Influence of a welding sequence on the welding

INSTITUTE OF PHYSICS PUBLISHING MODELLING AND SIMULATION IN MATERIALS SCIENCE AND ENGINEERING
Modelling Simul. Mater. Sci. Eng. 13 (2005) 553–565 doi:10.1088/0965-0393/13/4/006
Influence of a welding sequence on the welding
residual stress of a thick plate
S D Ji, H Y Fang, X S Liu and Q G Meng
State Key Laboratory of Advanced Welding Technology Production, Harbin Institute of
Technology, Harbin 150001, People’s Republic of China
Received 26 October 2004, in final form 17 March 2005
Published 27 April 2005
Online at stacks.iop.org/MSMSE/13/553
Abstract
The residual stress in one-groove welding using an ellipsoidal heat source is
analysed. The results show that it can ameliorate the residual stress distribution
and greatly decrease the peak value of residual stress after welding if the
converse welding method is adopted between adjacent layers in a multi-layer
weld, or between adjacent beads in every layer. Moreover, the numerical
simulation results of the double V-groove thick plate welding model show
that the residual tensile stress appears on the weld and nearby, the residual
compressive stress appears on the area far away from the weld and the peak value
of tensile stress appears on the surface of the weld. Differences in the welding
sequence influence the value and the distribution of the welding residual stress
greatly, and a more suitable welding sequence can be deduced. The reliability
of the numerical simulation results is proved by the experimental results.
(Some figures in this article are in colour only in the electronic version)
1. Introduction
In the past 100 years, welding has been one of the most widely used technologies. It is very
difficult to find another technology that can develop as quickly as welding and can compare
with welding in the field of scale and diversification [1–3]. Moreover, welding plays a very
important role especially in manufacturing metal structures. The quality of welding influences
the quality of the project directly. Therefore, in order to control the quality of the metal
structure, a proper welding sequence is essential [4–9].
During the welding process, metal structure engenders distortion in some areas due to
the violent temperature change. The weld’s shrinkage in the process of cooling after welding
produces the internal stress. Thus, the welding region will cause the phenomena of jut, twist
and so on. Not only is it difficult to guarantee the dimensions of the work piece, but also the
existence of welding residual stress will reduce the rigidity and fatigue strength of the metal
structure. Therefore, it is important to construct a feasible welding sequence before welding
in order to reduce the residual stress after welding, and, thus, guarantee the welding quality of
0965-0393/05/040553+13$30.00 © 2005 IOP Publishing Ltd Printed in the UK 553
554 S D Ji et al
Figure 1. Mesh generation of finite model.
the workpiece [10–14]. During production, researchers have accumulated a lot of experience
on how to arrange the welding sequence [15–22], but the experience usually aims at a certain
concrete work piece instead of a general suitability.
The plate’s welding mainly includes one-groove welding and double-groove welding.
This paper discusses the numerical simulation of a double V-groove thick plate’s welding
residual stress by studying the influence of the welding sequence on the residual stress of
one-groove welding. A more reasonable welding sequence is obtained, and the simulation
result is compared with the experimental result obtained for the plate. Because the rational
welding sequence that results from the numerical simulation is suitable for general welding
procedures, it has important significance for practical applications.
2. Residual stress of one-groove welding
2.1. Effect of the welding direction on residual stress in multi-layer welding
2.1.1. Establishment of multi-layer welding’s finite element model. Due to the symmetry of
the workpiece, only half of the model is analysed. The dimensions of this plate model are
600mm × 400mm × 15 mm, and this model is divided into three layers to weld. The length
of an element of weld perpendicular to the welding direction is 1.25mm and the length of
an element of weld in the welding direction is 2.5 mm. The mesh is shown in figure 1. The
welding region has been subdivided and the area far away from the weld is meshed relatively
wider. This not only reduces the calculation time but guarantees computational accuracy. This
model is divided into 12420 hexahedral elements and 15656 nodes.
2.1.2. Confirmation of welding material and welding condition. In the numerical simulation,
Martensite Stainless Steel, 1Cr12NiWMoNb is chosen as the base material and filler metal.
The welding method used is GMAW. The voltage of the electric arc is 25V, the electric current
is 210A and the welding speed is 5mms−1.
2.1.3. Confirmation of the welding heat source. Because the welding method is GMAW,
the moving ellipse heat source is adopted in numerical simulations. It can be described by
Welding sequence on the welding residual stress 555
(a) (b)
(c) (d)
Figure 2. Schematic of several welding directions.
the following expression:
q(x, y, z, t) = 6
√
3Q
πabc
√
π
e−3x2/a2 e−3y2/b2 e−3[z+v(∂−t)]2/c2
.
Here, v represents the welding speed and Q the heat input rate, while a, b and c are the
dimensions of the heat source, and their values are 0.005 m, 0.005mand 0.006 m, respectively.
∂ represents the time lag of the welding heat source.
2.1.4. Taking account of the martensitic transformation. In this paper, because the material
of all models is martensitic and it undergoes transformation during the welding process, this
transformation must be considered.
In Marc, the change in heat quantity during the phase change can be controlled by latent
heat and the volumetric change during the phase change can be controlled by the coefficient
of thermal expansion. Moreover, these can be realized by a subprogram that is designed by
the author.
2.1.5. Simulation result of residual stress. Several kinds of welding directions are designed,
as is shown in figure 2. The weld is divided into three layers, and the arrows represent the
moving directions of the welding heat source.
The residual stress distribution trend of the above-mentioned welding sequences is the
same as that of the single layer welding. All the residual tensile stress appears in the weld and
nearby area, while all their peak values appear on the weld, so no figures are given here.
The equivalent Von Mises stress is a rule to judge whether the material in elastically or
plastically deformed. When the equivalent stress is lower than the material’s yielding limit,
the material is in the elastic phase; when the stress reaches its yielding limit, it can be regarded
as being in the plastic phase. So, it is important to study the equivalent stress in the process of
the numerical simulation of welding. The expression for this is as follows:
σ = 1 √
2
(σ1 − σ2)2 + (σ2 − σ3)2 + (σ1 − σ3)2.
In the expression, σ represents the equivalent Von Mises stress, while σ1, σ2 and σ3
represent the three principal stresses of one point in the plate, respectively.
Because the equivalent stress distribution trends of the above-mentioned four schemes are
all the same, the result of scheme c is taken as an example. It is shown in figure 3. The distance
in figure 3 represents the distance from one point to the centre line of the weld.
From figure 3, it can be seen that the peak value of the equivalent Von Mises stress appears
on the weld, and the equivalent stress gradually decreases to nearly zero in the area far away
from the weld.
556 S D Ji et al
Equivalent stress (MPa)
Distance (m)
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
0
100
200
300
400
Figure 3. Equivalent stress distribution of the weld’s middle section.
Table 1. Residual stress’s peak in different welding directions in multi-layer welding.
Scheme serial number
a b c d
Equivalent stress (MPa) 546 528 422 521
Transverse stress (MPa) 453 351 336 422
Longitudinal stress (MPa) 463 363 356 431
Table 1 shows the peak values of transverse and longitudinal residual stress obtained by
the above-mentioned schemes.
It can be seen from table 1 that the third scheme (namely, the converse welding between
adjacent layers) can effectively reduce residual stress after welding.
When welding the first layer, because the area of heat emission of the striking arc region
or quenching arc region is relatively smaller, the temperature variation between these two parts
is relatively more significant than that in the weld’s middle part. Therefore, in the process of
welding, greater compressive plastic deformation is produced near the striking arc region and
near the quenching arc region than in the weld’s middle part. In the process of cooling after
welding, the striking arc region or the quenching arc region, consequently endures greater
tensile residual stress than the middle part of the weld. Moreover, because the intensity of
temperature variation on the striking or quenching arc region is different, two different peak
values appear on the two parts. When welding the second layer, if the converse welding method
is adopted, the trend of residual stress distribution is opposite to that in the first layer. Thus,
the two layers’ resultant effect makes the residual stress distribution relatively even, and then
the peak value is relatively low. This is why the converse welding method is better.
2.2. Effect of the welding direction on residual stress in multi-bead welding
The dimensions of the finite model are 200mm ×150mm ×4 mm, and the width of the weld
is 20.4 mm, which is packed with three beads. The mesh of the weld area is fine and the
gradational transition to the coarse mesh is gradually carried out. The length of an element
Welding sequence on the welding residual stress 557
Figure 4. Mesh generation of plate.
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
0
100
200
300
400
500
Equivalent stress (MPa)
Distance (m)
Figure 5. Equivalent stress distribution of the weld’s middle section.
of weld in the welding direction is 2.5 mm. This structure is divided into 3520 hexahedral
elements and 4736 nodes, as shown in figure 4.
The welding material, the welding condition and the welding heat source of this model
are all the same as those in the multi-layer welding.
Four kinds of different welding directions are designed like multi-layer welding, as shown
in figure 2. The arrows show the welding direction of the three beads, and the bottom arrow’s
position represents the centre line of the weld.
Because the stress distribution trends of the above-mentioned four schemes are the same,
the result of scheme b is taken as an example. Figure 5 is the distribution of equivalent stress
on top of the plate and figure 6 is the distribution of longitudinal residual stress on top of the
plate. Distance in the following two figures represents the distance from one point to the centre
line of the weld.
From figure 5, it can be seen that the peak value of equivalent stress appears on the weld
and the value of stress is nearly zero in the area far away from the weld. From figure 6,
it is seen that the longitudinal residual tensile stress appears on the weld and nearby while
the compressive stress appears on the area away from the weld. Moreover, because of the
severe change in temperature at the interface between the base metal and the weld, the peak
value of tensile residual appears in this area. Because the distribution trend of transverse stress
is similar to that of longitudinal stress, it is not discussed here.
558 S D Ji et al
Longitudinal stress (MPa)
Distance (m)
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
-300
-200
-100
0
100
200
300
400
500
Figure 6. Longitudinal residual stress distribution of the weld’s middle section.
Table 2. Peak residual stress of different welding directions in multi-bead welding.
Scheme serial number
a b c d
Equivalent stress (MPa) 602 595 561 601
Transverse stress (MPa) 500 525 489 522
Longitudinal stress (MPa) 559 552 500 551
200
150
36
14 2 20
40
Figure 7. Dimensions of the specimen for simulation in mm.
The peak values of stress of several different schemes are shown in table 2. From this
table, it can be seen that the peak values obtained from scheme c are the minimum. That is, in
multi-bead welding, the residual stress is the minimum when the converse welding method is
adopted.
3. Welding residual stress of a double V-groove thick plate
3.1. Establishment of the multi-layer welding model
Figure 7 is a model of the double V-groove plate. Figure 8 is the meshing sketch map of this
model.
The base material and filler metal material of the double V-groove welding model is
also Martensite Stainless Steel, 1Cr12NiWMoNb. The welding heat source is an ellipse heat
Welding sequence on the welding residual stress 559
Figure 8. Mesh generation of double V-groove model.
Figure 9. Schematic of weld beads’ delamination.
source model. The welding electric current is 200 A, the welding voltage is 25V, and the
welding speed is 5mms−1. Moreover, the model’s preheating temperature is 100˚C, while the
temperature between the layers is lower than 150˚C in the welding process.
In the course of numerical simulation, the weld is divided into nine layers, as shown in
figure 9. In this figure, 1, 2, 3, 4, 5, 6, 7, 8 and 9, respectively, express the serial number for
each layer of the weld.
In order to study the influence of welding sequences on residual stress, several kinds
of welding sequences are designed according to the above-mentioned weld. Moreover, the
position of each layer’s serial number represents the precedence order of each layer in the
welding process:
(a) 2 3 1 4 8 5 9 6 7;
(b) 2 3 4 1 5 8 6 9 7;
(c) 2 3 1 4 5 8 6 7 9;
(d) 1 2 8 3 9 4 5 6 7;
(e) 1 2 3 8 4 5 9 6 7;
(f) 1 2 3 8 9 4 5 6 7;
(g) 1 2 3 4 5 8 6 9 7;
(h) 2 3 4 5 1 6 8 7 9.
560 S D Ji et al
Figure 10. Distribution of equivalent stress.
Figure 11. Distribution of transverse residual stress.
3.2. Distribution of the residual stress field of the double V-groove thick plate
In each of the above-mentioned welding sequences, the distribution of residual stress is
substantially consistent. Therefore, only the residual stress of scheme f is discussed here.
Moreover, in the course of the numerical simulation, the converse welding method is adopted
between adjacent layers in a multi-layer weld or between adjacent beads in every layer.
Figure 10 shows the distribution of equivalent stress. Figure 11 shows the distribution of
the transverse residual stress, while figure 12 shows the distribution of the longitudinal residual
stress.
The following observations can be made from the figures: the peak value of equivalent
stress appears on the weld; the longitudinal or transverse residual tensile stress of the thick
plate is mainly distributed on the weld or near the weld and the peak value appears on the
weld; the longitudinal or transverse compressive residual stress appears far away from the
Welding sequence on the welding residual stress 561
Figure 12. Distribution of longitudinal residual stress.
Distance (m)
Equivalent stress (MPa)
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040
300
350
400
450
500
550
600
650
Figure 13. Distribution of equivalent stress along the thickness of the weld in the weld’s middle
section.
weld. Moreover, because no constraint exists, the longitudinal residual stress at the two ends
of the bead approaches zero.
Figure 13 is the distribution of equivalent stress along the thickness of the thick plate on
the weld’s middle section. Figure 14 is the distribution of transverse residual stress along the
thickness while figure 15 is the distribution of longitudinal residual stress along the thickness
of the thick plate in the weld’s middle section. In the following figures, ‘Distance’ represents
the distance from the point to the bottom of the thick plate.
From these figures, it is seen that the peak values of equivalent stress, transverse residual
stress or longitudinal stress appear on the surface of the plate, while the stress value is relatively
lower in the middle part.
3.3. Experimental validation
In order to validate the simulation results, experimental verification has been carried out in the
double V-groove plate. Moreover, in order to facilitate the analysis, the scheme f is chosen
562 S D Ji et al
Distance (m)
Transverse stress (MPa)
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040
0
50
100
150
200
250
300
350
400
450
500
550
Figure 14. Distribution of transverse stress along the thickness of the weld in the weld’s middle
section.
0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040
-200
-100
0
100
200
300
400
500
600
Distance (m)
Longitudinal stress (MPa)
Figure 15. Distribution of longitudinal stress along the thickness of the weld in the weld’s middle
section.
as the welding method. The material, the plate’s dimensions and the welding condition in the
experiment are all the same as those in the simulation.
The AST portable x-ray stress tester is adopted to test the residual stress of the specimen.
The x-ray method involves irradiating the specimen by x-rays after the specimen has been
treated, and then the diffraction is obtained. The internal stress is measured from the crystal
lattice distortion. This method is not only easy to carry out but also keeps the specimen intact.
Moreover, it is acceptable to measure surface residual stress.
Figure 16 is the contrast between the test result and the simulation result of transverse
residual stress on the surface of the plate’s middle section. Figure 17 is the contrast between
the test result and the simulation result of longitudinal residual stress on the surface of the
plate’s middle section. In these two figures, ‘Distance’ represents the distance from one point
to the centre line of the weld.
Welding sequence on the welding residual stress 563
Distance (m)
Transverse stress (MPa)
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
-200
-100
0
100
200
300
400
500
600
Experimental result
Simulation result
Figure 16. Experimental verification of transverse residual stress.
Distance (m)
Longitudinal stress (MPa)
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
-200
-100
0
100
200
300
400
500
600 Experimental result
Simulation result
Figure 17. Experimental verification of longitudinal residual stress.
It can be seen by analysing those two figures that the longitudinal or transverse residual
stress of simulation results is basically the same as the experimental results from the viewpoint
of changing trends. It proves that the simulation results are rational and reliable.
3.4. Simulation results of several welding sequences
Table 3 is the contrast of the residual stress’s peak values. It can be seen from the table that the
peak values of transverse residual stress or longitudinal residual stress that are obtained from
scheme c are the minimum. Therefore, this welding sequence is reasonable.
Abstractly, the distribution of welding temperature is not even along the thickness direction
of the thick plate. One part of the welded plate expands more while the other part expands less
or even not at all. Therefore, the expansion of the welded part is blocked and, then, transverse
564 S D Ji et al
Table 3. Residual stress’s peak value for several schemes.
Scheme serial number
a b c d e f g h
Equivalent stress (MPa) 701 712 571 885 653 658 921 912
Transverse stress (MPa) 623 544 405 718 505 511 829 857
Longitudinal stress (MPa) 634 636 507 823 592 607 879 865
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
-400
-200
0
200
400
600
Welding in same direction
Welding in inverse direction
Distance (m )
Transverse stress (MPa)
Figure 18. Comparison of transverse stress between welding in the same direction and welding in
the inverse direction.
compressive plastic deformation is generated. What is more, because of the phenomenon of
uneven shrinkage along the thickness direction of the plate after welding, angular deformation
appears on the workpiece and then the root of the weld carries high tensile stress. When the
tensile stress resulting from the welding angular deformation exceeds the tensile strength’s
critical value of the filler metal in the root of the weld, a crack appears on the weakest root of
the weld. Therefore, when the filler metal of the weld is more uniformly filled, the angular
deformation is less and then the residual stress is less. This is why the residual stress that is
obtained by scheme c is the minimum.
In order to validate whether the welding residual stress that is obtained by the converse
welding method between adjacent layers is the minimum, the double V-groove plate’s welding
stress is studied under the condition of synclastic welding or converse welding. The contrast
in welding transverse residual stress is shown as in figure 18, and ‘Distance’ in figure 18
represents the distance between points and the endpoint of the weld. The simulation result
shows that the peak value of the welding transverse residual stress by the method of converse
welding is 405.4MPa, which is 18.2% less than the stress by the method of synclastic welding.
Moreover, the contrast of the longitudinal stress is similar to this figure and the stress’s peak
value decreases by 16.9%.
4. Conclusions
(1) The simulation results of the plate butt welding show that the peak values of transverse
residual tensile stress and longitudinal residual tensile stress obtained by the method of
Welding sequence on the welding residual stress 565
converse welding between adjacent layers in a multi-layer weld, or between adjacent
beads in every layer, are the minimum. Therefore, the residual stress in the weld is greatly
reduced if this method is adopted. The weld’s strength is enhanced and the quality is
improved.
(2) The residual stress distribution in the double V-groove thick plate shows that the residual
tensile stress appears on or near the weld and the peak value appears on the weld.
Moreover, the simulation results are validated by the experimental results.
(3) The contrast between the simulation results of a double V-groove thick plate welded using
many different welding sequences shows that differences in welding sequence affect the
distribution and the peak value of the residual stress, and also affect the strength and the
crack resistance of the weld. Moreover, a kind of suitable welding sequence is obtained
by comparing simulation results; e.g. when two parts of the weld are filled more evenly,
the residual stress is less.
References
[1] Nakanishi Y 2003 J. Japan Weld. Soc. 72 5–11
[2] Wang J C 2001 Laser Technol. 25 48–51
[3] Song T 1994 China Mech. Eng. 5 61–3
[4] Teng T-L and Chang P-H 2003 Comput. Struct. 81 273–8
[5] Chang P-H 1998 Int. J. Pressure Vessels Piping 75 857–63
[6] Feng Z L, David S A, Zacharia T and Tsai C L 1997 Sci. Technol. Weld. Joining 2 11–19
[7] Deng D and Naqano S 2003 Proc. Int. Conf. on Offshore and Polar Engineering (Honolulu, HI, USA) pp 2403–9
[8] Masahiro T and Koji G 1998 Proc. Int. Conf. on Offshore Mechanics and Arctic Engineering (ASME, Fairfield,
NJ, USA) p 8
[9] Tsai C L and Park S C 1999 Weld. J. 78 156–65
[10] Farkas J and Jarmai K 1998 Weld. World 41 385–98
[11] Josefson B L and Wikander L O 1996 Trans. ASME 118 121–6
[12] Senkara J and Zhang H 2000 Weld. J. 79 194–201
[13] Quadrini E and Wohlfahrt H 1998 Comput. Aided Assess. Control 639–48
[14] Li Y and Frank S 2003 Proc. SPIE: The Int. Soc. Opt. Eng. 4997 30–9
[15] Mochizuki M and Hayashi M 2000 J. Pressure Vessel Technol. 122 27–32
[16] Li Y and Liu W 2002 Trans. Adv. Packaging 25 73–8
[17] Loureiro A and Pina J 2002 Mater. Sci. Forum 404 387–92
[18] Fukuda S and Yoshikawa K 1990 Eng. Anal. 7 78–82
[19] Lu A 2001 China Mech. Eng. 12 183–6
[20] Lin Y C and Chou C P 1995 J. Mater. Process. Technol. 48 693–701
[21] Li K and Yue H 2001 Trans. China Weld. Inst. 22 46–8
[22] Troive L and Nasstrom M 1998 J. Pressure Vessel Technol. 120 244–51