Influence of Welding Sequence on Welding Distortions in Pipes

ARTICLE IN PRESS

0308-0161/$ - se

doi:10.1016/j.ijp

�CorrespondE-mail addr

International Journal of Pressure Vessels and Piping 85 (2008) 265–274

www.elsevier.com/locate/ijpvp

Influence of welding sequence on welding distortions in pipes

I. Sattari-Far�, Y. Javadi

Faculty of Mechanical Engineering, Amirkabir University of Technology, Tehran, Iran

Received 1 December 2006; received in revised form 25 July 2007; accepted 26 July 2007

Abstract

This paper presents a three-dimensional thermo-mechanical analysis to investigate the effect of welding sequence on welding

deformations in pipe–pipe joints of AISI stainless-steel type. Single-pass TIG welding with V-joint geometry in pipes having a diameter

of 274mm and a thickness of 6.2mm is studied here. Nine different welding sequences are analysed. The finite element results are

compared with experimental data. It has been shown that selecting a suitable welding sequence can substantially decrease the amount of

welding distortions in this pipe geometry.

r 2007 Published by Elsevier Ltd.

Keywords: Finite element simulation; Welding distortion; Welding sequence; Welding modelling

1. Introduction

Pipe welding is widely used in a variety of engineeringapplications such as oil and gas industries, nuclear andthermal power plants and chemical plants. A non-uniformtemperature field, applied during the welding process,produces deformation and residual stresses in weldedstructures. In pipe welding, ‘‘diameter change’’ is the mostusual deformation type. After welding, the pipe diameter ischanged from the original diameter because of weldingshrinkage, as shown in Fig. 1. The diameter changes arenot uniform in the circumferential direction of the pipe,and thus the pipe sections would not be circlular after thewelding process. This non-uniformity of the pipe section iscalled ‘‘ovality’’, and is shown in Fig. 2.

The extent of deformations and residual stresses inwelded components depends on several factors such asgeometrical size, welding parameters, welding sequenceand applied structural boundary conditions.

Finite element (FE) simulation has become a populartool for the prediction of welding distortions and residualstresses. A substantial amount of simulation and experi-mental work focusing on circumferential welding withemphasis on pipe welding is available in the literature

e front matter r 2007 Published by Elsevier Ltd.

vp.2007.07.003

ing author. Tel.: +98 21 64543426; fax: +98 21 6641 9736.

ess: [email protected] (I. Sattari-Far).

[1–12]. To reduce computational power requirements,assumptions such as rotational symmetry and lateralsymmetry have been employed in numerical simulations[4–6]. These assumptions reduce the computational de-mand but may make the problem over-simplified bylimiting the analysis to one section of the completegeometry and eliminate modelling of the welding sequence.Therefore, these simplified models are not capable ofpredicting the effects of weld start/stop locations, weldingsequence and tack welds.Fricke et al. [10] investigated multi-pass welding on a

complete three-dimensional (3D) model for pipe weld, butnothing is mentioned about welding sequence. Tsai et al.[13] employed a 3D shell element and moving welding arcto simulate welding residual stresses in AISI 304 stainless-steel pipes. Li et al. [14] developed a full 3D FE model tosimulate a multi-pass narrow gap girth welding process.Recently, Jiang and co-workers [15] used a 3D FE model

to predict temperature distributions in a multi-pass weldedpipe branch junction. However, none of these works hassimulated a fully 3D model for comparing deferent weldingsequences in pipe welding.This paper presents a parametric study to determine the

effect of welding sequence on welding distortions. 3D FEsimulation of a single pass butt-weld joint is performedusing the FE code ANSYS [16]. Two stainless-steel pipeswith an outer diameter of 273.7mm, wall thickness of

www.elsevier.com/locate/ijpvp

dx.doi.org/10.1016/j.ijpvp.2007.07.003

mailto:[email protected]

ARTICLE IN PRESS

Welding Area

Welding Start

Fig. 2. Ovality after welding.

Fig. 1. Pipe diameter variation after welding.

Fig. 3. Welding start and tack welds position.

I. Sattari-Far, Y. Javadi / International Journal of Pressure Vessels and Piping 85 (2008) 265–274266

6.2mm and a length of 300mm are welded together in asingle-pass V-joint. Welding start locations and tack weldpositions are shown in Fig. 3. A total of nine differentsequences are analysed for the welding sequence of thispipe, as shown in Fig. 4. The case entitled as 1-seg, in whichthe weld is conducted entirely in one segment from the startto the final location, is chosen as the basic case here. Thiscase has four tack welds, and is validated experimentally inthis study. Any effects of tack welds on distortions andresidual stresses are neglected in the analysis.

2. Modelling of physical phenomena

Numerical simulation of residual stresses and distortionsdue to welding needs to accurately take account of theinteractions between heat transfer, metallurgical transfor-mations and mechanical fields.

The phenomena involved in the heat input such as arc,material interactions as well as fluid dynamics in the weldpool are not accurately described. From the thermo-mechanical point of view, the heat input can be seen as avolumetric or surfaced energy distribution, and the fluidflow effect, which homogenizes the temperature in themolten area, can be simply taken into account byincreasing the thermal conductivity over the fusiontemperature.The different phenomena involved and their couplings

are given in Fig. 5. As no metallurgical transformationoccurs in the 304 stainless steel considered in this paper, nodetailed modelling of the melting is considered here.

2.1. Heat transfer analysis

The heat transfers in solids are described by the heatequation

rdH

dt� divðlgradTÞ �Q ¼ 0, (1)

lgradTn ¼ qðT ; tÞ on qOq, (2)

T ¼ TpðtÞ on qOt,

where r, H, l and T are density, enthalpy, thermalconductivity and temperature, respectively. In Eq. (1), Q

represents an internal heat source. In Eq. (2), n is theoutward normal vector of domain dO and q the heat fluxdensity that can depend on temperature and time to modelconvective heat exchanges on the surface. Tp represents aprescribed temperature. The heat input is represented by aninternal heat source.In the present study, the double ellipsoid heat source

configuration proposed by Goldak et al. [17] is used, asshown in Fig. 6. As is seen, the front half of the heat sourceis the quadrant of one ellipsoidal source, and the rear halfis the quadrant of another ellipsoid. In this model, thefractions of ff and fr of the heat deposited in the front andrear quadrants are needed, where ff+fr ¼ 2. The powerdensity distribution inside the front quadrant is

qf ðx; y; zÞ ¼6ffiffiffi

3p

f fQ

afbcp3=2eð�3x2=a2

fÞeð�3y2=b2Þeð�3z2=c2Þ. (3)

Similarly, for the rear quadrant of the source the powerdensity distribution inside the ellipsoid becomes

qrðx; y; zÞ ¼6ffiffiffi

3p

f rQ

arbcp3=2eð�3x2=a2r Þeð�3y2=b2Þeð�3z2=c2Þ. (4)

Physically, these parameters are the radial dimensions ofthe molten zone in front, behind, to the side and under-neath the arc. If the cross-section of the molten zone isknown from the experiment, these data may be used to fixthe heat source dimensions.If cross-sectional dimensions are not available, the

experience data given by Goldak et al. [17] suggest that itis reasonable to take the distance in front of the heat source

ARTICLE IN PRESS

Fig. 4. Nine sequences for pipe welding investigated in this study.

Fig. 5. Physical phenomena involved and their couplings.

I. Sattari-Far, Y. Javadi / International Journal of Pressure Vessels and Piping 85 (2008) 265–274 267

equal to one-half the weld width and the distance behindthe heat source equal to twice the width. These suggestionsare used in this paper.

The internal heating due to the plastic dissipation can beneglected considering the small transformation ratesgenerated by a welding operation.

2.2. Mechanical analysis

The mechanical analysis is based on the usual equationsdescribing the static equilibrium. As the plastic dissipationis neglected in the thermal analysis, thermal and mechan-ical analyses can be treated separately. Thus, the mechan-ical calculation is achieved using the temperature fields

computed previously by the thermal analysis. The materialsare supposed to follow elastic–plastic behaviour withisotropic hardening. The material parameters Young’smodulus, Poisson’s ratio, yield stress, strain hardeningand heat expansion coefficient are temperature dependent.

3. Material modelling

Material modelling has always been a critical issue in thesimulation of welding because of the scarcity of materialdata at elevated temperatures. Some simplifications andapproximations are usually introduced to cope with thisproblem. These simplifications are necessary due to bothlack of data and numerical problems when trying to modelthe actual high-temperature behaviour of the material [18].The material properties for AISI 304 stainless steel areshown in Fig. 7. These data are taken from Lindgren [19].The pipe material and the filler metal are assumed to be ofthe same chemical compositions.Due to the lack of data on material properties of the

weld metal and heat-affected zone (HAZ), it is assumed inthis analysis that thermal and mechanical properties of theweld metal and HAZ are the same as those of the basemetal.

4. Finite element modelling

The problem is formulated as a sequentially coupledthermal stress analysis. First, a non-linear thermal analysis

ARTICLE IN PRESS

Fig. 6. Double ellipsoid heat source configuration.

Fig. 7. Material properties for 304 stainless steel used in this study [19].

Fig. 8. 3D finite element model.


is performed to calculate the temperature history of thewhole domain. Then, the results of the thermal analysis areapplied as a thermal body load in a non-linear structuralanalysis to determine distortions. The FE models for boththermal and structural analyses are the same. The general-purposed FE program ANSYS [16] is used for the analyses.During the analysis, a full Newton–Raphson iterativesolution technique with direct sparse matrix solver isemployed for obtaining a solution. During the thermalanalysis, the temperature and the temperature-dependentmaterial properties change very rapidly. Thus, a fullNewton–Raphson technique using modified material prop-erties is believed to give more accurate results.

A conventional element technique named ‘‘element birthand death’’ [20] is used for modelling of the deposited weld.A complete FE model is generated at the start of theanalysis. However, all elements representing the depositedweld except elements for the tack welds are deactivated byassigning them a very low stiffness. During thermalanalysis, all the nodes of deactivated elements (excludingthose shared with the base metal) are also fixed at roomtemperature till the birth of the respective element.Deactivated elements are reactivated sequentially whenthey come under the influence of the welding torch. For thesubsequent structural analysis, birth of an element takesplace at the solidification temperature. Melting andambient temperatures are set as the reference temperatures

(at which thermal strains are zero) for thermal expansioncoefficients of the filler and base metals. To avoid excessivedistortion, initial strains in the elements are set to zero atthe time of element reactivation.In thermal analysis, after extinguishing the arc, the FE

model was run without any load to return to the ambienttemperature of 27 1C. The load steps in the structuralanalysis are kept the same as in the thermal analysis.Linear elements are preferred to higher-order elements in

non-linear problems of this type [21]. Here, eight-noded-brick elements with linear shape functions are used in theFE modelling. Only one half of the pipes is modelled withassumption of symmetry. The basic FE model, used for allthe cases of nine sequences, is shown in Fig. 8.In order to facilitate data mapping between thermal and

structural analysis, the same FE model is used withrespective element types. For thermal analysis, the elementtype is SOLID70, which has a single degree of freedom,temperature, on each node. For structural analysis, theelement type is SOLID45 with three translational degreesof freedom at each node. Due to anticipated hightemperature and stress gradients near the weld, a relativelyfine mesh is used there. Element sizes increase progressivelywith distance from the weld centre line.

5. Experimental studies

5.1. Welding

For circumferential welding of the pipes, an automaticgirth-welding machine with advanced system control wasused. The welding machine was an automatic TIG weldingmachine with an advanced controller, which can simulta-neously control power source, gripper chuck, torch drivingvehicle, inert gas supplier and automatic wire feeder.The automatic circumferential welding system is shown inFig. 9.

ARTICLE IN PRESS

Fig. 10. Weld groove geometry.

Fig. 11. The welded pipe analysed in this study.


When the operator pushes the start button, thecontroller sends four simultaneous signals to the powersource, wire feeder, gripper chuck and inert gas supplier.Consequently, the electrical arc is turned on, the filler is fedinto the molten pool, the pipes are rotated with theassigned welding speed and the inert gas flows into the weldpool. In addition, if any oscillatory motion is needed, thecontroller sends a signal to the torch driving vehicle.

A single pass butt-weld joint geometry with a singleV-groove (601 included angle) and without root gap wasused, as shown in Fig. 10.

The weld joint contained four initial tack-welds atangular positions of 361, 1501, 2341 and 3061 from theweld start position, as shown in Fig. 3. The inert gaswas argon with 99.9999% purity, the welding current was230A, the welding voltage was 20V, the welding speed was16 cm/min and the wire-feeding speed was 90 cm/min. Thewelded pipes are shown in Fig. 11.

5.2. Measurement of the distortions

The diameters of the pipes were measured before andafter welding to determine diameter changes due towelding. Measurements were done in three sections ofany pipe, as shown in Fig. 12. This figure shows a frontview of the pipes and the scribed lines in this figure arediametrical measuring locations.

In each section, 10 diameter locations were measuredusing an accurate micrometer of a range of 250–275mmwith an accuracy of 0.01mm.

The measured results before welding showed that thepipe sections were not completely circular. The nominalpipe diameter is 273mm. After measurement of the pipe

Fig. 9. Automatic circumfe

diameter at different locations, the average pipe diameterwas set to be 273.7mm in this analysis. The measurementof the thickness at different locations of the two pipes

rential welding system.

ARTICLE IN PRESSI. Sattari-Far, Y. Javadi / International Journal of Pressure Vessels and Piping 85 (2008) 265–274270

showed that the thickness of the pipes varied between 6.07and 6.32mm. Here, an average value of 6.2mm wasassigned in the analysis.

The measurements of the diameters at the same locationswere repeated after welding. The differences in themeasured results are considered as the diameter changes

Section 3

10mm

150mm

290mm

Section 2

Section 1

Pipe2

Pipe1

Fig. 12. Measuring sections of welded pipes.

-2.5

-2

-1.5

-1

-0.5

0

0.5

0 5 10 1

Diameter No.

Dia

mete

r V

ariation(m

m)

Welding Start Position

Fig. 13. Comparison of FEM and e

due to welding. These results are compared with the resultsfrom the FE analysis.

6. Results and discussion

6.1. Verification of the FE modelling

The FE model was run for 1-seg sequence, and itsdiameter changes were calculated for the three measuringsections shown in Fig. 12. The FE results are comparedwith the experimental measurements in Figs. 13–15.Because of symmetry in welding the two pipes, any

difference between pipe-1 and pipe-2 measurements is errordue to experimental measurements.Fig. 13 shows the comparison results in section-1, which

is the nearest section to the welding area. It shows goodagreement between the FE results and the experimentalmeasurements, having a deviation of about 710%.It should be noted that this type of pipe is produced from

the rolling of plates. After rolling, the edges of the platesare axially welded together to obtain the final form of thepipes. This axial weld is not considered in the FE analysisdue to the lack of welding information. Consequently,more deviations are observed for locations near the axialwelds in the two pipes.Figs. 14 and 15 show good agreement between the FE

results and the experimental measurements, except forpoints having deviations in their experimental measure-ments of pipe-1 and pipe-2.Based on the results presented in Figs. 13–15, it can be

concluded that the developed FE modelling is suitable toestimate the distortions in the pipes welded with differentsequences.

5 20 25

FE Model

Experimental(pipe-1)


xperimental results in section-1.

ARTICLE IN PRESS

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0

Diameter No.

Dia

mete

r V

ariation(m

m)

FE model




5 10 15 20

Fig. 14. Comparison of FEM and experimental results in section-2.

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

0

Diameter No.

Dia

me

ter

Va

ria

tio

n(m

m)

FE model




5 10 15 20

Fig. 15. Comparison of FEM and experimental results in section-3.


6.2. Finding the best sequence

The FE model is used for the analysis of all ninesequences shown in Fig. 4. Fig. 16 shows a comparison ofthe calculated diameter changes in all nine cases.

For practical use of the results, we consider the followingtwo criteria:

(1)
maximum of diameter variations (2) average of diameter variations.
The maximum of diameter variations means the highestdiameter variation in a selected section. This gives a valueof the ovality of the pipe due to welding. For section-1, alldiameter variations were decreasing (see Fig. 13). For

section-2 and section-3, the diameter variations were bothdecreasing and increasing (see Figs. 14 and 15). Average ofdiameter variations indicates an average of all diametervariations in each section. Figs. 17–19 compare these twocriteria in all nine sequences in three sections analysed bythe FE modelling.For choosing the best sequence, we should first

determine which criterion is of the most interest. Forexample, if the maximum diameter variations insection-3 (farthest locations from the welding area)are critical in a structure, the best choice for weldingsequence is 8-seg-a, (see Fig. 19). In another case, if theaverage of diameter variations in section-2 (mid-position ofpipe) is important, welding with 6-segments is the bestchoice.

ARTICLE IN PRESS

-2.5

-2

-1.5

-1

-0.5

0

0.5

0 5 10 15 20 25

Diameter No.

Dia

me

ter

Va

ria

tio

n(m

m)

1-seg

2-seg

4-seg-a

4-seg-b

4-seg-c

6-seg

8-seg-a

8-seg-b

8-seg-c

Fig. 16. Comparison of diameter deformations of nine sequences in section-1.

1.92

1.79

2.02

1.43 1.40

1.09

1.501.58

1.43

0.640.57

0.75

0.320.23

0.28 0.260.32

0.26

0

0.5

1

1.5

2

2.5

1-seg 2-seg 4-seg-a 4-seg-b 4-seg-c 6-seg 8-seg-a 8-seg-b 8-seg-c

Maximum of Diameter Variations(mm)

Average of Diameter Variations(mm)

Fig. 17. Comparison f deformations of nine sequences in section-1.


Table 1 gives a guideline in choosing the best weldingsequence. Here, a value between 1 (poor) and 4 (best) isassigned for any welding sequence based on the specifiedcriteria. For example, if the maximum diameter variationin section-1 is critical, welding with 6-segments or 4-segments-c are the best, and welding with 4-segments-a or1-segment are the worst choices for a welding sequence.

From Table 1, it can be concluded that weldingaccording to 4-seg-c, 6-seg and 8-seg-c are very goodwelding consequences to minimize welding distortions. Theworse case in this study is 4-seg-a. This finding questionsthe common understanding that increasing the number ofsequences in the welding of pipes always leads todecreasing welding distortions. Here, for instance, weldingaccording to 4-seg-a causes more distortion than weldingaccording to 1-seg and 2-seg.

7. Conclusions

Based on the FE analysis of AISI 304 stainless-steelpipes welded with different welding sequences in thecircumferential direction, the following conclusions maybe made:

(1)
Predicted diametric distortions from three-dimensionalFE analysis are in reasonable agreement with experi-mental measurements.
(2)
Welding causes diameter variations (ovality) in thepipes depending on the welded sequence.
(3)
Pipe diameter variations in the welded section arenegative (diameter decreases in this section), but withincreasing distance from the welding centre line, thesevariations go to zero and afterwards become positive

ARTICLE IN PRESS

0.400.42

0.39

0.310.29 0.29

0.340.36

0.31

0.170.18 0.19

0.100.09

0.07

0.10 0.100.09

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45




Fig. 18. Comparison of deformations of nine sequences in section-2.

0.59

0.54

0.68

0.60

0.55

0.41

0.20

0.33

0.280.30

0.27

0.37

0.300.27

0.14

0.05

0.11 0.07

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1-seg



8-seg-c8-seg-b8-seg-a6-seg4-seg-c4-seg-b4-seg-a2-seg

Fig. 19. Comparison of deformations of nine sequences in section-3.

Table 1

A guideline in choosing the best welding sequence

Selecting criteria Welding sequence


Max. diameter variation in section 1 1 2 1 3 4 4 2 2 3

Ave. diameter variation in section 1 1 2 1 2 4 3 3 2 4





Average 1.33 1.67 1.17 2.17 3.17 3.50 2.83 2.50 3.67


ARTICLE IN PRESSI. Sattari-Far, Y. Javadi / International Journal of Pressure Vessels and Piping 85 (2008) 265–274274

(diameter increases in the sections that are far from thewelding section).

(4)
The common understanding that increasing number ofwelding sequences always leads to decreasing weldingdistortions of pipes may be questionable. Here, it isshown that, under certain conditions, welding with foursegments may cause more distortions than welding withone or two segments.
(5)
The maximum diameter variation in a section far fromthe welding area welded with one segment was0.59mm. This value could be decreased to 0.2mm byusing a sequence of eight segments, indicating thebenefits of welding sequence to substantially decreasethe welding distortions.
References

[1] Abid M, Nash DH. Risk assessment studies for gasketed and non-

gasketed bolted pipe joints. In: Proceedings of the international

pipeline conference (IPC2002), Calgary, Canada, IPC-27386, Sep-

tember 29–October 3, 2002. p. 1–11.

[2] Nash DH, Abid M. Surface sensitivity study of non-gasketed flange

joint. J Process Mech Eng Part-E 2004;E4:218.

[3] Brickstad B, Josefson BL. A parametric study of residual stresses in

multi-pass butt-welded stainless steel pipes. Int J Press Vessels Piping

1998;75:11–25.

[4] Rybicki EF, Schmueser DW, Stonesifer RW, Groom JJ, Mishaler

HW. A finite element model for residual stresses and deflections in

girth-butt welded pipes. Trans ASME J Press Vessel Technol 1978;

100:256–62.

[5] Rybicki EF, McGuire PA, Merrick E, Wert J. The effect of pipe

thickness on residual stresses due to girth welds. ASME J Press Vessel

Technol 1982;104:204–9.

[6] Rybicki EF, Stonesifer RB. Computation of residual stresses due to

multipass welds in piping system. Trans ASME J Press Vessel

Technol 1979;101:149–54.

[7] Karlsson L, Jonsson M, Lindgren LE, Nasstrom M, Troive L.

Residual stresses and deformations in a welded thin-walled pipe.

In: Rybicki E, Shiratori E, Widera GEO, Miyoshi T, editors. ASME

pressure vessels and piping conference 1989. Weld residual stresses

and plastic deformation, PVP-vol. 173. Hawai: Honolulu; 1989.

p. 7–14.

[8] Dong Y, Hong J, Tasi C, Dong P. Finite element modeling of residual

stresses in Austenitic stainless steel pipe girth welds. AWS Weld J,

Weld Res Suppl 1997;442:444–9.

[9] Karlsson RI, Josefson BL. Three dimensional finite element analysis

of temperature and stresses in single-pass butt-welded pipe. Trans

ASME J Press Vessel Technol 1990;112:76–84.

[10] Fricke S, Keim E, Schmidt J. Numerical weld modeling—a method

for calculating weld-induced residual stresses. Nucl Eng Des 2001;

206:139–50.

[11] Siddique M, Abid M, Junejo HF, Mufti RA. 3-D finite element

simulation of welding residual stresses in pipe-flange joints: effect of

welding parameters. Mater Sci Forum 2005;490–491:79–84.

[12] Jonsson M, Karlsson L, Lindgren LE. Simulation of tack welding

procedures in butt joint welding of plates. AWS Weld J, Weld Res

Suppl 1995;1:296s–3301s.

[13] Tsai CL, Dong Y, Dong P. In: 77th annual AWS, Chicago, USA,

21–25 April 1996. p. 51.

[14] Li M, Atteridge DG, Meekisho LL, West SL. Trends in welding

research. In: Fourth International Conference, Gaylingurg, TN,USA,

5–8 June 1995. p. 51–6.

[15] Jiang W, Yahiaoui K, Hall FR. Finite element predictions of

temperature distributions in a multipass welded piping branch

junction. Trans ASME J Press Vessel Technol 2005;127(1):7–12.

[16] ANSYS user’s manual. SAS IP inc.; 1998.

[17] Goldak J, Chakravarti A, Bibby M. A new finite element model for

welding heat sources. Metall Trans B 1984;15B:299–305.

[18] Lindgren LE. Finite element modeling and simulation of welding.

Part 2: improved material modeling. J Therm Stress 2001;24:

195–231.

[19] Lindgren LE, Hedblom R. Modelling of addition of filler material in

large deformation analysis of multipass welding. Commun Numer

Methods Eng 2001;17:647–57.

[20] Lindgren LE. Finite element modeling and simulation of welding.

Part 1: increased complexity. J Therm Stress 2001;24:141–92.

[21] Deng D, Murakawa H. Numerical simulation of temperature field

and residual stress in multi-pass welds in stainless steel pipe and

comparison with experimental measurements. Comput Mater Sci

2006;37(3):269–77.

Influence of Welding Sequence on Welding Distortions in Pipes

Documents