28 shows that near at the heat source a
value of 26 W/mK produces the highest peak temperature, with 120 W/mK
producing the lowest peak temperature. Model 2 gives a lower peak value due to
the high temperature at the heat source and the artificially raised value of 120
W/mK for this model. At 8mm from the source the lowest thermal conductivity still
produces the highest peak temperature and it is clear to see that the effect of high
thermal conductivity is to increase the rate of cooling of the material. At 31mm
from the heat source the peak temperature is now produced by the thermal
conductivity model of 120 W/mK and it shows that the high thermal conductivity
increases the rate of heating of the material. Clearly high thermal conductivity
increases the heat flow rate through the material and this is as expected. Thermal
conductivity will have the highest influence on the model as it is dominated by the
conduction mode of heat transfer.
54

Figure 28: Thermal conductivity variation 0mm




Figure 29: Thermal conductivity variation 4mm
55

Figure 30: Thermal conductivity variation 8mm




Figure 31: Thermal conductivity variation 11.51mm
56

Figure 32: Thermal conductivity variation 21.32mm




Figure 33: Thermal conductivity variation 31.8mm

Specific heat capacity was studied using 5 datasets. Temperature dependent
models represented with 3 models, shown in figure 34, with two constant values of
450 and 1400 J /kgK. The temperature dependent models take into account the
57
latent heat of transformation that occurs when the material transforms from
ferrite to austenite at 727C and latent heat of melting at the melting temperature.
Model 1 data was taken from source [31] whereby discrete values are used to model
the specific heat. Problems can occur due to the discontinuous jump in specific heat
values using this method. Often times the finite element solution will not converge
properly. The solution is to model the specific heat on the enthalpy change during
these transformation reactions. This allows a continuous function to be developed
and put into the model which is then used instead of the discontinuous jumps.
To model specific heat and take account of latent heat the equation becomes:
p
C H + (23) [27]
where H is the latent heat of transition and is given by the relationship:
( )
( )
2
2
exp /
m
T T dT
dT


(24) [27]
where Tm is the melting point of the material and dT represents half of the width
over which the transition takes place.
For model 3 the value of half-width (dT) of the transformation was 100K at a mid
temperature of 1998 K, while the latent heat of fusion was 247 kJ /kg. Model 2
used an artificially low value of latent heat of 8kJ /kg to study the effect of the
variable. Both models modelled the ferrite to austenite transformation with dT as
30K at a mid temperature of 998K, with a latent heat of transformation of
24kJ /kg.
58

Figure 34: Specific heat capacity models



The figures 35-40 show the effect of varied specific heat on the thermal profiles at
varied distances from the heat source on the top of the modelled plate. At the heat
source all temperature dependent models give the same temperature profile. Low
specific heat values give a higher peak temperature, while a high specific heat
value gives a lower peak temperature.
At 8mm from the heat source model 1 and 2 produce the lowest peak temperature,
while model 3, with the correct latent heat value, produces a higher peak
temperature. This peak is close to the value obtained from the constant specific
heat of 450 J /kgK. At 31mm from the heat source the trend persists with models 1
and 2 giving low peak temperatures, while the constant low value of 450 J /kgK is
close to the temperature profile of the enthalpy modelled specific heat. This shows
that possibly a single constant value of specific heat may suffice for simplicity.
59

Figure 35: Specific Heat variation 0mm





Figure 36: Specific Heat variation 4mm
60

Figure 37: Specific Heat variation 8mm





Figure 38: Specific Heat variation 11.51mm
61

Figure 39: Specific Heat variation 21.32mm




Figure 40: Specific Heat variation 31.8mm

Density was studied with 3 datasets, a low constant value of 5900 kg/m
3
, a high
constant value of 7800 kg/m
3
and a temperature dependent model shown in figure
41. I t can be seen that the density of the material falls as temperature increases.
62
The figures 42-47 show the effect of varied density on the thermal profile at varied
distance from the heat source. At all distances from the heat source the
temperature dependent model thermal profile is clearly matched by that of the
constant density value when 7800 kg/m
3
. This shows that a constant value of 7800
kg/m
3
could be used instead of the temperature dependent model for simplicity and
quicker solution time. The constant low value of 5900 kg/m
3
produces a higher peak
temperature for all of the thermal profiles at varied distances from the heat source.

Figure 41: Density and Emissivity models

Emissivity was investigated using 3 datasets, one low value of 0.2, one high value
of 0.6 and the temperature dependent model shown in figure 41. The results of
figures 48-53 show that there is no difference between the models in any of the
thermal profiles at varied distance from the heat source. Therefore the choice of
emissivity is large irrelevant.
63

Figure 42: Density variation 0mm






Figure 43: Density variation 4mm
64

Figure 44: Density variation 8mm






Figure 45: Density variation 11.51mm
65

Figure 46: Density variation 21.32mm






Figure 47: Density variation 31.8mm
66

Figure 48: Emissivity variation 0mm






Figure 49: Emissivity variation 4mm
67

Figure 50: Emissivity variation 8mm






Figure 51: Emissivity variation 11.51mm

68

Figure 52: Emissivity variation 21.32mm





Figure 53: Emissivity variation 31.8mm

Heat transfer coefficient of the material affects the heat loss of the welded plate
surfaces due to convection and losses. The heat loss of the model was studied using
a varied set of constant heat transfer coefficient values. The chosen constants were
69
5, 50 and 200. Figures 54-59 show the effect on the model thermal profiles at
varied distance from the heat source. Near the heat source very little effect of
varied heat transfer coefficient is visible. Peak temperatures are all the same. At
increased distance from the heat source the heat transfer coefficient has more
effect, showing that a high coefficient does lower the thermal profile. Refinement of
this material property will allow the cooling rate of the model to closely match the
experimental results.






Figure 54: Heat-transfer coefficient variation 0mm


70

Figure 55: Heat-transfer coefficient variation 4mm






Figure 56: Heat-transfer coefficient variation 8mm
71

Figure 57: Heat-transfer coefficient variation 11.51mm






Figure 58: Heat-transfer coefficient variation 21.32mm

72

Figure 59: Heat-transfer coefficient variation 31.8mm




The thermal efficiency of a process is the effectiveness of transforming electrical
power into heat input for welding. A thermal efficiency of 1 represents a process
that has no heat losses to the atmosphere and therefore converts all electrical
power to heat input. A heat input this high is not would be very difficult to achieve
given that heat losses will always occur to some extent.
The thermal efficiency effect on modelling was studied with 3 values of 1, 0.8 and
0.6. The results are shown in figures 60-65. I t can clearly be seen that the thermal
efficiency effectively just reduces the heat input into the model. I n turn the lower
heat input reduces peak temperature and overall thermal profile temperatures for
all distances from heat source.
73

Figure 60: Thermal efficiency variation 0mm






Figure 61: Thermal efficiency variation 4mm
74

Figure 62: Thermal efficiency variation 8mm






Figure 63: Thermal efficiency variation 11.51mm
75

Figure 64: Thermal efficiency variation 21.32mm




Figure 65: Thermal efficiency variation 31.8mm


Travel speed, like thermal efficiency, effectively varies the heat input per unit time
in to the plate. The effect of varied travel speed is shown in figures 66-71, with
values of 0.005, 0.01174 and 0.02 m/s. As expected the slower the travel speed the
76
more heat input into the plate and therefore the higher the peak temperatures.
This relationship is consistent with all distances from the heat source.

Figure 66: Travel speed variation 0mm




Figure 67: Travel speed variation 4mm
77

Figure 68: Travel speed variation 8mm






Figure 69: Travel speed variation 11.51mm
78

Figure 70: Travel speed variation 21.32mm






Figure 71: Travel speed variation 31.8mm

79
4. Model Validation through Experiment
Introduction

The previous section of this report has documented how the finite element model
was developed to simulate the thermal cycles present during welding of a narrow
groove pipe geometry. The material properties have been studied and suitable
parameters selected to characterise the X100 material as closely as possible given
that no material data is available. The other model input parameters, such as heat
input, thermal efficiency of the Tandem GMAW-P process and size of heat source
dimensions are still necessary in order to complete the model with sensible input
values. For this reason a number experiments had to be devised to measure the
aforementioned quantities.
Performing an actual pipeweld with the correct configuration of welding equipment
and joint geometry was seen as the best way to acquire temperature profiles. These
are used to validate the developed model using a number of sensibly placed
thermocouples. Test welds provide the input power used as input data to asses the
model validity. Heat source model dimensions matched to actual weld pool
dimensions for the Tandem GMAW-P process give greater accuracy to the model.

As the investigation of modelling variables has shown it is extremely important to
quantify the thermal efficiency of the process being modelled. The thermal
efficiency of the process defines the amount of actual heat input into the plate
being welded. Very often thermal efficiency for a welding process is taken as a
known value, when modelling welding phenomena. I n effect the efficiency can be
used to modify the model and refine results to match more closely the experimental
results. For this project it was decided that the variable efficiency approach would
not be adopted and the thermal efficiency for the Tandem P-MI G process would be
evaluated experimentally for later input into the model. This would hopefully have
the effect of producing a model with less uncertainty in input factors. The aim of
the thermal efficiency tests was to evaluate the thermal efficiency of the Tandem
GMAW-P process by placing a welded sample into a calorimeter filled with liquid
nitrogen. The heat input during welding a metal sample was shown to be
80
equivalent to a loss of liquid nitrogen evapourated from the calorimeter and lost to
the atmosphere. Further details are given in the Experimental Procedure.

The next section details the materials used in addition to the procedure carried out
for the experimental tests.

Materials

The project aim was to create a model for X100 pipeline steel. Therefore all welding
experiments used this material.
Groove sections were made from 23mm thick plates, that were then placed on a
backing bar with a 5mm gap at the base, throughout the length (figure 73). The
plates were then tack welded to the backing bar and left to cool in a clamping jig to
prevent distortion.
Bead on plate trials used the same X100, 23mm thick plates, without prior
welding.
The efficiency measurement tests were done with X100 material for the welding
trials, however, for calibration of the test procedure, a number of different sized
2mm thick mild steel samples were also used, shown in figures 83-85.
Welding Equipment

For all of the welding experiments done during the project two Fronius TPS
(TransPulse Synergic) 4000 Thermo power supplies were used. These units offer
the ability to perform Tandem Pulsed Gas Metal Arc Welding using programmed
synergic curves, which optimise pulse settings for a given wire feed speed. The unit
is fully digitised and can be used for Metal I nert Gas Welding, Metal Active Gas
Welding, Tungsten I nert Gas Welding and Shielded Metal Arc Welding. Transfer
modes are short circuit, spray and pulsed. The units main features are
Welding Current Range: 3 400Amps
Maximum OCV: 70Vdc
Pulse Voltage Range: 14 34Vdc
I nput Voltage: 400V
81
I nput Frequency: 50/60 Hz
The two independent power supplies can then be setup with the correct pulse
waveform profile and finally can be synchronised so that lead and trail arcs do not
interfere.

The synergic curve used was previously developed for tandem groove welding for
pipeline steel. This meant that, by using these curves, the entire project started
from a point that provided stable welding parameters. More information on the
optimisation of the welding equipment can be found in [5].

The Fronius Tandem torch was used for all welding. This torch has two sets of
contact tips, at a spacing of 5mm between the two. This torch enables two
electrodes to feed one weld pool for increased deposition. Figure 72 shows the torch
used.

During welding the torch remained stationary as it was mounted to the welding
rig. The material was moved underneath the torch at a set rate to achieve the
desired weld quality. An RMS type control pendant controlled the vertical and
horizontal positioning of the torch, allowing flexibility of movement into the
grooved joint. Torch oscillation could be set with this unit, prior to a welding run,
by specifying the speed of oscillation in beats per minute (BPM) and oscillation
width. The control unit was also used to trigger the welding cycle, thereby starting
and stopping the welding arc.

Lincoln Electric LA-100 filler wire was used for each welding test. The wire was 1.0
mm in diameter and had the nominal composition of: <5% wt Manganese, <5%wt
Nickel, 0.5% wt Molybdenum, <0.5% wt copper including plated coating, with the
balance of I ron.

One single shielding gas mixture was used during the project. The Trimix 82.5%Ar
12%CO2 5%He was selected as much work has been done with using this gas. The
arc characteristics have been studied by LARATZSI S [5] and HUDSON [4] and
quality of resulting welds are fully documented. Given more time the current study
82
could have investigated the influence of modelling different shielding gas
composition on the weld thermal cycle.


Figure 72: Twin contact tip tandem torch





Figure 73: Groove plate sample secured onto the welding rig

83

Figure 74: Welding rig with table motion control box



Recording Equipment

During setup of the welding equipment the arc length was measured to ensure
correct welding parameters. A normal speed camera and video cassette recorder
were used to record the arc length, using a filter, to reduce glare from the welding
arc. The Panasonic CCD Model F15 video camera was used. The Panasonic, NV-
FS88HQ video cassette recorder was used to record any passes for later playback.
This setup allows paused viewing of each frame and can be used for assessing the
arc characteristics throughout the entire welding pass. The video camera was
situated to record the view down the length of the grooved plate, allowing a view of
the complete pass.

I n order to correctly characterise the welding power supplies in the model it was
very important to capture the voltage and current levels during welding
experiments.
84
The current and voltage levels of each power supply were measured for each
experiment by the Yokogawa DL750 Scopecorder digital oscilloscope. This recorded
real-time both voltage and current of the two power supplies into data channels
with a high sampling rate to allow later investigation of each waveform. The setup
also allowed an accurate weld time to be measured for the welding efficiency test.

To record the temperature profiles two thermocouple types were used. For the
thermocouples placed near the weld pool the R type thermocouple was used. This
selection was made because of the high temperature in this region and the fact that
R type thermocouples can withstand temperature above 1500C. This thermocouple
type can also be placed directly in the weld pool and will not degrade. The K type
thermocouple was used for more remote temperature measurements at least 10mm
from the weld pool, where peak temperatures are expected to be much less than the
rated maximum of 1370C for this thermocouple type [5]. Alumina insulation rods
of 3mm were used to separate the two wire limbs. They also provided a more
definite contact point when in contact with the molten weld metal. Temperature
compensation cables, specific to each thermocouple type, were connected between
the thermocouples and the data acquisition hardware.
The National I nstruments thermocouple module provided an interface between the
thermocouples and the data acquisition software. For the tests only 8 channels
were available for data capture.
LabView software was used for thermocouple data capture.
Thermocouples were welded into place using a capacitor discharge machine
connected to create a circuit between the plate and the two terminals of the
thermocouple. Current was sufficient to allow melting of the thermocouple tip.

Metallographic Examination

The STARTRI TE V500H vertical band-saw was used to section the welded
samples. Grinding was done using the Sealley grinder, model SM14C, to greatly
save time in removing saw marks. The samples were subsequently polished using a
combination of P120, P240, P1200 silicon carbide grinding papers. Samples were
etched using 2% natal etchant.
85
Calorimeter Measurements

For calorimeter tests Liquid Nitrogen was used in combination with a stainless
steel Statebourne Cryogenics dewar of 5.25 litre capacity. Dewar inside diameter
was 150mm and the internal height was 297mm. Weight of the dewar full was 9kg.
The calorimeter was specifically designed for cryogenic purposes and meant that
with liquid nitrogen inside it was optimised to ensure that as little heat conduction
could be achieved through the container walls.
Measurements of combined weight of calorimeter, liquid nitrogen and sample were
made using a high accuracy scale. The scale accuracy was to within 1 gram.
A webcam was positioned to record the scales LCD display. This was connected to
a laptop computer and could record data for every calorimeter test performed.
A fluorescent light was placed above the camera to add illumination to the LCD
display, as no backlight was available.
An oven was used to raise the temperature of metal samples for test calibration.


Figure 75: Calorimeter experimental setup
86
Experimental Procedure

Welding Tests

The aim of the welding tests were as follows:
To find suitable welding parameters using the Fronius Tandem setup that
could be used to create a defect free multipass pipe weld of 23mm, X100 pipe
material.
Capture the voltage and current values of the welding power supplies
during welding to enable correct power input values for modelling.
Measure as accurately as possible the temperature profiles from
thermocouples placed near to the weld pool for each weld, in addition to
more distant thermocouples place either on the surface or through the
midsection of the sample.
Measure the weld pool width, length and depth to incorporate the data into
the heat source model.

Set up of Welding Rig

The moving table of the welding test rig had previously been calibrated for travel
speed conversion between the arbitrary dial number used and a value in m/min.
The calibration is shown in figure 76.
87
Travel speed calibration
y = 0.0011x + 0.2096
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1.10
1.20
100 150 200 250 300 350 400 450 500 550 600 650 700 750 800
Dial No
T
r
a
v
e
l

s
p
e
e
d

(
m
/
m
i
n
)

Figure 76: Welding table travel speed calibration

Prior to welding the following procedure was completed:
Alignment of the plate on the rig especially important for groove samples
where incorrectly positioned samples could damage the contact tip if there
was a collision. I f the contact tip struck the side wall of the groove this
would also produce a poor quality weld.
Ensuring the oscilloscope signal detector is switched on the oscilloscope
used an automatic trigger for data acquisition once welding had
commenced.
Check welding parameters, travel speed and oscillation width are correct.
Purge the gas
Check safety equipment such as welding visors.

Welding Test 1: Flat Plate Test 3 Thermocouples

Test 1 was designed to measure the temperature profiles on the top surface of a
plate at a distance from the weld pool.
A weld bead was put on the centre of a plate of dimensions 200mm x 150mm
and 23mm thick.
88
K type thermocouples were placed at 11.51mm, 21.32mm and 31.8mm from
the weld centreline.
Contact between the plate and the welding rig was minimised to reduce
conductive losses.
Data acquisition frequency was 10 Hz.
Wire feed speed was set to 11m/min, with synergic pulse parameters pre-
set.
Travel speed was set to 0.011743 m/s
Gas flow rate of Trimix was 26 l/min


Figure 77: Test 1 thermocouple locations

89

Figure 78: Test 1 thermocouple distances from the weld line

Welding Test 2: Flat Plate Test 4 Thermocouples

Welding test 2 was designed to measure the temperature profiles at varied depths
of a welded plate, without removal of plate material by hole drilling.
A weld bead was put on a plate of dimensions 200mm x 150mm and 23mm
thick, 20.41mm from one edge
K type thermocouples were placed:
o 10.71mm from the edge on the top surface
o 7.5mm from the top edge on the side surface
o 15.07mm from the top edge on the side surface.
o 22.23mm from the bottom edge on the bottom surface.
Data acquisition frequency was 10 Hz.
90
Wire feed speed was set to 11m/min, with synergic pulse parameters pre-
set.
Travel speed was set to 0.011743 m/s
Gas flow rate of Trimix was 26 l/min


Figure 79: Test 2 thermocouple locations

Welding Test 3: Bead on Plate Welds for Weld Wool
Measurement
Test 3 was designed to allow measurement of the molten weld pool during welding.
A camera was setup to record this for a number of travel speeds. The intense light
of the welding arc proved to obscure the view of the molten weld pool even through
a number of different filters.
Alternatively the final shape of the weld pool was measured from the cooled weld.
A number of weld beads were made for 3 travel speeds: 0.01321m/s, 0.01174 m/s
and 0.01028 m/s.
91

Welding Test 4: Multipass Groove Weld with Multiple
Thermocouples

The objective of Test 4 was to produce a complete groove weld of a 5 degree bevel
angle configuration shown in figure 82. For each pass the temperature profiles for
distant locations from the weld were measured, in addition to a thermocouple
placed near the weld bead. As the weld was built up the previous thermocouple
became further from the weld pool, giving thermocouples located at increasing
distances from the weld. This offered a good solution to providing temperature
profiles for comparison with model data.
Two methods of thermocouple placement were considered to measure the
temperature at or near the weld pool. Figure 80 shows a groove configuration with
two weld passes highlighted. Placement method 1 is designed to make a hole from
the bottom surface of the plate, and place the thermocouple contact point near the
bottom of the subsequent weld. Placement method 2 is designed to place the
thermocouple at the same height as the method 1, but the hole is made from the
top surface at an angle to the surface. The advantage of method 2 is that the plate
does not need to be removed from the welding rig. Movement of the plate during
welding greatly increases the amount of time to complete the weld, as more time is
needed to setup the plate between passes. The advantage of method 1 is that the
location of the thermocouples is more definite as the hole is made vertically and not
at an angle.
92

Figure 80: Bottom and top drill hole positions for thermocouples

Figure 81 shows that 9 thermocouples were used to measure the temperature
profiles during the multipass welding of the groove.

Figure 81: Test 4 multipass weld thermocouples locations
93
Thermocouples 1, 2 and 3 were K type thermocouples, at distances of 12.62mm,
23.35mm, 33.08mm from the weld groove centreline. The depth of these distant
thermocouples was approximately 11mm from the top surface.
Thermocouples 4 to 9 were R type thermocouples due to the higher peak
temperatures of the weld pool. During welding the thermocouples were added in
the following sequence:
Pass 1: Thermocouple 4 from the back of the plate, prior to welding
Pass 2: Thermocouple 5 from the top
Pass 3: Thermocouple 6 from the top
Pass 4: Thermocouple 7 from the top
Pass 5: Thermocouple 8 from the top
Pass 6: No additional thermocouples
Cap pass: Thermocouple 9 placed into the rear of the molten weld pool
The last thermocouple allowed measurement of the weld pool temperature should
the other thermocouples not be placed within the molten weld pool area.
A weld was made in 7 passes in a groove sample of dimensions as shown
below
Data acquisition frequency was 10 Hz.
To maintain a high deposition of weld metal the travel speed was varied to
give roughly 3 mm of deposition per pass.
Torch oscillation was set to within 1mm of the side walls for each pass.
Contact tip to work distance was set to 13.5mm.
Gas flow rate of Trimix was 26 l/min.

Figure 82: Groove geometry dimensions
5.00
9.00
22.8 All measurements in mm
94

Thermal Efficiency Measurements

The thermal efficiency tests comprised of an initial set of calibrations that
effectively proved the experimental procedure. The first test was to measure the
amount of liquid nitrogen lost to the atmosphere without any other substance
added. This was known as the normal evapouration rate. The procedure was:
Calorimeter was roughly 4/5 full (to a combined weight of roughly 8kg)
The total combined weight of the stainless steel dewar plus liquid nitrogen
was measured at 15 second intervals
A graph was plotted showing weight loss of liquid nitrogen to the
atmosphere.

The second group of tests was to measure the weight loss of liquid nitrogen after
samples of different weights, sizes and stabilised temperature were added to the
liquid nitrogen dewar. This was then compared to a theoretical value of energy
input for each sample. The process was as follows:
Sample weight was measured and recorded
Sample was placed either at room temperature or into an oven set to a
known temperature (measured with a thermocouple). Temperatures
measured were 21, 185, 400, 600 and 900C
Calorimeter was filled to roughly 4/5 full (to a combined weight of roughly
8kg)
The weight of liquid nitrogen plus dewar was recorded before sample was
added
Sample was dropped into the liquid nitrogen and combined weight of the
dewar, liquid nitrogen and sample were measured at 15 second intervals for
roughly 10 minutes
A graph was plotted of combined weight vs. elapsed time
Values were recorded of initial liquid nitrogen weight, peak weight and final
weight before nitrogen stabilisation (highlighted in figure 108 of the results
section).
Experimental energy was calculated by:
95
o
exp
.
v
E W L (25) [21]
where W is weight loss of liquid nitrogen during the test and Lv is
the latent heat of vapourisation of liquid nitrogen defined as 199 J /g
Theoretical energy was calculated by:
o
theo
E mc T (26) [21]
where m is the mass of the steel sample, c is the specific heat at an
average temperature (between liquid nitrogen temperature of -
196C and sample temperature) and T the temperature difference
(between liquid nitrogen temperature of -196C and sample
temperature)
These two values were then compared
The process was repeated for different sample weights and ambient
temperatures.
A number of different sample weights and sizes were tested and compared. Firstly
mild steel samples of 100 x 100 x 2.4mm were tested, each weighing roughly 180g.
Smaller samples with the same material were also tested weighing 27g. For more
relevant tests X100 material was tested. A sample of 100 x 30 x 23mm, was tested
in addition to two of these samples welded to a backing bar to create a grooved
geometry. This sample weight was roughly 1000g. Figure 83 shows this sample
after it has been removed from the liquid nitrogen.

Figure 83: Welded 1000g sample
96
I n order to allow better clamping during welding efficiency measurements the X100
samples had mild steel brackets welded to them. This increased the weight of the
samples to 1100g for the single X100 sample and 1600g for the welded groove
sample.
Figures 84-85 show the samples with added metal for better clamping during
welding.

Figure 84: Sample with added clamping feature



Figure 85: Groove sample with added clamping feature

97
To measure thermal efficiency of the chosen welding process the samples shown in
figures 84-85 were welded and put into the liquid nitrogen calorimeter. The process
was as follows:
Sample weight was measured and recorded
Sample was placed on the welding table and secured with mole grips
Welding speed was set to 0.01174 m/s with wire feed speed of 11 m/min,
contact tip to work distance was set to 13.5mm.
Calorimeter was filled to roughly 4/5 full (to a combined weight of roughly
8kg)
The weight of liquid nitrogen plus dewar was recorded before sample was
added
Oscilloscope set to trigger and acquire welding current and voltage for both
lead and trail power supplies for the duration of the test
Sample was welded for a set length of sample
Sample was unclamped and removed from the welding table as quickly as
possible and dropped into the liquid nitrogen and combined weight of the
dewar, liquid nitrogen and sample were measured at 15 second intervals for
roughly 10 minutes
A graph was plotted of combined weight vs. elapsed time
Values were recorded of initial liquid nitrogen weight, peak weight and final
weight before nitrogen stabilisation (to be highlighted in the results
section).
Welding experimental energy was calculated by:
o
exp
.
v
E W L
Once the sample had reached room temperature the sample was reweighed
and put back into the liquid nitrogen
Combined weight of the dewar, liquid nitrogen and sample were measured
at 15 second intervals for roughly 10 minutes
A graph was plotted of combined weight vs. elapsed time
Values were recorded of initial liquid nitrogen weight, peak weight and final
weight before nitrogen stabilisation (to be highlighted in the results
section).
Room temperature experimental energy was calculated by:
98
.
room v
E W L

Figure 86: Welding energy composition

Figure 86 shows how the energy input, due to welding, is the welding calorimeter
test energy in the first test minus the room temperature energy from the second
test.

The thermal efficiency of the process is given as:
.
weld
weld
v
inst
Qt Q
EIt EI
W L
P

(27) [21]
where, Q is heat input, E is welding voltage, I welding current, Pinst is the
instantaneous power input. Therefore the thermal efficiency is the welding energy
from the liquid nitrogen test divided by the power input of the welding supply.
The instantaneous power is given as:
1
.
n
i i
inst
i
I E
P
n


(28) [21]
As the literature review has shown this is the most accurate way to calculate power
input for pulsed welding power supplies.
99
To calculate Pinst for tandem process the recorded voltage and current was
multiplied for each instance and was averaged over the time of the welding test.
The power for each welding arc was summed to give a total power in Watts.
The efficiency tests were performed 3 times for bead on plate tests on top of the
1100g samples and 3 times inside the groove geometry for the 1600g samples.
Comparisons were made to the efficiency on the top of the plate and in the groove.
100
5. Experimental Results

Welding Tests 1&2: Bead on Plate Tests


Table 3: Welding parameters for bead on plate tests





Figure 87: Oscilloscope data for Test 1

101

Figure 88: Oscilloscope data for Test 2








Figure 89: Oscilloscope data for Test 3

102

Figure 90: Weld Test 1 thermocouple data






Figure 91: Weld Test 2 Thermocouple data




103
Weld Test 3: Weld Pool Measurement



Figure 92: Weld pool size dimensions, 0.617 m/min above, 0.793 m/min below






Figure 93: Bead on plate weld dimensions

104

Figure 94: Bead on plate with torch oscillation weld dimensions

Weld Test 4: Multipass Groove Weld




Table 4: Multipass groove weld input parameters and weld deposition








Table 5: Mutipass weld power input results

105

Figure 95: Pass 1 thermocouple






Figure 96: Pass 2 thermocouple data

106

Figure 97: Pass 3 thermocouple data






Figure 98: Pass 4 thermocouple data

107

Figure 99: Pass 5 thermocouple data






Figure 100: Pass 6 thermocouple data

108

Figure 101: Cap pass thermocouple data


Figure 102: Macro of multipass groove weld

109

Figure 103: Multipass weld location of thermocouple 4


Figure 104: Multipass weld location of thermocouple 5
110

Figure 105: Multipass weld location of thermocouples 1,2,3 & 6





Figure 106: Multipass weld location of thermocouple 8

111

Table 6: Multipass weld thermocouple location summary
Process Efficiency Tests



Table 7: Calorimeter test calibration summary




112

Table 8: Welding efficiency results




Figure 107: Liquid nitrogen normal evapouration rate
113

Figure 108: Calorimeter tests example weight loss profile







Figure 109: Calorimeter test data showing variation with temperature

114

Figure 110: Calorimeter test data showing variation with weight

115
6. Discussion of experimental results

The discussion in this section refers to the results given in the previous section. All
tables and figures can be found in the Experimental Results section of the report.
Weld Test 1

Test 1 ran a bead on plate weld on the top surface of an X100 steel plate, with
three thermocouples to measure the temperature profiles during welding. I n
addition an oscilloscope captured the current and voltage during welding. The
travel speed during the test was set to 0.704 m/min. Table 3 shows the
instantaneous power measurements for all bead on plate test welds. A power input
in the region of 8kW was measured for a wire-feed speed of 11 m/min at the travel
speed stated. Figures 87-89 show example plots of oscilloscope data gathered
during the plate welding tests. Peak current and voltage values are highlighted in
table 3. The current waveforms obtained are as expected given that a pulsed power
supply was used. The actual waveform characteristics, in terms of peak current
and time etc, were previously defined to produce high quality welds [5].
Figure 90 shows the thermal profiles obtained at roughly 10, 20 and 30mm from
the weld centreline. The plots have a common feature of an initial increase in
temperature as the arc passes the thermocouple. This is often accompanied by an
unstable period of temperature acquisition. The thermocouples were protected as
much as possible to prevent the welding arc destroying the thermocouples. The
results from weld test 1 show good thermal cycles that can be used for comparison
to those derived from the model, in addition to heat input values for the known
travel speed.

Weld Test 2

Weld test 2 was done to acquire thermocouple data at varied depths without the
need to drill holes into the plate material. I t was thought that drilled holes into the
plate would have some effect on the thermal conduction of the plate and therefore
this test would not have these problems. The fact that the weld was close the actual
116
edge of the plate may, however, cause inaccuracies for modelling. The
thermocouple data for test 2 is shown in figure 91. Once again the problem of
intense arc heat passing the thermocouple on the top face was evident. I n addition
the data acquisition equipment was quite temperamental and meant that noise
was intermittently present on some thermal cycle plots. The data gathered from
the test was, however, deemed sufficient to use for comparison purposes of
experimental and model cases.

Weld Pool Measurement

To characterise the heat source model it is recommended that the weld pool
dimensions are used, at least as a starting point [14]. I nitially to size the weld pool
a stationary camera was installed, with a filter, connected to a digital recording
device. The arc intensity prevented accurate measurement of the weld pool with
this method. A number of different filters were tried but the problem persisted
throughout.
An alternative method was used, whereby the final size of the weld pool was
measured, was done on the surface of a plate. Figure 92 shows that the weld pool
size is quite clear after welding. I t seems that the tandem welding technique makes
this possible by creating a very large weld pool and the same technique did not
produce the same quality results with a single arc welding supply.
Two travel speeds were used to evaluate weld pool dimensions. The lowest speed of
0.617 m/min gave a total pool length of 37mm and a width of 11mm. The fastest
speed of 0.793 m/min did not produce a longer weld pool as was expected. I nstead
the pool length was only 34mm. The width of the weld pool was, however, narrower
with a faster travel speed.
The figure shows there is a high degree of subjectivity when sizing the weld pool
visually and accuracy is not considered high. The test did, however, provide a ball-
park figure that could be used as a starting point for heat source modelling.

Figures 93-94 show the effect of oscillation on the fusion zone profile of bead on
plate welds with the same input settings as the previous tests. The oscillated
117
source has a much more uniform fusion zone that closely matches the ellipse shape
proposed by the model.

Weld Test 4: Multipass Groove Weld

Welding of the narrow groove joint geometry was completed in 6 passes with a final
cap pass. Wire-feed speed of 11 m/min was chosen due a high productivity rate and
has previously been shown to create good quality welds [5]. Table 4 details the
travel speed of each pass, which was chosen based on a deposition of roughly 3-
4mm of weld metal per pass. The table shows that the chosen travel speeds
achieved the intended deposition for each pass.
Torch oscillation was set to prevent lack of sidewall fusion, a common defect in
narrow groove pipe welds. Figure 102 shows the final weld cross-section macro,
providing evidence that there are no sidewall defects present. I n this respect the
welding test was seen as a success. The weld cap did not provide the required
coverage for and actual pipe weld, however for thermal cycle analysis it was
deemed acceptable.
Table 5 provides details of measured power input during welding for each pass.
This data is highly important for further modelling work and allows heat input
variables to be matched to travel speed for each weld pass. Power input ranged
from 8257 W, for a travel speed of 0.771 m/min, to 8480 W for a travel speed of 0.65
m/min during the cap pass. As expected the higher power input was seen with the
lower travel speed, although the highest heat input was not obtained with the
lowest welding speed. I t is thought that the groove geometry affects the amount of
power input due to welding arc dynamics. As the welding voltage is dependent on
arc length the interaction of the arc with sidewalls will complicate the measured
power input.
Thermocouple placement provided the greatest challenge during the weld test. The
three K type thermocouples at a distance from the welding heat source provided
the most stable results during all of the passes, with the R type thermocouples
being quite variable in effectiveness.
For Pass 1 the three K type thermocouples were inserted, with one R type
thermocouple located from the bottom face of the plate. This thermocouple location
118
is shown in figure 103. The figure clearly shows that the actual thermocouple
measurement point is lower than that intended due to a higher than expected
fusion zone during the pass. Further weld passes were measured with
thermocouples inserted down diagonal holes drilled into the plate from the top
surface. This technique had the problem that exact measurement location was not
guaranteed. Thermocouple welding to the bottom of each hole was successful if the
plug contacts were used as the contact point to the capacitor discharge welder. A
far better technique, for thermocouple placement, considered drilling holes from
the underside of the plate for thermocouple insertion. This method was not used as
it would have increased dramatically the time to complete welding. For each pass
the plate would require removal, drilling, thermocouple welding, and then
realignment of the plate before welding. This realignment of the plate was not
favoured and so the former technique of thermocouple placement was chosen.
Figures 103-106 show the final placement of thermocouples during the test. The
sectioning of the samples at correct locations allowed the evaluation to be made.
Table 6 provides a summary of thermocouple placement for all known locations.
Unfortunately thermocouple 6 could not be located due to problems with the saw
bench during sample sectioning. The accuracy of the thermocouple locations is not
very great and the distant K type thermocouples provided a more consistently
accurate measure for each pass due to known distance from the heat source.

Figures 95-101 show the thermal profiles obtained from thermocouples during each
weld pass. Pass 1 shows in figure 95 that thermocouple 2 did not function.
Unfortunately the data acquisition system used did not guarantee consistent
temperature measurement. The thermocouple placed in the groove measured a
weld pool temperature of 2400C and provided a good temperature profile. Pass 2,
shown in figure 96, gives working thermocouples for all 5 used. The placement of
thermocouple 5 was evidently, however, too close to thermocouple 4 as the
temperature profile is very similar and the peak temperature very closely matches.
This further highlights the problems of thermocouple placement with the employed
technique. Pass 3, shown in figure 97 shows that the 6
th
thermocouple was placed
in the molten pool area of the weld pool, this is evident by the high peak
temperature and good thermal profile. I t can be seen that, as expected, the peak
temperatures of the mid-thickness thermocouples at a further distance from the
119
heat source are increasing as the weld is deposited higher in the groove. Pass 4,
shown in figure 98, shows that firstly thermocouple 6 no longer provides sensible
information, which may be due to arc heating degrading the thermocouple as it
passes. Thermocouple 7, used to measure the molten temperature of pass 4, was
not placed well and only reached a peak temperature of 400C. The cooling rate of
this thermocouple is also not consistent with the other temperature profiles,
further indicating incorrect placement. Thermocouples 1-5 still function correctly
and provide good temperature profiles. Pass 5, detailed in figure 99, once again
shows that thermocouple 7 was misplaced. Thermocouple 8, added during the pass,
did not function correctly and was not included in the plot. Pass 6 did not include
any extra thermocouples, while the cap pass shown in figure 101 highlights the
peak temperature of 1400C for the thermocouple harpooned into the passing
welding pool.
For the entire groove weld thermocouples 1-5 represent the most reliable measure
of the thermal cycles. The advantage of using so many thermocouples during the
test is that data was still gathered even though some thermocouples did not
function. The resulting temperature profiles are adequate to use for weld model
validation.

Welding Efficiency Calorimeter Tests

Calorimeter tests to evaluate welding efficiency of the Tandem GMAW-P were
carried out, with results shown in tables 7-8 and figures 107-110.
The first test measured the rate at which the liquid nitrogen evapourated from the
dewar, with no external influence other than atmospheric temperature. The graph
in figure 107 shows the rate to be roughly 1g every 15 seconds. This was seen to be
affected by the time the liquid nitrogen had been in the dewar. I f the dewar had
just been filled the evapouration rate was higher due to cooling of the dewar
material. The nominal evapouration rate was not considered further during the
efficiency measurements as the loss of nitrogen was very small in comparison to
when a large welded sample was added. I n addition the violent nature of the liquid
nitrogen boiling that occurred during the weld tests meant it was not apparent
120
whether the normal evapouration rate would be relevant, as it was measured in
stable conditions.
Table 7 summarises the test calibration, whereby the experimental energy taken
from a sample was compared to the theoretical energy in the sample. This test
encompassed a wide range of sample weights and ambient test temperatures. The
results produced are very encouraging. Figure 110 shows a linear relationship
between stored energy in the sample and the sample weight. The linearity of the
experimental points shows that the test was very consistent. The relationship
shows that a sample held stable at a known temperature contains double the
energy if the sample weight is doubled. I nterestingly, very little difference in stored
energy is seen between the two sample materials of mild steel and X100, with both
having the same linear relationship as described.
The calibration procedure of the test was really more of a proof that the test
method was correct, by comparing experimental energy stored in a sample with
theoretical calculated values. Unfortunately the theoretical calculation, with
equation 26 given in the experimental description, contains more uncertainties
than the experimental value. The choice of specific heat for the steel must be
correct, in addition to correct temperature difference from ambient temperature to
liquid nitrogen temperature. The liquid nitrogen temperature was measured to be
-196C and -189C for two consecutive tests. The quoted value of -196C was used
for theoretical calculations. Given the linearity of the results the test procedure
was thought to be very satisfactory.
The welding efficiency test results are shown in table 8. I t is clear to see that
efficiency values are not very consistent across the tests and the overall values are
far too high, exceeding 100% on a number of occasions. There are a number of
reasons why results are not as expected. Firstly the size of the samples used for
welding efficiency were quite large, meaning there was a high stored energy. Upon
placement of the sample in the dewar it was difficult to prevent liquid nitrogen
form spilling slightly due to fast movement from the welding rig. I n addition the
high stored energy meant the evapouration of liquid nitrogen as quite violent and
this action may have added to nitrogen losses, overestimating the welding heat
input.
121
Results of room temperature tests, used to calculate welding energy, gave good
consistency. Samples were at room temperature meant it was far easier to place
the sample without loss of liquid nitrogen, as more time could be taken.
I t is not known whether the power input calculation, in equation 28, and intended
for pulsed power supplies, is correct for a tandem process. Further work needs to be
done with tandem power supplies to investigate the accuracy of the relationship.
The efficiency of the process does seem to be consistently higher in the groove
geometry than the bead on plate trials. This may be due to the fact that heat is
reflected back in the groove by the sidewalls, whereas the heat is lost to the
environment with a bead on plate weld. This agrees with Rykalin [29] who
developed an arc efficiency factor for analytical modelling of welding in groove
geometries.

Unfortunately for modelling purposes the efficiency values obtained from the
experimental work could not be used given that results over 100%. The trend of the
tests does show, however, that the Tandem GMAW-P is a high efficiency process
and so can be modelled as such.
122
7. Incorporation and comparison of Experimental
Results

The section details the development of modelling results, by the incorporation of
earlier detailed experimental input variable such as input power, travel speed and
heat source size.
I t was natural to use the best material properties available as a starting point for
generation model thermal profiles. The temperature dependent material property
models were used as a starting point for modelling as these should have been the
most representative properties available. As the section will show, a number of
material properties were modified to match the model thermal profiles to those
taken from experimental data. All material models referred to are described in the
modelling section of this report. Choice of model datasets are discussed in the
following section.
Weld Test 1
The model for weld test 1 first used the following input variables, known as
Dataset 1:
Thermal conductivity- Temperature dep. model 2
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 100
Efficiency - 0.9
Power input - 4137 W
Travel speed - 0.011743

Plate model 1 used the following size for heat source dimensions regardless of the
material properties used:
Front length - 11 mm
Rear length - 26 mm
Width (half width) - 5.5 mm
Depth - 5 mm
123

Figure 111: Weld test 1 model data comparison dataset 1


Figure 111 shows the results of the plate 1 model using the above dataset in
comparison to the experimental results.

Dataset 2 used the following variables:
Thermal conductivity- 40 W/mK
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 50
Efficiency - 0.8
Power input - 4137 W
Travel speed - 0.011743
124

Figure 112: Weld test 1 model data comparison dataset 2

Figure 112 shows the results of the plate 1 model using dataset 2 in comparison to
the experimental results.



Dataset 3 used the following variables:
Thermal conductivity- 20 W/mK
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 50
Efficiency - 0.9
Power input - 4137 W
Travel speed - 0.011743

125

Figure 113: Weld test 1 model data comparison dataset 3

Figure 113 shows the results of the plate 1 model using dataset 3 in comparison to
the experimental results.

The Heat Affected Zone (HAZ) size was evaluated by creating a temperature filter
in the COMSOL software. The edge of the HAZ was defined as any area that had a
peak temperature of at least 600C. Figures 114-115 shows the results obtained
from this temperature filter.


Figure 114: Heat Affected Zone model size for 5 mm heat model depth
126
For HAZ correlation to experimental values the heat source dimensions were
varied as the material properties had little local effect. Figure 114 shows the
results given heat source dimensions stated previously, a depth of 5 mm. Figure
115 shows the updated HAZ size with heat source model depth of 2 mm.
The HAZ dimensions compare to experimental values as follows:
Experimental HAZ size for weld test 1, given in Figure 93:
Width - 6.15 mm
Depth - 4.33 mm
Compared to HAZ size for heat source depth of 5 mm:
Width - 6.62 mm
Depth - 6.52 mm
HAZ dimensions for a heat source depth of 2 mm:
Width - 6.34 mm
Depth - 6.04 mm


Figure 115: Heat Affected Zone model size for 2 mm heat model depth

Weld Test 2

The model for weld test 2 first used the following input variables, known as
Dataset 1:
Thermal conductivity- Temperature dep. model 2
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 100
Efficiency - 0.9
127
Power input - 4137 W
Travel speed - 0.011743


Figure 116: Weld test 2 model data comparison dataset 1

Figure 116 shows the results of the plate 2 model using the above dataset in
comparison to the experimental results.

Dataset 2 used the following variables:
Thermal conductivity- 20 W/mK
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 50
Efficiency - 0.9
Power input - 4137 W
Travel speed - 0.011743

128

Figure 117: Weld test 2 model data comparison dataset 2

Figure 117 shows the results of the plate 2 model using the above dataset in
comparison to the experimental results.

Weld Test 4: Multipass Groove Weld

The experimental data taken from the welding test was used as input parameters for
each pass of the groove welding model. Five models were created to closely match the
first five passes of groove welding. This was seen to provide a good estimation of
whether the model thermal profiles matched closely enough the experiment
thermocouple data. Table 9 gives the geometric data for each pass in addition to the heat
input into the model and heat source dimensions. The heat source width was modelled
as being 1mm wider than the width of the groove.


Table 9: Model input data for Weld Test 4
129
Material properties for modelling were the same for all passes. The initial values
used were dataset 1.
Dataset 1:
Thermal conductivity- Temperature dep. model 2
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 100
Efficiency - 0.9
Power input - Table 9
Travel speed - Table 9

Comparison of model results to experimental thermal plots are given in the
following figures 118 to 127.

Figure 118: Multipass weld test 4 model data comparison dataset 1, pass 1
130

Figure 119: Multipass weld test 4 model data comparison dataset 1, pass 1







Figure 120: Multipass weld test 4 model data comparison dataset 1, pass 2
131

Figure 121: Multipass weld test 4 model data comparison dataset 1, pass 2







Figure 122: Multipass weld test 4 model data comparison dataset 1, pass 3


132

Figure 123: Multipass weld test 4 model data comparison dataset 1, pass 3






Figure 124: Multipass weld test 4 model data comparison dataset 1, pass 4

133

Figure 125: Multipass weld test 4 model data comparison dataset 1, pass 4






Figure 126: Multipass weld test 4 model data comparison dataset 1, pass 5

134

Figure 127: Multipass weld test 4 model data comparison dataset 1, pass 5

After refinement of material properties the follow dataset provided the closest
match to the experimental results:
Dataset 2:
Thermal conductivity- 20 W/mK
Specific heat - Enthalpy calculated model 3
Density - Temperature dep. model
Emissivity - Temperature dep. model
Heat transfer coeff. - 50
Efficiency - 0.9
Power input - Table 9
Travel speed - Table 9

Figures 128-137 following show the best comparison between model data and
experimentally derived thermal profiles.

135

Figure 128: Multipass weld test 4 model data comparison dataset 2, pass 1






Figure 129: Multipass weld test 4 model data comparison dataset 2, pass 1
136

Figure 130: Multipass weld test 4 model data comparison dataset 2, pass 2







Figure 131: Multipass weld test 4 model data comparison dataset 2, pass 2
137

Figure 132: Multipass weld test 4 model data comparison dataset 2, pass 3






Figure 133: Multipass weld test 4 model data comparison dataset 2, pass 3
138

Figure 134: Multipass weld test 4 model data comparison dataset 2, pass 4






Figure 135: Multipass weld test 4 model data comparison dataset 2, pass 4
139

Figure 136: Multipass weld test 4 model data comparison dataset 2, pass 5




Figure 137: Multipass weld test 4 model data comparison dataset 2, pass 5

The Heat Affected Zone (HAZ) size was again evaluated by creating a temperature
filter in the COMSOL software. The edge of the HAZ was defined as any area that
140
had a peak temperature of at least 600C. Figures 138-139 shows the results
obtained from this temperature filter.


Figure 138: Heat Affected Zone size with width of 7.9 mm

The HAZ width from the oscillated heat source is shown in figure 138 for the model
of Pass 3. The source width dimension of 7.91mm was initially used. The resulting
HAZ width was 7.9mm. Using a wider heat source width of 10mm produced an
HAZ width of 8.3mm shown in figure 139.



Figure 139: Heat Affected Zone size with width of 10 mm
141
8. Discussion of Model results

After evaluation of the experimental results the heat input and travel speed were
defined and therefore remained constant throughout the model refinement.
Efficiency and material properties were available for tweaking the model results.
The efficiency was set to a high number given that the calorimeter tests results
seemed to show a high thermal efficiency, even if results were not conclusive. This
was especially true in the groove sample. The efficiency was also the only factor
available to affect the peak temperature of the thermal profiles by increasing the
heat input in the model.
I nitial choice of material properties was to use the models that seemed most
appropriate for the X100 steel. Given that finite element methods can incorporate
temperature dependent material properties it was thought that these represented
the most sophisticated solution and therefore more closely described the properties
of the material being modelled.
The actual properties of X100 steel are, as previously stated, not known and
therefore the validity of using the temperature dependent material models was not
proven.

During model refinement previous experience of how a particular material
property varied the thermal profile of a model was used to match the model
temperatures to those of the experimental results.

Weld test 1, shown in Figure 111, compares the thermal profiles from the model
with those taken from three thermocouples at varied distance from the welding arc
on the surface of the 23 mm thick plate. The results of the dataset1, utilising
temperature dependent properties, show that peak temperatures are reached
earlier than the experimental values. The cooling rate is also higher for the model
thermal profiles. The thermal conductivity was subsequently changed, to decrease
the rate of heating, by applying a constant value of 40 W/mK. To decrease the
cooling rate in the model the heat transfer coefficient was reduced from 100 to 50.
I n addition the effect of a lower process efficiency was investigated. The results
comparing dataset 2 with the experimental results are shown in Figure 112. The
figure shows that the model profiles, although close in peak magnitude to the
142
experimental values, are still peaking too early. The cooling rate displays a closer
match. Dataset 3 uses and artificially low thermal conductivity value of 20 W/mK
in the model. The results in Figure 113 show a very good comparison between the
model thermal profiles and the experimental thermocouple data. The peak of the
profile at 11.51mm from the source is not as high as the experimental value and
the model also underestimates the profile of the 31mm thermocouple, even with a
high efficiency model value of 0.9. The general result is, however, a reasonable
match between datasets.

The variation of HAZ size was investigated by varying the heat source dimensions
in the model. I t can be seen in Figures 114-115 that the dimensions can easily
influence the overall size of the model HAZ to closely match any experimentally
derived HAZ width. The aspect ratio, or the width to height ratio, was not greatly
affected by variations in heat source depth and therefore depth comparisons were
not accurate.

Weld test 2, using four thermocouples around a 23mm plate, had two
thermocouples specially placed on the side surface of the plate to reduce the effect
of holes in the material. Figures 116-117 show that, as with the previous model,
the original temperature dependent material properties do not produce a good
match with experimental results, while the low thermal conductivity of 20 W/mK
does produce good results.

For evaluation of the multipass weld models experimentally derived values were
used as input parameters for the five passes. Five model passes seemed enough to
show whether the model closely matched the static thermocouple data, with the
variation of the heat source height for each successive pass.
I nput parameters that were not changed for the model refinements were travel
speed, power input, groove dimensions and thermal efficiency. I t is thought that
the thermal efficiency may decrease as the torch moves higher in the groove, but
this consideration was not included in the model. Heat source dimensions were
taken from the weld pool measurements made previously.
As with previous model refinements, the first dataset used included the
temperature dependent dataset defined as the closest available. Figures 118-127
143
show the comparison between the model thermal profiles at the same location as
the actual thermocouples, as accurately as can be known for each weld pass.
Results show that peak temperatures are comparable for the distant
thermocouples 1-3 located at the mid-depth in the plate. The heating and cooling
rate is, however, as has been displayed before far too high. Peak temperatures for
model profiles near the heat source are shown not to match the thermocouple
placed near the arc during welding. This highlights that the modelling around the
heat source is more complex than the model produced. Dataset 2, including an
artificially low constant thermal conductivity value of 20 W/mK and a heat transfer
coefficient of 50 was the final refinement for each weld pass model. The results are
shown in Figures 128-137 and clearly show that, with these material properties,
there is good agreement between the model and experimental thermal profiles.
This is true with regard to the heating and cooling rates of the thermal cycles. I n
addition, although the peak temperatures do not match for all thermocouples, the
general fit is good. Clearly a great deal more work could be done to fit the
temperature profile exactly by further modification to the model and material
properties, but given the timeframe of the project the result provided present the
best results possible.

Modification of heat source parameters and subsequent variation in HAZ size is
shown in figures 138-139. As previously found the heat source can easily be
modified to fit the width of the experimental HAZ, however the depth is not equally
as adaptable. Further investigation of the HAZ size did not seem sensible as the
model produced was essentially too simple to accurately characterise the thermal
cycles very close to the modelled heat source.

The low constant value of 20 W/mK used to model the thermal conductivity was
previously found by Moore et al. [14]. Moore used this value with an analytical
model, however, and concluded that the value was necessary to compensate for
other material properties, that were not temperature dependent, in addition to
neglecting the latent heat effects. The low value used for the current model may be
a result of the numerous holes drilled into the plate to locate thermocouples during
welding. This would not, however, explain why the low value also gave good
comparisons for the bead on plate models, where no holes were drilled for
144
thermocouple placement. The low apparent conductivity may be due to the fine
grain size present in the X100 steel. The grain size gives the steel its high strength
while also maintaining a high toughness. High grain density may inhibit heat
conduction through the material. Alternatively the same thermal profiles may have
been produced by variation of another set of material properties, which may be
closer to the actual properties of the material. I t is clear that without the X100
temperature dependent properties there is no way to evaluate this.
145
9. Conclusions and Further Work

The aim of this project was to create a model that would accurately predict the
thermal cycles in X100 steel during welding. The model considered the Tandem
Pulsed GMAW process and the narrow groove geometry currently being developed
for industrial use. The modelling approach purposefully included complex features
such as groove geometry and an oscillated heat source in order to more accurately
match the real welding process. Utilisation of these complex variables, in addition
to temperature dependent material properties, meant the model was developed
with a finite element package. The main assumption was to simplify the model by
considering the weld metal deposited during a pass as having the same material
properties as those of the base metal, and therefore the entire geometry was one
single homogeneous domain. Modelling weld metal addition has been quite a
challenge to researchers, with many using a technique of additive mesh generation
as the model heat source passes [10].
The results produced from the model show that good agreement can be obtained,
even with a very simplistic model, between model and experimental thermal cycles
that are not close to the heat source being modelled. Clearly the model does not
adequately describe the phenomena around the heat source correctly due to
neglecting the variation of metal properties in this area. I n addition it is known
that the interactions in the weld pool are very complex and many models simplify
this analysis. The finite element package used is capable of creating a second
domain, to model the deposited weld, and a different set of material properties
could be applied to this second domain. Another common way to describe the heat
source in welding is to use a prescribed temperature for at the weld pool boundary.
This is used in place of the distributed heat source and may produce a more
accurate result closer to the heat source.
The lack of temperature dependent material data for the X100 steel has made it
difficult to be sure that the results obtained are not merely an exercise in curve
fitting and it is clear that better characterisation of the welded material is
required. This means either material data for X100 steel must be sourced or tests
must be done to gather the required properties.
146
An alternative approach would be to use the same welding process to weld a
different base metal, one which has detailed material property data available. This
would allow more refinement of the welding model, should the thermal profiles
produced not match the experimental results, instead of any change in the material
property data.

Given the timescale of the project both modelling and experimental work had to be
brief. Further work could be done on each aspect of the project for more
comprehensive results. Measurement of the thermal cycles during multipass
welding proved extremely difficult and refinement of the method would provide
more reliable experimental data in the future. The final location of the
thermocouple, with the method used, proved to be low in accuracy. More defined
thermocouple placement would allow better comparison with model results. The
thermal efficiency measurements also proved to be far from simple and the test
requires more time to overcome the problems encountered in this project. Once a
stable efficiency has been found more emphasis can be placed on the variation of
thermal efficiency as the torch moves higher in the groove.

Given more time the effect of varied heat input on thermal cycles of X100 can be
appreciated to a greater extent, but the model developed provides a suitable
starting point for further work.
147
10. References

1. Aristotele, R., Di Vito, L. F., Barsanti, L., Welding X100 steels for gas
pipelines, Welding I nternational 2004, 18(11) 877-882
2. Richter, K., Hanus, F., Wolf, P., Structural Steels of 690 MPa Yield Strength
a State of Art , web document
www.sintef.no/units/matek/PRESS/22_Hanus.doc, last visited 22/06/2007
3. Hillenbrand, H., Development of large-diameter pipe in grade X100, web
document http://www.europipe.de/www/download/EP-TP_36-00_en.pdf, last
visited 22/06/2007
4. Hudson, M., Blackman, S.A., Hammond, J ., Dorling, D.V., Girth welding of
X100 pipeline steels, Proceedings of the I nternational Pipeline Conference, I PC
Hudson yr:2002 vol:A pg:525 -532
5. Laratzis, T., Tandem Gas Metal Arc Pipeline Welding, PhD thesis 2007,
Cranfield University
6. ASM Metals Handbook, Volume 6, Welding, Brazing and Soldering.
7. Kou, S., Welding Metallurgy, 2nd Edition, I SBN: 0471460931 Publisher: J ohn
Wiley & Sons, I nc.
8. Komanduri, R., Hou, Z.B., Thermal analysis of the arc welding process: Part I I .
Effect of variation of thermophysical properties with temperature,
Metallurgical and materials transactions. B, Process metallurgy and materials
processing science, yr:2001 vol:32 iss:3 pg:483 -499
9. Lindgren, L.-E., Finite element modeling and simulation of welding part 1:
I ncreased complexity, J ournal of thermal stresses, yr:2001 vol:24 iss:2 pg:141 -
192
10. Goldak, J ohn A., Computational Welding Mechanics, I SBN10: 0387232877,
Publisher(s): Springer Verlag
11. Eagar, T.W., Tsai, N.-S., Temperature fields produced by travelling distributed
heat sources, Welding journal, yr:1983 vol:62 iss:12 pg:346 -355
12. Wahab, M.A., Painter, M.J ., Numerical models of gas metal arc welds using
experimentally determined weld pool shapes as the representation of the
welding heat source, The I nternational journal of pressure vessels and piping,
yr:1997 vol:73 iss:2 pg:153 -159
148
13. Gery, D., Long, H., Maropoulos, P., Effects of welding speed, energy input and
heat source distribution on temperature variations in butt joint welding,
J ournal of materials processing technology, yr:2005 vol:167 iss:2-3 pg:393 -401
14. Moore, J ., Bibby, M., Goldak, J ., A comparison of the point source and finite
element schemes for computing weld cooling, Welding Research: The State of
the Art, Proc. of 1985 I nt. Welding Congress in junction with ASM Materials
Week 85, p.1, 1985
15. Wu, C., Wu, L., Microcomputer-aided system for selecting arc-welding process
parameters, Computer-aided engineering journal, yr:1991 vol:8 iss:3 pg:122 -
125
16. Okada, A., Kasugai, T., Hiraoka, K., Heat source model in arc welding and
evaluation of weld heat-affected zone, Transactions of the I ron and Steel
I nstitute of J apan, yr:1988 vol:28 iss:10 pg:876 -882
17. Yurioka, N., Kasuya, T., Prediction of HAZ hardness of transformable steels,
Metal construction , yr:1987 vol:19 iss:4 pg:217r -223r
18. Yurioka, N., Kojima, K., A predictive formula of weld metal tensile strength,
Quarterly journal of the J apan welding society, yr:2004 vol:22 iss:1 pg:53 -60
19. Odanovic, Z., Numerical Modelling of Microstructure in Heat Affected Zone of
GMA Welded HY-100 Steel, Mathematical Modelling of Weld Phenomena 5,
Cerjak, I SBN: 1861251157, Publisher: I OM Communications Ltd
20. Goncalves, C.V., Vilarinho, L.O., Scotti, A., Guimaraes, G., Estimation of
heat source and thermal efficiency in GTAW process by using inverse
techniques, J ournal of materials processing technology, yr:2006 vol:172 iss:1
pg:42 -51
21. J oseph, A., Harwig, D., Farson, D.F., Richardson, R., Measurement and
calculation of arc power and heat transfer efficiency in pulsed gas metal arc
welding, Science and technology of welding and joining, yr:2003 vol:8 iss:6
pg:400 -406
22. Glowacki, M., The mathematical modelling of thermo-mechanical processing of
steel during multi-pass shape rolling, J ournal of materials processing
technology, yr:2005 vol:168 iss:2 pg:336 -343
23. Frewin, M.R., Scott, D.A., Finite element model of pulsed laser welding,
Welding journal, yr:1999 vol:78 iss:1 pg:15 s
149
24. Zhu, X.K., Chao, Y.J ., Effects of temperature-dependent material properties on
welding simulation, Computers & structures, yr:2002 vol:80 iss:11 pg:967 -976
25. Sabapathy, P. N., Wahab, M. A., Painter, M. J ., Numerical models of in-
service welding of gas pipelines J ournal of Materials Processing
Technology, Volume 118, I ssues 1-3, 3 December
26. Radaj, D., Heat Effects of Welding, I SBN10: 038754820, Publisher: Springer-
Verlag
27. COMSOL Multiphysics, Heat Transfer Module, User Guide, 2005
28. Kou, S., HEAT FLOW DURI NG THE AUTOGENOUS GTA WELDI NG OF
PI PES, Metallurgical transactions. A, Physical metallurgy and materials
science, yr:1984 vol:15 A iss:6 pg:1165 -1171
29. Ramirez, A.J ., Brandi, S.D., Application of discrete distribution point heat
source model to simulate multipass weld thermal cycles in medium thick
plates, Science and technology of welding and joining, yr:2004 vol:9 iss:1 pg:72
-82
30. Hirata, Y., Pulsed arc welding, Welding research abroad, yr:2003 vol:49 iss:4
pg:1 -18
31. Frewin, M.R., Scott, D.A., Finite element model of pulsed laser welding,
Welding journal, yr:1999 vol:78 iss:1 pg:15 s
32. Patankar, S., Numerical Heat Transfer and Fluid Flow, I SBN10: 0891165223,
Publisher: Taylor & Francis

Anda mungkin juga menyukai