Ihab F. Z.

Fanous
Mem. ASME
3-D Finite Element Modeling of
e-mail: fanous@aucegypt.edu

Maher Y. A. Younan
the Welding Process Using
Mem. ASME
Professor of Mechanics and Design,
Element Birth and Element
e-mail: myounan@aucegypt.edu

Mechanical Engineering Department,

Movement Techniques
American University in Cairo,
Cairo, Egypt
The modeling and simulation of the welding process has been of main concern for differ-
ent fields of applications. Most of the modeling of such a problem has been mainly in 2-D
forms that may also include many sorts of approximation and assumptions. This is due to
limitations in the computational facilities as the analysis of 3-D problems consumes a lot
Abdalla S. Wifi of time. With the evolution of new finite element tools and fast computer systems, the
Professor
analysis of such problems is becoming in hand. In this research, a simulation of the
e-mail: aswifi@hotmail.com
welding process with and without metal deposition is developed. A new technique for
Mechanical Design and Production Department,
metal deposition using element movement is introduced. It helps in performing full 3-D
Cairo University,
analysis in a shorter time than other previously developed techniques such as the element
Cairo, Egypt
birth. 关DOI: 10.1115/1.1564070兴

Introduction applies only for a straightline welding. However, Dong’s tech-

nique could give a general overview of the residual stresses due to
Many models for the welding simulation have been developed
welding.
in the past few years. Most of the models had to include some Hibbit and Marcal 关6兴 have performed an advanced modeling
approximations in order to avoid long computing time and geo- procedure for a complete 3-D simualtion of the welding process.
metrical nonlinearity. Also, most of them were intended for spe- The loading of the welded structure due to operation conditions,
cial applications in which reducing the model from 3-D to 2-D, temperature dependent material properties and phase transforma-
for example, could be a valid assumption. On the other hand, tion were accounted for in their model.
some 3-D models were developed in which approximations were Friedman 关7兴 developed a comprehensive two-dimensional
applied to the material behavior at elevated temperatures, and analysis. Noting that the temperature profile does not vary with
without including the metal deposition. time but moves at constant speed along the welding line, the prob-
Nguyen, Ohta, Matsuoka, Suzuki and Maeda 关1兴 have devel- lem size was reduced so as to evaluate the temperature profile at a
oped an analytical procedure for evaluating the transient tempera- section perpendicular to the welding line.
ture profile during the welding process. They used Goldak’s 关2兴 Detailed analyses for 2-D and 3-D welding using the ADINA
formulation of the double-ellipsoidal heat source and compared software were shown by Wilkening and Snow 关8兴. They consid-
the results with an experiment that he conducted. They assumed ered most of the non-linear aspects of the analysis including
that there is no heat loss neither through convection nor radiation temperature-dependent material properties and metal deposition.
from the surfaces of the plate, which lead to some discrepancy in The element birth technique was used to simulate the metal depo-
temperature predictions. sition. The moving heat source was modeled using a time curve
To calculate the residual stresses, Ueda and Yuan 关3兴 and Mo- for nodal heat flux.
chizuki and Hattori 关4兴 used the inherent strain method. It is based In the present work, a new technique of element movement is
on the assumption that the inherent strain of a complicated welded developed for full 3-D simulation of the welding process. The
structure can be approximated by another of a similar simpler standard commercial code ABAQUS is used to model this prob-
structure. The inherent strain is affected by other parameters such lem as it has been used extensively in such a highly nonlinear
as the material of the base metal, the difference between the ma- problem in previous researches. In the first model considered in
terial of the base metal and that of the filler, the welding speed, the this research, the dimensions and thermal load values are acquired
amount of heat input, etc. Hence, to use the inherent strain tech- from the research done by Friedman 关7兴 for comparison and veri-
nique, a database of the profile for different parameters should be fication. In this case, simple butt welding of two coplanar plates
developed. However, for a complicated structure, the use of inher- without considering metal deposition is simulated. In the other
ent strain method becomes inaccurate. two models that are considered in this research, larger thicknesses
Dong 关5兴 has developed a model utilizing the element birth are assumed to include the effect of adding filler material in the
technique to simulate the metal deposition that is valid in the weld pool between the plates.
cases with no thermal analysis of the sudden large temperature
variation. This model is a combined two 2-D model with a small Finite Element Models
allowable variation in the results along the thickness. This concept In developing a general purpose model for the welding process,
eliminates the effect of the heat transfer through the thickness of it is important to consider the moving heat source, heat loss,
the plate, which is of great effect in case of the thick plate weld- temperature-dependent material properties and metal deposition.
ing. The technique used for developing a cross-sectional model A moving heat source is modeled by setting a heat flux distribu-
tion that varies with time applied to the top surface of the weld
Contributed by the Pressure Vessels and Piping Division and presented at the pool zone. Using ABAQUS, it is simulated by developing a user-
Pressure Vessels and Piping Conference, Vancouver, BC, Canada, August 4 – 8, 2002,
of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript received by
defined subroutine to which time and position of interest are
the PVP Division April 26, 2002; revised manuscript received February 5, 2003. passed as parameters and returns the heat flux accordingly. Simi-
Associate Editor: S. Y. Zamrik. larly, another subroutine is developed to simulate the heat lost

144 Õ Vol. 125, MAY 2003 Copyright © 2003 by ASME Transactions of the ASME

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 1 The diagram of the welding process of case 1

from the top surface of the body. These subroutines are verified in
‘‘Model 1’’ which simulates the welding process of two plates, Fig. 3 The moving heat source of Model 1
without considering metal deposition, and compares to the results
of Friedman 关7兴. Also, the material properties are entered to the
model for different temperature values as described later in this
The total heat input Q is evaluated according to the type of heat
paper. The verified subroutines of the heat load and heat loss and
source. For example, in electric arc welding
material data are then used to simulate another welding process
that includes the addition of filler material to the base plate. Ini- Q⫽ ␩ VI (2)
tially, ‘‘Model 2’’ is developed to simulate the process utilizing
This distribution, according to 关7兴, represents 95% of the total heat
the element birth technique to simulate the metal deposition as has
Q when applied within a circle with radius r b . The distance r in
been used in previous researches. This model shall be used as a
Eq. 共1兲 shown in Fig. 4 is the distance from the center point of the
reference for verification of ‘‘Model 3’’ utilizing the element
heat source to the point for which the heat flux is being calculated
movement technique that is developed in this research.
and is given by
Verification Model „Model 1…. The model simulates basic
arc welding of two coplanar plates along the parting line as illus- r⫽ 冑共 x⫺x h 兲 2 ⫹y 2 (3)
trated in Fig. 1. The model is developed similar to that of 关7兴 so as where
to be able to verify the subroutine of the heat source and heat loss
in comparing the thermal history, and the structural boundary con- x h ⫽ 共 t⫺t 0 兲v (4)
ditions in comparing the residual stresses. The value of t 0 is the time taken for the center point of the heat to
Each plate has a length of 100 mm (x direction兲, width of 50 reach the first node along the welding line that has a value of zero
mm (y direction兲 and height of 2.5 mm. The welding speed is 2 in Model 1. This way, as the time increases, x h increases simulat-
mm/s. The electric input is 24 V and 30 A, and the arc efficiency ing the motion of the circle of the heat load zone as shown in Fig.
is assumed to be 90%. The process is modeled using one plate 3. Therefore, when the value of r is less than or equal to r b , the
upon which symmetry loading and boundary conditions are ap- heat flux is calculated according to 共1兲. Otherwise, the heat load is
plied. Parametric meshing is used in order to easily track the set to zero. r b is set to 5 mm according to 关7兴.
results along a certain predefined path in any direction. The thermal boundary conditions include the radiation and con-
The element used is an 8-noded brick element that can perform vection to the environment from all sides of the welded plate
a coupled displacement-temperature analysis. For thermal symme- except the symmetry surface and the area upon which the heat is
try, the heat flux passing across the surface of symmetry shown in applied. For all sides of the plate that lose heat, the heat lost is
Fig. 2 is assumed to be zero, and, for structural symmetry, the calculated by
translation in the y direction of the same surface is also zero.
Besides, for structural stability of the model, fixture point 1 is q⫽h convection共 T i ⫺T a 兲 ⫹␧ em␴ bol共 T i4 ⫺T 4a 兲 (5)
constrained in the x and z directions, and fixture point 2 is con- 2
The connection coefficient is 8 W/m °C, which is assumed to be
strained in the z direction. constant as it depends primarily on the ambient temperature, and
The moving heat load is applied as distributed heat flux to the the emissivity is 0.5.
top surface of the model, Fig. 3. The region within which the heat Another subroutine called FILM is developed in 关9兴 in conjunc-
is applied has a circular shape assuming the heat source is applied tion with DFLUX to account for the variation of the heat loss
perpendicularly to the plate without any inclination. A user sub- coefficients with time for the top surface. In this code, the location
routine named DFLUX is developed in 关9兴 using the FORTRAN under consideration is checked if it lies within the circle of appli-
language and included in the model to calculate the heat flux at a cation of the load using Eqs. 共3兲 and 共4兲, and, if it is true, there is
certain time and location within the surface of elements upon no heat loss. Otherwise, heat is lost by the same coefficients as the
which the load is applied according to Eq. 共1兲. other sides from the area of the top surface other than that of the
3Q 2
q共 r 兲⫽ e ⫺3(r/r b ) (1)
␲ r 2b

Fig. 2 General boundary conditions Fig. 4 Zones of heat load and heat loss

Journal of Pressure Vessel Technology MAY 2003, Vol. 125 Õ 145

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 6 General boundary conditions

The heat source in this model is similar to that used in Model 1.

However, it initially stays a while at the beginning of the welding
process for the filler material to start melting. Then the heat flux
starts to gradually impose its effect as shown in Fig. 5. Note that
there is no heat load at the initial state. This effect is modeled by
using the DFLUX and FILM subroutines as developed in Model 1
having the initial time t 0 greater than zero,
Fig. 5 The moving heat source of Model 2 rb
t 0⫽ (8)
v
At time zero, x h equals to ⫺r b . This makes the heat source just
heat load. The coefficient calculated in the FILM subroutine is a out of the model at the beginning of the analysis with the first
combination of both the convection and radiation coefficients group of elements being active. As the heat source moves and
which is given by starts to go out of the current active element slice, the next slice of
elements is activated, as illustrated in Fig. 5. The total heat input
h total⫽h convection⫹␧ em␴ bol共 T i3 ⫹T i2 T a ⫹T i T 2a ⫹T 3a 兲 (6) rate is 1300 W.
such that the total heat loss in Eq. 共5兲 becomes The analysis procedure is divided into several steps. In each
step, a new group of elements is activated. This means that every
q⫽h total共 T i ⫺T a 兲 (7) step performs analysis over a time period t i ⫽x i v where x i is the
length of the group of elements 共along the x direction兲 to be
Element Birth Technique „Model 2…. To broaden the appli- activated in that step. A step is added when all elements are acti-
cation of the modeling procedure, the problem of Model 2 simu- vated to allow for heat loss only over a long period of time 共40
lates arc welding of two coplanar plates with the addition of a min兲 simulating cool down. Finally, an extra step is added in
filler material between them. By developing several trial models, which the fixation of the far surface is released, as mentioned
it was found that a length of 100 mm (x direction兲, a width of 50 before, to be able to check for the residual stresses at no load.
mm (y direction兲 and a thickness of 5 mm (z direction兲 are ad- It is important to note that the automatically estimated time
equate for a single pass welding process. The welding speed is increment in the analysis drops to a very small value due to the
assumed to be 1 mm/s. vast difference in the temperatures of consecutive nodes in the
The metal deposition of the filler material is considered using reactivated elements. This effect can be dramatic for small ele-
the element birth technique. This technique is based on deactivat- ment sizes of the weld pool at the fusion surface 共the parting
ing and reactivating the elements of the weld pool as the welding surface兲. In addition, the part of the heat passed to the base plate
progresses. The meshing of the base plate and the weld pool has a is neglected in this model. Therefore, it is important to gradually
clear parting surface between them. That way, when the elements transfer heat from the filler material to the base plate to smoothen
of the weld pool are deactivated, the remaining elements would the contact between the two parts.
have the initial shape of the base plate with its modified edge that
forms the cavity of the weld pool. The meshing in the weld pool is Element Movement Technique „Model 3…. The meshing in
fine enough to account for the high temperature gradient calcula- this model is similar to that in Model 2. However, the elements of
tions and that in the base plate has similar meshing characteristics the weld pool are separated from those of the base plate so as to
described in Model 1 with the elements near the fusion surface be free to move as shall be discussed in the next section. Paramet-
having a size close to the ones in the weld pool. ric meshing is used in both parts in order to be able to handle
The elements defining the weld pool are grouped to form slices. nodes along a certain path or on a certain plane within the two
Having the elements of the weld pool initially deactivated, every parts. In order to impose the gradual heat transfer effect, the part
slice is then reactivated as the heat sources moves along the weld- of the weld pool is shifted in the z direction a certain distance
ing line as shown in Fig. 5. The nodes that appear due to the from the base plate as shown in Fig. 7. This way, the thermal and
reactivation of the elements have an initial temperature above the
liquidous temperature.
The structural boundary conditions described in Model 1 are
not sufficient for the base plate since it is free to deform in the y
direction. The welding process is assumed to be in a large struc-
ture that can allow for slight movement of the far surface in Fig.
6, but with no rotation about any axis which is simulated by
forcing the displacement of all the nodes on the far surface to be
the same. This is modeled by using a constraint equation that
couples the y component of the displacement of each two con-
secutive nodes on the surface. Therefore, with reference to Fig. 6,
the y displacement of node 1 is set to be equal to that of node 2
and that of node 2 is set to be equal node 3, etc... Fig. 7 Gap clearance

146 Õ Vol. 125, MAY 2003 Transactions of the ASME

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 8 „a… The effective area, and „b… thermal conductivity ver-
sus the gap clearance of the gap elements

structural interaction between the two parts is made dependent on

the distance d between them.
To avoid large skewing of the elements due to the moving ones
simulating metal deposition, the initial gap between the weld pool
and the base plate is set to be small compared to the element size
in the z direction. The type of element used is the same as that in
the previous models with structural and thermal degrees of free-
dom in order to perform coupled displacement-temperature Fig. 9 The steps of the element movement technique
analysis.
Having the weld pool and the base plate parts separated re-
quired the introduction of gap thermal and structural interaction
weld pool are allowed to penetrate the elements of the base plate
between the two bodies. The gap links, named GAPUNIT in
共a criteria know as over-closure in ABAQUS explained in 关10兴兲
ABAQUS, are used between the two parts to take care of the
when they are moved towards it.
interaction by joining adjacent nodes together. These links are
When modeling the element movement, the nodes lying on the
two-noded elements and are modeled such that they allow heat to
same y-z plane move together towards the base plate by applying
be transmitted between nodes of the weld pool and the adjoining
translational boundary condition equivalent to the initial gap
nodes of the base plate, which implies that the meshing of the
clearance. With reference to Fig. 9, the nodes that are yet to move
adjoining surfaces must be the same in the horizontal and vertical
in the following steps are held in position in order not to allow
directions of the surfaces as illustrated in Fig. 7. The gap elements
any deformation in the remaining portion of the weld pool. In this
are assigned thermal conductivity, which is at the room tempera-
case, the elements formed by the nodes that moved towards the
ture. Its effective area which is the contact area that each gap
base plate and the next group of nodes posses some strain. There-
element represents on the contact surfaces is shown in Fig. 8(a).
fore, the group of nodes that has just reached the base plate is held
To avoid heat to be transmitted from the weld pool to the base
in position until those of the next step follow them. This way, the
plate before any deposition, the thermal conductivity of the gap
strain in the translated elements shall tend back to zero. When the
elements are initially set to zero. However, since the base plate is
second group of nodes is lowered, the first one is released to
at room temperature and the weld pool is at the melting point, the
deform freely without being affected with the strain previously
conductivity cannot be set to its maximum value just at the time of
generated in neighboring element. In order to reduce the strain
contact since there will be coincident points having temperatures
generated in the element, the nodes move only half of the gap
greatly different. This may cause problems in the automatic incre-
distance every step, which implies that the nodes reach the base
menting process since a major parameter for calculating the time
plate after two subsequent steps. The welding speed is the length
increment is the maximum change in temperature per increment.
This should not exceed a certain limit selected to be 200°C 共larger
value of the temperature change limit may cause loss of accuracy
and smaller value would lead to longer analysis time兲. Therefore,
the thermal conductivity of the gap elements gradually increases
from the initial value of zero at a certain gap clearance to the
maximum value at zero clearance according to the values shown
in Fig. 8(b). This way, when the elements of the weld pool reach
those of the base plate, the temperature of the two coincident
nodes shall become approximately equal. This early interaction
accounts for the heat lost from the molten metal to the base plate
while falling into the weld pool. Also, by increasing the value of
thermal conductivity in the range before contact, the transferred
amount of heat can be increased and, thus, account for heat being
transmitted from the heat source to the base plate directly. This is
important in the model simulating arc welding where the heat is
generated through the electric arc between the electrode 共the filler
material兲 and the base plate. Also, in gas welding, there is some
heat not subjected to the filler material and is applied directly to
the base plate. However, it is necessary to keep the conductivity at
zero clearance to be at a value similar to that of the base plate
material so that the flow of heat across the gap element would be
equal to that through the material itself.
Due to the possible expansion of the base plate, elements of the Fig. 10 The moving heat source of Model 3

Journal of Pressure Vessel Technology MAY 2003, Vol. 125 Õ 147

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 11 Zones of heat load and heat loss

of each element divided by the time of each step. When the nodes Fig. 13 Stress distribution along the midsection
of the weld pool reach the base plate, a coupling equation is
activated between the coincident nodes of both parts to simulate
the fusion process. These coupling equations force the deforma-
tion of the coincident nodes to be equal. Hence, the weld pool and with the base plate. In other words, since the first group of nodes
the base plate act as one body at this point. A user subroutine reaches the base plate in two consecutive steps, the value of t 0
MPC is developed in 关9兴 to activate the coupling between every shall be twice the time taken for each step, which is 2 s for the
group of nodes that come into contact at a specific time according model in hand. Figure 10 shows the motion of the heat load whose
to the welding speed. In this subroutine, the x-coordinate of every center 共the darkest point兲 reaches the first node of the welding at
group of nodes is checked if it is less than the x-coordinate of the time⫽4 s.
center point of the heat source, and, if it is true, the coupling Finally, the thermal heat loss is modeled such that heat is lost
equation is activated. Usually, the fusion at different points be- from all sides of the plate and the weld pool except those surfaces
tween the weld pool and the base plate depends on the peak tem- that are in contact. Just as it is mentioned in the previous case, the
perature and the time during which the coincident points stay in heat from the top surface is modeled using the previously devel-
the liquid state. However, this criterion is not included in the oped FILM subroutine with the value of t 0 identical to that just
research and all coincident points on the contact surfaces are as- calculated for DFLUX. However, since the topmost elements of
sumed to have full fusion because the model is designed to check the weld pool that are yet to be deposited are considered part of
for the residual stresses. the top surface, they must not be considered in the heat loss area.
The heat source has the same concept as that of Model 2 in Therefore, the FILM subroutine is slightly modified so that ele-
which the heat source must stay a while at the start of the welding ments with x-coordinate larger than that of the center point of the
process before it moves along the welding. However, when apply- heat source must have a heat loss coefficient of zero. Figure 11
ing the DFLUX subroutine developed earlier, the value of t 0 will shows the heat load and heat loss regions in Model 3 according to
be different to match the deposition process. It is assumed that the the modifications in the FILM subroutine.
center point of the heat source should be at the first node of the The analysis steps in this case are similar to that of Model 2
welding line just as the first group of nodes comes into contact with two extra steps added to the procedure in order to account for
the depositing groups of nodes, and that the deposition of each
group is done in two steps
Table 1 Material properties versus temperature

Temperature (°C) 20 1550 1650 2000

Young’s modulus 共GPa兲 200 0.2 2⫻10 ⫺5
2⫻10⫺5
Poisson ratio 0.25 0.25 0.25 0.25
Yield strength 共MPa兲 290 1 0.01 0.01
Yield strength at strain 1.0 共MPa兲 314 1 0.01 0.01
Thermal expansion (1/°C⫻10⫺6 ) 10 15 15 15
Thermal conductivity (W/m.°C) 50 30 30 30
Specific heat (J/kg.°C) 450 400 400 400
Latent heat 共J/kg兲 260,000

Fig. 14 Temperature history at the top and bottom surfaces of

Models 2 and 3

Table 2 Comparison of analysis steps

Element Element
birth movement
Min. time increment 0.000002872 0.00011
No. of increments 1753 966
No. of iterations 4758 3471
Total analysis time ⬃48 h ⬃24 h
Fig. 12 Temperature history

148 Õ Vol. 125, MAY 2003 Transactions of the ASME

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 Fig. 15 Comparison between the element birth and element movement techniques for „a… ␴ x „longitudinal stress… and „b…
␴ y „transverse stress… history at the monitoring points, „c… ␴ x and „d… ␴ y distribution along the midsection, „e… ␴ x and „f …
␴ y distribution along the welding line, and „g… and „h… along the fusion line

Material Properties. The material used in Model 1 is Inconel Results and Discussion
Alloy 600 that is used by Friedman 关7兴. The material properties
used with metal deposition in Model 2 and Model 3 were acquired Verification Study. Good agreement appeared between the
from Brown 关11兴 whose properties are shown in Table 1. The thermal and structural results of Model 1 and those of Friedman
latent heat indicated is included by ABAQUS in the specific heat 关7兴. Figure 12 shows the temperature history at 3 monitoring
variation with temperature between the liquidous and solidous points indicated in Fig. 2 along the mid transverse line on the top
levels. and bottom surface. Figure 13 shows the history of the variation

Journal of Pressure Vessel Technology MAY 2003, Vol. 125 Õ 149

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
 of the longitudinal stress along the transverse direction (y direc- be very effective. It allowed for early thermal interaction between
tion兲 along the top of the midsection. the weld pool and the base plate reducing the analysis time. Also,
the stress history and the residual stress distribution resulting from
Element Birth Versus Element Movement Technique. The
both techniques compared well, with an acceptable difference
results of the element movement technique 共Model 3兲 had a very
when evaluated versus the computing cost.
close match with those of the element birth 共Model 2兲. Figure 14
shows a comparison of the temperature history at the monitoring
point near the fusion surface shown in Fig. 6. The slight reduction Nomenclature
in the peak temperature predicted using the element movement is q(r) ⫽ heat flux at distance r from center of heat source
due to the gradual flow of heat from the weld pool to the base. (W/m2 )
Besides, due to the difference in the node arrangement between Q ⫽ total heat input rate 共W兲
the two techniques in having two separate bodies in Model 3 r b ⫽ radius at which total heat input is 95% of actual
instead of one in Model 2, there is a difference in the arrangement value 共m兲
of the integration points representing similar nodes. This caused a h convection ⫽ convection coefficient (W/m2 °C)
difference in the temperature of the monitoring points due to the ␧ em ⫽ emissivity
high temperature gradient. In addition, it can be observed that the ␴ bol ⫽ Stefan-Boltzman constant (5.669⫻10⫺8 W/m2 K)
temperature of the monitoring point at the top is lower than that at
T i ⫽ expected temperature for current increment 共K兲
the bottom since the distance of the former from the welding line
T a ⫽ ambient temperature 共K兲
is larger than that of latter.
t i ⫽ time of step i (s)
Table 2 shows a comparison between the two techniques in the
v ⫽ welding speed 共m/s兲
analysis steps and timing. This illustrates the huge reduction in the
time spent to perform the simulation using the element movement
technique. References
Figure 15 illustrates a detailed comparison of the longitudinal 关1兴 Nguyen, N. T., Ohta, A., Matsuoka, K., Suzuki, N., and Maeda, Y., 1999,
( ␴ x ) and transverse ( ␴ y ) stresses between the two models. A very ‘‘Analytical Solutions for Transient Temperature of Semi-Infinite Body Sub-
close match in the history of the stress components at the moni- jected to 3-D Moving Heat Sources,’’ Weld. J. 共Miami兲, Aug., pp. 265–274.
关2兴 Goldak, J., 1990, ‘‘Keynote Address: Modeling Thermal Stresses and Distor-
toring point 共Fig. 6兲 is shown in Fig. 15(a) and (b). The mid tions in Welds,’’ Recent trends in welding science and technology, ASM Inter-
section 共Fig. 6兲 was selected for monitoring since it gives an over- national.
view of the behavior of the whole welded structure. A very close 关3兴 Ueda, Y., and Yuan, M. G., 1993, ‘‘Prediction of Residual Stresses in Butt
match can be observed between the two techniques in the distri- Welded Plates Using Inherent Strains,’’ ASME J. Eng. Mater. Technol., 115,
Oct., pp. 417– 423; 116, July 1994, pp. 285.
bution of the residual stresses along the mid section’s top and 关4兴 Mochizuki, H., and Hattori, H., 1999, ‘‘Residual Stress Analysis by Simplified
bottom paths as illustrated in Fig. 15(c) and (d). Finally, the Inherent at Welded Pipe Junctures in a Pressure Vessel,’’ ASME J. Pressure
welding line, being the hottest part, and the fusion line, where a Vessel Technol., 121, Nov., pp. 353–357.
great variation in the residual stress distribution is expected to 关5兴 Dong, P., 2001, ‘‘Residual Stress Analyses Multi-Pass Birth Weld: 3-D Special
Shell versus Axisymmetric Models,’’ ASME J. Pressure Vessel Technol., 123,
occur, were also selected for monitoring. The variation of the May, pp. 207–213.
residual stresses along these paths of Model 2 compared well with 关6兴 Hibbit, Hugh D., and Marcal, Pedro V., 1973, ‘‘A Numerical, Thermo-
those of Model 3 as shown in Fig. 15(e) – (h). Mechanical Model for the Welding and Subsequent Loading of a Fabricated
Structure,’’ Comput. Struct., 3, pp. 1145–1174.
关7兴 Friedman, E., 1975, ‘‘Thermomechanical Analysis of the Welding Process Us-
Conclusion ing the Finite Element Method,’’ ASME J. Pressure Vessel Technol., 97, Aug.,
In the verification model, the results of the 3-D model simulat- pp. 206 –213.
关8兴 Wilkening, W. W., and Snow, J. L., 1993, ‘‘Analysis of Welding-Induced Re-
ing the welding process compared well with that of 关7兴. The sub- sidual Stresses With The Adina System,’’ Comput. Struct., 47共4/5兲, pp. 767–
routines in that model can be used confidently to simulate most of 786.
the moving heat source used in the welding processes. It can be 关9兴 Fanous, I. F. Z., ‘‘3D Modeling of the Welding Process Using Finite Ele-
further developed to simulate inclined heat sources to cover a ments,’’ M.Sc. thesis, The American University in Cairo, February 2002.
关10兴 Hibbitt, Karlsson, and Sorensen, ‘‘ABAQUS/Standard User’s Manual,’’ 6.2,
wider range of applications. 2001.
In comparing the element movement technique versus the ele- 关11兴 Brown, S., and Song, H., 1992, ‘‘Finite Element Simulation of Welding Large
ment birth technique, it can be observed that the former showed to Structures,’’ ASME J. Eng. Ind., 114, Nov., pp. 441– 451.

150 Õ Vol. 125, MAY 2003 Transactions of the ASME

Downloaded 14 May 2011 to 203.110.246.230. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm