Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

International Journal of Pressure Vessels and Piping 87 (2010) 424e432

Contents lists available at ScienceDirect

International Journal of Pressure Vessels and Piping


journal homepage: www.elsevier.com/locate/ijpvp

Numerical simulation and experimental investigation of temperature distribution


in the circumferentially butt GTAW of Incoloy 800H pipes
H. Purmohamad, A. Kermanpur*, M. Shamanian
Department of Materials Engineering, Isfahan University of Technology, Isfahan 84156-83111, Iran

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 27 January 2008
Received in revised form
25 March 2010
Accepted 26 March 2010

The multi-pass circumferential butt GTAW process of Incoloy 800H pipes was modelled with the FEM in
3D. The element birth and death technique was used for the addition of ller material. Goldak model was
used to simulate the distribution of arc heat source. The validation of the simulation model was carried
out based on the precise temperature measurements within the HAZ of the welds by thermocouples as
well as metallographic characterisation of the cross section of the welds. A good agreement was found
between the simulation and experimental results for both thermal eld and weld zone shape. The
present model showed that increasing the heat input resulted in a wider weld zone as well as a higher
HAZ peak temperature. These effects were related to the net heat input and not to either welding current
or welding speed, individually. The developed simulation model is a useful tool to investigate the
welding thermal regime and the weld pool prole.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
The circumferential, butt-weld is a common type of joint in
piping systems which is often constructed of several weld passes.
Incoloy 800H is widely used in chemical and petrochemical
industry, power plants and industrial furnaces in which pipes,
headers and thin sheets are butt-welded by utilising Gas Tungsten
Arc Welding (GTAW) process [1]. Due to intense concentration of
heat in the GTAW process, the regions near the weld line undergo
severe thermal cycles. The temperature distribution within material
during the welding process affects directly the mechanical properties, distortion and residual stresses that will be present in the
material after cooling to room temperature. Very limited experimental data regarding temperature distribution during the multipass welding of pipes is available in the literatures.
In current industrial practice, welding processes are developed
largely based on trial and error experiments incorporating engineers knowledge and experience of previous similar designs.
Computer simulation tools based on nite element method (FEM)
are very useful to predict (a) welding distortions and residual
stresses at the early stage of product design [2,3], (b) formation of
defects and weldability [4], and (c) welding process development
[5,6]. However, the complexity of welding processes and the

* Corresponding author. Tel.: 98 (0) 311 3915738; fax: 98 (0) 311 3912752.
E-mail address: ahmad_k@cc.iut.ac.ir (A. Kermanpur).
0308-0161/$ e see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2010.03.023

complex geometry of real engineering components have made the


prediction of defects, welding distortions and residual stresses
a very difcult task. The metallurgical and mechanical consequences of phase transformations during welding as well as the
heat input distribution and the addition of ller material into the
weld pool should all be considered in the modelling procedure.
The research on welding heat source models dates back to early
1940s. Rosenthal [7] rst proposed a mathematical model of the
moving heat source under the assumptions of quasi-stationary
state and concentrated point heating in three dimensions (3D). In
the late 1960s, Pavelic et al. suggested a circular disc heat source
model with Gaussian distribution of heat ux on the surface of the
workpiece [8]. Goldak et al. further developed a double ellipsoidal
power density distribution of heat source model below the welding
arc, which can accurately simulate different types of welding
processes with shallow and deep penetration [9]. These heat source
models and some simplied models have been widely used in
welding simulation for prediction of the distortions and residual
stresses [2,3, and 10]. Recently, Kermanpur et al. presented a 3D
thermal simulation model based on the FEM in which both surface
and volumetric heat ux distributions were considered separately
for the GTAW process [11].
This paper presents a 3D thermal simulation model for the
GTAW process of circumferentially butt-welded Incoloy 800H
petrochemical pipes by using FE transient heat transfer analysis. A
moving heat source model based on the Goldak model was applied
to the thermal model in a discontinuous manner, assuming
a constant welding speed. The computer programme developed

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

425

Table 1
Chemical composition of Incoloy 800H pipes (base) and Inconel 82 (ERNiCr-3 ller material) (%wt.).
Material

Fe

Ni

Cr

Ti

Al

Mn

Si

Nb

Incoloy 800H
Inconel 82

39.5
3.0

30.0e35.0
67.0

19.0e23.0
18.0e22.0

0.15e0.6a
0.75

0.15e0.6a
0.3

1.5
2.5e3.5

1.0
0.5

e
2.0e3.0

0.05e0.1
0.1

0.015
0.015

e
0.03

%(Al Ti) 0.85e1.20.

was veried by thermocouple measurements recorded during the


GTAW of several pipes. The transient temperature distribution in
the workpiece as well as the weld pool shapes were predicted
under different process conditions. The effect of the heat input on
temperature distribution was investigated.
2. Experimental procedure
2.1. Welding experiments
The iron-based Incoloy 800H pipes were circumferentially buttwelded. Table 1 shows the chemical composition of Incoloy 800H.
Pipes with inner diameter of 60.3 mm and thickness of 5.54 mm
were used [12]. The single-V groove weld joint design was prepared
as shown in Fig. 1. The ller material ERNiCr-3 with chemical
composition shown in Table 1 was used. To obtain a quality weld,
correct joint preparation and accurate tacking are necessary. Faulty
tacking will cause defects in the nal welding. So, four tack-welds
were used to maintain the uniform gap of 3 mm. The sequence of
tacking and the pass schedules are shown in Fig. 2 and Table 3. The
pipes were welded in two passes without preheating. The welding
conditions for two passes are shown in Table 2. To protect the inside
face of the rst welding pass, one end of two tack welded pipes was
obstructed with a piece of scotch tape and argon inert gas was
blown from another end side of specimen.
2.2. Temperature measurement tests
In order to verify the present thermal model, temperature
measurements were conducted at several points of the pipes with
different distance from the welding line. To record the measured
thermal cycles, a personal computer with an integrated PCI-773
measuring card (Eagle Technology) and Lab View 5.1 software was
used. Fig. 3a and b shows the real picture and the schematic of the
test specimen along with the thermocouple locations, respectively.
Five K-type thermocouples with diameter 0.3 mm were installed at
the locations 3, 7, 11, 15 and 19 mm distance from the groove edge
as shown in Fig. 3b.

Fig. 1. Geometry of the single-V groove two-pass weld joint design (not to scale).

3. Model theory
3.1. Governing equations
The spatial and temporal temperature distribution T(x, y, z, t)
satises the following differential equation for 3D heat conduction:









v
vT
v
vT
v
vT
vT
vT
kx

ky

kz
Q_ rCp
v
vx
vx
vy
vy
vz
vz
vt
vx

(1)

where kx, ky and kz are thermal conductivity in x, y and z directions


(Wm1K1), Q_ power generation per unit volume in the domain
(Wm3), r density (kgm3), Cp specic heat (Jkg1K1), and v
relative velocity of workpiece (here in x direction) (ms1). The
initial condition is:

Tx; y; z; 0 T0 forx; y; z

(2)

The essential boundary condition is either Dirichlet:

Tx; y; z; t T0

(3)

or natural:

kn



vT
 q hc T  TN s3F T 4  Tr4 0
vn

(4)

where kn is thermal conductivity normal to the surface, q is the heat


ux (Wm2), hc is the convective heat transfer coefcient
(Wm2K1), s is StefaneBoltzmann constant for radiation
(5.67  108 Wm2K4), 3 is the emissivity, F is the conguration
factor, TN is the surrounding temperature and Tr is the temperature

Fig. 2. The pass schedules of the welding experiments and sequence of taking.

426

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

Table 2
The welding conditions of the pipes.
Specimen No.

Pass No.

I (A)

V (V)

v
(mm/s)

Net Heat Input


(J/mm/s)

Polarity

1
2
1
2
1
2

80
80
100
100
120
120

10
10
10
10
10
10

0.77
0.47
1.03
0.79
1.41
1.11

1038
1702
970
1266
851
1081

DCEN
DCEN
DCEN
DCEN
DCEN
DCEN

2
3

of the radiation heat source. The inclusion of temperature-dependent thermo-physical properties along with a radiation term in the
above boundary condition makes this type of analysis highly
nonlinear. However, the radiation heat exchange can be treated as
convection, expressed by the following equivalent heat transfer
coefcient:



hr s3F T 2 Tr2 T Tr

(5)

in this work, a constant equivalent heat transfer coefcient was


assigned over the weld bead to treat the radiation boundary
condition.
3.2. Modelling the moving heat source
In the present analysis, a moving heat source based on the
Goldak model was developed to represent the heat generated by
the torch in the GTAW process. A feature of this model is that
a double ellipsoid energy is distributed over a volume described by
two ellipsoid regions, one ahead and one after the weld torch
(Fig. 4). The power density distribution inside the front quadrant
becomes [9]:

!
p
6 3Qff
3x2 3y2 3z vs  t2
pexp
qf x;y;z
2
a2
b
abcf p p
c2f

results of the weld pool shape. The following values were used in
the present investigation: a b cr 3 mm and cf 9 mm.

(6)
3.3. Numerical simulation

similarly, for the rear quadrant of the source, the power density
distribution inside the ellipsoid becomes:

qr x;y;z

!
p
6 3Qfr
3x2 3y2 3z vs  t2
pexp

a2
b2
c2r
abcr p p

Fig. 3. (a) real picture and (b) schematic of the test specimen along with the thermocouple locations (dimensions in mm but not to scale).

(7)

The GTAW process was simulated with the FEM using the ANSYS
commercial code [13]. Due to the symmetry, only half of the welded
pipes were modelled. In order to take into account the effect of
thermal convection within the weld pool, thermal conductivity of
the alloy above the solidus temperature was ctitiously increased
by the order of 2 [14]. The boundary conditions allow for both

where Q hVI/y is the total net heat input, h is the arc efciency
that is assumed 0.6 for GTAW process, V is arc voltage, I is the arc
current, y is the welding speed, and s is the incubation time. The
energy deposited in the front and rear quadrants are dened by the
fractions ff and fr where fffr 2. Constants a, b, cf and cr are the
shape parameters determining the shape and size of the weld pool.
These constants were estimated according to the experimental

Table 3
The pass sequence parameters used in the welding experiments. (H and P stand for
half and pass, respectively).
Specimen No.

Pass
number

1
2
1
2
1
2

2
3

Welding
current, A
80
100
120

Time, s
H1

H1eH2

H2

P1eP2

100
200
75
120
55
85

25
25
25
25
25
25

100
200
75
120
55
85

35
35
35
35
35
35

Fig. 4. The Goldak heat source distribution model.

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

427

Fig. 7. Mesh sensitivity analysis based on the maximum temperature of the pipe.

Fig. 5. Thermo-physical properties of the Incoloy 800H alloy as a function of


temperature used in the present simulation.

the boundary condition expressed by Eq. (8) was applied for all
outside free boundaries of the pipe including the successive
boundaries created for each new weld pass. For the inside free
boundaries, because of blowing argon gas, heat transfer coefcient
was assumed as a constant value h2 20 (Wm2K1). The
temperature-dependent thermo-physical properties of the Incoloy
800H used in the simulation are shown in Fig. 5 [16].
The eight-node 3D thermal element SOLID70 with a single
degree of freedom (temperature) was used for the bulk material
(both the base and weld zones). The FE mesh is shown in Fig. 6. It
consists of 5700 element and 6720 nodes. The used mesh was
based on a mesh sensitivity analysis performed for several mesh
renements. The maximum temperature of the pipe achieved

during the welding process was considered as the basis of


comparing different mesh. Fig. 7 shows that the maximum
temperature is not further increased when the element number is
more than 5700. This mesh (shown by an arrow in Fig. 7) was then
considered for simulations in order to get mesh-independent
results with a reasonable CPU time.
The welding process of the pipes was done in two passes. The
power density distribution functions (Eqs. 6 and 7) were applied on
the top side of the weld pool in the rst and second passes. It was
assumed that the GTAW torch was attached to the origin of the
moving coordinate system. The coordinate system was therefore
translated through the descretised domain based on the welding
speed and the mesh size in a discontinuous manner. The element
birth and death technique was used to simulate the weld ller
material variation with time in the multi-pass butt-welded joints
[13]. All elements must be created, including those weld llers to be
born in later stages of the analysis. The method deactivates these
elements by multiplying their stiffness by a severe reduction factor
(e.g. 106). Although zeroed out of the load vector, element loads
associated with deactivated elements still appear in element-load
list. Similarly, mass, damping, specic heat, and other such effects
are set to zero for deactivated elements. The mass and energy of
deactivated elements are excluded from the summations of the
model. When elements are born, they are not actually added to the
model, but are simply reactivated: its stiffness mass, element loads,
etc. return to their full original values.

Fig. 6. Finite element mesh of the pipe.

Fig. 8. Thermal welding cycles measured during GTAW at different distances from the
weld bead: TC1: 3, TC2: 7, TC3: 11, TC4: 15, and TC5: 19 mm.

convection and radiation. Radiation losses are dominating for


higher temperatures near the weld and convection losses for lower
temperatures away from the arc. A combined boundary condition
was used here, which accounted for both convection and radiation.
The resulting expression for the temperature-dependent heat
transfer coefcient h1 is given by Eq. (8) [15].



0 < T < 500 C
h1 0:0668T Wm2 K 1


500 C < T
h1 0:231T  82:1 Wm2 K 1

(8)

428

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

Fig. 9. Comparison between the experimental thermal cycles for different thermocouples with the corresponding simulation results for the welding current of 100 A: (a) TC1, (b)
TC2, (c) TC3, (d) TC4, and (e) TC5.

4. Results and discussion


4.1. Experimental results
Temperature measurements at different locations were performed to verify the computed temperature eld. Typical thermal
welding cycles for I 100 A and V 10 V are shown in Fig. 8. The
plots show time-dependent thermal cycles measured at different
locations in the workpiece shown in Fig. 3. These measurements
were performed along the pipe axis, 3 mm beneath the surface
based on the path schedule shown in Fig. 2. As it can be seen from
Fig. 8, all thermocouples have experienced their maximum
temperatures when the torch passed the thermocouple line.
Therefore the two rst peaks belong to the rst pass and the two
other peaks belong to the second pass. The maximum measured
values have decreased with increasing the distance from the weld
bead. As it was already suggested, the measurements also show
that the maximum temperature for the second pass is greater than
the rst pass due to the thermal history of the rst pass.

Fig. 10. Comparison between the experimental thermal cycle for thermocouple TC4
with the corresponding simulation results for the welding current of 120 A when using
the Gaussian distribution function [11].

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

429

Fig. 11. Comparison between the weld pool shape in the longitudinal cross section of the pipe welded under 100 A: (a) experimental macrograph, (b) simulated weld pool in the
rst and second passes (left and right, respectively). Note that the colour counters in the rst and second passes in the left side of Fig. 11b are not real (see text).

Fig. 12. The 3D simulated temperature elds of the circumferentially butt-welded pipe at different times after the start of welding: (a) t 11.4 s, (b) t 99.8 s, (c) t 209.4 s and (d)
t 353.4 s.

430

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

Fig. 13. The effect of heat source efciency on the temperature cycle of TC1.

4.2. Experimental validation of the simulation model


The present simulation model was validated against the
experimentally measured thermal cycles as well as the shape and
size of the weld zone determined from the longitudinal section of
the pipes. Fig. 9 compares the experimental thermal cycles of
different thermocouples with the corresponding simulation
results for the welding current 100 A. It can be seen that a good
agreement was achieved for the temperature eld throughout the
whole welding process. This shows that the Goldak model is
a useful representation of the heat input distribution function of
the GTAW process. The precision of the above simulation results
would be better understood when they compares with the other
distribution functions. As an example, Fig. 10 shows the thermal
cycle at thermocouple TC4 when using a Gaussian distribution
function instead of the Goldak model [11]. As it can be seen, the
difference between the simulation results and the experimental
one will increase during the welding process. Nevertheless, the
disagreement shown in Fig. 9 for the second pass of all
thermocouples during the cooling stage is probably due to the

approximations of the convective heat transfer coefcient as well


as the thermo-physical properties of the pipe.
Fig. 11a shows a macrograph of the longitudinal cross section of
the pipe welded under 100 A. The depth and diameter of the weld
zone can be measured from this macrograph. Simulation predictions of the weld zone are also shown in Fig. 11b for the rst and
second passes (left and right sides, respectively) of the welding
route. The colour contours represent the positions that have been
heated up above the solidus temperature (e.g. 1357  C). It is shown
that a good agreement can be seen between the two weld zone
proles. It should be noted that since an element birth and death
technique was used in this simulation, the colour counters corresponding to both the rst and second passes are presented in the
left side of Fig. 11b (e.g. for the rst pass). This is because during the
rst pass, all elements corresponding to the second pass were killed, but they can still be viewed due to the continuity of the results
in the rst pass. This will not signicantly affect the simulation
results.
4.3. 3D simulation results
The simulated 3D temperature elds of the circumferential
welded pipe at different times of 11.4, 99.8, 209.4 and 353.4 s after
the start of welding are illustrated in Fig. 12. Note that only half of
the workpiece is modelled and the isotherms and the welding bead
are symmetrical with respect to the welding path. As it can be seen,
the transient temperature distribution varies with respect to the
moving arc. Because of the locally concentrated heat source, the
temperature near the weld bead and heat-affected zone rapidly
changes with the distance from the centre of the heat source.
Therefore, the highest temperature is limited to the domain of the
heat source from which lower temperature zones fan out.
4.4. Effects of process parameters
4.4.1. Heat source efciency
A wide range of variation has been reported for the value of
heat source efciency for the GTAW process [17]. Fig. 13 compares

Fig. 14. Comparison between the weld pool shape in the longitudinal cross section of the pipe welded under 80 A: (a) experimental macrograph, (b) simulated weld pool shape.

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

431

Fig. 15. Comparison between the weld pool shape in the longitudinal cross section of the pipe welded under 120 A: (a) experimental macrograph, (b) simulated weld pool shape.

the temperature cycle of TC1 under different heat source efciency values of 40, 50, 60, and 70% with the experimental
measurement. It can be seen that the heat source efciency of
approximately 60% shows the best agreement with the experiments carried out in this work. All simulation in this work was
carried out using this efciency. However, it should be noted that
the present simulation was done with the assumption of using
a perpendicularly placed torch. This condition was aimed at
during the welding experiments.
4.4.2. Heat input
Effect of the heat input, as one the most important parameters in
the GTAW process, can be easily studied using the present simulation model. In this study, only the welding current was changed to
alter the heat input. Fig. 14a and b shows the macrograph and the
simulated results of the longitudinal cross section of the pipe
welded under 80 A (i.e. 1702 J/mm/s), respectively. The corresponding results for the welding current of 120 A (i.e. 1081 J/mm/s)
are also shown in Fig. 15a and b. Values of the weld bead width
(WBW), distance between the centre line and the edge of weld

Fig. 16. The effect of heat input on the weld bead width (WBW) parameter for both the
experimental and simulation results.

bead, for three samples welded with different heat inputs were
measured. The experimental and simulated values of the WBW
parameter are shown in Fig. 16. Both the experimental and simulation results show that the WBW parameter increases with
increasing the welding heat input. As it can be seen, increasing the
heat input causes the increase of the weld pool dimensions (Fig. 15
vs. Fig. 14). One should note that although both the welding current
and welding speed in Fig. 15 is more than that of Fig. 14 (see Table
2), but the net heat input in Fig. 15 with the current of 120 A is lower
than that of Fig. 14 with the current of 80 A. This would cause
a lower WBW parameter for the higher welding current of 120 A (e.
g. a more concentrated weld zone). These results suggest that the
net heat input value is more important that of either welding
current or welding speed individually.
Fig. 17 shows the maximum temperature (Tpeak) at TC1 thermocouple location on the pipe welded under different heat input
values in two passes. It is shown in this gure that the value of Tpeak
increases with the increasing of the heat input. Also, as the welding
speed of the pass 1 is higher than that of the pass 2 (see Table 2), the

Fig. 17. The effect of heat input on the maximum temperature at thermocouple TC1 in
two passes.

432

H. Purmohamad et al. / International Journal of Pressure Vessels and Piping 87 (2010) 424e432

values of Tpeak for the pass 1 is lower. The lower welding speed of
the pass 2 would also cause a less sensitivity for the variation of the
Tpeak with the heat input.
The above simulation results clearly show that using the 3D
simulation model developed in this work, it is possible to investigate the effect of different process parameters involved in the
GTAW process of the circumferentially butt-welded pipes as
a useful tool to control and optimise the thermal regime and the
weld pool prole of the weldments. The extension of this work is in
progress to develop a thermo-mechanical model in order to
simulate the thermal stress generated during the process and
dimensional accuracy of the welded pipes [18].
5. Conclusions
A 3D nite element model of the heat ow during the GTAW
process of the circumferentially butt-welded pipes was developed
in this study. The Goldak model was used for the heat input
distribution of the welding torch on the pipe. Simulation results
compared to experiments showed that the Goldak model would
represent more precise results than the Gaussian model. Using this
model, a parameter study performed for the process variables such
as the welding current (heat input), welding efciency and welding
velocity. The model showed that increasing the heat input
decreased the weld bead width parameter. This would also increase
the peak temperature of the pipe with a decreasing effect for the
lower welding speeds.
Acknowledgements
The authors would like to acknowledge the Oil Renery
Company of Isfahan, Iran for their partial nancial support and
permission for publishing this paper.

References
[1] Lippold JC. An investigation of weld cracking in alloy 800. Weld J 1984;63
(3):91se103s.
[2] Bachorski A, Painter MJ, Smailes AJ, Wahab MA. Finite element prediction of
distortion during gas metal arc welding using the shrinkage volume approach.
J Mater Process Tech 1999;92-93:405e9.
[3] Chang PH, Teng TL. Numerical and experimental investigation on the residual
stresses of the butt-welded joints. Comp Mater Sci 2004;29:511e22.
[4] Dye D, Hunziker O, Reed RC. Numerical analysis of the weldability of superalloys. Acta Mater 2001;49:683e97.
[5] Mackerle J. Finite element analysis and simulation of welding e an addendum:
a bibliography (1996-2001). Model Simul Mater Sci 1996;4:501e33.
[6] Grey D, Long H, Maropoulos P. Effects of welding speed, energy input and heat
source distribution on temperature variations in butt joint welding. J Mater
Process Tech 2005;1678(2e3):393e401.
[7] Rosenthal D. Mathematical theory of heat distribution during welding and
cutting. Weld J 1941;20:220se34s.
[8] Pavelic V, Tanbakuchi R, Uyehara OA, Myers PS. Experimental and computed
temperature histories in gas tungsten-arc welding of thin plates. Weld J
1969;48:295se305s.
[9] Goldak J, Chakravarti A, Bibby M. A new nite element model for welding heat
source. Metal Trans B 1984;15:299e305.
[10] De A, Maiti SK, Walsh C, Bhadeshia HDKH. Finite element modelling of laser
spot welding. Sci Technol Weld Joi 2003;8(5):377e84.
[11] Kermanpur A, Shamanian M, Esfahani Yeganeh V. Three-dimensional thermal
simulation and experimental investigation of GTAW circumferentially buttwelded Incoloy 800H pipes. J Mater Process Tech 2008;199:295e303.
[12] ASTM B407-87/ASME SB407, Specication for nickel-iron-chromium alloy
seamless pipe and tube.
[13] ANSYS users manual, Version 10, Swanson Analysis Systems, Inc.
[14] Lee CH, Chang KH. Three-dimensional nite element simulation of residual
stresses in circumferential welds of steel pipe including pipe diameter effects.
Available online. Mater Sci Eng A; 10 October 2007.
[15] Brickstad B, Josefson BL. A parametric study of residual stresses in multi-pass
butt-welded stainless steel pipes. Int J Pres Ves Pip 1998;75:11e25.
[16] Little GH, Kamtekar AG. The effect of thermal properties and welding efciency on transient temperatures during welding. Comput Struct
1998;68:157e65.
[17] Kou S. Welding metallurgy. Willey; 2002.
[18] Purmohamad H, Kermanpur A, Shamanian M. Numerical simulation and
experimental investigation of residual stresses in the circumferential butt
GTAW of Incoloy 800H pipes. J Mater Eng Perfor 2010;19:13e21.

You might also like