Content uploaded by Obeid Obeid
Author content
All content in this area was uploaded by Obeid Obeid on Jul 10, 2019
Content may be subject to copyright.
Content uploaded by Obeid Obeid
Author content
All content in this area was uploaded by Obeid Obeid on Jul 10, 2019
Content may be subject to copyright.
1
Thermo-mechanical analysis of a single-pass
weld overlay and girth welding in lined pipe
Obeid Obeid1, a, Giulio Alfano1,b, Hamid Bahai1,c
1College of Engineering, Design and Physical Sciences, Brunel University, UB8 3PH
Uxbridge, UK
aobeid.obeid@brunel.ac.uk, bgiulio.alfano@brunel.ac.uk, chamid.bahai@brunel.ac.uk
ABSTRACT
The paper presents a nonlinear heat-transfer and mechanical finite-element (FE) analyses of a
two-pass welding process of two segments of lined pipe made of a SUS304 stainless-steel
liner and a C-Mn steel pipe. The two passes consist of the overlay welding (inner welding) of
the liner with the C-Mn steel pipe for each segment and the girth welding (outer welding) of
the two segments. A distributed power density of the moving welding torch and a non-linear
heat transfer coefficient accounting for both radiation and convection have been used in the
analysis and implemented in user-subroutines for the FE code ABAQUS. The modelling
procedure has been validated against previously published experimental results for stainless
steel and carbon steel welding separately. The model has been then used to determine the
isotherms induced by the weld overlay and the girth welding and to clarify their influence on
the transient temperature field and residual stress in the lined pipe. Furthermore, the influence
of the cooling time between weld overlay and girth welding and of the welding speed have
been examined thermally and mechanically as they are key factors that can affect the quality
of lined pipe welding.
Keywords: Finite-element, Lined pipe, weld overlay, girth welding, thermal analysis,
temperature field, residual stress.
1. Introduction
As the increasing energy demand forces the offshore industry to explore more reservoirs in
increasingly harsher environments, larger volumes of corrosive products which contain CO2
and H2S need to be transported through pipes from the well to the processing facility. Lined
pipes are a relatively new and increasingly practical solution for transporting these corrosive
products in the oil and gas production industry. Recent marketing studies have shown there is
2
a massive demand for lined pipes as they result in reduced maintenance costs and increased
life cycle [1].
A double-wall, bimetallic CRA-lined pipe, consists of an inner layer (the liner) made of
corrosion resistant alloy (CRA) such as Alloy625, Alloy825, 304, 316L stainless steel and an
outer layer made of carbon steel.
Incompatible plastic strains are generated during welding due to the high temperature
gradients experienced in the vicinity of both the weld overlay and the girth weld when the
carbon-steel and the stainless-steel cool down from their melting temperatures. The
susceptibility of a weld to fatigue damage, fracture and stress-corrosion cracking increases
due to the presence of tensile residual stresses which are generally considered detrimental [2].
Due to the complexity of the welding process, experimentally validated and reliable models
for predicting welding residual stresses are desirable. Accordingly, the application of the
finite-element method (FEM) has become extremely useful for predicting the thermal history
and residual stresses. Goldak et al. [3, 4] developed a nonlinear transient finite-element heat-
flow model where a double ellipsoidal geometry was proposed to simulate the heat torch.
With their approach, the size and shape of the heat source can be easily changed to model
both the shallow penetration of conventional welding processes and the deeper penetration of
laser and electron beam processes. For the thermal analysis of welds, Brickstad and Josefson
[5] used a technique called ‘element birth’ to represent the sequence of weld beads, which
avoids the displacement or strain mismatch at the nodes connecting the weld elements to
those of the base material. Karlsson and Josefson [6] studied temperature, stress and
deformation in a single-pass butt welded joint made of C-Mn steel with a full three-
dimensional finite element (FE) model developed in ADINA. In addition, they verified their
numerical model by comparing temperature and strain values with experimental ones
measured via thermocouples and strain gauges. Likewise, Deng and Murakarwa [2]
experimentally verified their numerical simulations of a multi-pass weld in a stainless-steel
pipe, which has the same outer diameter as in [6]. Subsequently, Deng et al. [7] tried to
validate welding simulations of dissimilar materials by studying a joint between a low alloy
steel pipe and an austenitic stainless steel pipe. Some discrepancy between the numerical and
the experimental results were found because the cladding was not taken into account in the
dissimilar metal pipe. Instead, very good agreement between experimental and simulation
results was found by Attarha and Sattari-Far [8] for dissimilar thin butt-welded joints in
which one joint is made of St37 and the other one is S304. Lee and et al. [9] found that there
3
is no symmetry in the temperature field in dissimilar steel girth-welded pipes. Moreover,
residual stresses are considerably different from their counterparts in similar steel pipe welds.
From the foregoing studies, it appears that no work has thus far addressed the three-
dimensional features of the thermal fields induced by the traveling arc and welding start/stop
effects during overlay weld to fix the liner at the pipe ends and girth welding of lined pipes.
Hence, in this study, using the ABAQUS software, a three-dimensional FE model is
developed to simulate the evolution of the temperature field and residual stresses in a lined
pipe made of a SUS304 stainless-steel liner and a C-Mn steel pipe. The proposed method
uses non-linear modelling of the heat flux through exposed metal surfaces and accounts for
the moving heat source during welding. These two features are implemented by coding two
separate user subroutines in ABAQUS.
The presented numerical procedure is validated against previously published experimental
results for stainless steel and carbon steel welding separately. The model has been then used
to predict the transient temperature field and residual stress distributions during the weld
overlay (inner welding) and the girth welding (outer welding) of a lined pipe (case A).
Furthermore, a sensitivity analysis to determine the influence of the cooling time (case B)
between weld overlay and girth welding and of the welding speed (cases C and D) has been
conducted thermally and mechanically, as these are key parameters that can be fine-tuned to
improve the welding quality.
2. Description of manufacturing and welding conditions
Lined pipes are typically manufactured using hydraulic expansion, thermal shrinking or a
combination of the two, in order to guarantee a reliable interference fit between the liner and
the backing steel pipe [1]. If thermal shrinking is used, the carbon-steel pipe is first heated up
to about 300-400°C; the liner is then inserted in it, possibly having been first cooled down
itself in cold water. Thermal shrinking of the carbon-steel pipe then ensures the required fit.
Alternatively, if hydraulic expansion is used, the liner has sufficiently lower dimensions so
that it can be inserted into the carbon-steel pipe and is then expanded hydraulically.
Typically, in this case the liner is expanded beyond yielding while the carbon-steel pipe
remains elastic, so that the elastic spring back of the outer pipe ensures a tight fit. An overlay
welding between the pipes is normally used also as a way to seal the gap between them, and
therefore avoid that moisture, grease and dirt penetrate the gap, as shown in Fig. 1.
4
Fig. 1. Weld preparation to seal the pipe at ends.
The weld overlay filler material is chosen to have corrosion resistance which preferably
exceeds that of the liner material.
3. Thermal Analysis
3.1. Modelling of the heat source
The physical phenomena associated with the interaction between the welding torch and the
weld pool are complex [10].
Goldak [11, 12] developed a three-dimensional model in which the heat source as a
function of position is represented with a Gaussian distribution of the power density in an
ellipsoid (welding pool) with centre that is taken as:
(1)
where , and are the semi-axes of the ellipsoid in directions , and , respectively, as
illustrated in Fig. 2.
5
Fig. 2. Ellipsoidal weld bead with semi-axes, and .
Notice that total heat input can be related to the applied voltage and current in the heat
torch as follows:
(2)
where is the welding efficiency.
In some studies [11, 12], a modified double ellipsoidal heat source model is utilised with a
slight readjustment of the heat distribution equation according to the circumferential moving
whereby the fractions and of the heat deposited in the front and rear quadrants are
needed, where . However, this modification is not used in this work because other
authors [4, 13] found good correlation with experiments taking and equal to 1.
To account for the rotational movement of the welding front along the circumference, the
power density can be given as a function of position and time as follows:
(3)
where is the radial distance of the heat torch centre from the pipe axis, is the angle that
the torch has travelled around the pipe, starting from a starting point where . Denoting
by the angular velocity used in welding, it results , where is the current
time and is the initial time of the analysis.
Equation (3) has been implemented in ABAQUS by coding the FORTRAN DFLUX user-
subroutine [14]. The position of the weld torch is calculated first in DFLUX according to the
welding time. The power density q is then computed at each integration point.
3.2. Thermal properties
6
A transient heat-transfer analysis is conducted to evaluate the temperature field history during
welding. In this case, the energy balance for each domain is governed by the classical energy
balance equation given as [11]:
(4)
where denotes the density of the materials, is the enthalpy (per unit volume), is the
time, is the temperature, is the material thermal conductivity, assumed to be
isotropic, and is the welding volume heat input (defined earlier in Section
3.1).
The specific enthalpy in Eq. (4) is defined as:
(5)
where and are the latent heat and heat capacity, respectively, is an arbitrary reference
temperature, and is the volumetric liquid fraction known as a characteristic function of
temperature, defined as:
(6)
where and are the solidus and liquidus temperatures, respectively.
The initial condition to Eq. (4) is in our case given by:
(7)
where is the initial temperature of the pipe, that with a good approximation can be taken as
constant in space and equal to the ambient temperature.
The boundary conditions on the outer and inner surfaces are given by:
(8)
where , and are the direction cosines of the normal to the boundary, is the heat-
transfer coefficient, that is defined as a function of temperature as discussed below, is the
current temperature at the pipe surface, whereas is the ambient temperature.
7
Since we exploit the symmetry of the problem, we need to enforce on the plane of symmetry
that the heat flux is zero, which leads to this other boundary condition on this plane:
(9)
In this work, both radiation and convection are taken into account for the boundary
conditions during the thermal analysis. During a thermal cycle, radiation and convection take
place from all the surfaces exposed to the environment. In particular, radiation heat losses are
dominant in and nearby the weld pool whereas convection heat losses are dominant at lower
temperatures away from the weld pool [5, 8, 9, 15]. As there are two different base materials,
two heat transfer coefficients are considered. Each heat coefficient includes a combination of
convection and radiation effects. For the carbon steel surfaces, the total heat transfer
coefficient can be written as [9]:
(10)
where is the convective heat transfer coefficient, is the effective radiation
emissivity, is the current temperature at the pipe whereas is the ambient temperature,
and is the Boltzman constant. Following [16], in the present study, the convective heat
transfer coefficient is assumed to be 8 W/m2°C whereas the emissivity is set to be
0.51.
For the stainless steel surfaces, we used the following widely used bilinear law [5]:
(11)
because for this material it is a good approximation of the actual cubic expression that would
be obtained using Equation (11) for typical values of the emissivity of stainless steel of 0.5-
0.75 and a range of temperature between ambient and 2400°C.
A FILM user subroutine [14] was used to implement the above expressions of the heat-
transfer coefficient in ABAQUS. It is worth noting that ABAQUS allows one single user-
subroutine to be written for both materials by simply specifying which surface each condition
applies to.
8
Fig. 3. Effect of radiation and convection in Lined pipe.
As can be seen from Fig. 3, the radiation and convection take place from all sides of the
welded lined pipe exposed to the environment except the area at which the heat flux is
applied [15].
To account for material melting and for the heat transfer due to the fluid flow in the weld
pool, two methodologies are used [2]. An artificially increased thermal conductivity, which is
several times larger than the value at room temperature, is assumed for temperatures above
the melting point. The thermal effects due to solidification of the weld pool are modelled by
taking into account the latent heat, which is the heat energy that the system stores and
releases during the phase transformation.
4. Validation
In order to validate the FE procedure for the girth welding, the approach outlined above is
used to simulate the problem studied by Karlsson and Josefson [6], and our numerical results
are compared with the experimental ones reported in [6]. The pipe studied has an outer
diameter of 114.3 mm and a wall thickness of 8.8 mm, with a 5.5mm-deep V-groove for
welding, and the pipe material is C-Mn steel (Swedish standard steel SIS2172). The welding
material is MIG (Metal Inert Gas) deposited from the outside into the groove in a single pass
with a speed equal to 6 mm/s as shown in Fig. 4.
9
Fig. 4. Karlsson and Josefson FE model [6]
Our numerical results of the thermal and mechanical analyses have been compared with the
experimental measurements at various axial locations, where the circumferential angle
from the welding start/stop position is 150°. In the thermal analysis, those points are located
on the outer and inner surfaces with respect to the weld centreline CL. It can be seen from the
plot shown in Fig. 5(a) that the thermal simulation results correlate well with the
experimental red contour lines. In the mechanical analysis, the residual stress distributions at
150° from start/stop welding location on the inner surface along the axial direction correlate
also well with those obtained by the validated experiment performed by Karlsson and
Josefson [6] as shown in Fig. 5(b).
(a)
(b)
-100
-50
50
100
150
200
250
300
350
400
010 20 30 40
Hoop Sress (MPa)
Distance from WCL (mm)
Exp.
FEM
10
Fig. 5. Distributions of (a) temperatures (°C) and (b) inner hoop residual stress at =150° numerically computed
in this work and experimentally validated in [6].
In a similar way, to validate the weld overlay FEA approach, the experiment conducted by
Deng and Murakawa [2] has been simulated. The material used in this work was stainless
steel (SUS304) and the pipe model has a 114.3 mm outer diameter and 6 mm thickness. Gas
Tungsten Arc (GTA) welding was used in the experiments to fill a U-groove by two welding
passes with 80 mm/min as welding speed as shown in Fig. 6.
Fig. 6. Deng and Murakawa experiment [2].
The thermal history findings have been numerically compared with their experimental
counterparts at three axial locations, points 1, 2 and 3 as shown in Fig. 6. These points are
placed on the outer surface with respect to the axial distance from the weld centre line where
the circumferential angle φ from the start/stop position is 180°. Fig. 7(a) shows there is a
good match between our thermal FEA results and the experimental ones obtained from [2].
Moreover, Fig. 7(b) shows a good correlation between the results of the hoop residual
stresses along the axial distance on the inner surface which are taken from our FE mechanical
model and the experimental results in [2] where the angular location from the welding
start/stop point is 180°.
11
(a)
(b)
Fig. 7. Comparison of (a) thermal histories and (b) inner hoop residual stresses at =180° numerically computed
in this work and experimentally validated in [2].
5. Finite element modelling of the lined pipe
5.1. Description of the lined pipe joint and welding conditions
Using ABAQUS [14], the FE computational procedure described in Section 3, and validated
in Section 4 for the separate cases of a carbon-steel pipe and a stainless-steel pipe, has been
implemented to calculate the transient temperature field and residual stresses during welding
of two segments of a lined pipe, in which a one-pass weld overlay and a one-pass butt-welded
joints are used. The configuration of the lined-pipe joint has an outer diameter of 114.3 mm
and a wall thickness of 6 mm, of which 4.5 mm is the thickness of the C-Mn outer pipe and
1.5 mm is the liner thickness, as schematically shown in Fig. 8. Only one-half of the pipe,
which is 200 mm long, is analyzed due to symmetry around the weld line.
0
200
400
600
800
1000
100 150 200 250 300 350 400
Temperature (°C)
Time (s)
1(Ex.)
2(Ex.)
3(Ex.)
1(FEM)
2(FEM)
3(FEM)
-300
-200
-100
100
200
300
400
050 100 150
Hoop Stresses (MPa)
Distance from WCL (mm)
Exp.
FEM
12
Fig. 8. Dimensions of analysis model.
The outer pipe material is C-Mn steel with a composition of 0.18%C, 1.3%Mn, 0.3% Si,
0.3%Cr, 0.4%Cu (Swedish standard steel SIS2172) and the temperature-dependent thermo-
mechanical material properties, namely density, specific heat, latent temperature, thermal
expansion, yield stress, Young’s modulus and conductivity used for the outer pipe are taken
from the work of Karlsson and Josefson [6] as reported in Table 1. The thermo-mechanical
properties for the SUS304 SS liner are obtained from the study of Deng and Murakawa [2] as
shown in Table 2. Moreover, the MIG-welding is implemented for girth weld whilst the
GTA-welding process is used in filling the weld overlay groove. In absence of specific data,
we follow [2, 6] so that base metals and weld metals are defined as different materials in
ABAQUS-code but having the same thermo-mechanical properties corrospending to thier
base materials except the yield stress, as illustrated in Table 1 and 2. This is an approximation
of reality that in some cases may be inaccurate, but in this work we are using the cases
studied in [2, 6] for the two different materials as a way of validating the separate models,
and therefore we follow these articles in this simplified assumption and also when analysisng
the lined pipe.
Table 1. Thermo-mechanical properties of C-Mn [6].
Temperature
(°C)
Density
(Kg/m3)
Specific
heat
(J/Kg°C)
Conductivity
(W/m°C)
Thermal
expansion
(x10-5°C-1)
Yield stress (MPa)
Young’s
modulus
(GPa)
Possion’s
ratio
Base
Weld
0
7860
444
50
1.28
349.45
445.42
210
0.26
100
480
48.5
1.28
331.14
441.29
200
0.28
200
503
47.5
1.30
308.00
416.49
200
0.29
300
518
45
1.36
275.00
376.18
200
0.31
400
555
40
1.40
233.00
325.54
170
0.32
600
592
35
1.52
119.00
172.59
56
0.36
800
695
27.5
1.56
60.00
43.41
30
0.41
13
1000
700
27
1.56
13.00
14.47
10
0.42
1200
700
27.5
1.56
8.00
9.30
10
0.42
1400
700
35
1.56
8.00
9.30
10
0.42
1600
700
122.5
1.56
8.00
9.30
10
0.42
Table 2. Thermo-mechanical properties of SUS304 [2].
Temperature
(°C)
Density
(kg/m3)
Specific
heat
(J/kg°C)
Conductivity
(W/m°C)
Thermal
expansion
(x10-5°C-1)
Yield stress
(MPa)
Young’s
modulus
(GPa)
Possion’s
ratio
Base
Weld
0
7900
462
14.6
1.70
265
438.37
198.50
0.294
100
7880
496
15.1
1.74
218
401.96
193
0.295
200
7830
512
16.1
1.80
186
381.5
185
0.301
300
7790
525
17.9
1.86
170
361.25
176
0.310
400
7750
540
18.0
1.91
155
345.94
167
0.318
600
7660
577
20.8
1.96
149
255.71
159
0.326
800
7560
604
23.9
2.02
91
97.41
151
0.333
1200
7370
676
32.2
2.07
25
28.41
60
0.339
1300
7320
692
33.7
2.11
21
16.23
20.00
0.342
1500
7320
700
120
2.16
10
12.17
10
0.388
The numerical values for the variables in the power density distribution Eqs. (2) and (3) are
illustrated in Table 3 for each welding materials.
Table 3. Heat source parameters and welding parameters.
SUS304
C-Mn (SIS2172)
Half-length of arc (mm)
2.765
3.26
Depth of arc (mm)
2.575
3.2
Half-width of arc (mm)
1.5
3
Welding current (A)
120
170
Voltage (V)
8
20
Welding speed (mm/s)
1.33
6.25
Welding efficiency
70%; Gas Tungsten Arc
(GTA) [19]
85%; MIG (Metal Inert
Gas) [19]
Based on the heat torch parameters presented in Table 3, the power density distributions of
Goldak ellipsoidal heat source along the welding directions for SUS304 and C-Mn (SIS2172)
are depicted in Fig. 9.
14
Fig. 9. Power density distributions of Goldak ellipsoidal heat source models, liner and carbon steel.
The latent heat for C-Mn steel (SIS2172) is set to be 247kJ/kg between the solidus
temperature 1440 °C and the liquidus temperature 1560 °C. For stainless steel (SUS304), the
latent heat is assumed to be 260kJ/kg between 1340 °C and 1390 °C, solidus and liquidus
temperature, respectively. Consequently, the melting point for carbon steel is 1500 °C while
it is 1365 °C for SUS304. The initial temperature of the lined pipe and the weld bead is set at
room temperature, namely 20 °C.
5.2. Finite element mesh
Only one half of the lined joint is modelled due to symmetry. The three-dimensional FE
model contains a total of 51840 nodes associated with 10560 elements. Among these, 17400
nodes and 2400 elements represent the liner geometry whereas the remaining of 51840 nodes
and 10560 elements represent the backing pipe geometry.
Fig. 10. Three-Dimensional FEM.
15
A fine mesh has been used in the fusion zone (FZ), the zone where the temperature reaches
values beyond the melting point, and its vicinity, i.e. in the heat affected zone (HAZ),
because of the higher temperature and flux gradients. The element size increases with the
distance from the welding centreline (WCL) for both the C-Mn pipe and the liner. The
number of divisions in the circumferential direction is 120. Furthermore, there are 4 layers of
elements through the thickness direction, three of them for the C-Mn pipe and one for the
liner as shown in Fig. 10.
Uncoupled thermo-mechanical analyses have been developed to simulate the welding.
Therefore, the thermal analysis is simulated first to acquire the thermal history at each node
through the lined pipe. This thermal history is then transferred to the mechanical analysis as
an input to determine the temperature-dependent mechanical properties. In this case, the FE
mesh of the mechanical analysis should have the same mesh associated with the same
arrangement of nodes and elements used in the thermal analysis. In the thermal analysis, 20-
node quadratic hexahedral heat-transfer elements, named DC3D20 in ABAQUS, have been
employed. In the mechanical analysis, 20-node reduced integration elements, named C3D20R
in ABAQUS, are employed to minimise the simulation time.
5.3. Thermal Analysis
5.3.1 Moving of filler metal
A moving heat source combined with the element-birth technique is used to simulate the
deposition of the elements of weld bead incrementally. In other words, to represent the
transient nature of weld metal deposition [17], a number of element sets is created, so that the
elements forming each weld bead belong to a specific set. In this way each bead in a weld
pass can be deposited independently during simulation. At the beginning of the thermal
simulation, the element sets of both weld passes are made ‘inactive’ by assigning very low
conductivity to them. The deposition of each bead is then modelled using a sequence of
‘steps’. In each step, the element set that is in the current position just reached by the weld
torch is re-activated in the FE mesh. The sequential steps of the developed procedure to
simulate the welding are described in Table 4.
16
Table 4: Simulation procedure.
Step(s)
Initial step
time (s)
Final step
time (s)
Description
1
0
1×10-10
All of the weld passes’ elements, SUS304 and C-Mn,
are deactivated.
2
1×10-10
2×10-10
The first section of the weld overly pass is added.
3
2×10-10
2.067
The heat source begins to move, applying the heat
flux corresponding to liner heat flux equation.
4
2.067
2.067+10-10
The second bead of the liner weld pass is added.
5
2.067
4.133
The heat source continues its motion.
6-241
4.133
248
Steps 4-5 are repeated adding new sections of the
weld overlay pass and applying the liner heat flux.
242
248
518
The torch is removed and the two pipes cool down
until the maximum temperature is below 100°C.
243
518
518+10-10
The first section of the girth welding pass is added.
244
518
518.479
The heat source moves, applying heat flux
corresponding to carbon-steel heat flux equation.
245-482
518.479
575.5
Steps 243-244 are repeated adding new sections of the
C-Mn welding pass and then applying girth welding
heat flux.
483
575.5
3575.5
The torch is removed and the two pipes cool down
almost to room temperature
5.3.2 Conditions of the thermal analysis
For an optimally designed welding process, the results provided by the thermal FE analysis
should satisfy the following three conditions:
1- All integration points in target FZ should reach at least the melting temperature. This
guarantees that the FZ melts entirely before cooling down.
2- Because welding parameters such as current, voltage, speed and welding pool
geometries have constant magnitudes during welding, the temperature history for
every node located on the same circumferential line should be close to identical after a
relatively short initial transient part of the analysis, except for a time shift.
17
3- The boundary of the HAZ should remain about 2-3 mm from the FZ boundary
whereby the net heat input plays a crucial role. The problem is that, even for the cases
where weld specifications exist in codes of practices such as API 1104, EN ISO
15609, ASME IX, the data in terms of current, voltage and welding speed are
generally given with such wide limits that the net heat input can easily vary by a
factor of 4 and still be inside the allowed limits for welding process [5].
Brickstad and Josefson [5], who studied a case where the material is stainless steel, and
Karlsson and Josefson [6], who considered carbon steel, found out the typical boundary of the
HAZ is located approximately 2-3 mm from the FZ boundary where the temperature is
between 800-900°C.
5.4. Mechanical Analysis
In the mechanical analysis, body forces and surface tractions are assumed to be neglected
according to the definition of residual stresses which are the self-equilibrating internal
stresses [18, 19].
The only load considered in the structural model is the load generated by the transient
thermal field at each node during the thermal analysis. This induces non-uniform thermal
strain through the entire lined pipe because: a) two base materials with their welding
materials have accordingly different coefficients of thermal expansion, b) the initial
temperatures of welding and its base material are different and c) high temperature gradients.
Furthermore, the symmetry plane is constrained, which has an effect on the mechanical
strain. In general, the total strain is composed of three components given as:
(12)
where the three components on the right hand side of Eq. 12) refers to elastic, plastic and
thermal strains, respectively. The mechanical strain is the sum of the elastic and plastic
strains.
Isotropic linear elasticity has been assumed with temperature-dependent Young’s modulus
and Poisson’s ratio. To obtain the thermal strain field, isotropic thermal expansion is assumed
with a temperature-dependent expansion coefficient. To get the plastic strain, the Von Mises
yield criterion with an associate flow rule and linear kinematic hardening rule have been
used. Kinematic hardening is assumed to consider the thermal loading and unloading during
welding. Yield stress and Young’s modulus decrease exponentially with increasing
temperature to be near zero as temperature approaches melting point. Therefore, the filler
18
material flows through welding groove with almost free stress and strain. Fig. 11 illustrates
the temperature-dependant yield stress as the plastic strain of C-Mn and SUS304 is equal to
1% [16, 20].
Fig. 11. Yield strength of C-Mn steel and SUS304 corrosponding to 1% hardening [16,20].
In the mechanical analysis, initial boundary conditions are performed to just prevent lined
pipe motion. Due to the symmetry of model, the symmetric plane is fixed in the axial
direction, the Z-direction. On the lined pipe end, lateral and transversal restrictions are
applied at the lined pipe end in the X and Y-directions.
6. Results of the thermal analysis
The thermal cycles due to welding induce metallurgical changes in the FZ and HAZ. These
changes influence the final microstructure of the welded pipe and, therefore, the resistance of
the pipe to creep and fracture during service [17]. Furthermore, the HAZ is the most
vulnerable part of the pipe because of the accumulation of creep damage at the inter-critical
zone near the boundary of the HAZ where the peak temperature during welding is 800-
900°C, A1- A3, at which the austenitic transformation starts and finishes, respectively,
associated with reduction in volume. A1 is a cementite disappearance temperature whereas A3
is an α-ferrite disappearance temperature. During rapid cooling, an increase in the volume
happens in the HAZ because the austenite transforms to martensite. In this work, the
austenitic and martensitic transformations are neglected.
As a rule, the HAZ extends up to approximately 2-3 mm from the FZ edge. Consequently, the
peak temperatures on the integration points throughout the FE model can be related to the
0
100
200
300
400
500
600
0 200 400 600 800 1000 1200 1400 1600
Yield Stress (MPa)
Temperature (°C)
Yield stress of BM (C-Mn)
Yield stress of WM (C-Mn)
Yield stress of BM (SUS304)
Yield stress of WM (SUS304)
19
final mechanical and material features, such as residual stresses or phase transformations in
the FZ or HAZ. For that reason, the peak temperatures predicted by the FE model are the key
output of this analysis and are indicated in Fig. 12 to show the transient temperature
distributions during the welding process for the liner and the carbon steel backing. The
maximum temperature, which definitely exceeds the melting point, is attained in the middle
of the welding pool coloured in light grey in Fig. 12. The region located in the middle of
welding pool is the last one eligible to cool down from the melting temperature because there
is no direct contact with the air or other external surfaces.
(a)
(b)
Fig. 12. Maximum temperatures in (a) weld overlay and (b) girth welding (°C).
6.1. Thermal history during the weld overlay
The thermal history profiles were predicted at 6 locations along the axial direction. Points T1,
T2 and T3 are located on the CRA liner and T4, T5 and T6 are located on backing steel, as
sketched in Fig. 13. For each of these locations, a number of points at different
circumferential angles were considered.
20
Fig. 13. Schematic illustration showing temperature measurement positions on axial direction (mm).
Fig. 14 shows the temperature history for points T1, T2 and T3, at three different
circumferential angles 90°, 180° and 270° from the start/stop position. The plots represent the
change in temperature as a function of time. As expected, the temperatures start out at the
ambient temperature of 20°C. Once the heat source reaches the particular point, the
temperature rises very rapidly, especially at point 1 because it is located upon the WCL of
weld overlay welding where the torch is moving. The temperature at point T1 reaches a
maximum peak temperature of 2084°C as a balance between the flux and the heat losses. It
can be seen that the peak temperature stays on the point at a single time instant, due to the
constant velocity of the heat source. Once the heat source has passed point T1, a rapid drop in
the temperature occurs.
21
Fig. 14. Temperature histories during weld overlay at (a) 90°, (b) 180° and (c) 270°.
The peak temperature at point 2, located on the FZ boundary of the SUS304 welding, reaches
1405°C which is higher than the melting point of SUS304, 1365°C. As pointed out earlier,
Brickstad and Josefson [5] proved that an optimal power should lead to an extension for the
HAZ of up to approximately 2-3 mm from the FZ boundary, where the temperature reaches
800-900°C. In our work, point 3 is located after 2 mm from the FZ boundary where the
maximum temperature at this point is around 910°C. In other words, the thermal history
predicted by our FE model has met this optimal condition.
The combination effect of incorporating radiation and convection into the heat transfer on the
inner and outer surfaces justifies the rapid drop in the temperature in the FZ during cooling
where the radiation is the dominant mechanism of heat loss. Away from the weld, heat
transfer is dominated by conduction into remote regions of the base pipe.
The temperature history and peak temperature obtained at three different circumferential
angles of 90°, 180° and 270° from the start/stop position during weld overlay for the previous
axial points are practically identical. This shows that steady state, defined here as the
condition of self-similar moving of the temperature field, is well established before a
circumferential angle of 90° is reached.
6.2. Thermal history during girth welding
22
Once the overlay weld is terminated, the whole lined pipe cools down for 270 seconds to
reduce the maximum pipe temperature to just below 100°C, which is called inter-pass
temperature [5]. The thermal history of three different axial points located at the top surface
of the C-Mn backing steel is illustrated in Fig. 15. Point T4 is placed on the WCL of the girth
welding where the heat source is applied. Consequently, the maximum temperature, 2430°C,
on the outer pipe is attained at point T4. To check the fluidity is complete in the V-groove, all
the integration points there should reach the melting point of SIS2172 which is 1500°C. Point
T5, on the FZ boundary is the outmost point from the WCL, 3.2 mm, where the maximum
temperature is 1550°C. About 2.55 mm away from the FZ boundary, the temperature has
reached 953°C at point T6.
Fig. 15. Temperature histories during girth welding at (a) 90°, (b) 180° and (c) 270°.
Again the curves in Fig. 15 relative to points at the same axial distance and different
circumferential angles are almost identical, showing that steady state is well established
during moving the heat source around the pipe. The thermal results reported above for points
T1-T6 belong to a typical case, which will be referred to as case A (reference case) in the
next section.
7. Effect of welding factors
In general, the welding parameters play a decisive role in the weld pass quality and affect the
total shrinkage. In turn, this affects residual stresses and, ultimately, the probability of crack
23
occurrence [21]. Therefore, it is important to investigate the influence of two important
different welding parameters, interval time and welding speed, thermally and mechanically.
7.1. Effect of welding factors on thermal history
7.1.1. Effect of interval time on thermal history
As discussed in Section 6.2., an interval time is needed between the welding passes to reduce
the maximum temperature to an appropriate value, which is called inter-pass temperature.
Depending on the type of steel, the weld specifications normally specify the allowable range
for the inter-pass temperature between 100-180°C. A lower inter-pass than this range does
not have a considerable effect on residual stresses [5]. An inter-pass temperature over 250-
300°C can lead to excessive increase in residual stresses.
In case A, the lined pipe was allowed to cool down to around 100°C after the weld overlay
and before starting the girth weld as shown in Fig. 16. As can be seen from Fig. 16, the
thermal history distributions for all points except point T4, because not deposited yet, drop
down sharply from their peak temperatures accordingly because of cooling down to reach the
inter-pass temperature of 100°C in 270 s before all curves rapidly rise up once the girth
welding torch reaches the prescribed bead. For example, the temperature at point T5 drops
from 1118°C to 100°C before rising up very rapidly to 1550°C again.
Fig. 16 Thermal history at 270° for case A.
The result of changing the interval time is that, the less interval time is applied, the higher
temperature is obtained, as expected. In particular, reducing the interval time to 1 second
(case B) increases the minimum interval temperature to 244°C. At point T1, the maximum
0
500
1000
1500
2000
2500
150 250 350 450 550 650
Temperature (°C)
Time (s)
T1
T2
T3
T4
T5
T6
24
temperature during the girth welding increases somewhat from 1500°C in case A to 1545°C
in case B as shown in Fig. 17. In the same way, the thermal history at point T5 during interval
cooling also drops to 244°C before rapidly heading up to 1615°C.
Fig. 17 Thermal history at 270° with 1 second interval time in case B.
7.1.2. Effect of welding speed on thermal history
Welding speed plays a decisive role in determining the quality of weld (e.g. a lack of fusion
leads to porosity in welding surfaces which could be visually inspected) but there is also a
wide range of possible speeds that can be used [5, 22]. Hence, two extra cases have been
discussed by doubling (case C) and halving (case D) the welding speed, whereas the other
parameters are kept equal to those in Table 3.
In the first case, case C, the overlay and girth welding speeds are doubled to 2.66 and 12.5
mm/s, respectively. At point T2 at 270° from start/stop position, the maximum temperature
during overlay welding drops to a value lower than the melting point of C-Mn to reach
1046°C before heading down to 77°C during interval cooling. Likewise, the peak temperature
reaches 902°C at point T5 during girth welding which is lower than the melting point for C-
Mn as shown in Fig. 18.
0
500
1000
1500
2000
2500
150 250 350
Temperature (°C)
Time (s)
T1
T2
T3
T4
T5
T6
25
Fig. 18. Thermal history for a doubled welding speed, case C.
In the second case, case D, the welding speeds are halved to 0.665 and 3.125 mm/s during
overlay and girth welding, respectively. As expected, the temperature is beyond the melting
point of SUS304 and C-Mn in point T2 and T5 to be 1540°C and 1961°C as shown in Fig.
19, respectively.
Fig. 19. Thermal history with respect to reduce the speed to half , case D.
7.2. Effect of welding factors on residual stresses
0
500
1000
1500
2000
80 180 280 380 480
Temperature (°C)
Time (s)
T1
T2
T3
T4
T5
T6
0
500
1000
1500
2000
2500
3000
250 350 450 550 650 750 850 950
Temperature (°C)
Time (s)
T1
T2
T3
T4
T5
T6
26
7.2.1. Effect of interval time on residua stresses
The influence of reducing the interval time to 1 second on the residual stresses has been
discussed herein. Fig. 20(a)-(d) compares the simulated axial and hoop residual stress results
on the inner and outer surface at 270° central angle in case A (basic case) and case B (1
second interval time).
On the inner surface, Fig. 20(a) shows the axial residual stresses along the longitudinal
distance starting from the WCL. It is observed that case B is characterised by a larger value
of the axial stress on the WCL than its counterpart in case A, namely 624 and 517 MPa,
respectively. Furthermore, the extent of the tensile-stress zone in case A is narrower than that
in case B, namely 34.6 and 43.5 mm, respectively. Likewise, the maximum hoop residual
stress is located at the centreline of the weld overlay, Z=1.89 mm, whereas case B is
characterised by a somewhat larger tensile stress relatively to case A, namely 277 and 275
MPa, respectively. Starting from the WCL, the extent of the tensile-stress in case B is larger
than in case A, namely 49.16 and 39 mm, respectively, as shown in Fig. 20(b).
On the outer surface, the axial residual stresses are compressive in the FZ and HAZ, whereas
the maximum magnitude of compressive stress is located at the WCL for both cases as
depicted in Fig. 20(c). The maximum compressive stress in case B is larger than its
counterpart in case A, namely -511 and -464 MPa, respectively. Moreover, case B is
characterised by a slightly larger extent of the compressive-stress zone than case A, namely
26.9 and 20.75 mm, respectively. Returning to the hoop stress on the outer surface, in both
cases the stress distributions have a waved shape, whereas case A is characterised to some
extent by larger absolute residual stress values than in case B along the longitudinal distance,
as portrayed in Fig. 20(d).
-100
100
200
300
400
500
600
700
050 100 150 200
Axial Stress (MPa)
Distance from WCL (mm)
Axial-Case A
Axial-Case B
-600
-500
-400
-300
-200
-100
100
200
050 100 150 200
Axial Stress (MPa)
Distance from WCL (mm)
Axial-Case A
Axial-Case B
27
(a)
(c)
(b)
(d)
Fig. 20 Comparison of residual stresses at 270° central angle in case A and case B: (a) axial stress distributions
on the inner surface, (b) hoop stress distributions on the inner surface, (c) axial stress distributions on the outer
surface and (d) hoop stress distributions on the outer surface.
7.2.2. Effect of welding speed on residual stresses
The effects on residual stresses of the doubled and halved welding speed with respect to the
typical welding speed in case A have been investigated in this section. Fig. 21(a)-(d) portrays
the comparison between the numerically computed axial and hoop residual stresses on the
inner and outer surface at 270° central angle in case A (basic case), case C (doubled welding
speed) and case D (halved welding speed).
On the inner surface, the axial residual stress distributions for the three cases along the
longitudinal direction starting from the WCL are plotted in Fig. 21(a). It could be seen that
reducing the welding speed to half (case D) of a typical one (case A) leads to significantly
increase in the axial residual stress in the WCL, 832 MPa. The axial residual stress
magnitudes in the FZ and its vicinity, Z ≤ 20.7mm, in cases A and C are close to some extent.
The extent of the tensile-stress zone in case D is the longest one comparing to other cases
whereas case C has the narrowest extent of the tensile-stress zone. Similarly, as for the hoop
residual stresses, Fig. 21(b) shows that case D has the largest magnitude of hoop stress at the
WCL, 400 MPa. Up to a distance of about 18mm from the WCL, a good correlation between
the results of cases A and C is found. Also, the extent of the tensile-stress zone of case A is
between the longest one (case D) and the narrowest one (case C).
On the outer surface, the axial residual stresses are compressive in the FZ and HAZ as shown
in Fig. 21(c). However, the largest compressive stress at the WCL, equal to 578 MPa, is
-200
-100
100
200
300
400
050 100 150 200
Hoop Stress (MPa)
Distance from WCL (mm)
Hoop-CaseA
Hoop-CaseB
-300
-200
-100
100
200
300
400
050 100 150 200
Hoop Stress (MPa)
Distance from WCL (mm)
Hoop-CaseA
Hoop-Case B
28
found in case D. Up to a distance of 13.75 mm from the WCL, the compressive stresses of
case A are slightly lower than their counterparts in case C. Furthermore, the compressive-
stress zone in case C is slightly longer than in cases A and D. Returning to the hoop stress on
the outer surface, in all cases the stress distributions have a waved shape, whereas case D has
the largest absolute residual hoop stress value at the WCL in comparison with its counterparts
in cases A and C, as clarified in Fig. 21(d).
(a)
(c)
(b)
(d)
-200
200
400
600
800
1000
050 100 150 200
Axial Stress (MPa)
Distance from WCL (mm)
Axial-Case A
Axial-Case C
Axial-Case D
-700
-600
-500
-400
-300
-200
-100
100
200
050 100 150 200
Axial Stress (MPa)
Distance from WCL (mm)
Axial-Case A
Axial-Case C
Axial-Case D
-200
-100
100
200
300
400
500
050 100 150 200
Hoop Stress (MPa)
Distance from WCL (mm)
Hoop-CaseA
Hoop-Case C
Hoop-Case D
-300
-200
-100
100
200
300
400
050 100 150 200
Hoop Stress (MPa)
Distance from WCL (mm)
Hoop-Case A
Hoop-Case C
Hoop-Case D
29
Fig. 21Comparison of residual stresses at 270° central angle in case A, C and D: (a) axial stress distributions on
the inner surface, (b) hoop stress distributions on the inner surface, (c) axial stress distributions on the outer
surface and (d) hoop stress distributions on the outer surface.
9. Conclusions
In this study, a 3-D FE model has been developed to predict temperature fields and residual
stress distributions induced by two circumferential welds for a lined-pipe, namely the overlay
welding of the stainless steel liner with the C-Mn steel pipe and the girth welding of two
segments of pipe. Two user-subroutines have been coded to model a distributed power
density of the moving welding torch and to use a non-linear heat transfer coefficient
accounting for both radiation and convection. The temperature and stress variations in space
and time have been reported in both the axial and the circumferential directions. The model
procedure has been validated against experimental results in the literature related to two
different cases involving the welding of a C-Mn pipe and a stainless steel pipe. Furthermore,
a sensitivity analysis to determine the influence of the cooling time between weld overlay and
girth welding and of the welding speed has been conducted thermally and mechanically.
This paper has shown that the welding parameters, namely interval time and welding speed,
play a vital role in the weld pass quality. According to the results in this work, we can draw
the following specific conclusions.
(1) Based on the thermal results, it is clear that the temperature distribution around the
heat source reaches steady state when the welding torch moves to fill the weld overlay
and girth welding. From the outcome of our investigation it is possible to conclude
that the temperature history is not sensitive to the variations of the circumferential
angles, except for a small angle close to the starting point, associated with a short
initial transient.
(2) Increasing the inter-pass temperature leads to a significant increase in the temperature
at the boundary of FZ. As expected, the less interval time is applied, the higher
temperature is obtained during the girth welding.
(3) Doubling or halving the weld overlay and girth welding speeds leads to significantly
decrease or increase the width of the FZ, respectively.
(4) The largest tensile and compressive axial residual stresses occur at the FZ and its
vicinity on the inner and outer surfaces, respectively. Beyond the FZ and its vicinity,
compressive and tensile residual stresses are produced on the inner and outer surfaces
30
of lined pipe, respectively. The hoop residual stress results are affected to some extent
by the axial residual stress results.
(5) Minimizing the inter-pass time to 1 second leads to a reasonable increase in the
absolute magnitudes of axial residual stresses at the WCL about 21% on the inner
surface and 10% on the outer surface.
(6) Halving the weld overlay and girth welding speeds has more influence in increasing
the absolute values of hoop and axial residual stresses at the WCL whereas doubling
speed does not have that effect on the results at the FZ according to the basic case,
case A.
References
[1] De Koning, A. and Nakasugi, H. (2004) 'Li Ping (2004),“TFP and TFT Back in Town (Tight Fit CRA Lined
Pipe and Tubing)”', Stainless Steel World, , pp. 53-61.
[2] Deng, D. and Murakawa, H. (2006) 'Numerical simulation of temperature field and residual stress in multi-
pass welds in stainless steel pipe and comparison with experimental measurements', Computational materials
science, 37(3), pp. 269-277.
[3] Goldak, J., Chakravarti, A. and Bibby, M. (1984) 'A new finite element model for welding heat sources',
Metallurgical transactions B, 15(2), pp. 299-305.
[4] Goldak, J.A. and Akhlaghi, M. (2006) Computational welding mechanics. Springer Science & Business
Media.
[5] Brickstad, B. and Josefson, B. (1998) 'A parametric study of residual stresses in multi-pass butt-welded
stainless steel pipes', International Journal of Pressure Vessels and Piping, 75(1), pp. 11-25.
[6] Karlsson, R.I. and Josefson B.L. (1990) 'Three-dimensional finite element analysis of temperatures and
stresses in a single-pass butt-welded pipe', Journal of Pressure Vessel Technology, 112(1), pp.76–84.
[7] Deng, D., Kiyoshima, S., Ogawa, K., Yanagida, N. and Saito, K. (2011) 'Predicting welding residual stresses
in a dissimilar metal girth welded pipe using 3D finite element model with a simplified heat source', Nuclear
Engineering and Design, 241(1), pp. 46-54.
[8] Attarha, M. and Sattari-Far, I. (2011) 'Study on welding temperature distribution in thin welded plates
through experimental measurements and finite element simulation', Journal of Materials Processing
Technology, 211(4), pp. 688-694.
[9] Lee, C., Chang, K. and Park, J. (2013) 'Three-dimensional finite element analysis of residual stresses in
dissimilar steel pipe welds', Nuclear Engineering and Design, 256, pp. 160-168.
[10] Pavelic, V., Tanbakuchi, R., Uyehara, O. and Myers, P. (1969) 'Experimental and computed temperature
histories in gas tungsten-arc welding of thin plates', WELD J, 48(7), pp. 295.
[11] Goldak, J., Chakravarti, A. and Bibby, M. (1984) 'A new finite element model for welding heat sources',
Metallurgical transactions B, 15(2), pp. 299-305.
[12] Goldak, J., McDill, M., Oddy, A., House, R., Chi, X. and Bibby, M. (1986) 'Computational heat transfer for
weld mechanics', Proceeding of International Conference on Trends in Welding Research. , 15-20.
31
[13] Fachinotti, V.D., Anca, A.A. and Cardona, A. (2011) 'Analytical solutions of the thermal field induced by
moving doubleellipsoidal and doubleelliptical heat sources in a semiinfinite body', International Journal for
Numerical Methods in Biomedical Engineering, 27(4), pp. 595-607.
[14] Dassault Systèmes, ‘ABAQUS Example Problems Guide, Volume I: Static and Dynamic Analyses’, USA,
1.0-1-2.4.1-13, 2014.
[15] Akbari, D. and Sattari-Far, I. (2009) 'Effect of the welding heat input on residual stresses in butt-welds of
dissimilar pipe joints', International Journal of Pressure Vessels and Piping, 86(11), pp. 769-776..
[16] Malik, A.M., Qureshi, E.M., Dar, N.U. and Khan, I. (2008) 'Analysis of circumferentially arc welded thin-
walled cylinders to investigate the residual stress fields', Thin-Walled Structures, 46(12), pp. 1391-1401.
[17] Yaghi, A., Tanner, D., Hyde, T., Becker, A. and Sun, W. (2011) 'Abaqus Thermal Analysis of the Fusion
Welding of a P92 Steel Pipe', SIMULIA Customer Conference, Barcelona. , 17-19.
[18] Deus, A.M. and Vilar, R., 1996. One-dimensional thermal model including the dependence of absorptivity
on temperature using Hagen-Rubens equation. In Laser Processing: Surface Treatment and Film Deposition
(pp. 195-201). Springer Netherlands.
[19] Noyan, I.C., Huang, T.C. and York, B.R., 1995. Residual stress/strain analysis in thin films by X-ray
diffraction. Critical Reviews in Solid State and Material Sciences, 20(2), pp.125-177.
[20] Deng, D., Murakawa, H. and Liang, W. (2008) 'Numerical and experimental investigations on welding
residual stress in multi-pass butt-welded austenitic stainless steel pipe', Computational Materials Science, 42(2),
pp. 234-244.
[21] Aloraier, A., Almazrouee, A., Shehata, T. and Price, J.W. (2012) 'Role of welding parameters using the flux
cored arc welding process of low alloy steels on bead geometry and mechanical properties', Journal of materials
engineering and performance, 21(4), pp. 540-547.
[22] Holloway, G., Zhang, Z. and Marshall, A. (2008) 'Weld metals for USC power plant: properties of T/P92
CrMo weld metals for ultra super critical (USC) power plant', Metrode Products Ltd, UK,
