Finite Element Simulation of

Welding of Large Structures

S. Brown
Richard P. Simmons Associate Professor of
Materials Manufacturing.
Department of Materials Science and
Engineering,
Massachusetts Institute of Technology,
Cambridge, MA 02139
H. Song
Anadrill, Schlumberger,
105 Industrial Boulevard,
Sugarland, Texas 77478
Current simulations of welding distortion and residual stress have considered only
the local weld zone. A large elastic structure surrounding a weld, however, can
couple with the welding operation to produce a final weld state much different from
that resulting when a smaller structure is welded. The effect of this coupling between
structure and weld has the potential of dominating the final weld distortion and
residual stress state.
This paper employs both two- and three-dimensional finite element models of a
circular cylinder and stiffening ring structure to investigate the interaction of a large
structure on weld parameters such as weld gap clearance (fitup) and fixturing. The
finite element simulation considers the full thermo-mechanical problem, uncoupling
the thermal from the mechanical analysis. The thermal analysis uses temperature-
dependent material properties, including latent heat and nonlinear heat convection
and radiation boundary conditions. The mechanical analysis uses a thermal-elastic-
plastic constitutive model and an element "birth"procedure to simulate the dep-
osition of weld material. The effect of variations of weld gap clearance, fixture
positions, and fixture types on residual stress states and distortion are examined.
The results of these analyses indicate that this coupling effect with the surrounding
structure should be included in numerical simulations of welding processes, and that
full three-dimensional models are essential in predicting welding distortion. Elastic
coupling with the surrounding structure, weld fitiip, and fixturing are found to
control residual stresses, creating substantial variations in highest principal and
hydrostatic stresses in the weld region. The position and type of fixture are shown
to be primary determinants of weld distortion.

1 Introduction
Recent developments in nonlinear finite element methods
have provided the opportunity for simulation of complex weld--
ing processes. The vast majority of finite element simulations,
however, have concentrated on resolution of processes occur-
ring within the weld zone. These include investigations of such
field variables as state of stress (including both residual and
process), local distortion, temperature, microstructure, and
composition. Coupling of these variables has resulted in much
of the recent success in welding simulations. For example,
advances by Goldak [11] and Easterling [7, 5] in coupling
temperature fields with microstructure have provided some
initial steps in addressing the fully coupled welding process,
as well as in extending simulations into three-dimensions. The
finite element simulations associated with these investigations
involve very dense meshes with particular emphasis on reso-
lution within the weld zone, with the consequence that pre-
dicted gradients may be resolved with very high accuracy.
In contrast to these studies of local weld zone processes,
very little has been done to understand the significance of
interactions between the weld zone and the overall, elastic
Contributed by the Production Engineering Division for publication in the
JOURNAL OF ENGINEERING FOR INDUSTRY. Manuscript received Oct. 1989; As-
sociate Technical Editor: S. Lu.
welded structure. This is surprising, since the extent of the
interaction may be large enough to dominate the final distor-
tion and residual stress states, both in the structure and in the
local weld zone. Practical welding technology must deal with
problems associated with distortion in large structures. Ther-
mal expansion, lack of appropriate tolerances, variation in
weld gap clearance (or fitup), and location of welding fixtures
can all affect weld quality. The distortion during welding is a
particular problem, for the coupled thermo-mechanical char-
acter of welding becomes pronounced when thermal expansion
and distortion lead to changes in both the orientation of parts,
and the separation of weld gaps.
An experimental investigation of structure interactions with
welding [18] produced large deformations and local buckling
during the welding process. Such deformations would not have
developed in the butt welding of small plates with low ratios
of overall structure size to weld zone dimension. Welding of
large structures, therefore, involves problems which cannot be
investigated simply by examining the weld zone.
We have characterized the coupling between the local weld
zone and the surrounding structure according to the following
categories. These categories are not usually considered in weld-
ing analyses that, as mentioned previously, concentrate on local
weld zone phenomena:

Journal of Engineering for Industry NOVEMBER 1992, Vol. 114/441

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Copyright © 1992 by ASME
 (1) Elastic Coupling-Type 1
As the weld cools behind the travelling electrode, the newly
deposited weld material couples the two previously inde-
pendent parts, thereby changing the response of the structure
to the ensuing deposition of material further down the weld
path. The full effect of this behavior can be simulated only
through the use of three-dimensional models.
(2) Elastic Coupling-Type 2
The effect of heating and cooling of the overall structure
during welding may introduce large elastic distortions which
cannot be anticipated when examining only the weld zone.
These distortions are strongly dependent on the geometry
of the structure and greatly influence the final residual stresses
and state of distortion, within both the weld and the overall
structure.
(3) Large Deformations
Given the potentially large coupling between the weld and
other parts of a structure, portions of the structure may
move dramatically relative to each other, particularly during
the initial heating of the first weld pass. Large displacements
are an example of how large structure effects may give rise
to particular problems in the numerical analysis of welding
in large structures: These deformations can be large enough
to invalidate the uncoupling of the thermal and mechanical
analyses, since the thermal analysis based on the original,
undeformed configuration is no longer applicable to the
deformed shape.
This paper will examine the two structure/weld interaction
effects: weld clearance gap (fitup) and fixture setup. The results
of our work suggest that these effects have been underappre-
ciated as important determinants of distortion and residual
stress. A ring-stiffened cylinder will be used as the model struc-
ture in this investigation. We begin by looking at the general
model of finite element analysis of the welding process using
simulations that decouple heat transfer analysis from stress/
strain analysis, and discuss large structure effects. We then
present two-dimensional simulations of welding the cylinder
and examine the fixturing and fitup effects. Next is the three-
dimensional simulation of welding the same ring, this time
focusing on the fixturing effects. (The special interface ele-
ments necessary to analyze fitup effects in three-dimensional
model are not available.) Experimental confirmation of fix-
turing effects is then presented. Finally, we discuss remaining
issues and future work.
2 Finite Element Modeling of Welding Large Struc-
tures
The finite element models presented here address structure/
weld interactions. The article will present the results of both
two-dimensional and three-dimensional analyses. Although
three-dimensional analyses are preferable for large, complex
structures, we resorted to two-dimensional simulations in some
instances because of limitations in current three-dimensional
finite element modeling. Both the two- and three-dimensional
analyses illustrate the extent of coupling between the weld zone
and the overall structure.
The unresolved difficulties with three-dimensional finite ele-
ment analyses, which have retarded the analysis of large struc-
tures welding, concern primarily the implementation of the
three-dimensional model: A three-dimensional analysis dra-
matically increases the size of the model due to the increase
in degrees of freedom; a simple three-dimensional model can
overwhelm the computational capabilities of a mainframe
which could handle a two-dimensional geometry with relative
ease. In addition, finite element software frequently does not
easily accommodate special elements in three dimensions. The
absence of special elements to accommodate deforming, con-
ducting, and sliding interfaces, for example, inhibits the mod-
eling alternatives available for three-dimensional simulations.
442 / Vol. 114, NOVEMBER 1992
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Finally, there are difficulties in modeling the motion of the
welding electrode and deposition of weld material in three-
dimensional simulations. A separate paper will describe pos-
sible strategies to make three-dimensional analyses more tract-
able.
Whether a two- or three-dimensional model is used, we sim-
ulate the welding process using three process steps. The base
metal is first heated, and then at the end of the heating, weld
material is deposited; finally, the whole structure cools through
conduction, convection, and radiation. The deposition of ma-
terial is modeled by an element birth technique available in
the finite element code used for this investigation, ABAQUS
[1]. Elements of weld material deactivated at the beginning of
the heating are reactivated at the melting temperature of the
metal and left to cool after the weld electrode passes each weld
element [22].
The geometry of the weld structure is modeled in two parts-
the local weld zone and the rest of the structure, (denoted as
the far field). A denser mesh is used for the local zone, and
a coarser mesh for the far field. The particular mesh used here
represents a fairly coarse discretization. A finer mesh density
was also tested, where the side of each element in the coarser
model was halved, generating in the case of the axisymmetric
model four elements in place of one. The magnitude and dis-
tribution of stresses were found to be very close to those re-
sulting from the coarser mesh. Predicted distortions were
virtually identical. We therefore chose to use the coarser mesh
for computational convenience. All the analyses were per-
formed using the nonlinear finite element code ABAQUS [1]
on either a VAX-3500 workstation or an Apollo 10000 work-
station.
Once the model is constructed, we analyze deformation and
stress states of the structure by decoupling the analysis into
first a heat transfer process followed by deformation analysis.
The temperature fields from each time increment of the heat
transfer analysis provides the loading increments for the de-
formation analysis. This decoupling is only valid when the
relative displacements within the welding portions of the struc-
ture are small, assuring that displacements do not shift the
welding electrode position and consequently the location of
heat flux applied to the model.
2.1 Details of Heat Transfer Analysis. The following pa-
rameters and procedures were used in the welding heat transfer
analyses.
(1) Material Behavior: We model both the weld and base
metals as continuous solids with temperature-de-
pendent properties. The same properties are used for
both metals. The variations of thermal conductivity,
specific heat, thermal emissivity, and convective heat
transfer coefficient with temperature are given in Table
1. Different convective coefficients are used for the
vertical and the upper and lower horizontal surfaces as
discussed later. An averaged value of thermal emissivity
is used since ABAQUS does not permit the use of tem-
perature-dependent thermal emissivity. Rate-dep-
endent effects are currently ignored.
(2) Heat input: Heat transfer to the base metal from the
electrode arc is modeled by a surface heat flux on ele-
ments near the weld. The net power input is chosen
from the gross power and the arc efficiency. The power
distribution is generally assumed to be a Gaussian curve
following Tsai [24].
(3) Latent heat effect: The release or absorption of energy
upon solidification or melting is included through a
latent heat model available in ABAQUS, which mod-
ifies the specific heat to accommodate the extra heat
absorption or release [9]. The latent heat effect asso-
ciated with the solid-solid phase change, which is much
smaller than that associated with the solid-liquid phase
Transactions of the ASME
 Table 1 Thermal material constants used in finite element model. Ma- Table 2 Mechanical materials properties used in model. Material con-
terial constant values are interpolated when values are not given for a stant values are interpolated when values are not given for a temperature.
temperature.
ay E ay at
k C
P e h Temperature Yield Young's at strain Thermal expansion
Temperature Thermal Specific heat Emissivity Convective heat (Celsius) stress modulus of 1.0 coefficient
(Celsius) conductivity transfer coeff. (MPa) (GPa) (MPa) (10- 6 l/'C)
(W/m - ° K) (J/kg -' K) (W/m2 -° K) 0 290 314
200 10
0 51.9 450 .2 2.5 100 260 200 349 11
75 - 486 - 4.4 300 200 200 440 12
100 51.1 - .4 5.4 450 . 150 150 460 13
175 - 519 - - 550 120 110 410 14
200 49.0 • - .45 6.4 600 110 330 14
88
225 - 532 - - 720 9.8 20 58.8 14
275 - 557 - - 800 9.8 20 58.8 14
300 46.1 - .475 7.0
325 - 574 - -
1200 - 2 - 15
375 - 599 - -
1550 .98 .2 1.0 -
400 42.7 - - -
1600 .0098 .00002 .01 -
475 5000 .0098 .00002 .01 15.5
- 662 - -
500 39.4 - .54 7.7
575 - 749 - -
600 35.6 - - -
675 following assumptions and procedures were used in the me-
- 8.46 - - chanical, stress/strain analysis.
700 31.8 - - -
725 - 1432 - - (1) Material behavior: Material is modeled as thermo-elas-
750 - - .58 8.2 tic-plastic with isotropic strain hardening. Material
775 - 950 - - properties are assumed to be temperature-dependent,
800 26.0 - - -
27.2
including Young's modulus, yield stress, hardening,
1000 - .59 8.5
and coefficient of thermal expansion [12]. Table 2 shows
1500 29.7 400 - -
1540 - 400 - - the dependencies of material properties on temperature.
1590 - 847 .60 9.1 Young's modulus and yield stress are given small, finite
1840 - 847 - - values at high temperatures to avoid difficulties with
1890 - 400 - - numerical convergence.
2050 - 400 - -
5000
(2) The deposition of weld material: This is modeled, as
- 400 .65 10.0
described earlier, by reactivating elements which are
deactivated at the beginning of the analysis. The ac-
tivation procedure in the stress/strain analysis deforms
change, is included in the specific heat values given in the deactivated element from the original position so
Table 1. that it will be compatible to other parts of the structure
(4) Boundary heat loss: Boundary heat transfer is modeled and in a state of force equilibrium. Although this pro-
by heat convection and radiation. Convection follows cedure creates a stress/strain field in the newly bom
Newton's law, where the coefficient of convective heat element, the high temperatures of the element reduce
transfer is assumed to vary with both temperature and the effect of this virtual stress on the structure. Similar
orientation of the boundary [15]: treatment of deposited welding material can be found
in Tekriwal [23, 22],
kNu
K. (1) (3) Rate dependence effects: Rate dependence and micro-
L structure evolution are not considered in this analysis
Here k is the thermal conductivity, L the characteristic length because few high homologous temperature constitutive
of the plate (or surface), and Nu the Nusselt number defined models and material constants exist for these materials.
by: Research in this area is also very active [3, 6, 4, 9].
Ni = O.WP>/3Gi/3 Future work may include such effects since they are
relatively significant when the material is at a high
(Horizontal plate, facing upward) (2) homologous temperature and experiencing rapid tem-
iV„ = 0.54Ff 4 G; /4 perature changes.
(4) Plasma pressure: Although experiments indicate that
(Horizontal plate, facing downward) (3) pressure from the highspeed plasma on the base metal
/4 /4
7V„ = 0.517/>> G; affects the shape of the molten weld pool [16, 17], it
(Vertical plate) is ignored in this analysis because we do not model the
(4)
shape of the weld pool. Also, the effect of plasma
Pr is the Prandtl number and Gr the Grashof number, both pressure on the deformation of the whole structure is
of them are functions of ambient air properties and temper- relatively small.
ature differences between the wall and the environment [21].
Radiation is modeled by the standard Boltzman relation
3 Two-dimensional Simulation
gr=eo(T4-Tt) (5) Although full simulation of welding requires a three-di-
where e is the heat emissivity, a the Stefan-Boltzman constant, mensional simulation, certain aspects of weld/structure inter-
and Ta the ambient temperature. Radiation is assumed to be actions are usefully demonstrated using two-dimensional finite
from surface to ambient air; no radiative interaction between element models. The two-dimensional model described below,
surfaces is considered. The effect of radiation is smaller than although an approximation to the three-dimensional process,
that of convection except near the melting temperature. accentuates certain interactions, such as weld gap clearance
effects, that are currently difficult to demonstrate using three-
2.2 Details of Stress/Strain Analysis. Solution of the dimensional models.
temperature-induced stress/strain problem is derived from the
temperature history obtained from the thermal analysis. The 3.1 The Two-dimensional Finite Element Model. The ax-

Journal of Engineering for Industry NOVEMBER 1992, Vol. 114/443

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Temperature °C
1 200
2 520
3 840
4 II 6 0
5 1480
6 1800

152.4 mm

101.6 mm

457.2mm-

Fig. 2 Transient temperature distribution immediately after deposition

6.35mm of first weld on two-dimensional model.

is assumed to be 800 Watts/cm , applied to a surface of 0.4

Jib cm2 near the weld groove. The duration of heating is 15 seconds
Fig. 1 Axisymmetric FE mesh of the ring and cylinder structure. Radius which is equivalent to a travelling velocity of 5 cm/min. A net
of cylinder inner surface is 457.2 mm.
power of about 400 Watts/cm2 is also required to melt the
electrode continuously. Such values of working parameters are
typical for a welding process with a gross power of 2,000 Watts/
isymmetric finite element model of a ring-stiffened cylinder cm2 and an arc efficiency of 60 percent. Material properties
with boundary conditions is shown in Fig. 1. The continuum are those of mild steel and are given in Tables 1 and 2.
elements are four node linear elements. Such a model assumes Figure 2 shows the transient temperature distribution after
symmetry of the welding process where the weld is performed weld material is deposited. Discontinuities in the temperature
simultaneously about the circumference of the structure. This contour are artifacts of the contour-plotting software. These
assumption is rigorously valid only when the traveling velocity temperature/time histories then provide the loading histories
of the electrode is high enough to make the three-dimensional for the stress/strain analyses.
effect negligible. Such electrode velocity rarely, if ever, occurs
in actual welding. In spite of this approximation, features of 3.3 Two-dimensional Analysis of Stress/Strain. Plastic
structure/weld interactions can be investigated in two-dimen- deformation near the weld develops due to.the temperature
sional models. spatial gradients and material softening at elevated tempera-
The interaction between the ring and cylinder before any tures. After cooling, the structure retains a distorted config-
material is deposited is modeled by interface elements which uration and a state of high residual stress, particularly near
allow no interaction before contact and stick together after the weld path.
touching. Thermal expansion and deformation of the ring and Figure 3 shows the distorted geometry of the ring-stiffened
cylinder may cause the weld gap to close and the two com- cylinder predicted by the axisymmetric model after one weld
ponents to touch. The use of such interface elements and a has been deposited. A 3.2 mm weld gap clearance (fitup) is
slide line prevent any material overlapping or penetration, used in this model. One point on both the cylinder and ring
while sliding between the contacting surfaces is permitted. is pinned in the axial direction to remove rigid body displace-
ments. The pinned point on the ring is located at a distance
3.2 Two-dimensional Analysis of Heat Transfer. The heat corresponding to 3 element lengths (about 11 millimeters) from
transfer analysis for the welding of the cylinder follows the the weld center line. The upward movement of the ring has
general modeling procedures discussed in Section 2.1. Each of been confirmed by experiment [18]. The angular deflection of
the weld passes involves two steps, heating and cooling. The the ring results from the elastic deflection of the ring during
structure is heated near the notch with the weld material deac- heating, where the ring deflects up to relieve circumferential
tivated (dead); at the end of heating, elements inside the weld stresses developed by thermal expansion at the inner ring ra-
region are reactivated at a temperature close to the melting dius. The deposition of weld metal then fixed the structure in
point (1600°C in these cases) to simulate the deposition of that deflected orientation, and the subsequent contraction due
molten material. During the cooling step, the deposited ma- to the cooling of the molten metal is not sufficient to reverse
terial cools with the rest of the structure through conduction, the deflection. This deflection is opposite to that normally
boundary convection, and radiation. Conduction, however, obtained in a T-weld, particularly with flat plates, and is due
dominates the process. It takes several minutes for conduction to the axisymmetric form of the model. The thermoelastic
to equilibrate the cylinder and ring to a uniform temperature, distortion of this geometry therefore dominates over the near-
and a couple of hours for boundary heat transfer to bring the weld effects.
structure to room temperature. Boundary radiation contributes Figure 4 shows the corresponding residual Mises stress and
less than convection, except near the melting point. highest principal stress contours. The highest principal stress
The average net power input or heat flux to the base metal corresponds closely to the circumferential stress. High tensile

444 / Vol. 114, NOVEMBER 1992 Transactions of the ASME

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 residual stress components develop in the local area around
and inside the weld, particularly in the gap between the cylinder
and ring.
3.4 The Effect of Fitup. The relative orientation and
proximity of components to be welded has a significant effect
on the distribution of stresses near the weld. Welded com-
ponents are frequently asymmetric, and distort either due to
asymmetric elastic distortion or asymmetry of the welding
process. If interaction between the components (contact, for
example) occurs before deposition of weld material, the de-
formation pattern due to such interaction will obviously be
different from that when no contact occurs.
The ring-stiffened cylinder considered here illustrates the
sensitivity to fitup. The cylinder is more flexible radially than
the ring, and when heated, it expands more in the radial di-
rection. If contact occurs during heating, further expansion
of the cylinder is confined by the ring. Such incompatibilities
Displacement
Magnification
Factor
1 material properties, and geometry of the structure.
Fig. 3 Distorted geometry after first weld. 2D model, ring pinned at
node located 11.1 mm (7/16 in.) from the weld center line as shown.
M ses SI esstMPal
1 150
2 (BO
3 2 10
A 240
5 270
6 300
Area of
Detail
(a) Mises stress
Fig. 4 Residual stress distributions after first welding pass with fixture
removed. Two-dimensional model and 3.2 mm (1/8 in.) weld fitup.
Journal of Engineering for Industry
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
result in deformation and stress/strain fields different from
the case when no contact occurs. These differences are accen-
tuated after cooling. We present two cases here, with fitup
between the ring and cylinder being 3.2 mm (1/8 in.) and 0.4
mm (1/64 in.). There is no contact during heating for the
former case, while contact occurs in the latter before the end
of heating.
Figure 5 shows the stresses at integration points near the
free surface of the weld. Although there is little difference in
the Mises equivalent tensile stresses, large and important dif-
ferences result in other aspects of the final residual state of
stress in the weld.
Significant differences develop in principal stress state due
to fitup contact. The highest principal stress corresponds to
the circumferential stress components in this problem. The two
cases differ most in highest principal stresses along the free
surface of the weld. The highest principal stress values for the
case of 0.4 mm fitup is about 50 to 100 percent higher than
those for the case of 3.2 mm of fitup gap. We include highest
principal stress as an indicator of susceptibility to various fail-
ure modes, including hot cracking, brittle fracture, and fatigue
[8].
Mean normal or hydrostatic stresses in the 3.2 mm fitup
case are lower than those in the 0.4 mm case. More signifi-
cantly, the mean normal stresses are either negative or close
to zero in the former case but positive in the latter case. The
change of signs in the hydrostatic stress will change the state
of quality of the weld since a positive hydrostatic stress (neg-
ative pressure) favors void nucleation.
A fitup of 3.2 mm therefore appears significantly less sus-
ceptible to cracking or void initiation and growth than does a
fitup of 0.4 mm. The major difference between the two fitups
is the interaction in the gap between the cylinder and ring
during heating. The fitup case with 0.4 mm gap generates
contact between the ring and cylinder during heating. Other
FEM calculations show that the stress fields are similar in cases
with no gap contact. Cases where contact occurs develop sim-
ilar stress fields. The critical fitup gap that determines the
occurrence of contact or no contact is in the range of 0.8 mm
(1/32 in.) to 1.6 mm (1/16 in.) for this particular structure.
The value of critical fitup gap depends on the heat input,
The large difference in residual stress components in the case
of 0.4 mm of fitup may be explained by the contact of the
welded parts which occurs during heating. A portion of the
resulting elastic incompatibility remains after the structure
cools, which results in a state of higher residual stress. We
believe that the role of fitup on residual stresses and distortion
Highest Principal
Stress.fMPa)
100
140
180
220
260
300
340
380
420
460
500
(b) Highest principal Btress
NOVEMBER 1992, Vol. 114 / 445
 has not been recognized. Given the significant variation of
residual stresses, it is important to include fitup variation as
an integral part of weld simulation.
3.5 Fixturing. Although the analysis of fixturing may be
considered a subset of the larger topic of elastic coupling, we
consider here an example of fixturing to illustrate how ob-
stensibly small changes in the portion of the structure that
remains linear elastic (in other words far from the weld) can
lead to potentially very large variations in deformation and
stress response. Fixturing limits the deformation of a structure
300
200
S inder results when the outside of the ring is pinned. The mag-
100
0 displacement also alter the state of residual stress in the cooled
-100
1 2 3 to more accurately simulate a clamping fixture which restricts
Integration points close to free surface of tile weld
— a — Mises stress, 3.2 mm ( j inch) Fitup
" = 0 = Mises stress, 0.4 mm ( JJ inch) Fitup
• • • • •
Highest principal stress, 3.2 mm (|) Fitup
-•-
_ .•_ Highest principal stress, 0.4 mm ( J J inch) Fitup
Hydrostatic stress, 3.2 mm ( | inch) Fitup
Hydrostatic stress, 0.4 mm (gj inch) Fitup
Fig. 5 Values of residual stresses at integration points near the free
surface of the first weld, two-dimensional model. Integration points are
indicated in the weld elements in the inset drawing.
(a) Pinned near the weld (b)
as indicated
Fig. 6 The deformed shapes of two-dimensional models at end of heat-
ing for different fixture locations.
446 / Vol. 114, NOVEMBER 1992
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
during welding by the imposition of what can be considered
to be very stiff elastic constraints on the structure. The ap-
propriate location of fixtures plays an important role in the
final quality of welds, particularly in terms of distortion and
residual stresses. The following paragraphs consider different
aspects of the influence of fixtures.
Pinning of axial rigid body motion. Since an axisymmetric
finite element model requires that at least one node on each
independent structure be constrained in the axial direction to
prevent rigid body motion, we simulate the positioning of a
fixture via the required removal of one node's axial degree of
freedom. We fix both the cylinder and the ring in the axial
direction (direction 2 in the figures), but permit radial dis-
placement. After the first welding pass the cylinder and ring
are joined, requiring only one constraint.
The fixture or constrained node on the cylinder is placed at
its bottom edge, and the fixture on the ring is placed at different
positions along its lower surface. Figure 6 illustrates the relative
motion of the ring and cylinder at the end of the first heating
sequence for three positions of the ring fixture: near the ring,
at the middle of the ring, and at the outside edge of the ring.
Heat is applied in this simulation first to the upper weld.
Maximum relative displacement between the ring and cyl-
nitude of this motion carries implications for both the welding
process and modeling of that process. Large displacements like
that of Fig. 6(c) obstruct the use of automatic welding equip-
ment. Such large displacements also invalidate the use of a
decoupled thermal and mechanical analyses, necessitating cou-
pled thermo-mechanical procedures. The large differences in
weld zone due to the return of the elastic structure from its
distorted geometry. Relatively minor variations in fixture lo-
cation, therefore, carry significant consequences for the state
of deformation and stress in the final welded structure.
Clamping both sides of the ring. The model was modified
movement of both sides of the ring. The effect of this greater
constraint, imposed again through the elimination of the axial
degree of freedom on opposing nodes on the upper and lower
ring surfaces, is to reduce rotation of the ring.
Figure 7 illustrates the deformed shape of the cylinder/ring
structure after simulating the deposition of one weld pass.
Much less distortion results relative to that illustrated in Fig.
6, but greater residual stresses result.
Figure 8 contains principal stress contours for the clamped
Area of
Detail
A
(o>
Pinned at middle of web (c) Pinned at outside
edge of web.
Transactions of the ASME
 of a welding electrode in a three-dimensional structure con-
tinually alters the elastic coupling between the components
being welded, thereby affecting subsequent distortion. The
two-dimensional model used in this article to illustrate the
effect of structure/weld coupling accentuates distortions in the
plane of the model given that welding occurs simultaneously
around the entire circumference. This distortion should be
reduced in the actual three-dimensional structure since the
electrode more closely approximates a point source and since
<& the structure before and behind the electrode will constrain
motion. The three-dimensional model predicts the variation in
weld gap clearance as the electrode moves-a prediction not
possible in the two-dimensional model.
ratefe"

4 Three-dimensional Simulations
Area of As stated in the introduction, three-dimensional models are
Detail required to obtain realistic simulations; yet it is computation-
Displacement ally difficult to accommodate the large number of degrees of
Magnification im* freedom and the large number of iterations associated with
Factor: nonlinear material properties. Three-dimensional models of
1 welding processes, therefore, historically have been simple
models of butt welding of plates and cylinders.
Dashed Line: Original Mesh Argyris approximated a three-dimensional welding process
Solid Line: Distorted Mesh first by solving the two-dimensional problem for the thermal
after welding
fields associated with the motion of an electrode along a plate.
Fig. 7 Relative distortion of two-dimensional model for clamped ring He than applied the temperature-time history associated with
fixturing. Distortion represents condition after first weld pass and re-
moval of fixtures. a particular node of the plate model as the thermal history of
a two-dimensional analysis of the cross-section normal to the
weld path [2]. Karlsson [13] modeled the circumferential weld-
ing of a single pass, butt-welded pipe and found a substantial
circumferential variation of residual stresses and displace-
ments. Mahin et al. reported successful three-dimensional ther-
mal and stress/strain simulations of welding process on plate
[20]. Goldak et al. [11] examined the three-dimensional tem-
Highest Principal
Stress, (MPa)
perature, stress, and strain fields associated with the butt weld-
1 100
ing of a bar.
2 140
3 180 4.1 The Model. Figure 9 shows a mesh representative of
4 220 the three-dimensional analyses discussed in this article. The
5 260
6 300 weld path in this case is very short, approximately 2.5 centi-
7 340 meters long. The major portions of the cylinder and ring struc-
8 380 5 4 32 tures are modeled using shell elements, while the local weld
9 420 J_L zone is modeled by three-dimensional, eight-node brick ele-
10 4 6 0
1 1 500 ments. The model comprises approximately 4000 degrees of
freedom once the far field structure is represented by substruc-
Area of rp tures.
Detail

4.2 The Heat Transfer and Stress/Strain Analyses. Heat

Fig. 8 Residual highest principal stress contours after first weld pass.
Two-dimensional model with ring clamped.
fixture case. Higher principal residual stresses develop near
the free surface of the weld metal. The less constrained fixture
geometries, on the other hand, have reduced highest principal
stresses near the free surface, as shown in Fig. 4(b). Increased
constraint reduces the distortion while increasing the residual
stress by limiting how much the structure accommodates weld
pool shrinkage.
The two-dimensional analysis of the full three-dimensional
problem suggests that the structure to be welded can only be
fully represented by a full three-dimensional model. For ex-
ample, the angular deflection between the ring and cylinder in
Fig. 3 is a consequence of the model's axisymmetry. Motion
flux input is assumed to follow a two-dimensional Gaussian
distribution. Motion of the electrode is simulated in a step-
wise fashion, with the center of distribution being translated
by element increments. Material deposition is modeled using
an element birth technique, i.e., by activating the originally
deactivated elements. Element birth/death creates a tempo-
rary, unbalanced state and increases the number of required
iterations, adding a substantial computational penalty for spa-
tial resolution of electrode motion. Heat conduction across
the gap is ignored until the activation of weld elements in the
gap. Figure 10 contains transient temperature distributions in
the local weld zone on the outer surface of the cylinder and
the top of the ring at the time when the last weld element was
activated.
The mechanical analysis uses the same rate-independent con-
stitutive models used in the two-dimensional analysis. The
mechanical analysis for this short weld was still computation-
ally demanding, requiring 40 CPU hours on an Apollo 10,000
workstation. Strategies to reduce the computational require-
ments of large structure analyses will be described in a separate
article.

Journal of Engineering for Industry

NOVEMBER 1992, Vol. 114 / 447

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 -101.6mm

Temperature
(•C)
1 300
2 900
3 1500
4 2100
5 2400

Area of ( *L i i ./*
Detail l_
1 ' I

Fig. 10 Distribution of temperature on the outside cylinder surface from

Fig. 9 Finite element mesh of three-dimensional ring and cylinder three-dimensional thermal analysis, with net power of 2.5KW and elec-
structure. trode velocity of 17.5 cm/min.

4.3 Fixturing Effects. Special interface elements, which (a) Case I (renting on four locations 90 degrees apart) j ! 3 attached to the web
simulate the interactions between different parts in a large
structure, are required to examine the fitup effects in three-
dimensional models. Currently, these elements are not avail-
able in ABAQUS. The fitup effects are thus not examined in
three-dimensional models. —*— Permanent Constraints
C Removable Constraints
Three fixturing arrangements were analyzed for the follow- rrr\ Nonlinear Springs
ing three geometries: Case I, both the cylinder and ring rest ivw\ Linear Springs
on supports at 90 degree intervals; Case II, both are clamped (b) Case II (clamped at four locations) f 4 attached to the web
at the same 90 degree positions; and Case III, both are clamped
only in the immediate vicinity of the weld. These arrangements
were selected both to simulate actual fixturing arrangements
and to investigate extremes of fixturing constraint on the cyl-
inder and ring structure. These three cases are illustrated in
Fig. 11 where the different fixturing constraints are indicated
on the model mesh. Case I simulates the cylinder and ring
sitting on supports, free to move upward. Nonlinear springs (c) Case III (clamped close to the weld)
were used to simulate these constraints, with the upward spring
stiffness (5,000 N/m) approximately ten orders of magnitude
less than the downward stiffness (1.0 x 1012 N/m). Case II
fixes the degrees of freedom at the same points, simulating
rigid clamping. Case III represents a situation of strong con-
straint near the weld, by fixing both the cylinder and ring on
both sides of the weld and using nonlinear springs to simulate
supports around the rest of the structure. Fig. 11 Fixture setups of three different cases: (a) Case I, boundary
conditions 5, 6, and 7 initially prevent rigid body motions of the ring.
These models, identical except for the fixturing, were sub- These boundary conditions and springs 1, 2, 3, 4, and 10 are removed
jected to the same thermal history to simulate welding and at the end of the analysis, (6) Case II, boundary conditions are active
subsequent cooling to room temperature. The constraints were until their removal at the end of the analysis except 15 which is required
to prevent rigid body motions. Springs 12 and 13 will become active
then removed, leaving each model with identical constraints after the boundary conditions are removed at their corresponding nodes,
that removed rigid body modes. Figure 12 illustrates the re- (c) Case III, boundary conditions and springs are active until their removal
sulting distortion between the cylinder and the ring, magnified at the end of the analysis. Springs 23 and 24 are never removed.
to show the effect of the different fixturing schemes. The
greater distortion occurred in Case III where the fixturing
provides the most constraint about the weld. The least dis- this situation where the weld path is very short, and the fix-
tortion occured in Case I where the fixtures were least con- turing is symmetric about the weld. Second, the results of the
straining. distortion do not suggest that increased fixturing acts to reduce
The requirement of full, three-dimensional simulation is clear distortion: We would expect Case III to yield the least distor-
in these results. First, the distortion is asymmetric, even for tion since the fixtures provide the most constraint to the weld.

4 4 8 / V o l . 114, NOVEMBER 1992 Transactions of the ASME

Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 (a) Case I
Magnification
factor
40.0
i
§
Hfc
i-fftlllfl- i
(c) Cai
Fig. 12 Final displacement between the ring and cylinder at the indi-
cated locations when removable constraints are deactivated at analysis
end: (a) Case I-0.635 mm. (b) Case 11-1.90 mm. (c) Case HI-2.88 mm.
Instead, Case I with the least constraint also provided the least
distortion.
4.3.1 Experimental Correlations. A welding experiment
was performed to compare with the conclusions of the nu-
merical simulations. Similar to the numerical model, a cylinder
and ring stiffener were joined with a 2.5 centimeter weld de-
posited along the circumference of the cylinder-ring interface.
Two sets of fixture setups were used: In the first case (Case
I), both the cylinder and the ring sat on four supports on top
of a table. In the second (Case III in the numerical model),
the cylinder and the ring were tack welded at four points around
the local weld zone, similar to the numerical models.
The fitup gap between the cylinder and the ring is approx-
imately 5 millimeters, so that there is no contact between the
cylinder and the ring before the weld deposition, and the side
of the ring opposite to the weld can move freely up and down.
The power supply used in the welding was 26 Volts at 140
Amperes, which resulted in a gross power of 3.64 KW. The
electrode speed was controlled by an experienced welder so
that the total time elapsed for the weld was about 10 seconds,
approximately the same used in the numerical analyses. The
size of the weld was also controlled to match that of the model.
A depth gauge was used to measure the distance between
the top of the cylinder and the top of the ring at 36 locations
around the circumference of the cylinder-ring interface. Since
the ring was not perfectly flat before the welding, we measured
these distances before and after the welding process and cal-
culated the differences to obtain the distortion.
Gravity was not modeled in the numerical simulations since
gravity did not influence distortion during the experiment. In
the actual welding, there are two stages where gravity may
influence the distortion pattern. The first stage occurs during
the welding and the second stage occurs after the welding but
during the measurement of the distortions. In the first stage,
the gravity effect is not significant for two reasons: First, when
the local zone is hot, the weld metal has a low flow strength;
the local thermal expansion or contraction will therefore not
cause significant displacement far from the weld. Gravity only
assists the structure to sit on the supports. Second, when the
weld cools and contracts, the section of the ring opposite the
weld will tend to displace downward, prevented again by the
supports. Gravity again assists the structure to sit on the sup-
ports and does not affect the deformation pattern.
Journal of Engineering for Industry
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
ingular position from start of weld (degree)
Fig. 13 Welding distortion of the ring around the circumference of the
cylinder-ring interface. Four locations near the weld were tack welded
for a locally constrained fixture setup (Case III).
After the welding process and the removal of fixtures, gravity
can introduce elastic distortion of the structure that is com-
parable to the distortion expected from welding. If the cylinder
is oriented horizontally while the ring cantilevers freely about
the weld, the weight of the ring will generate a bending moment
on the weld and the ring will thus rotate about the weld. The
welded cylinder-ring structure was oriented vertically after the
welding to reduce this effect of gravity. Any remaining effect
of gravity was removed by first measuring the rotational stiff-
ness of the cylinder-ring structure about the weld, then deter-
mining the remaining moment due to the alignment of the ring
from vertical, and finally subtracting the calculated deflection
due to gravity from the total measured deflection.
Figure 13 shows the distortions in the case of locally con-
strained welding (Case III). The general trends of distortion
match each other closely. The discrepancy in the exact mag-
nitude of distortion can be attributed to many factors: the
differences between the model and the actual welding in power
in input, electrode speed, and the size of the weld as well as
the idealization of fixtures in the model. The measured dis-
tortion possesses a slight positive, upward distortion as well
as a less gradual downward distortion. The difference in down-
ward motion between the experiment in model suggests a local
stiffness in the region of the actual weld that inhibited a ro-
tation of the ring about the weld. The cause of this discrepancy
is not immediately obvious. One possible explanation is that
the model fixtures did not precisely duplicate the fixturing in
the experiment, the experiment fixtures introducing unmodeled
distortions.
For the case when the structure sits on four supports on top
of a table, the measured distortion shows about 50 percent
less than in the case when the local zone is tack-welded, similar
to that predicted in the numerical modeling. The only differ-
ence between these two cases was the arrangement of fixtures.
Two major conclusions can be drawn from the experimental
results. First, the fixture setup is very important in welding
distortion of large structures. Attempts to reduce distortion
through increased fixturing may instead act to increase dis-
tortion. Second, welding distortion of a large structure is com-
plicated and nonintuitive; numerical simulations are necessary
to predict such a distortion.
5 Conclusions
The effect of the coupling of the larger elastic structure to
the weld zone appears to be underappreciated in the welding
literature. The examples of fitup or weld gap clearance and
fixture variation discussed above illustrate the significant effect
of thermal distortion outside the weld zone. The contribution
of the elastic structure may indeed dominate the state of dis-
NOVEMBER 1992, Vol. 114/449
 tortion and stress in the final welded structure, producing a
final state much different from the one predicted by simulation
of the weld zone alone. The implication of such results is that
a reasonable simulation of welding must include a represen-
tation of the structure away from the weld.
The representation of the complete welded structure is nec-
essary to reproduce several phenomena. First, the elastic
boundary conditions on the weld zone represented by the larger
elastic structure may be much different from a simple clamped
or pinned boundary. Deformation within the weld zone de-
pends heavily upon these boundary conditions, and therefore
an accurate accounting of these boundary conditions is re-
quired. Second, the thermal elastic distortion of the structure
near the weld can dominate the final state of residual stress
and distortion in the welded structure. Inclusion of the elastic
structure in the thermal and mechanical analyses is therefore
needed to include the potentially controlling effect of these
thermal distortions. Third, simple boundary conditions such
as location and type of fixtures and weld gap variation may
dramatically affect the final state of the weld. Evaluation of
the true effect of these boundary conditions thus requires rep-
resentation of the entire structure.
Our evaluation of the two-dimensional approximation to
the full three-dimensional problem suggests that a structure to
be welded can only be completely represented with a full three-
dimensional model. Motion of a welding electrode in a three-
dimensional structure continually alters the elastic coupling
between the components being welded, thereby affecting sub-
sequent distortion. The two-dimensional model used in this
article to illustrate the effect of structure/weld coupling not
only cannot represent this effect, but it also accentuates dis-
tortions in the plane of the model, since welding is assumed
to occur around the entire circumference.
We draw the following conclusions from the finite element
analyses described in the preceding sections:
(1) The structure around a welded joint should be incor-
porated into a simulation of the welding process. Dis-
tortion and state of residual stresses resulting from
welding depend primarily on the character of the struc-
ture surrounding the weld. Much of this structural con-
tribution derives from thermoelastic distortion of the
metal surrounding the weld region.
(2) Variations in weld fitup can produce wide variations
in residual stresses and distortion. These variations re-
sult from closure of the weld gap due to thermal ex-
pansion of the overall structure. The variation in
residual stress state and distortion is binary in character
(closed versus unclosed gap), and selection of the ap-
propriate fitup gap during a welding design process can
make large differences in the condition of the final
weld. Analysis of this fitup effect should be included
in any analysis of a weld for design applications.
(3) Modeling of the entire welded structure is useful for
design of fixturing. Fixturing, like fitup, can alter dis-
tortion and residual stresses dramatically. Structural
models allow fixture design and positioning, conse-
quently permitting design of a complete welding proc-
ess.
(4) Full representation of this and other welding processes
requires a three-dimensional analysis. A two-dimen-
sional model, although tractable computationally, ex-
aggerates certain modes of distortion and ignores others.
A three-dimensional model also accommodates the ef-
fect of the elastic coupling of the welded components
as the weld metal cools behind the moving electrode.
The issue of computationally tractable three-dimen-
sional analyses remains problematic. The effect of
structure, however, indicates that finite element anal-
ysis of welding processes will not be a valid design
450 / Vol. 114, NOVEMBER 1992
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
procedure until three-dimensional analyses can be ef-
fectively performed.
6 Remaining Issues and Future Work
Although significant progress has been made in the finite
element modeling of welding, several areas of investigation
remain unsettled. The areas listed below represent our sense
of where the most promise exists for future modeling efforts:
9
Three-dimensional modeling
Inevitable technical compromises result from the two-di-
mensional models used in the welding simulations, compro-
mises which may prevent prediction of distortions or stresses
within actual welding processes. The interaction of the welding
electrode motion and component geometry, for example, can-
not be accommodated through two-dimensional models. Issues
which arise and are still active in three-dimensional modeling
include meshing compromises to make the analysis tractable,
the use and necessity of fully coupled thermo-mechanical anal-
yses, and procedures for simulating the motion of a welding
electrode and the sequential deposition of weld material.
• Thermal modeling
Although welding-induced temperature field simulation is
further advanced than the corresponding mechanical defor-
mation problem, open issues remain in the simulation of tem-
perature distribution in the weld zone. Thermal analyses of
the weld pool depend on a better understanding of how material
and energy are both transferred to and circulated within the
molten material. Convective and electro-dynamic effects on
pool size and geometry are not easily simulated.
A large component of the uncertainty associated with more
accurate thermal models comes from inadequate material data,
particularly regarding convective, radiative, and contact heat
transfer coefficients. Increased modeling accuracy is therefore
coupled to improve experimental methods to measure and rep-
resent boundary conditions.
• Constitutive models for deformation
Although the constitutive behavior of high temperature weld
metal is inherently rate dependent, most FE simulations of
welding distortion employ rate-independent constitutive
models, and are frequently elastic/perfectly plastic. Numerical
complications arise in the implementation of elastic/visco-
plastic models with hardening which more accurately represent
the rate-dependent deformations occurring in the weld zone.
In addition, there are very few models which encompass the
range of constitutive behavior in a weld zone, where defor-
mation mechanisms include power-law and exponential stress
dependence, Nabarro-Herring creep, and low temperature
plasticity. There is, therefore, a strong need for constitutive
models wide enough in their range of application, yet simple
enough for incorporation into a finite element analysis.
Coupled with the requirement for higher accuracy consti-
tutive models is the need for accurate material data. Rate
dependent deformation data is rarely available for common
engineering base alloys.
• Evolution of Microstructure
It is not possible to separate the usefulness of more accurate
thermal and mechanical constitutive models from the issue of
the evolution of weld zone microstructures. The distribution
of phases and the kinetics of their formation dramatically
affect the thermal and mechanical response of the weld zone.
More sophisticated constitutive models for material properties,
whether they pertain to thermal or mechanical properties, will
inevitably have to include both the kinetics of microstructure
evolution and the coupling between microstructures and those
properties. One obvious example of this coupling is the vari-
ation in the rate dependent deformation behavior of different
regions of a weld depending on the developed microstructures.
Another more subtle effect is that of transformation plasticity
Transactions of the ASME
 resulting from phase transformations during cooling within
the weld zone [10, 14].
8 7 Easterling, K. E., 1986, "Predicting Heat-affected Zone Microstructures
Design Technology
The use of finite element modeling and other simulation
methods to assist in the design of welding processes has been
ignored to a large extent. Much of the attention in the finite
element modeling of welding is directed towards greater re-
finement and resolution in simulating the processes occurring
within the zone of weld metal. The consequence of this con-
tinued increase in model complexity is to produce models which
continually strain current computational limits. Little work
has been directed towards appropriate sensitivity analyses to
determine the level of complexity appropriate to design anal-
yses, where computational complexity is exchanged for the
ability to use the models in normal design activities. Perhaps
the greatest possibility for practical advances exists in this area,
where the benefits of finite element modeling can be used in
the design of actual welding processes.
Acknowledgment
This research project has been funded by the Office of Naval
Research under contract number N00014-89-J-1187. Support
was also provided by the Materials Processing Center at MIT
and the Apollo Computer Company. The authors also express
appreciation for the suggestions and criticism of Prof. Thomas
W. Eagar of the Department of Materials Science and Engi-
neering at MIT.
References
1 ABAQUS Finite Element Code, Hibbett, Karlsson, and Sorenson, Inc.,
Providence, Rhode Island, version 4. edition.
2 Argyris, J. H., Szimmat, J., and William, K. J., 1982, "Computational
Aspects of Welding Stress Analysis," Computer Methods in Applied Mechanics
and Engineering, Vol. 33, pp. 635-666.
3 Brown, S. B., Anand, L., and Kim, K. H., March 1989, "An Internal
Variable Constitutive Model for the Hot Working of Metals," International
Journal of Plasticity, pp. 95-130.
4 Denis, S., Sjostrom, S., and Simon, A., 1987, "Coupled Temperature,
Stress, Phase Transformation Calculation Model, Application to the Prediction
of Phase Transformation During Cooling of an Eutectold Carbon Steel," Re-
sidual Stresses in Science and Technology, pp. 595-602.
5 Easterling, K. E., 1983, Introduction to the Physical Metallurgy of Weld-
ing, Butterworths.
Journal of Engineering for Industry
Downloaded From: https://manufacturingscience.asmedigitalcollection.asme.org on 01/16/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use
6 Easterling, K. E., 1987, "Modeling the Weldability of Structural Steels,"
Welding Metallurgy of Structural Steels, pp. 45-62.
and Properties in Fusion Welds," Advances in Welding Science and Technology,
pp. 177-185.
8 Goldak, J., May 1989, "Modeling Thermal Stresses and Distortions in
W'elds," Second International Conference on Trends in Welding Research, Gat-
linburg, Tennessee, USA, pp. 71-82.
9 Goldak, J., Bibby, M., Moore, J., House, R., and Patel, B., September
1986, " Computer Modeling of Heat Flow in Welds,'' Metallurgical Transactions
B, pp. 587-600.
10 Goldak, J., Oddy, A., Gu, M., Ma, W., Mashaie, A., and Hughes, E.,
June 1991, "Coupling Heat Transfer, Microstructure Evolution and Thermal
Stress Analysis in Weld Mechanics," In L. Karlsson, ed., Mechanical Effects
in Welding, International Union of Theoretical and Applied Mechanics.
11 Goldak, J., Oddy, A., McDill, M., and Chakravarti, A., May 1986, "Prog-
ress in Computing Residual Stress and Strain in Welds," In International Con-
ference on Trends in Welding Research, Gatlinburg, Tennessee, USA, pp. 523-
527.
12 Iwamura, Y., and Rybicki, E. F., February, 1973, "A Transient Elastic
Plastic Thermal Stress Analysis of Flame Forming," ASME JOURNAL OF E N -
GINEERING FOR INDUSTRY, pp. 163-171.
13 Karlsson, R. I., and Josefson, B. L., 1988, "Three-dimensional Finite
Element Analysis of Temperature and Stresses in a Single-pass Butt Welded
Pipe," Supplied by Author.
14 LeBlond, J. B., Devaux, J., and Devaux, J. C , 1991, "Mathematical
Modelling of Transformation Plasticity in Steel," International Journal of Plas-
ticity, in press.
15 Lewis, R. W., andMorgan, K., 1985, Numerical Methods in Heat Transfer,
Vol. 3, John Wiley & Sons.
16 Lin, M. L., and Eagar, T. W., June 1985, "Influence of Arc Pressure on
Weld Pool Geometry," Welding Journal, pp. 163-169.
17 Lin, M. L., and Eagar, T. W., September 1986, "Pressures Produced by
Gas Tungsten Arcs," Metallurgical Transactions B, Vol. 17B, pp. 601-607.
18 Machuca, B., June 1988, "Distortion and Residual Stresses in a Welded
Stiffened Ring," Master's thesis, Massachusetts Institute of Technology.
19 Mahin, K. W., June 1991, "Prediction and Verification of Residual Stress
Distributions in Fusion Welds," In L. Karlsson, ed., Mechanical Effects in
Welding, International Union of Theoretical and Applied Mechanics.
20 Mahin, K. W., and Shapiro, A. B., and Hallquist, J., May 1986, "As-
sessment of Boundary Condition Limitations on the Development of a General
Computer Model for Fusion Welding," In International Conference on Trends
in Welding Research, Gatlinburg, Tennessee, USA, pp. 215-223.
21 Rohsenow, W. M., andHartnett, J. P., 1973, Handbook of Heat Transfer,
McGraw-Hill Book Company.
22 Tekriwal, P., and Mazumder, J., July 1988, "Finite Element Analysis of
Three-Dimensional Transient Heat Transfer in gma Welding,'' Welding Journal,
pp. 150-156.
23 Tekriwal, P., and Mazumder, J., May 1986, "Finite Element Modeling
of Arc Welding Processes," Advances in Welding Science and Technology, pp.
71-80.
24 Tsai, N. S., and Eagar, T. W., Dec. 1985, "Distribution of the Heat and
Current Fluxes in Gas Tungsten Arcs," Metallurgical Transactions B, Vol. 16B,
pp. 841-846.
NOVEMBER 1992, Vol. 114 / 451