Distortion Analysis of Welded Stiffeners

O. A. Vanli ,∗ P. Michaleris †

June 19, 2001

Abstract

This paper presents a welding distortion analysis approach for T-stiffeners with a particular

emphasis on welding-induced buckling instabilities. 2-D thermo-mechanical welding process

simulations are performed to determine the residual stress and angular distortion. The critical

buckling stress along with the buckling mode and bowing distortion are computed in 3-D

eigenvalue and linear stress analyses. The effects of the stiffener geometry, weld sequence, weld

heat input and mechanical fixturing on the occurance of buckling and the distortion pattern are

investigated.
∗
Graduate Student, Department of Mechanical and Nuclear Engineering, 307 Reber Building, Pennsylvania State

University,University Park, PA 16802, USA Tel : (814) 865-0059, Email: oav100@psu.edu

†
Assistant Professor, Department of Mechanical and Nuclear Engineering, 232 Reber Building, Pennsylvania State

University, University Park, PA 16802, USA Tel : (814) 863-7273 Fax : (814) 863-4848 Email: pxm32@psu.edu

1
1 Introduction

Welding, among all mechanical joining processes, has been employed at an increasing rate for

its advantages in design flexibility, cost savings, reduced overall weight and enhanced structural

performance. In this research, welding is being evaluated as the manufacturing method for stiffeners

as an alternative to the present fabrication method of ”cutting” or ”stripping” from standard I-

beams. Use of welding in stiffener fabrication may introduce considerable savings by eliminating

scrap parts.

Thinner section components made of higher strength steels are being commonly utilized in

shipbuilding, railroad and aerospace industries in fabricating large structures to achieve reduction

in overall weight and more controllable manufacturing. However, for the structures made of

relatively thin components, welding can introduce significant buckling distortion which causes loss

of dimensional control and increased fabrication costs due to poor fit-up between panels. Flame

straightening is the commonly used technique to correct the out-of-plane distortion resulting from

welding processes, and is a labor intensive and costly process.

Finite element techniques have been used in the prediction of welding residual stress and

distortion for more than two decades. Due to the nature of the process, additional complexities are

involved in the FEA of welding compared to traditional mechanics, such as temperature and history

dependent material properties; high gradients of temperature, stress and strain fields with respect to

both time and spatial coordinates; large deformations in thin structures and phase transformation

and creep phenomenon.

Most of the currently performed welding simulations, both 2-D and 3-D, are based on small

2
deformation assumption and are limited to simpler structures and weld geometries (e.g. butt joints)

or focusing only to the heat affected zones, ignoring the surrounding structure. A small deformation

analysis assumes infinitesimal displacements and loads being applied to the undeformed geometry.

The interaction between the weld zone and the structure is effective on the accumulated distortion,

and large deformation modes in unrestrained structures may not be captured with this type of

analysis [1], [2]. Brown and Song [1] have performed 2-D axisymmetric and 3-D weld simulations

of a ring stiffened cylinder structure, and concluded that 2-D analysis overestimated the rotation

of the ring during the heating segment, and it was very sensitive to model modifications, such as

joint clearance and location of constraints. Michaleris et. al. [2] studied the effects of the restraints

and the solidified portions of the weld on the residual stress and distortion profiles by comparing

the performance of 2-D and 3-D weld simulations.

Earlier studies of welding accounted for the non-linearities due to temperature dependent

material properties and plastic deformations [3, 4, 5]. The majority of those analyses were

limited to two-dimensions on the plane perpendicular to welding direction, but good correlations

have been observed between the numerical predictions and experimental results [6, 7, 8, 9], and

especially for residual stress predictions, 2-D models provided accurate estimations comparable to

3-D analyses, since the stress field exhibits a fairly uniform distribution through the length of the

work-piece. Argyris et. al. [6] computed the thermo-mechanical response using 2-D models in a

staggered solution strategy to combine and integrate the thermal and mechanical computational

steps. Rybicki et. al. [7] performed thermo-elasto-plastic analysis on a 2-D axisymmetric finite

element model for a two-pass girth-butt welded pipe problem, and verified the numerical results

with the experimentally obtained temperature history and residual stress distributions. Papazoglu

and Masubuchi [8] solved the multipass GMAW process problem by performing uncoupled 2-D heat

3
transfer and stress-strain analyses, incorporating the phase transformation strains.

2-D models, as mentioned above, have been particularly useful with their high efficiency and

accuracy in determining the solution in the analysis plane and reduced computational requirements.

However, for welding practices where tack welding or fixturing allow out-of-plane movement 2-D

analyses may not be accurate, particularly, in distortion predictions [1]. Furthermore, longitudinal

heat transfer and instability aspects, and end effects (i.e. due to initiation and termination of the

heat source) cannot be realized in two dimensional formulations.

Oddy et. al. [10] examined the butt welding of a bar via 3-D FEM, and computed the

temperature, strain and stress fields. Tekriwal and Mazumder [11, 12] simulated thermal and

elasto-plastic response of the butt-welded plates through 3-D models, considering filler material

addition. Multi-pass welding simulation of plates and experimental validation have been addressed

in [13, 14].

Welding-induced buckling of thin-walled structures has been investigated in greater detail by

[15, 16, 17]. Tsai et. al.[15] studied the distortion mechanisms and the effect of welding sequence on

panel distortion. Ueda et. al. [16, 18] presented a methodology to determine the buckling behavior

of plates by large deformation elastic FEA and employing inherent strain distributions.

For the welding practices where tack welds or fixturing are used to restrict the movement of the

welded parts, the structural response may be evaluated by means of decoupled 2-D welding and

3-D buckling simulations. When mechanical fixturing on the structure prevents the longitudinal

shrinkage during welding, the out-of-plane structural behavior doesn’t have influence on the in-

plane welding response, and buckling is only observed after the restraints are removed and the

structure cools down. Exploiting this fact, Michaleris et. al. [17] proposed a method to predict

4
welding-induced buckling by uncoupling the weld simulation and the structural buckling analysis.

They expressed the residual stress profile from the 2-D welding simulations as buckling stress on

the 3-D structural model. This approach is analogous to the work by Ueda et. al. [18], where the

concept of inherent strain is used to generate the welding residual stresses by applying a prescribed

thermal strain field using empirical methods. In the former study, however, residual stresses are

calculated with weld process simulations, which provides improved estimations for buckling analysis

compared to empirically determining the residual stress.

Phase transformations and transformation plasticity have also been incorporated in the analysis

as recent developments [10, 19, 20]. The primary objective there is to more accurately model the

residual stress distribution, microstructure and local distortion in the area immediately adjacent

to the weld.

In this work the decoupled 2-D and 3-D finite element analysis technique by Michaleris et.

al. [21, 17] is applied to evaluate welding-induced buckling of fabricated stiffeners. Effects of the

following process and design parameters are investigated,

• stiffener cross-section (small section, large section),

• fixturing and future stiffener straightening

(as weld, and rammed-down configurations),

• Gas Metal Arc and Submerged Arc Welding processes,

• simultaneous (offset-torch) and sequential welding processes.

5
 Figure 1: Distortion Modes in fabrication of stiffeners

2 Analysis Approach - Modelling the Welding Distortion

Following the work of Michaleris et. al. [21, 17], the response of the stiffener is evaluated in two

steps by combining two-dimensional welding simulations with three-dimensional structural analyses

in a decoupled approach.

2-D Thermo-mechanical Weld Simulation :

A two dimensional thermo-elasto-plastic analysis is performed to determine the angular

distortions, residual stresses, and plastic strain fields during the welding process ignoring the

structural response. Residual stresses are caused by the negative plastic strains resulting from

the welding thermal cycle.

3-D Eigenvalue and 3-D Linear Stress Analyses :

The buckling distortion and critical buckling stresses are consequently determined by an

6
eigenvalue and a linear stress analysis applying the, mostly uniform and compressive, longitudinal

plastic strain field of the 2-D weld model on the 3-D structural model as equivalent load.

A constant, negative thermal load is applied at the weld region to introduce the effects of welding

into the 3-D structure. Thermal loading is used rather than mapping the plastic strain field, which

would require a complex analysis procedure. An eigenvalue analysis is performed to determine

the critical residual stresses and buckling distortions, and a linear stress analysis is performed to

compute the bowing distortion.

2.1 Modes of Welding Distortion

The structural response of the stiffener is evaluated in terms of angular, bowing and buckling

distortions as illustrated in Figure 1. The overall distortion of the structure will be a combination

of these distortion forms.

Angular Distortion: The angular distortion is the change in the included angle between the

web and the flanges on both sides (θ1 , θ2 , Figure 2 ) and is computed by the 2-D welding simulation.

Bowing Distortion: The bowing distortion is the displacement of the web in y-direction (d,

Figure 3 ) due to bending of the stiffener about the x-axis, and is computed by the 3-D linear stress

analysis.

Buckling Distortion: The critical buckling and the corresponding mode shape of the structure

are determined in the 3-D eigenvalue buckling analysis.

7
 Figure 2: Angular Distortion of the stiffener flanges

Figure 3: Bowing Distortion of the stiffener

2.2 Welding Simulation

The welding simulation involves a thermal and a mechanical analysis. The effect of mechanical

response is assumed to be negligable on the thermal behavior, thus the temperature field is solved

independently from the mechanical solution. To determine the temperature history profile, a non-

linear, transient heat-flow finite element analysis is performed on the plane perpendicular to the

welding direction.

The numerical implementation of the history dependent (transient) heat transfer problem

involves an incremental scheme with several small time increments. The solution at a given time

increment is obtained by using the solution at the previous time increment as an initial condition.

This problem is addressed in detail in references [4, 22, 11].

The governing energy balance equation for transient heat transfer analysis is given as follows,

dT
ρCp (r, t) = −∇r · q(r, t) + Q(r, t) in the entire volume Vr of the material (1)
dt

8
 Weld Qb [W] η v [mm/s]

GMAW 9750 0.75 11.85

SAW 9000 0.90 10.16

Table 1: Weld Parameters

where ρ is the density of the body ([kg/mm3 ]), Cp is the specific heat capacity ([J/kg o C]), T is the

temperature ([o C]), q is the heat flux vector, Q is the internal heat generation rate, t is the time,

r is the coordinate in the reference configuration and ∇r is the spatial gradient operator.

The nonlinear isotropic Fourier heat flux constitutive relation is enforced; using the temperature-

W
dependent thermal conductivity, k ([ o Cmm2 ]),

q = −k∇r T [W/mm2 ] (2)

Convection boundary conditions are assigned for all free surfaces. The internal heat generation

rate by the welding torch, modeled with a ”double ellipsoid” heat source model [23], is given as,
√
6 3Qb (η)f −[ 3x22 + 3y22 + 3(z+vt) 2
]
Q= √ e a b c2 [W/mm3 ] (3)
abcπ π

where Qb is the welding heat input; η is the welding efficiency, x, y, and z are the local coordinates of

the double ellipsoid model aligned with the weld fillet; a is the weld width; b is the weld penetration;

c = 4a is the weld ellipsoid length, and f = 0.6 when the torch is behind the analysis plane, and

f = 1.4 after the torch passes the analysis plane; v is the torch travel speed; and t is time. Material

properties for high-strength steel (HSLA-65) are used in this study. Table 1 lists the values of Qb ,

η and v that are used for the GMAW and SAW processes.

The subsequent history dependent stress analysis is performed by modelling the stress problem

as a quasi-static process in a Lagrangian frame. This problem has been covered by several

9
investigators [6, 14, 5, 12, 10]. Similar to the heat transfer analysis, the numerical implementation

of the quasi-static analysis involves an incremental scheme with several small static increments. The

solution at a given time interval is obtained by using the solution at the previous time increment

as an initial condition.

The temperature values solved for in the previous thermal analysis are imported to the

mechanical analysis as loading. Generalized plane-strain conditions are assumed to account for

the out-of-plane expansion in the structure. The longitudinal (out-of-plane) strain is assumed to

vary linearly with x- and y- coordinates in the analysis plane:

²z = e − xφy + yφx (4)

where e is the z-component of the strain at the coordinate origin and the constants φx and φy

represent the strain variations in the y and x axes, respectively.

The stress equilibrium equation is given by,

∇r σ(r, t) + b(r, t) = 0 in Vr (5)

where σ is the stress, b the body force, and t is time. The mechanical constitutive law is :

σ̇ = C (²̇ − ²˙p − ²˙t ) (6)

²˙p = ²˙q · a (σ, ²q , T) (7)

f = σe − σy ≤ 0 (8)

where T is temperature, C is the material stiffness tensor, a is the plastic flow vector, ², ²p and ²t

are the total, plastic and thermal strains and ²q is the equivalent plastic strain. In Equation 8, f

is the yield function, σe is the Von Misses stress, and σy is the yield stress. Active yielding occurs

when f = 0.

10
 Figure 4: FEA mesh for 2-D welding process analysis

2.3 Structural Analysis

In shipbuilding, stiffeners are joined to panels with fillet-welding along the free edge of the web,

thus, the straightness of this edge after the stiffener fabrication process is important for the quality

of the fit-up between the stiffener and the panel. The structural analysis is carried out for the

stiffener web to compute bowing distortions, and the critical welding residual stress that will cause

buckling.

The longitudinal residual stress distribution (σ r ) computed in the 2-D analyses are compared

to the critical buckling stresses (σ cr ) of the structure from the 3-D structural analysis to determine

if the structure will buckle.

The structural analysis is composed of elastic eigenvalue and linear stress analyses. Incremental

11
 Figure 5: FEA mesh for 3-D buckling analysis

large deformation analyses may also be performed to determine the onset of buckling, buckling and

post buckling stages in response to increasing stress, but they are computationally intensive and

are usually used for validating the predictive methodology [17].

3-D Eigenvalue Buckling Analysis:

The elastic instability problem is defined as an eigenvalue problem as follows

det ( K + λKG ) = 0 (9)

where K and KG are the linear and non-linear strain stiffness matrices, and λ is the eigenvalue,

respectively.

A 3-D eigenvalue analysis is performed on the structural model with a unit negative thermal

load applied in the weld region (T = −1.0) to model the uniform compressive longitudinal plastic

12
strain field occurring in welding. The eigenvalues (λi ) represent the multipliers (scaling factors)

which result in the critical buckling stress field (σ cr )i when multiplied with the stress field resulting

from the unit thermal load (σ L ). Equation (10) shows the computation of the critical residual stress

distribution at the plate midspan.

(σ cr )i = λi · σ L [MPa] (10)

The scalar stress values σL , σcri and σr , that are used in the buckling criteria, are computed at the

free edge of the stiffener web.

The buckling analyses may yield negative eigenvalues, which often cannot be explained by

”physical” behavior. Those situations can be avoided by applying enough preload Tp , to load the

structure just below the buckling load before performing the eigenvalue analysis. In such a case,

the critical buckling stress in equation (10) is determined as

(σ cr )i = ( λi + | Tp | ) · σ L [MPa] (11)

Buckling distortion is determined from the eigenvectors (mode shapes) of the structure. The

structure may buckle in any of the modes with critical stresses lower than the residual stress field

due to welding. It will prefer to buckle with the permissible buckling mode having the lowest

critical stress. The permissibility of the modes are determined by the constraints on the structure.

Certain buckling modes may be suppressed by the mechanical fixturing applied, then the structure

will tend to buckle the next available (higher) mode. The weight of the structure might have an

influence of causing even higher buckling modes.

3-D Linear Stress Analysis:

A 3-D linear stress analysis is performed using the negative unit thermal load applied in the

13
weld region (T = −1.0). The bowing distortion, d, at the plate midspan for a given longitudinal

welding residual stress field (σ r ) is obtained by scaling the unit bowing distortion, dL , and linear

longitudinal stress profile, σ L , as follows,

σr
d=( ) dL [mm] (12)
σL

2.4 Validation

The predictive buckling analysis approach presented in this work has been verified to be accurate

by several researchers [17, 24]. The results from the experimental tests involving thermo-couple,

blind-hole drilling and out-of-plane distortion measurements have been reported to be in close

agreement with the computational results.

3 Numerical Implementation

The stiffener is fabricated by joining two plates longitudinally in a T-joint configuration. Two

different stiffener geometries are considered, as illustrated in Figure 6 and Table 2. The fillet welds

of size 1/4 in. are performed on both sides with dual-torches by either gas metal arc (GMAW)

or submerged arc welding (SAW). The influence of weld sequence is evaluated for simultaneous

(3.5” offset torches) and sequential welding cases. In addition to the geometry, weld sequence, and

process parameters, the effects of fixturing and mechanical restraint are investigated as the case

studies given in Table 3.

The boundary conditions used to implement the fixturing effects are:

14
 Figure 6: Dimensions of the T-type fillet welded stiffener

BW [in] tW [in] 2BF [in] tF [in] l [ft]

Small Section 6.9375 1/8 4.5 5/16 20.0

Large Section 15.625 5/16 10.5 5/8 20.0

Table 2: Dimensions of Small and Large sections

• Free boundary (As-weld) condition

• Restrained boundary (Rammed down) condition

The free boundary condition analysis allows the stiffener to bend during welding. This is the

most typical fixturing used in welding and the stiffener bending manifests itself as bowing distortion

(see Figure 3). The fixed boundary condition is formulated to model the future mechanical

straightening of the stiffener to be welded on a panel.

The finite element solutions are performed by utilizing the ABAQUS software both for the 2-D

and 3-D models. The implementation details pertaining to those problems; the type of elements,

15
boundary and loading conditions, are presented to allow convenient reproduction.

2-D Welding Simulation

The translating torch for the welding heat input is programmed using the user-subroutines. The

2-D finite element mesh used in the heat transfer analysis of the small section is illustrated in Figure

4. The model is made up of heat conduction, quadratic(8 node), quadrilateral elements(ABAQUS

DC2D8). The small section is composed of 716 quadrilateral elements and 2401 nodes, and the

large section is composed of 2256 quadrilateral elements and 7280 nodes.

The quasi-static mechanical problem, following the heat transfer analysis, is discretized into

a generalized plane strain finite element model, using ten node, quadratic, reduced integration,

quadrilateral elements(ABAQUS CGPE10R), with the mesh identical to that used for thermal

analysis.

To model the restriction of the supporting plate, the downward motion of the plates should

be restrained. This is implemented by placing two nonlinear dashpot elements (ABAQUS

DASHPOT1) at the two outermost nodes of the 2-D model, to exert high damping forces to resist

the downward motion. The dashpot damping coefficient is defined as a nonlinear function of the

y-displacement as follows,





 106 y<0
c= [N·s/mm] (13)



 0 y≥0

The rammed-down condition is implemented by fixing the x-axis rotation degree of freedom of

the generalized plane strain elements in the 2-D model.

16
 3-D Buckling and Linear Stress Analyses

The 3-D structural finite element models are developed using shell and truss elements. The

shell elements are representing the stiffener web and flange, and truss elements are representing the

weld joint along the crossing of the flanges and the web. The negative plastic strain resulting from

the weld process is applied to the truss elements in the form of negative thermal load. Trusses only

have axial degrees of freedom, thus, they will only experience orthotropic thermal expansion, and

the longitudinal compressive strains will accurately be modelled.

The model for small stiffener contains 8400 shell elements(ABAQUS S4R5), 350 truss

elements(ABAQUS T3D2), and a total of 8775 nodes; the large stiffener model has 18900 shell, 350

truss elements and 19305 nodes. The 3-D finite element mesh of the small stiffener is illustrated in

Figure 5. The use of structural elements rather than continuum elements is aimed at providing a

more robust design tool. With structural elements, geometric features such as the plate thickness

and the truss cross sectional area can be modified conveniently without changing the whole model.

For the free boundary condition case, the structural analysis, for both the small and large

sections, required a preload of Tp = - 0.05 be applied to the structure; whereas for the restrained

case, the reduced number of degrees of freedom eliminated the need to apply a preload (Tp = 0.0).

3.1 Results

The longitudinal residual stress distributions have come out to be as expected in both boundary

settings; for the as-weld configuration the regions close to the torch have tensile stresses, and the

areas away from the torch close to the flange edges, the stresses become compressive; only near the

17
 Case Weld Boundary Condition

1 GMAW (3.5in offset) free

2 GMAW (sequential) free

3 SAW (3.5in offset) free

4 SAW (sequential) free

5 GMAW (3.5in offset) restrained

6 GMAW (sequential) restrained

7 SAW (3.5in offset) restrained

8 SAW (sequential) restrained

Table 3: Weld Cases

free span of the web they become tensile to equilibrate the stress field in the web. Stress profiles

remain the same for the restrained boundary, except that the areas close to web free span now have

compressive stresses due to the fixed displacement conditions. The residual stress distributions

along the centre line of the stiffener web for the free and rammed-down cases are shown in Figure

7 to illustrate the influence of the boundary conditions.

The analysis was repeated for the 2-D welding simulation of the SAW process with the same

boundary conditions and the welding schemes. The results are compatible to those of GMAW

process and they have been tabulated in Tables 4 and 5 for small section; and Tables 7 and 8 for

large section.

3.1.1 Buckling in Small Stiffener

Tables 4 and 5 give the critical and residual longitudinal stress values required for buckling

prediction, as well as the bowing and angular distortions occurring in the eight different weld

cases. Table 6 presents the eigenvalues for the first 10 buckling modes obtained from the 3-D

18
 Case Weld Angular distortion Bowing distortion Residual Stress Critical Stress Buckling

θ1 (deg) θ2 (deg) d (mm) σr [MPa] σcr [MPa]

1 GMAW (3.5in offset) 89.290 89.462 42.948 79.876 458.90 No

2 GMAW (sequential) 89.572 89.212 28.839 58.248 458.90 No

3 SAW (3.5in offset) 88.835 89.466 46.219 93.378 458.90 No

4 SAW (sequential) 89.290 89.021 33.804 68.275 458.90 No

Table 4: Results for the Free Boundary Condition cases for Small Section

Case Weld Angular distortion Bowing distortion Residual Stress Critical Stress Buckling

θ1 (deg) θ2 (deg) d (mm) σr [MPa] σcr [MPa]

5 GMAW (3.5in offset) 89.262 89.448 0.0 -168.309 -71.40 Yes

6 GMAW (sequential) 89.590 89.156 0.0 -123.512 -71.40 Yes

7 SAW (3.5in offset) 88.838 89.436 0.0 -205.709 -71.40 Yes

8 SAW (sequential) 89.284 88.983 0.0 -146.997 -71.40 Yes

Table 5: Results for the Restrained Boundary Condition cases for Small Section

buckling analysis.

Free Boundary (As-weld) Condition

The critical stress (σcr ) in Table 4 corresponds to the first eigenvalue. Figures 8 and 9 illustrate

the residual stress distribution for the 3.5” offset and sequenced GMAW processes with free

boundary conditions, respectively. A possible mode of buckling for the as weld case is shown

in Figures 10. As listed in Table 4, in none of the cases, does the longitudinal stress at the edge

of the web exceed the critical stress value (458.90 MPa), thus, this condition will not cause any

buckling.

Restrained Boundary (Rammed-down) Condition

The residual stress values (-123.51 to -205.71 MPa) in the welding cases considered are greater

19
 Boundary Condition σL (MPa) dL (mm) λi

Free 6541.72 3238.90 0.020149

(Tp = - 0.05) 0.020149

0.020396

0.020399

0.020477

0.020483

0.020612

0.020614

0.020782

0.020803

Restrained -12184.75 0.0 0.000896

(Tp = 0.0) 0.003532

- 0.00553

0.005660

0.005663

0.005698

0.005708

0.005760

0.005788

0.005860

Table 6: 3-D Buckling Analysis Results of the Small Stiffener

20
 Case Weld Angular distortion Bowing distortion Residual Stress Critical Stress Buckling

θ1 (deg) θ2 (deg) d (mm) σr [MPa] σcr [MPa]

1 GMAW (3.5in offset) 89.95 89.70 3.796 19.950 528.844 No

2 GMAW (sequential) 90.00 89.73 1.990 10.455 528.844 No

3 SAW (3.5in offset) 89.84 89.68 5.173 27.191 528.844 No

4 SAW (sequential) 89.94 89.70 2.604 13.688 528.844 No

Table 7: Results for the Free Boundary Condition cases for Large section

Case Weld Angular distortion Bowing distortion Residual Stress Critical Stress Buckling

θ1 (deg) θ2 (deg) d (mm) σr [MPa] σcr [MPa]

5 GMAW (3.5in offset) 89.95 89.70 0.0 -34.531 -13.244 Yes

6 GMAW (sequential) 90.00 89.73 0.0 -19.421 -13.244 Yes

7 SAW (3.5in offset) 89.83 89.68 0.0 -45.797 -13.244 Yes

8 SAW (sequential) 89.94 89.69 0.0 -25.337 -13.244 Yes

Table 8: Results for the Restrained Boundary Condition cases for Large Section

than the critical stress of the restrained small stiffener (-71.40 MPa)(see Table 5). The critical

stress corresponds to the 10th eigenmode of the structure. This condition will result in buckling

distortions with a probability of buckling in a mode higher than the 10th eigenmode.

Figures 11 and 12 are the residual stress distributions for the simultaneous and sequenced

GMAW processes, respectively. The 4th and 10th eigenmodes for the rammed-down case are given

in Figures 13 and 14, to illustrate possible buckling modes. The higher modes have the same

characteristic shape due to the closely spaced eigenvalues, and the stiffener may buckle in any of

those shapes depending on the mechanical restraints.

21
 Boundary Condition σL (MPa) dL (mm) eigenmodes

mode eigenvalue (λ)

Free 4946.17 47.05 1 0.05692

(Tp = - 0.05) 2 0.057013

3 0.0574

4 0.057873

5 0.058315

6 0.059409

7 0.059976

8 0.060931

9 0.062619

10 0.06343

Restrained -7407.27 0.0 1 0.001788

(Tp = 0.0) 2 0.007010

3 0.011079

4 0.011119

5 0.011367

6 0.01167

7 0.012006

8 0.012395

9 0.01288

10 0.01337

Table 9: Results for the 3-D Buckling Analysis of the Large Stiffener

22
3.1.2 Buckling in Large Stiffener

Similar to small stiffener results, Tables 7 and 8 summarize the welding simulation, linear stress

and buckling analysis results.

Free Boundary (As-weld) Condition

The longitudinal residual stress distributions for the 3.5” offset and sequenced GMAW processes

are plotted in Figures 15 and 16, respectively. The first critical stress value (528.84 MPa) for the

as-weld large stiffener structure, corresponding to the first eigenvalue (0.05692) in Table 9, is an

order of magnitude higher than the residual stress values(10.46 to 27.19 MPa) for the four cases

(Table 7), which indicates that the structure in this configuration will resist buckling. A possible

mode of the free large stiffener is given in Figure 17.

Restrained Boundary (Rammed-down) Condition

Figures 18 and 19 are the residual stress values for the two welding sequences. For all four

welding cases, ramming-down will result in buckling distortions, since the residual stresses (Table

8) are higher than the critical stress (-13.24MPa). Moreover, the residual stress values for those cases

fall between the first and second critical stresses (-13.24MPa and -51.92MPa), thus the stiffener is

predicted to buckle in its 1st eigenmode shown in Figure 20 with the restrained boundary condition.

The weight of the structure may have an effect of causing a higher buckling mode, such as the one

given in Figure 21 (3rd mode).

23
4 Conclusions

This paper has presented a numerical analysis technique to predict the welding-induced distortion

in fabricated stiffeners. In particular, angular, bowing, and buckling distortion were evaluated

using thermo-mechanical and eigenvalue finite element analyses. The following conclusions can be

drawn based on the results obtained,

• Moderate angular distortions are predicted for all cases. The angular distortions computed

for GMAW and SAW are equivalent.

• For the small section geometry, bowing is large in magnitude (28.8 to 46.2mm, in Table 4). It

is considerably reduced in the large stiffener (2.0 to 5.2 mm, in Table 7), and can be ignored

in the scope of many applications.

• Sequenced welds lead to reduced distortion and residual stresses (Reductions from 46.2mm

to 33.8mm, and 93.38MPa to 68.28MPa), and possibility of buckling, in turn, is also reduced.

• Although it may cause large bowing distortion, the as-weld condition did not cause any

buckling distortion for both stiffeners.

• Buckling is predicted to occur when the stiffener is pressed (rammed) down on a straight

panel for future welding. As a consequence of straightening the web edge to correct the

bowing distortion, buckling becomes inevitable in both the small and large stiffeners. The

buckling pattern introduced particularly in the restrained small stiffener case (Figure 13), is

detrimental, and will make the future welds difficult to track along the distorted web.

• Employing large section rather than small cross section stiffener reduces the bowing distortion

and the probability of buckling by enhancing the structural rigidity. The residual stress values

are lower and the critical stresses are higher than those of the small section, both of which

24
 are in favor of reducing the susceptibility to buckling. The buckling pattern of the restrained

case for the large section as illustrated in the Figure 20 is also more tolerable than that of

the small section (Figure 14.

Figure 7: Residual Stress Distribution along the centerline (small stiffener)

4.1 Recommendations

Manufacturing process modifications can be utilized to minimize bowing distortion. Conventional

techniques such as reducing the welding heat input, weld size, or modifying the structural

parameters, might be employed to eliminate the occurrance of buckling [25, 26, 27]. When the design

considerations don’t permit such modifications, welding-distortions are remedied by utilizing special

manufacturing procedures during the fabrication process, such as thermal tensioning [28, 29, 30].

25
Auxiliary heating may also be applied to minimize the residual stresses; that will enable the

application of restrained boundary conditions to eliminate the bowing distortion without causing

any buckling.

References
[1] S.B. Brown and H. Song. Implications of Three-Dimensional Numerical Simulations of Welding of Large
Structures. Welding Journal, 71(2):55s–62s, 1992.
[2] P. Michaleris, Z. Feng, and G. Campbell. Evaluation of 2D and 3D FEA Models for Predicting Residual
Stress and Distortion. In Pressure Vessel and Piping Conference. ASME, 1997.
[3] A. P. Chakravati, L. M. Malik, and J. A. Goldak. Prediction of Distortion and Residual Stresses in
Panel Welds. In Computer modelling of fabrication processess and constitutive behaviour of metals,
pages 547–561, Ottawa, Ontario, 1986.
[4] J. Goldak and M. Bibby. Computational Thermal Analysis of Welds: Current Status and Future
Directions. In A. F. Giamei and G. J. Abbaschian, editors, Modeling of Casting and Weldin Processes
IV, pages 153–166, Palm Coast, FL, 1988. The Minerals & Materials Society.
[5] H. Hibbitt and P. V. Marcal. A Numerical, Thermo-Mechanical Model for the Welding and Subsequent
Loading of a Fabricated Structure. Computers & Structures, 3(1145-1174):1145–1174, 1973.
[6] J. H. Argyris, J. Szimmat, and K. J. Willam. Computational Aspects of Welding Stress Analysis.
Computer Methods in Applied Mechanics and Engineering, 33:635–666, 1982.
[7] E. F. Rybicki, D. W. Schmueser, R. B. Stonesifer, J. J. Groom, and H. W. Mishler. A Finite-Element
Model for Residual Stresses and Deflections in Girth-Butt Welded Pipes. Journal of Pressure Vessel
Technology, 100:256–262, 1978.
[8] V.J Papazoglou and K. Masubuchi. Numerical Analysis of Thermal Stresses during Welding including
Phase Transformation Effects. Journal of Pressure Vessel Technology, 104:198–203, 1982.
[9] H. Murakawa N. X. Ma, Y. Ueda and H. Maeda. FEM Analysis of 3-D Welding Residual Stresses and
Angular Distortion in T-type Fillet Welds. Transactions of JWRI, 24(2):115–122, 1995.
[10] A. S. Oddy, J. A. Goldak, and J. M. J. McDill. Numerical Analysis of Transformation Plasticity in 3 D
Finite Element Analysis of Welds. European Journal of Mechanics, A/Solids, 9(3):253–263, 1990.
[11] P. Tekriwal and J. Mazumder. Finite Element Analysis of Three-dimensional Tranient Heat Transfer
in GMA Welding. A.W.S. Welding Journal, Research Supplement, 67:150s–156s, 1988.
[12] P. Tekriwal and J. Mazumder. Transient and Residual Thermal Strain-Stress Analysis of GMAW.
Journal of Engineering Materials and Technology, 113:336–343, 1991.
[13] Y. Yeda and H. Murakawa. Applications of Computer and Numerical Analysis Techniques in Welding
Research. Transactions of JWRI, 13(2):165–174, 1984.
[14] E. F. Rybicki and R. B. Stonesifer. Computation of Residual Stresses due to Multipass Welds in Piping
Systems. Journal of Pressure Vessel Technology, 101:149–154, 1979.
[15] C.L. Tsai, S.C. Park, and W.T. Cheng. Welding Distortion of a Thin-Plate Panel Structure. A.W.S.
Welding Journal, Research Supplement, 78:156s–165s, 1999.
[16] Y. Ueda X. M. Zhong, H. Murakawa. Buckling Behavior of Plates under Idealized Inherent Strain.
Transactions of JWRI, 24(2):87–91, 1995.
[17] P. Michaleris and A. DeBiccari. Prediction of Welding Distortion. Welding Journal, 76(4):172–180,
1997.

26
[18] Y. Ueda, Y.C Kim, and M.G Yuan. A Predictive Method of Welding Residual Stress Using Source of
Residual Stress (Report I) Characteristics of Inherent Strain (Source of Residual Stress). Transactions
of JWRI, 18(1):135–141, 1989.
[19] D.F.Watt, L.Coon, M.Bibby, and C.Henwood. An algorithm for modeling microstructural development
in weld heat affected zones (part a) reaction kinetics. Acta metall., 36(11):3029–3035, 1988.
[20] J.B LeBlond and J.Devaux. A new kinetic model for anisothermal metallurgical transformations in
steels including effect of austenite grain size. Acta metall., 32(1):137–146, 1984.
[21] P. Michaleris and A. DeBiccari. A Predictive Technique for Buckling Analysis of Thin Section Panels
due to Welding. Journal of Ship Production, 12(4):269–275, 1996.
[22] D.F.Watt, L.Coon, M.Bibby, and C.Henwood. Coupled transient heat transfer-microstructure weld
computations (part b). Acta metall., 36(11):3037–3046, 1988.
[23] J. Goldak, A. Chakravarti, and M. Bibby. A New Finite Element Model for Welding Heat Sources.
Metallurgical Transactions B, 15B:299–305, 1984.
[24] P. Michaleris and X. Sun. Finite Element Analysis of Thermal Tensioning Techniques Mitigating Weld
Buckling Distortion. Welding Journal, 76(11):451–457, 1997.
[25] P. Michaleris and A. DeBiccari. A Predictive Technique for Buckling Analysis of Thin Section Panels
due to Welding. In 1996 Ship Production Symposium, 1996.
[26] K. Masubuchi. Analysis of Welded Structures. Pergamon Press, Oxford, 1980.
[27] K. Terai. Study on Prevention of Welding Deformation in Thin-Skin Plate Structures. Technical
Report 61, Kawasaki, 1978.
[28] Ya. I. Burak, L. P. Besedina, Ya. P. Romanchuk, A. A. Kazimirov, and V. P. Morgun. Controlling the
longitudinal plastic shrinkage of metal during welding. Avt. Svarka, (3):27–29, 1977.
[29] Ya. I. Burak, Ya. P. Romanchuk, A. A. Kazimirov, and V. P. Morgun. Selection of the optimum fields
for preheating plates before welding. Avt. Svarka, (5):5–9, 1979.
[30] P. Michaleris and X. Sun. Finite Element Analysis of Thermal Tensioning Techniques Mitigating Weld
Buckling Distortion. In Residual Stresses in Design Fabrication, Assessment and Repair. ASME, 1996.

27
Figure 8: Residual stress for the as-weld small section σzz (MPa) for case 1 (20X magnification)

Figure 9: Residual stress for the as-weld small sectionσzz (MPa) for case 2 (20X magnification)

28
Figure 10: Buckling mode of free small stiffener (3rd mode)

29
Figure 11: Residual stress for the rammed-down small section σzz (MPa) for case 5 (20X

magnification)

Figure 12: Residual stress for the rammed-down small sectionσzz (MPa) for case 6 (20X

magnification)

30
Figure 13: Buckling mode of rammed-down small stiffener (4th mode)

31
Figure 14: Highest buckling mode of rammed-down small stiffener (10th mode)

32
Figure 15: Residual stress for the as-weld large section σzz (MPa) for case 1 (50X magnification)

Figure 16: Residual stress for the as-weld large section σzz (MPa) for case 2 (50X magnification)

33
Figure 17: Buckling mode of the free large stiffener (1st mode)
34
Figure 18: Residual stress for the rammed-down large section σzz (MPa) for case 5 (50X

magnification)

Figure 19: Residual stress for the rammed-down large section σzz (MPa) for case 6 (50X

magnification)

35
Figure 20: Buckling mode of the rammed-down large stiffener (1st mode)
36
Figure 21: Highest buckling mode of rammed-down large stiffener (10th mode)

37