Thermal Analysis of The Arc Welding Process Part I. General Solutions

Thermal Analysis of the Arc Welding Process: Part I.
General
Solutions
R. KOMANDURI and Z.B. HOU
An analytical solution for the temperature-rise distribution in arc welding of short workpieces is
developed based on the classical Jaegers moving heat-source theory to predict the transient thermal
response. It, thus, complements the pioneering work of Rosenthal and his colleagues (and others who
extended that work), which addresses quasi-stationary moving heat-source problems. The arc beam
is considered as a moving plane (disc) heat source with a pseudo-Gaussian distribution of heat
intensity, based on the work of Goldak et al. It is a general solution (both transient and quasi-steady
state) in that it can determine the temperature-rise distribution in and around the arc beam heat source,
as well as the width and depth of the melt pool (MP) and the heat-affected zone (HAZ) in welding
short lengths, where quasi-stationary conditions may not have been established. A comparative study
is made of the analytical approach of the transient analysis presented here with the finite-element
modeling of arc welding by Tekriwal and Mazumder. The analytical model developed can determine
the time required for reaching quasi-steady state and solve the equation for the temperature distribution,
be it transient or quasi-steady state. It can also calculate the temperature on the surface as well as
with respect to the depth at all points, including those very close to the heat source. While some
agreement was found between the results of the analytical work and those of the finite-element method
(FEM) model, there were differences identified due to differences in the methods of approach, the
selection of the boundary conditions, the need to consider image heat sources, and the effect of
variable thermal properties with temperature. The analysis presented here is exact, and the solution
can be obtained quickly and in an inexpensive way compared to the FEM. The analysis also facilitates
optimization of process parameters for good welding practice.
I. INTRODUCTION
ARC welding is one of the most common manufacturing

operations for the joining of structural elements for a myriad
of applications, including bridges, building structures, cars,
trains, farm equipment, and nuclear reactors, to name a few.
In these applications, it is desirable, and oftentimes critical,
to determine the temperature-rise distribution in relation to
the location, time, and welding conditions, for it affects
the metallurgical conditions at and near the weld and the
consequent strength and reliability of the joint. In this investigation, an analytical solution for the temperature rise distribution in arc welding of short workpieces is presented based
on the moving heat source theory of Jaeger[1] and Carslaw
and Jaeger[2] to predict the transient thermal response.
Rosenthal laid the foundation for the analytical treatment
of the heat distribution in welding during the late 1930s and
in the mid-1940s.[3,4,5] Using the Fourier partial differential
equation (PDE) of heat conduction, he introduced the moving coordinate system to develop solutions for the point and
line heat sources and applied this successfully to address a
wide range of welding problems. His analytical solutions of
the heat flow made possible for the first time the analysis of
the process from a consideration of the welding parameters,
namely, the current, voltage, welding speed, and weld geometry. To facilitate solution of the PDE, Rosenthal assumed
R. KOMANDURI, Professor and MOST Chair in Intelligent Manufacturing, and Z.B. HOU, Visiting Professor, are with the School of Mechanical
and Aerospace Engineering, Oklahoma State University, Stillwater, OK
74078.
Manuscript submitted September 13, 1999.
METALLURGICAL AND MATERIALS TRANSACTIONS B
quasi-steady-state conditions that can be justified experimentally when the length of the weld is long compared to the
extent of heat. This means that an observer stationed at the
point heat source fails to notice any change in the temperature around him as the source moves on. Rosenthal also
gave an alternate analogy for this, wherein the temperature
distribution around the heat source is represented by a hill
that moves as a rigid body on the surface of the plane without
undergoing any modification either in size or shape.
Starting from the following PDE of heat conduction,
Rosenthal applied it for welding (by assuming the heat
source to be a moving point or a moving, infinitely long
line heat source) by considering a moving coordinate system.
2u 2u 2u 1 u
1
1 25
x2 y2
z
a t
[1]
When the origin of the moving coordinate system coincides with the moving heat source and moves along with it
at the same speed (with its X-axis coinciding with the x-axis
of the original absolute coordinate system), the relationship
between the coordinates of the point where the temperature
rise is concerned along the X- (or x-) axis at any time t is
given by X 5 x 2 vt. Substituting this in Eq. [1], the general
PDE of heat conduction in a moving coordinate system can
be obtained as
2u
2u 2u
v u 1 u
1
1 252
2 1
X
y2
z
a X a t
[2]
Equation [2], even for a unidirectional heat flow, e.g., along

the X-axis, involves three variables, namely, X, u, and t.
Hence, solution of this equation by the separation of variables or other similar techniques would not be feasible.
VOLUME 31B, DECEMBER 20001353
In order to solve this problem, Rosenthal incorporated an

experimental observation,[4,5,6] namely, when the working
time of the moving heat source is sufficiently long, the
temperature-rise distribution around the heat source (in the
moving coordinate system) would reach quasi-stationarystate conditions, i.e., u/t 5 0. The PDE of heat conduction
in a moving coordinate system for the quasi-steady-state
condition is, thus, given by
2u
2u 2u
v u
1
2
2 1
2 5 2
X
y
z
a X
[3]
By experimentally observing the shape of the weld pool on

the surface, which is of a tear-drop-like shape, Rosenthal
considered the final solution of Eq. [3] as a product of two
separate functions, given by
u 5 e2vX/2a w(X, y, z)
[4]
The first part of Eq. [4], namely, e2vX/2a, is an asymmetric

function along the X-axis. Here, e2vX/2a , e2v(2X)/2a, i.e., the
larger the value of the term .vX/2a., the higher the asymmetry. The second part, namely, w(X, y, z), is considered as a
symmetric function. As a whole, u is an asymmetric function
along the X-axis. This consideration is acceptable, for it is
closer to the real situation. In practical cases of moving heatsource problems, the rise of the temperature in front of the
heat source (where X is positive) is steeper than the fall of
the temperature behind the heat source (where X is negative),
i.e., the temperature distribution is asymmetric along the Xaxis relative to the heat source. The larger the value of the
term v/2a, the higher the asymmetry. When v/2a 5 0 (i.e.,
5 0), it becomes a symmetrical function (only the second
part exists), which is the case for a stationary heat-source
problem. With this substitution, one needs to solve only the
symmetric function, w(X, y, z). Thus, the moving heat-source
problems are greatly simplified. For solving the moving,
infinitely long line heat-source and moving point heat-source
problems, Rosenthal changed the coordinate systems into
cylindrical and spherical ones, respectively, to reduce the
number of variables, and developed solutions for a number
of quasi-stationary moving heat-source problems.
It is well known at the outset that Rosenthals solutions,
while valid at locations farther away from the heat source,
are subjected to considerable error at or near the heat source,
due to the assumption of a point or line heat source (zero
area). In fact, Rosenthal and Schmerber[6] cautioned the
application of the point or line heat-source model for locations at distances of less than a few millimeters ('6 to 8
mm) from the heat source, due to the finite size of the heat
source. Based on the analysis of the temperature distribution
during welding, Rosenthal came to the following important
conclusions: (1) the rise of the temperature in front of the
heat source is steeper than the fall of the temperature behind
the source, (2) the metal being welded is more quickly heated
than cooled from a given temperature, (3) increasing the
current density widens the heat-affected zone (HAZ) without
much change in the shape of the isotherms, (4) the speed
of welding affects most the shape of the isotherms, (5) the
higher the welding speed, the more elongated the isotherms,
and, finally, (6) the greater the heat diffusion of the work
material, the more circular the shape of the isotherm.
1354VOLUME 31B, DECEMBER 2000
II. REVIEW OF LITERATURE

Since the pioneering work of Rosenthal, considerable
interest in the thermal aspects of welding was expressed by
many researchers, as evidenced by numerous publications[737].
They can be classified into three categories. The first one
deals with the analytical and experimental work; the second
one deals with numerical methods, such as the finite-difference method and finite-element method (FEM); and the third
one deals with the effect of thermal properties at various
temperatures on the temperature distribution in the welded
plate. As the literature is numerous in this field, only literature pertinent to the present investigation will be briefly
reviewed. The readers are referred to numerous contributions
made over the years in the Welding Journal and to other
learned publications and conference proceedings.
In the following text, the literature covering the first two
aspects, namely, analytical and experimental studies, as well
as the numerical methods, will be briefly reviewed. The
third aspect, dealing with the effect of thermal properties at
various temperatures on the temperature distribution in the
welded plate, will be discussed in a companion article, Part
II[38] of this two-part series. It may be noted that much of
the intense analytical work on this subject was undertaken
starting in the 1930s, with the pioneering work of Rosenthal,
and continued to about the mid 1970s, after which the numerical methods (both the finite-difference method and FEM)
predominated. Consequently, one would see relatively fewer
references to the analytical approach in recent literature, and
this was reflected somewhat in the list of References in
this article.
A. Analytical and Experimental Studies on the
Temperature Distribution in Welding
Christensen et al.,[10] based on Rosenthals equation for
the point heat source moving across the surface of a semiinfinite body, developed generalized plots of the nondimensional temperature-rise distribution (on the surface as well
with respect to the depth), which can be used for the estimation of the width and depth of the HAZ, cooling rate, and
time of residence between given temperatures, which, in
general, can be used for all combinations of materials and
welding conditions. They, however, cautioned the use of
these plots for specific cases, except for the general guidance
of the temperature distribution. Wells[11] considered a twodimensional (2-D) moving rectangular heat source with a
uniform distribution of heat intensity. While Rosenthals
solution predicts the shape of the melt pool (MP) behind
the heat source, Wells proposed the inverse problem of estimating the heat input and welding speed by examining the
finished weld. Apps and Milner[12] investigated heat flow in
argon-arc welding (without a filler metal) for aluminum,
lead, nickel, copper, and Armco iron. They compared the
theoretical and experimental weld-pool shapes for these metals for various heat inputs and welding speeds.
Barry et al.[13] developed an analytical model by modifying the concept of a strip heat source, wherein the heat
input is visualized as being uniformly distributed over a
band whose width is about twice the weld bead. They measured the surface temperatures at key locations to determine
the peak-temperature distributions and heat-transfer efficiency. Paley et al.[14] investigated the heat flow in welding
heavy (thick) steel plates. They used the characteristic etching boundaries of the steels used (HY-80, T-1, and maraging
steels) to identify specific peak temperatures. In considering
the effects of the boundary surfaces, they assumed that all
surfaces (bottom or edge of the plate) are adiabatic. These
are surfaces of mirror-image symmetry with respect to the
distribution of heat sources. In analyzing the peak temperatures at any point of interest, the nearest heat sources would
have a significant effect, while the contribution from any
other heat source would decrease as the square of the distance
from the point under consideration.
Tsai[15] evaluated various mathematical models of the thermal behavior of metals during welding and summarized
their applicability in solving practical welding problems.
The objective is to assess a quick solution, or to formulate
a meaningful experimental procedure for quantitative results.
Nunes[16] developed an extended Rosenthal weld model in
a multipolar expansion form. Within a multipolar-expansion
content, he modeled phase changes by thermal dipoles and
circulation in the molten weld pool by thermal quadrupoles.
The model proposed is anticipated to provide insight into the
heat flow in welds. Eager and Tsai[17] modified Rosenthals
model to include a 2-D Gaussian distributed heat source and
developed a solution for a traveling distributed heat source
on a semi-infinite plate to provide the size and shape of the
arc weld pools. Their assumptions include the absence of
convective or radiative heat flow, constant average thermal
properties, and a quasi-steady-state semi-infinite medium.
The welding parameters, namely, current, arc length, and
traverse speed and material properties (namely, thermal diffusivity) have significant effects on the weld shape. They
compared the theoretical predictions with experimental
results on carbon steels, stainless steel, titanium, and aluminum, with good agreement.
Zacharia et al.[18,19] developed thermal models for both
autogeneous and nonautogeneous welding. They developed
a transient, three-dimensional (3-D) computer simulation
model for autogeneous welding, taking into account the
conditions of heat transfer, including convection of a gas
tungsten arc (GTA) weld pool. They also developed a similar
model for nonautogeneous, moving-arc GTA welding process. Both the models incorporate the complete set of interrelated thermophysical phenomena.
Boo and Cho[20] developed an analytical solution to predict
the transient temperature distribution in a finite-thickness
plate during arc welding, using a 3-D heat-conduction equation with convection boundary conditions at the surface of
welding. Due to the flow of the shielding gas, a forcedconvection boundary condition was assumed at the top surface of the weldment beneath the welding torch, and a natural
convection condition was assumed at the bottom surface.
The analytical results were compared with the experimental
results obtained by GTA bead-on-plate welding on a
medium-carbon steel under various welding conditions. Tsai
et al.[21] developed a semiempirical, 2-D finite-element heattransfer model to investigate thermal-related welding problems. They developed a conduction model that treats the
melting interface as an inner boundary to calculate the quasi
steady state temperature field and cooling rate in the weld
HAZ.
Kasuya and Yurioka[22] developed analytical solutions for
heat conduction for three cases of welding. First, a quasisteady-state model for a moving heat source to predict the
shape of the HAZ and cooling times between 800 8C and
500 8C in wide plates; second, an instantaneous line heatsource model to predict the cooling time from solidification
to 100 8C; and third, an unsteady heat model to predict local
preheating. They compared the experimental results with the
analytical results. Jeong and Cho[23] developed an analytical
solution for the transient temperature distribution in fillet
arc welds using an energy equation and compared it with
GTA and flux-cored arc experiments under various conditions. Using the conformal mapping technique, they transformed the solution of the temperature field in the plate of
a finite thickness to the fillet welded joint. Nguyen et al.[24]
developed analytical solutions for the transient-temperature
field of a semi-infinite body subjected to a double-ellipsoidal
power density moving heat source with conduction only.
They compared the analytical results with experimental
bead-on-plate specimens and found good agreement.
B. Numerical Techniques for the Temperature-Rise
Distribution in Welding
On the numerical side, several investigations[79,2532] were
made using finite-difference method and FEM analyses, in
view of certain advantageous features associated with these
methods. For example, Pevelic et al.[25] developed a finitedifference method to determine the temperature distribution
in a 2-D plate using the line heat source. The shape of the
melt pool was correlated with the welding variables, and
this isotherm was used as a boundary condition. Thus, in
this numerical method, experimental work is required to
determine the boundary conditions. Better agreement of the
peak temperatures was found between the analytical and
experimental values. Paley and Hibbert[26] conducted another
elegant study of the computation of the temperatures in
actual welds by comparing the computed values from the
heat-conduction equation with those obtained from the metallurgical sections of the actual welds. Good correlation of
the fusion temperature and A1 temperature isotherms were
reported. Thus, using a computer program, they produced
graphical displays of both the maximum temperatures and
the moving temperature field on the surface, as well as
in the vertical plane, of the welded plate. Friedman and
Glickstein[27] developed a FEM analysis for transient heat
conduction to investigate the effect of a number of welding
parameters, including the magnitude of heat input from the
arc, the distribution of the heat input over the surface of the
weldment, and the duration of the heat input on the thermalresponse characteristicsin particular, the weld bead shape
and the depth of penetration. They demonstrated the potential
for calculating the optimum combination of welding parameters for a given weld joint. Wilson and Nickell[28] applied the
FEM to heat-conduction analysis. They applied a variational
principle to the transient heat-conduction analysis of complex solids of arbitrary shape with temperature and heatflux boundary conditions. Krutz and Segerlind[29] used a
nonlinear finite-element model to optimize the welding
parameters for weld-joint strength when a certain desired
metallurgical structure is achieved. Friedman[30] developed
a thermomechanical analysis of the welding process using
the FEM. The model enables calculation of the temperatures,
stresses, and distortions resulting from the welding process.
Goldak et al.[7] developed a mathematical model for weld

heat sources based on a Gaussian distribution of the power
density. They developed a nonlinear, transient FEM heatflow program for the thermal-stress analysis of welds. They
computed the results of temperature distributions for submerged arc welds in thick workpieces and compared them
with the experimental values of Christensen et al.[10] and
reported excellent agreement. Tekriwal and Mazumder[8,9]
developed a 3-D transient heat-conduction model for arc
welding using the FEM analysis software ABAQUS. They
compared the numerically predicted sizes of the melt pool
and the HAZ with the experimental results obtained by the
United States Army Construction Engineering Research
Laboratory (Champaign, IL) and found good agreement.
Na and Lee[31] conducted 3-D finite-element analysis of
the transient temperature distribution in GTA welding. They
introduced a solution domain that moves with the welding
heat source to minimize the number of elements and, consequently, the computational time. As the solution domain
moves with the progress of welding, new boundary conditions and new elements were generated in front of the heat
source, while some elements disappear in the rear of it. The
numerical analysis was verified with GTA welding experiments on a medium-carbon steel under various welding conditions. However, since the moving-solution domain is small
compared to the rest of the weld structure, two kinds of
boundaries, namely, a solid metalatmosphere boundary and
a solid metalsolid metal boundary, have to be considered.
Silva Prasad and Sankara Narayanan[32] developed a similar
technique involving finite-element analysis of the temperature distribution during arc welding using a transient adaptive grid technique. It gives a finer mesh around the arc
source, where the temperature gradients are high, and a
coarser mesh in other places. This way, both the accuracy and
computational efficiency of the analysis can be increased.
It is clear from this brief review of literature that, although
significant efforts were made in developing analytical models for the quasi-stationary conditions, limited efforts were
expended to predict the transient thermal response except
by numerical techniques, such as the finite-difference
method and FEM. While the FEM has many advantages in
that it can analyze transient and quasi-steady-state conditions
as well as take into account the variable thermal properties
with temperature, it has some limitations, including the need
for identifying a priori the boundary conditions and the
dependence of accuracy on the size of the meshthe smaller
the mesh size, the more accurate the solution but the longer
the computational time to reach truncation. Thus, the choice
of a suitable mesh is vital to the accuracy and economy of
the FEM results. Sometimes, this involves approximations,
and considerable skill is required in order to model the
process. Also, a finer mesh is needed near the weld centerline
to determine accurately the peak temperatures, and a coarser
mesh is needed for the areas farther away, which is adequate
to reduce the time and cost of the analysis. Oscillations may
occur in the numerical values because of the coarse mesh
size, even though the method is stable in time.[8,9] Also, a
sudden change in the slope of the curves is obtained at the
transition from a finer to a coarser mesh, which, in reality, is
an artifact. There may be other limitations with the numerical
techniques (as will be discussed in this article), based on a
comparison of the analytical and the numerical (FEM)
results. This, however, does not restrict the applicability of

the FEM, for it offers other advantages, including the ability
to incorporate variable thermal properties with temperature,
the determination of the residual stresses, etc. Thus, the use
of analytical and computational methods should be considered as complementary and not to the exclusion of each
other. It would have been illuminating had the FEM analysis
presented by Tekriwal and Mazumder[8,9] also been done, in
addition to using variable thermal properties with temperature, at average thermal properties, so that the analytical
results presented here could be compared directly with the
FEM under the same conditions.
While there are numerous applications where quasi-stationary conditions are valid, there are other applications
where this may not be the case, e.g., when the length of the
weld is short or the velocity of the arc beam is low. To the
knowledge of the authors, no analytical solution for this case
exists. The objective of this investigation is to develop an
analytical solution (both transient and quasi-steady state) for
the temperature-rise distribution in and around the arc-beam
heat source, as well as for the width and depth of the MP
and the HAZ. The results of this model were compared with
those of the finite element modeling of arc welding reported
by Tekriwal and Mazumder,[8,9] which, in turn, were compared with the experimental results obtained. In their work
as well as in the present investigation, the length of the weld
is deliberately considered to be short (25.4 mm, or 1 in.)
where quasi-stationary conditions may not have been actually established. The work material, the welding conditions,
the geometry of the weld, etc., were taken to be essentially
the same as in Tekriwal and Mazumder to enable direct
comparisons. The analytical model developed here is used
to determine the time required for reaching the quasi-steady
state and to calculate the temperatures at all points on the
surface as well as with respect to the depth, including the
points close to the heat source, be it transient or quasi-steady
state. Similarly, by considering the length of the weld to be
long enough for quasi-steady-state conditions to prevail (for
example, ten times the original length), the temperature distribution around the arc-beam heat source, as well as the
width and depth of the MP and the HAZ under quasi-steadystate conditions in welding, were calculated, to demonstrate
that the solutions developed are, in fact, general solutions
and are applicable for quasi-steady-state conditions.
III. THERMAL MODELING OF THE ARC
WELDING PROCESS
Figure 1 is a schematic of conventional arc welding of
two short (25.4 mm (1 in.)), thin (5.8 mm) mild steel plates.
The width of the plates is considered large enough such that
the effect of the widthwise boundaries is small or negligible.
As the length of the weld is short, it will be shown that the
thermal process is transient under the conditions of welding.
The data used for the thermal analysis (Table I) in this
investigation are essentially the same as the values used by
Tekriwal and Mazumder,[8,9] so that comparison can be made
between the analysis presented here and their FEM analysis.
However, the feed rate of the filler wire given by them
appears to be somewhat low, with the result that the volume
of the molten filler wire will not be sufficient to completely
fill the V-groove at the joint. Thus, in this analysis, the feed
rate of the filler wire (vfeed ) is increased to 8.50 cm/s. For
(a)
Fig. 1Schematic of conventional welding of two short (25.4 mm), thin
(5.8 mm) mild steel plates.
Table I. Data for the Arc Welding Process

Welding voltage
Welding current
Arc efficiency, h
Welding arc speed, v
Arc beam radius, rb
Filler wire
Filler droplet temperature
Ambient temperature
MP temperature
HAZ temperature
Latent heat of fusion
Density of mild steel
Thickness of steel plate
Thermal conductivity, lavg
Thermal diffusivity, aavg
28 V
400 A
67 pct
5.08 mm/s
5.56 mm
d 5 1.6 mm, vfeed 5 85.0 mm/s
2300 K (2027 8C)
25 8C
1750 K (1477 8C)
1000 K (727 8C)
276.3 J/g
7.87 g/cm3
5.8 mm
0.6 J/cm ? s ? 8C
0.118 cm2/s
a similar reason, the welding current (I ) is also increased

from 380 to 400 A.
The next issue is at what temperature should the thermophysical properties be taken for the analysis? Several investigators have attempted to address this difficult problem.[3337]
It is without doubt that a consideration of variable thermal
properties with temperature would be the ideal approach.
Unfortunately, the analytical technique, unlike the numerical
methods, such as the finite-difference method or FEM, cannot consider variable thermal properties with temperature,
as it would complicate the mathematical analysis. Consequently, only a constant value can be taken in the analysis.
There are several alternatives, including (1) consideration
of room-temperature properties, (2) consideration of maximum-temperature properties, and (3) judicious consideration
of properties at some intermediate temperatures. It appears
that both the room-temperature properties or the maximumtemperature properties do not match the values obtained
experimentally or by numerical methods. It, therefore,
appears that judicious consideration of properties at some
intermediate temperature is to be the most logical alternative.
The best solution would, of course, be the one that is close
to the results using variable thermal properties with temperature or, better still, using a value at which the analytical
solution agrees with the experimental results. The choice
also depends on the nature of the variation of thermophysical
properties with temperature for a given work material (e.g.,
(b)
Fig. 2Variation of thermal conductivity and specific heat of the mild
steel plates with temperature.[8,9]
copper with superior thermal conducting properties vs stainless steel with inferior thermal conducting properties).
Figures 2(a) and (b) show the variation of thermal conductivity and specific heat with temperature for mild steel,
respectively, based on the data of Tekriwal and Majumdar.[8,9]
It can be seen that the specific heat increases linearly up to
'400 8C and remains constant above this temperature. In
contrast, the thermal conductivity decreases linearly up to
'1500 8C, followed by a sudden increase in temperature
up to 1700 8C, and remaining constant thereafter. If we are
interested in the values of the thermal properties in the
temperature range from '400 8C to 1500 8C, then we can
assume the specific heat to be a constant and the thermal
conductivity to decrease linearly with temperature (Figures
2(a) and (b)). Similarly, if we are interested in the values
of the thermal properties in the temperature range from room
temperature up to '400 8C, then we can assume the specific
heat to increase linearly with temperature and the thermal
conductivity to decrease linearly with temperature (Figures
2(a) and (b)).
In this investigation, the following thermophysical properties were used in the analysis: thermal conductivity (lavg)
of 0.60 J/cm ? s ? 8C, specific heat (c) of 0.660 J/g 8C,
density (r) of 7.87 g/cm3, and thermal diffusivity (aavg) of
0.118 cm2/s (Table I).[8,9] It can be seen from Figure 2(a)
that thermal conductivity l is constant (0.60 J/cm ? s ? 8C)
in the temperature range from 1700 8C to 3000 8C and is
nearly the average (slightly higher) in the range from 0 8C
to 1500 8C. This value of thermal conductivity corresponds
to a temperature of '400 8C. It is also the maximum temperature at which the specific heat increases with temperature
or minimum temperature, above which the specific heat is
constant. Since the maximum widths and depths of the MP
Fig. 3Schematic of the heat-transfer model of the welding process showing primary (HS0) as well as image heat sources (IHS1-IHS5).
and HAZ are nearly the same from '400 8C to 1300 8C,[38]
we have taken the lower temperature, namely, '400 8C, as
the temperature at which the thermophysical properties are
to be taken for the analysis. This is also the temperature at
which the experimental results reported by Tekriwal and
Majumdar[8,9] seem to agree well with the analytical values.
The objective of the analysis is to determine the temperature-rise distribution at and near the arc-beam heat source,
as well as the width and the depth of the MP and the HAZ
using the classical Jaegers moving heat source theory. Figure 3 is a schematic of the heat-transfer model of the welding
process used, showing the primary (HS0) as well as the
image heat sources (IHS1 through IHS5). The heat source
formed by the arc beam is considered to be a moving circular
disc heat source of radius r0 (arc-beam radius) and with a
heat liberation rate of qpl (in J/s). As welding progresses,
the circular disc heat source moves along the X-axis on the
upper surface of the plate with a velocity of v. The distribution of the heat intensity over the arc-beam heat-source area
is generally nonuniform and may be represented by a radially
symmetric pseudo-Gaussian distribution similar to the models presented by Goldak et al.[7] and Tekriwal and
Mazumder.[8,9] As the thickness (H ) and the length (L) of
the plates being welded are rather small, the thermal effects
at the bottom-boundary surface CD, as well as the two
lengthwise boundary surfaces, AC and BD, cannot be
neglected. When these boundary surfaces are considered
adiabatic, relevant image heat sources (IHS1 through IHS5) of
the same shape, size, heat liberation rate, and heat-intensity
distribution and moving at the same velocity but in an opposite direction to the primary heat source (HS0) should be
considered. The temperature rise at any point M is the sum
of the effects from the primary as well as the image heat
sources. At any instant of welding, the heat sources are
located from point M at distances of R0, R1, R2, R3, R4, and
R5, respectively (Figure 3). They can be expressed in terms
of the dimensions of the workpiece, the velocity of motion
of the arc, the duration up to that instant, and the coordinates
of point M.
It is well known that the estimation of the actual heat
input from an arc heat source to the workpiece is rather
complex. It depends significantly on the type of welding
process used. Christensen et al.[10] extensively investigated
various arc sources. They found that for the submerged arc,
the arc efficiency is represented by a narrow band, in the
range from 91 to 99 pct. In the case of a.c.-shielded metal
arc welding of mild steel, they observed the arc efficiency
to be in the range from 66 to 85 pct. The values of the arc
efficiencies for shielded metal arc welding were in the range
from 66 to 69 pct on mild steel. The GTA welding, however,

was characterized by a wide range of low arc efficiencies
(21 to 48 pct). Recent advances in welding technology have
improved the arc efficiency significantly.[39] For example,
the values for most arc welding processes have been reported
to be generally in the range from 70 to '100 pct, with the
submerged arc welding process to be most efficient, with
efficiency values reaching '100 pct, and the GTA welding
process being the least efficient, with efficiency values reaching '70 pct. Unfortunately, no rigorous study similar to that
reported by Christensen et al. is available in the literature,
at least to the knowledge of the authors. In this investigation,
an arc efficiency of 67 pct was taken (the same as the
value used by Tekriwal and Mazumder[8,9]), mainly so that
comparision between the analytical and numerical methods
could be made; this efficiency, according to the results of
Christensen et al., is representative of the a.c. metal arc or
gas metal arc welding of mild steel. The analysis is equally
valid at any efficiency, as it is only a numerical value in
the analysis.
Usually, different protective gases are used in welding,
either to prevent oxidation or to promote exothermic reactions, as the case may be. Many researchers have reported
that the contributions due to convection and radiation are
small or negligible. For example, Well[11] estimated the emission by radiation from a black body at 1500 8C to be '1.3
cal/cm2 and the heat transfer through a stagnant atmosphere
of 1 mm thick, for a temperature gradient of 10,000 8C/
cm, to be only 0.5 cal/cm2, whereas the heat flow due to
conduction is several orders of magnitude greater than this
value. Tekriwal and Mazumder[8,9] estimated that heat loss
due to convection accounts for ,5 pct of the total heat
involved in the process. Recent investigations, however,
have shown that convection heat losses may not be negligible
and depend on the nature of the shield gas used, whether
convection is natural or forced. In this investigation, the
losses due to convection and radiation are not directly
accounted for in the analytical model; instead, they are
lumped in the arc-beam efficiency (h) used for simplicity
of analysis. They can, however, be condidered, although not
precisely as follows.
When the adiabatic boundary effects are considered by
using the image heat sources, the intensity of the image heat
source for an adiabatic boundary is taken to be the same as
the principal heat source (i.e., n 5 1). When the heat source
is at the boundary, the primary heat source and the image
heat source should coincide. The strength of the heat source
is equal to (1 1 n)q. For an adiabatic boundary, the strength
of the heat source is 2q. The actual value of n depends on
the nature of cooling. For very good cooling, i.e., at the
boundary, there is no temperature rise, n 5 21. For natural
convection, n , 0.7 to 0.9; for moderate forced convection,
n , 0.2 to 0.7; for strong forced convection, n ,20.5 to
0.2; and, for extremely strong forced convection, n ,21.0 to
20.5. In the case of welding, n is taken to be 0.4 (moderately
strong convection), or (1 1 n) 5 1.4. This means that nearly
30 pct of the heat is carried away by convection. Considering
other heat losses, such as Ohmic heating of the tip of the
electrode, radiation, etc., the total heat losses would be
higher, ,32 to 33 pct. The value of the efficiency used in
this investigation, which is the same as that used by Tekriwal
and Majumdar, is 67 pct, which closely compares with the

previously estimated value.
The effective power of the arc is given by the following
(refer to Table I for relevant parameters):
to determine the differential temperature rise at any point

M caused by this amount of heat being liberated instantaneously by the moving ring heat source at the instant of ti.
The solution for an instantaneous ring heat source is given by
P 5 V ? I ? h 5 28 ? 400 ? 0.67 5 7504 W

A part of it is consumed for melting the filler wire, raising
its temperature, and the remaining part is used for the temperature rise of the workpiece. But, since the high-temperature
molten droplets of the filler wire are finally filled in the
groove at the joint of the steel plates being welded, the
thermal energy in the molten filler material will be utilized
by the workpiece. Hence, for simplicity, it is reasonable to
consider that during welding, when the filler material is still
in its liquid state, a part of the power delivered by the heat
source is consumed as the heat of fusion of the filler material
and stored in the molten material as internal energy (latent
heat of fusion), which does not facilitate a further temperature rise of the molten material. Only the remaining part
contributes toward an increase in the temperature of the
material around the arc heat source, including the MP. This
power (delivered by the welding arc) is used as the effective
heat liberation rate (qpl) of the heat source in the thermalanalysis model shown in Figure 5. The total heat for fusion
of the filler wire per unit of time (Lhtotal) is given by
Lhtotal 5
p ? 0.162
? 8.5 ? 7.87 ? 276.3 5 371.7 J/s
4
uM 5
In the following text, the solution for a moving ring source

is derived first based on the classical Jaegers[1] moving heat
source method. This is followed by the solution of a disc heat
source with a pseudo-Gaussian distribution of heat intensity
moving on the surface of a semi-infinite conduction medium.
Finally, the solution for a moving disc heat source with a
pseudo-Gaussian distribution of heat intensity, considering
the boundary effects of the bottom and lengthwise side surfaces, was applied to the problem of welding short workpieces.
I0
duM 5
Figure 5 is a schematic of the moving ring heat source,

showing the relationships between the spatial and temporal
parameters.[41] The x-, y-, and z-axes are in the absolute
coordinate system, while the X-, y-, and z-axes are in the
corresponding moving coordinate system, which has its origin coinciding with the center of the moving ring heat source
and moving with it at the same velocity along the X-axis. As
shown in Figure 5, a moving ring heat source continuously
liberates heat (qrg, in J/s) and moves along the X-axis with
a velocity of v (in cm/s). It is the objective of this investigation to determine the temperature rise at any point M(x, y,
z) and at any time t after the initiation of the heat source.
Consider the small time interval dti at time ti (refer to
Figure 5). In this time interval, the heat liberated by the ring
heat source is given by qrgdti (in Joules). As the time interval
dti is infinitesimally small, this amount of heat can be considered to be liberated instantaneously. Thus, the available solution for an instantaneous ring heat source can be used[1,2]
ro
!x2 1 y2
2at
[5]
qrgdti
r 2o 1 (x 2 vti)2 1 y2 1 z2
exp
2
cr(4pat)3/2
4at
? I0
[6]
12at !(x 2 vt ) 1 y 2
ro
Integrating Eq. [6] from ti 5 0 to ti 5 t, the solution for

the moving ring heat source in an infinite medium is given by
uM 5
qrg
cr(4pa)3/2
r 2o 1 (x 2 vti)2 1 y2 1 z2
dti
exp 2
3/2
ti50 t
4at
ro
!(x 2 vti)2 1 y2
2at
[7]
ti5t
? I0
A. Solution of a Ring Heat Source of Radius r0, Moving
Along the Radial Direction with a Velocity of v
Here, I0( p) is the modified Bessel function of the first kind,

order zero (refer to the Nomenclature). From tables available
in mathematical handbooks,[40] it can be shown that when
p 0, I0( p) 1. Thus, when ro 0, Eq. [5] is identical
to the solution for an instantaneous point heat source. The
term !x2 1 y2 1 z2 in Eq. [5] is the distance (R) between
the center of the instantaneous ring heat source and point
M, while the term !x2 1 y2 is the projection of R on the x0-y plane (Figure 3).
At the instant of ti , the ring heat source has moved a
distance of vti (Figure 5). So, the distance between point M
and the heat-source center at this instant is
!(x 2 vti)2 1 y2 1 z2. The projection of it on the x-0-y
plane is !(x 2 vti)2 1 y2. Thus, the differential temperature
rise at any point M caused by this amount of heat liberated
(Qrg 5 qrgdti) from the moving ring heat source at the instant
of ti in the differential time interval dti (using Eq. [5]) is
given by
Thus, the effective heat liberation rate is given by

qpl 5 7504 2 371.7 5 7132.3 J/s
Qrg
r 2o 1 x2 1 y2 1 z2
3/2 exp 2
cr(4pat)
4at
For changing Eq. [7] into a moving coordinate system,

substitute X 5 x 2 vt and ti 5 t 2 t:
uM 5
qrg
cr(4pa)3/2
r 2o 1 (X 1 vt)2 1 y2 1 z2
dt
exp 2
3/2
t50 t
4at
t5t
[8]
r
? I0 o !(X 1 vt)2 1 y2
2at
uM 5
qrg
cr(4pa)3/2
r 2o 1 X 2 1 y2 1 z2
dt
exp 2
3/2
t50 t
4at
t5t
[8a]
Xv
v2t
ro
? exp 2
? I0
? exp 2
!(X 1 vt)2 1 y2
2a
4a
2at
1 2
1 2 1
The following substitutions are made to transform the

integral part of Eq. [8a] into a nondimensional form:
v
2v
v2t
5 v, R2h 5 r 2o 1 X 2 1 y2 1 z2, 5 V, RhV 5 u, vt 5
4a
2a
V
The solution of a moving ring heat source in an infinite

medium can be expressed as
uM 5
qrgv
e2XV
16lap 3/2
? I0
dv
u2
3/2 exp 2v 2
v
4v
v2t/4a
v50
2v
X1
V
1 !1
roV 2
2v
2 1y2
2
[9]
The solution of a moving ring heat source in a semiinfinite medium is twice that in an infinite medium, i.e.,
uM 5
qrgv 2XV
e
8lap 3/2
? I0
v2t/4a
v50
dv
u2
exp
2
v
2
v3/2
4v
1 !1
roV 2
2v
2v
X1
V
2 1y2
[10]
It can be noted from Eqs. [9] and [10] that the temperature
rise is time dependent, and the time variable t appears in
the upper limit of integration. The longer the time, the larger
the value of the integral part; thus, the higher the temperature
rise. So, it basically corresponds to transient conditions. The
integral part can be solved by numerical integration using
a personal computer. When v 5 0 (the lower limit of integration), v3/2 is zero and, hence, the value of the function to
be integrated is infinity. Strictly speaking, this makes the
numerical integration impossible to solve. However, it can
be addressed by considering the nature of the function f(v)
to be integrated, namely,
1
u2
f(v) 5 3/2 exp 2v 2
v
4v
? I0
2v
X1
V
1 !1
roV 2
2v
2
2 1y2
2
[11]
Figure 6 shows the variation of f(v) with v.[41] Consider

the case of ro 5 5 mm. It can be seen (from Figure 6) that
for different values of X, the form of the curves is different,
but the curves always converge as v 0 and v 5. That
is, the value of the function f(v) is very small or negligible
(0.001) for all values of v greater than 5 and for negligible
values of v. For example, when X 5 4.8 mm, value of v is
5 ? 1026, and when X 5 1 mm, v is 2.5 ? 1023. It can also
be noted that the function f(v) increases initially for X 5
1, 3, and 5 mm (approaching closer and closer to the ring
heat source) and then decreases with a further increase of
X to 7 or 9 mm (moving farther and farther from the heat
source). When (v2t/4a) (the upper limit of v) is larger than
5, no matter how large (even to `), the areas under the
corresponding curves remain the same and, hence, the results
of the integration remain the same. For all practical purposes,
when the upper limit of integration of Eqs. [10] or [9] is .5,
one can consider the upper limit as `. Once the integration is
from 0 to 5, the variable v becomes merely a dimensionless
real number. Thus, in Eqs. [9] and [10], the time variable
disappears or the temperature rise will no longer be time
dependent. This means, at all points in the moving coordinate
system, the temperature rise reaches a steady state, i.e., (du/
dt) 5 0. This is the so-called quasi-steady state, and Eqs.
[9] and [10] can also be used for the quasi-steady state with
the upper limit of integration as 5 instead of `. It can also
be seen from Figure 6 that, by considering for the lower
Fig. 4Schematic of a moving circular disc heat source with pseudoGaussian distribution of heat intensity.
limit of integration a very small but a finite value, perhaps

0.00001 instead of zero, the numerical integration of the
integral part of Eq. [9] or [10] becomes possible with negligible error. Based on the previous analysis, the time required
for establishing the quasi-steady state (tqs) can be calculated
as follows:
v2tqs /4a 5 5 or tqs 5 20a/v2
[12]
It can be seen from Eq. [12] that the time required for
establishing quasi-steady state is proportional to the thermal
diffusivity of the conduction medium and inversely proportional to the square of the velocity of the moving heat source.
The higher the velocity, the shorter is the time required for
establishing quasi-steady-state conditions. When v 5 0 (i.e.,
for a stationary heat source), tqs 5 `, i.e., it needs a very long
time to establish steady-state conditions. For the example
considered in this investigation, the time required for establishing quasi-steady state is given by tqs 5 20a/v2 5 20 ?
0.118/(0.508)2 ' 9.1 s.
B. Solution of a Disc Heat Source with a PseudoGaussian Distribution of Heat Intensity Moving on
the Surface of a Semi-Infinite Conduction Medium
The moving disc heat source can be considered to be a
combination of a series of segmental moving ring heat
sources of different radii (ri) where ri varies from 0 to r0
(refer to Figure 4(a)). The solution of a moving ring heat
source can be considered as a starting point for further
analysis. Here, the heat sources (including the primary and
image heat sources) are moving at the boundary surfaces,
and all the heat liberated from the heat source is conducted
into the material (the heat loss is included in the arc efficiency). Hence, the boundaries are considered adiabatic and
The coefficient C can be determined by integrating the

heat liberation rate of all the segmental ring heat sources
from ri 5 0 to ro and, by equating it to the total heat liberation
rate of the disc heat source, we get
qpl 5
ri5ro
ri50
dq 5
ri5ro
ri50
1 1 2 2 2pr dr
r
C exp 23 i
ro
[14]
Here, 2ri dri 5 d(r 2i ) 5 r 2o d(r 2i )/r 2o 5 r 2od (r 2i /r 2o) 5r 2o dz

(where z 5 (ri /ro)2). Substituting z for (ri /ro)2 and r 2o dz for
2ri dri in Eq. [14], we get
qpl 5 pr 2oC
5
C5
z51
z50
Z Z
z51
e23z
e23z dz 5 pr 2oC 2
3 z50
pr 2oC
(2e23 1 1) ' 0.3167 pr 2oC
3
qpl
0.3167 pr 2o
Substituting for C in Eq. [13], we get

Fig. 5Schematic of a moving ring heat source showing the relationships
between spatial and temporal parameters.
1 1 22
qpl
r
q0 5
exp 23 i
0.3167 pr 2o
ro
[15]
Considering a pseudo-Gaussian distribution of heat intensity, the heat liberation rate of the segmental ring heat source
(Figure 4(a)) is given by
qrg 5
5
1 1 2 2 ? 2pr dr
qpl
r
exp 23 i
0.3167 pr 2o
ro
1 1 2 2 ? r dr
2qpl
r
exp 23 i
0.3167 r 2o
ro
The differential temperature rise at any point M caused

by the segmental ring heat source is given by (using Eq. [10])
duM 5
?
Fig. 6Variation of the function f(v) with v for different values of x.
qplv
2
e23(ri/ro) ? ri dri ? e2XV
0.3167 ? 4lap3/2 r 2o
v2t/4a
v50
u2i
dv
r V2
exp
2
v
2
? I0 i
3/2
v
4v
2v
[16]
2 1 !1X 1 2Vv2 1 y 2
2
The temperature rise at any point M caused by the entire

moving circular disc heat source is given by
the conduction medium to be semi-infinite. The semi-infinite
solution of a moving ring heat source (Eq. [10]) is, thus,
used as the starting point.
The heat liberation rate of the segmental ring heat source
(Figure 4(a)) is given by
qrg 5 qo2pri dri
where qo (in J/cm s) is the heat-liberation intensity of the
ring segment. For a uniform circular disc heat source it is
a constant and is given by qo 5 qpl/(pr 2o). For a pseudoGaussian distribution, it is a function of ri. The larger the
value of ri , the smaller the value of qo . When ri 5 0, qo has
its maximum value. When ri ro, qo 0 (Figure 4(b)). The
functional relationship for a pseudoGaussian distribution is
given by
1 1r 2 2
ri
3.1576qplv 2XV
e
4lap3/2 r 2o
v2t/4a
v50
ri5ro
ri50
e23(ri/ro) ? ri dri
2
[17]
u2i
dv
r V2
? I0 i
3/2 exp 2v 2
v
4v
2v
2 1 !1X 1 2Vv2 1 y 2
2
where ui 5 V !r 2i 1 X 2 1 y2 1 z2.
qo 5 C exp 23
uM 5
(13)
C. Solution for a Moving Disc Heat Source with a

Pseudo-Gaussian Distribution of Heat Intensity,
Considering the Boundary Effects of the Bottom and
Lengthwise Side Surfaces
As mentioned previously, the bottom surface and the two
lengthwise side surfaces can be considered to be adiabatic.
Thus, five image heat sources (IHS1 through IHS5) of the
same shape, size, moving velocity, heat liberation rate, and
heat-intensity distribution as the primary heat source (HS0)
are considered (Figure 3). The temperature rise at any point

M is the sum of the effects from all the primary and the
image heat sources which are located away from the point
M at distances of R0, R1, R2, R3, R4, and R5, respectively.
The values of these distances and their projections on the
X-axis (direction of motion)X0, X1, X2, X3, X4, and X5 are
given by the following (refer to Figure 3):
R0 5 !X 2 1 y2 1 z2, the projection of R0 on the Xaxis is X0, X0 5 X
(a)
R1 5 !X 2 1 y2 1 (2H 2 z)2, the projection of R1 on

the X-axis is X1, X1 5 X
R2 5 !(X 1 2vt)2 1 y2 1 z2, X2 5 2(X 1 2vt)
R3 5 !(X 1 2vt)2 1 y2 1 (2H 2 z)2,
X3 5 2(X 1 2vt)
(b)
R4 5 !(2(L 2 vt) 2 X )2 1 y2, 1 z2,

X4 5 2(L 2 vt) 2 X
R5 5 !(2(L 2 vt) 2 X )2 1 y2 1 (2H 2 z)2,
X5 5 2(L 2 vt) 2 X
And the relevant values of u for these distances are
given by
u0 5 V !r 2i 1 R20; u1 5 V !r 2i 1 R21; u2 5 V !r 2i 1 R22;
(c)
u3 5 V !r 2i 1 R23; u4 5 V !r 2i 1 R24; u5 5 V !r 2i 1 R25

The temperature rise at any point M caused by the primary
heat source is given by Eq.[17], and the solutions for the
temperature rise at point M caused by each of the image heat
sources can be obtained by substituting the aforementioned
values of the distances and their relevant projections on the
X-axis instead of R (so the ui) and X in Eq.[17]. Thus, the
total temperature rise at any point M is given by
uM 5
3.1576qplv
4lap3/2r 2o
v2t/4a
v50
ri5ro
ri50
e23(ri/ro) ? ri dri ? on50

2
u2n
dv
r V2
? I0 i
3/2 exp 2v 2
v
4v
2v
n55
ee
(d)
2XnV
[18]
2 1 !1X 1 2Vv2 1 y 2
2
IV. RESULTS OF THE ANALYSIS AND

DISCUSSION
The temperature-rise distribution in the vicinity of the
welding arc heat source can be calculated using Eq. [18]
and the welding data given in Table I. The effective heat
liberation rate of the moving disc heat source considered is
7132.3 J/s. Figures 7(a) through (e) show isotherms of the
two temperatures of interest, namely, the MP temperature
(the liquidus temperature of mild steel at 1477 8C) and the
temperature of the HAZ (the critical temperature (Ac1) at
727 8C) at different times in the cross-sectional plane parallel
to the weld seam (in the x-0-z plane) and at y 5 3.4 mm
(refer to Figure 1). Figure 7(a) shows that the depth of the
MP does not reach the bottom surface of the work at 1
second after the initiation of the welding process. However,
soon thereafter, the MP extends to such an extent that it not
only touches the bottom surface but also exceeds it. This
means that for '80 pct of the weld length, the depth of the
(e)
Fig. 7(a) through (e) Isotherms of the temperature rise of the MP (1477
8C) and the HAZ (727 8C) at different times in the cross-sectional plane
parallel to the welding seam (i.e., in the x-o-z plane) at y 5 3.4 mm.
MP is at all times larger than the thickness of the steel plate

being welded (refer to the isotherms of the temperature for
1477 8C in Figures 7(b) through (e)). Consequently, during
most of the welding process, the molten material of the filler
wire as well as the work material will drain away from the
kerf along the weld seam, thus affecting the quality of the
weld joint. This indicates that the arc power is either too
high or the weld speed too low, or both. It can also be seen
that the isotherm of the MP or HAZ expands to such an
extent that the lower part of it approaches the bottom-surface
(a)
(b)
Fig. 8(a) and (b) Isotherms of the temperature distribution of the MP
(1477 8C) and the HAZ (727 8C) at different times on the top surface (in
the x-o-y plane, z 5 0) of the plate along the welding seam.
boundary. The effect of the bottom-surface boundary (nearly

adiabatic) changes the direction of the isotherm to be nearly
perpendicular to the bottom surface, for the isotherms are
always perpendicular to the adiabatic boundaries when they
go across.
The dotted lines in Figures 7(a) and (b) are the same
isotherms obtained for this example by Tekriwal and
Mazumder[8,9] using the FEM. It can be seen that while the
results by the two methods are somewhat close for the area
near the heat-source boundary (namely, the top surface) the
contours deviate quite significantly near the bottom surface.
This appears to be due to an overestimate of the boundary
effect from the bottom surface, in the case of the FEM. The
isotherms of the HAZ at t 5 1 and 2 by the FEM appear
almost to be straight lines perpendicular to the top and bottom surfaces. This means that the temperature gradient along
the z-axis is almost zero and the heat flow along the z-axis
is nearly zero, which appears to be somewhat unrealistic.
Figures 8(a) and (b) show the growth of the MP and HAZ,
respectively, on the top surface of the plate (in the x-0-y
plane) (z 5 0) at different times. They show that the halfwidth of the MP increases gradually from '5 mm (at t 5
1 s) to a maximum of 7.6 mm (by the end of the welding
process, at t 5 5 s), and the HAZ increases from '8 (at t
5 1 s) to 10.5 mm (at t 5 5 s). The solid lines are the
analytical results and the dotted lines (in Figures 8(a) and
(b)) are the same results obtained for this example by Tekriwal and Mazumder[8,9] using the FEM. It may be noted that
Tekriwal and Mazumder[8,9] have plotted their FEM results
only up to y 5 4 mm. A reasonably good agreement can be
seen between the analytical and the FEM results during the
initial period, t 5 1 to 3 s. However, for the later stages,
namely, t 5 4 or 5 s, the analytical results show that the
left-hand side of the MP curve is bent downward and reaches
the midplane ( y 5 0). It, thus, gradually approaches the
tear-droplike shape (Figure 8(a), which is a typical top
view of the isotherms of various temperatures, including
the MP), for moving heat-source problems, as Rosenthal
elegantly demonstrated in his pioneering work. It can be
seen that while the analytical results show this trend, the
FEM results do not.
For the cases without the aforementioned boundaries, the
isotherms of various temperatures always have the tear-droplike shape. The existence of the boundary shown as AC in
Figure 3 may cause a certain distortion on the left-hand side
of the curve of the isotherms. When the heat source is very
near to the boundary, the distortion is significant. During
welding, the heat source is moving away from the boundary
AC. The longer the distance between them, the smaller its
effect and consequent distortion. In the analytical heatsource method described in Section II, the effect of each
boundary is represented by a corresponding image heat
source. It can be seen from Figure 3 that the longer the time,
the larger the distance between the primary and the image
heat sources (for the boundary AC) and the less is the influence of the image heat source. Consequently, the shape of
the contour of the MP approaches a tear-drop-like shape.
Thus, the use of image heat sources in the analysis closely
reflects the real effects of the boundaries. In the FEM, the
boundary conditions are determined somewhat arbitrarily,
although judiciously. Usually, a reasonable value of the temperature or temperature gradient at the boundary is assumed
by trial and error, based on experience.[8,9] Actually, the
temperature at the boundary AC, for this case, is not a
constant but a complex function of time and the spatial
variable z. For the FEM, to determine or assume the boundary conditions of any form a priori with reasonable accuracy
is somewhat difficult. The dotted curves of the isotherms
obtained by the FEM computations show that the effect of
boundary conditions does not gradually decrease with
respect to time. Thus, it appears that the effect of the boundary AC for t 5 3, 4, or 5 s during the computation by
the FEM is somewhat overestimated. Thus, while it gave
somewhat reasonable results initially, it deviated significantly later on, and, as a result, the tear-drop-like shape
contour was not obtained. Similarly, while the results for
the HAZ by the two methods are close for t 5 1, 2, or 3 s
(Figure 8(b)), the results for t 5 4 or 5 s differ quite significantly for the same reasons outlined earlier. It can again be
seen that the assumed boundary effect does not gradually
decrease with time. Thus, the width of HAZ by the FEM
appears to have been overestimated.
Figure 9(a) shows the analytical results of the temperaturerise distribution along the weld seam at the top edge of the
V-groove on the top surface of the plate under transient
conditions. At t 5 5 s, the maximum temperature rise on
the right-hand side of the work material is '2925 8C. At
(a)
(a)
(b)
Fig. 9(a) and (b) Temperature rise along the weld seam at the top edge
of the V-groove on the upper surface of the plate and the bottom edge
of the V-groove, respectively, under transient conditions obtained by an
analytical method.
(b)
other times during the welding process, the maximum temperature rise is in the range from 1500 8C to 1750 8C. Figure
9(b) shows the temperature rise along the weld seam at the
bottom edge of the V-groove under transient conditions. At
t 5 5 s, the maximum temperature rise on the right-hand
side of the work material is '3300 8C. At other times during
the welding process, the maximum temperature rise is in
the range from 1850 8C to 2400 8C. In the absence of a
boundary, the temperature-rise distribution curve always
rises up gradually with respect to time during the transient
stage. In the presence of the boundary AC, the boundary
effect can be very significant initially, with its effect deceasing rapidly with time. The combined effect of these two
factors makes the temperature-rise distribution curves from
t 5 2 to 4 s quite similar and nearly constant, as shown in
Figures 9(a) and (b). When t 5 s, the heat source
approaches the boundary BD (Figure 3), and the effect of
the boundary BD becomes stronger and significant. Thus, the
temperature-rise distribution curve rises sharply, as shown.
Here, it can be seen that the boundary BD has an effect only
when the heat source is approaching it. This is because this
boundary is located in front of the moving heat source and
is approaching with respect to time. Usually, in front of the
moving heat source, the temperature rise drops rather steeply
with respect to distance. Thus, when t 5 2, 3, or 4 s, the
distance between the primary heat source and the boundary
BD (or the image heat source) is rather large. Thus, the
Fig. 10(a) and (b) Temperature rise along the weld seam at the top edge
of the V-groove on the upper surface of the plate and the bottom edge
of the V-groove, respectively, under transient conditions obtained by the
FEM method.[8,9]
effect of the boundary (or the image heat source) would

be very small or unnoticeable. That is the reason why the
boundary BD has an effect only when the heat source
approaches it, or for example, when t 5 s, in this case.
Figures 10(a) and (b) are the corresponding temperaturerise distribution curves obtained by the FEM based on the
data of Tekriwal and Mazumder.[8,9] Comparing the results
of the two methods, it can be seen that when t 5 5 s, the
corresponding curve for the temperature rise along the top
edge of the V-groove, calculated by the FEM, does not rise
as sharply (compare Figures 9(a) and 10(a)). It, thus, appears
that the effect of the boundary BD was largely underestimated during the FEM computation. The widthwise boundary is located at a distance of 12.7 mm from the heat-source
boundary, as shown in Figure 1 (with the data from Table
I), which is farther away from the heat source. So, the effect
of the boundary is negligible. Thus, along this direction, it
can be considered as though there is no boundary. For the
analytical method, it is a simple problem, and the temperature-rise distribution there can be calculated easily, no matter
how complicated the function is (a function of time and the
spatial variable y). But for the FEM, an assumption of the
temperature distribution at that boundary should be made a

priori to the computation. Without some prior analysis, it
is a difficult task to determine the boundary conditions (as
a function of time and y). It seems that the results shown
in Figures 10(a) and (b) are also based on an overestimation
of the effect of the widthwise boundary. This, together with
the overestimation of the effect of the boundary AC and
underestimation of the effect of the boundary BD (as discussed previously), can result in significant errors in the
calculations for the points near these boundaries, but, for
the points near the heat-source boundary (rather far from
these boundaries just discussed), the error would be small.
Other discrepancies can also be noted for the temperature
along the bottom edge of the V-groove in the FEM results.
For example, it is not clear why the temperature is low at
the left end of the weld at the bottom edge at t 5 1 s when
the effect of the image heat source would be significant, or
why it would not decrease with distance when the effect of
that image heat source diminishes. It is also not clear why,
at t 5 5 s, the temperature is consistently lower than at t 5
1 to 4 s along the length of the weld and why the temperature
has not increased significantly at the right end of the weld
at t 5 5 s, when the image effect would be significant. One
would expect the nature of the variation of the temperature
from t 5 1 to 5 s for the top edge, as well as the bottom
edge, to be somewhat similar, although the magnitude could
be different due to the effect of image heat sources. The
analytical method shows such a relationship, while the FEM
shows a significant difference. Also, the curves in Figure
10(b) show more symmetry, while they are expected to
be asymmetric.
It can also be seen that the entire process (i.e., from the
beginning to t 5 5 s) is under transient conditions. Using Eq.
[12] (the time taken to reach quasisteady state conditions), t
can be estimated to be '9 seconds. It may be noted that
the arc heat source is shut off at t 5 5 s, i.e., prior to reaching
the quasisteady state conditions (the time for the heat
source to go through the entire length of the work material
being welded is given by t 5 L/v 5 2.54/.508 5 5 s). After
the heat source is shut off, the temperature rise at the central
area around the heat source is still considerably high, much
higher than the rest of the work material. Thus, the heattransfer process proceeds continuously, even though the arc
is shut off, resulting in a decrease in the temperature at the
central part and a corresponding increase in the rest of the
work material. Consequently, the widths of the MP and HAZ
would increase further, even after the shut-off of the arc
heat source (Figures 11(a) and (b)).
Figures 11(a) and (b) show the change in the half-width
of the MP and the HAZ, respectively, on the surface after
the shut-off of the arc heat source at different times. The
temperature-rise distribution in the work material after the
shut-off of the arc heat source can be calculated using Eq.
[18], but considering the lower limit of integration of its
second integral as v 5 v2(t 2 ts)/4a (instead of zero), where
ts is the time when the arc heat source is shut off. Figure
11(a) shows that the half-width of the MP after the shut-off
of the heat source increases initially from 7.6 mm (at the
time of shut-off of the heat source) to about 8.0 mm at t 5
6 s and decreases gradually thereafter. At t 5 7 s it is '6.1
mm, and at t 5 8 s it is '2.1 mm. Figure 11(b) shows that
the half-width of the HAZ after the shut-off of the heat
(a)
(b)
Fig. 11(a) and (b) Variation of the half-width of MP and HAZ, respectively, after the shut-off of the arc heat source, i.e., during the cooling period.
Fig. 12Variation of the width of the MP and HAZ with respect to time
including the duration during welding and after the shut-off period. Also
shown in dotted lines are the results of the FEM.[8,9]
source increases initially from 10.5 mm (at the time of shutoff of the heat source) to '12.5 mm at t 5 7 s and gradually
decreases thereafter. At t 5 8 s it is '12.3 mm, and at t 5
9 s it is '11.4 mm.
Figure 12 shows the variation of the width of the MP and
HAZ with respect to time during the welding process as
well as after the shut-off period. The dotted curves of the
widths of the HAZ and MP are the same results obtained
for this example by Tekriwal and Mazumder[8,9] using the

FEM. The curves for the width of the MP by the two methods
are rather close, and the differences between them are small.
But the differences in the widths of the HAZ by the two
methods are quite significant. This is because the curves for
the MP (Figure 8(a)) are comparatively closer to the heatsource boundary, and those for the HAZ (Figure 8 (b)) are
closer to the widthwise boundary. As discussed previously,
if the effects of boundaries AC and BD and the widthwise
boundary are not properly estimated, a significant error for
the estimation of the HAZ can occur because it is close to
those boundaries, but, for the MP, the error is small or
unnoticeable because it is near the heat-source boundary.
V. QUASI-STEADY-STATE ANALYSIS
Consider the case where the work material being welded
in the previous example is much longer, (perhaps 10 times)
longer, where L 5 254 mm (10 in.) instead of 25.4 mm (1
in.). The time required for completing the welding process
is t 5 L/v 5 254/5.08 5 50 s. It was shown earlier that the
time required for reaching quasi-steady state is '9.1 s.
Hence, the heat transfer during the welding process for this
rather long workpiece would be mostly under quasi-steadystate conditions. For this case, Eq. [18] can be used for the
calculation of the temperature-rise distribution in the work
material around the moving arc heat source by substituting
v 5 5 for the upper limit of integration. It was shown earlier
that the arc power used initially was rather high. So, for this
analysis, the effective heat liberation rate of the heat source
is reduced from 7132.3 J/s (used in the previous example)
to 4280 J/s ('60 pct). The methodology for the selection
of the optimal value of power will be presented in the
next section.
Figure 13(a) shows the isotherms of the various temperatures on the surface under quasi-steady-state conditions, i.e.,
from t 5 11 to 47 s. During this period, the shape and size
of the isotherms are nearly constant. It can be seen that the
half-width of the MP is '5 mm and the half-width of the
HAZ is '7.6 mm (measured from the isotherms of 1477
8C and 727 8C, respectively). The points of the maximum
width of the various isotherms of relevant temperatures on
the top surface of the plate are located on the curve n-n,
which is curved backward. These curves are very similar to
the plots presented by Rosenthal for the quasisteady state
conditions.[5,6] As pointed out by Rosenthal, this is due to
the finite speed of heat flow in metals, which delays the
occurrence of the maximum temperature parallel to the direction of welding. Figure 13(b) shows the temperature-rise
distributions along the X-axis on the top surface for different
values of y ( y 5 0, 2, 4, 6, 8, 10, or 12 mm). The dotted
line in this figure shows that the maximum temperature rise
is shifted backward with respect to y. Similar results were
reported by Rosenthal in his pioneering work,[4,5,6] thus confirming the validity of the analysis presented here for the
quasi-steady state.
Figures 14(a) and (b) show the change in the shape and
size of the HAZ and MP, respectively, on the top surface
of the work material (isotherms of 727 8C and 1477 8C,
respectively) with time during the welding process (from
transient to quasi-steady state). It can be seen from Figure
14(a) that the contours at t 5 11 s, 13 s, 15 s. . . are almost
the same, for they are all under quasi-steady-state conditions.
(a)
(b)
Fig. 13(a) and (b) Temperature distribution under quasi-steady-state conditions (a) on the surface and (b) along the x-axis on the top surface at
different values of y.
The boundary AC at that time is rather far away from the

moving heat source; thus, it does not have a noticeable effect.
But at t 5 9 s, the distance between the boundary AC and
the heat source is still not far enough, and the left-hand side
of the HAZ contour is still rather close to the boundary AC,
so that the boundary effect in this part is still noticeable.
However, at this time, the quasi-steady state has already
been established (the contour of the HAZ should be nearly
the same as that at t 5 11 s, 13 s, 15 s. . .), so the left-handside of the contour of the HAZ at t 5 9 s is stretched
toward the boundary AC and appears longer compared to
the contours of the HAZ at t 5 11 s, 13 s, 15 s. . . . The
right-hand side of the contour at t 5 9 s in the range of X
'210 to 15.4 mm is rather far from the AC boundary, so
the effect of the boundary is almost unnoticeable. As far as
the contour of the MP at t 5 9 s is concerned (Figure 14(b)),
it is much smaller than that of the HAZ, so the MP contour
at t 5 9 s (under quasi-steady state) is entirely out of the
influence of the boundary AC. On comparing Figures 14(a)
and (b), the MP contour at t 5 9 s can be seen to be entirely
located in the range of X '29 to 1 4 mm. Thus, the contour
(a)
(a)
(b)
(b)
Fig. 14(a) and (b) Plan view of the top surface of the weld showing
changes in the shape and size of MP and HAZ, respectively, from transient
to quasi-steady-state conditions
of the MP at t 5 9 s is the same as that at t 5 11 s, 13 s,

15 s. . . .
VI. OPTIMIZATION OF THE WELDING
PROCESS
Equation [18] is the analytical solution for the heat conduction during welding. Hence, it can be used to determine
the optimal combination of welding parameters, mainly, the
power and the moving velocity of the arc beam. It can be
noted that the parameters used in the example of welding a
short workpiece, discussed previously, are far from optimal.
The heat liberation rate of the heat source considered (qpl
5 7132.3 J/s) appears to be too high and/or the velocity too
low. The net result is that, after '1 second, the entire thickness of the plate at the joint area is all melted. So, most of
the molten filler as well as the molten work material at the
joint would drain out through the wide kerf (wider than 2
3 3.4 mm). This makes the welding joint weaker. Thus, the
use of excessively high power results in not only higher
manufacturing costs but also poor-quality products. Using
Eq. [18], an optimum value of the heat liberation rate of the
heat source can be determined.
Figure 15(a) shows the isotherms of the MP temperature
in the x-z plane at y 5 3.4 mm, when the heat liberation
rate of the heat source is qpl 5 3940 J/s ('55.2 pct of the
power used in the example considered). It shows that at this
power level, even at time t 5 5 s, a part of the work material
at the bottom surface is still solid. Consequently, all molten
filler material is retained in the groove. However, it can also
be seen that at time t 5 1, 2, or 3 s, the temperature rise at
(c)
Fig. 15(a) through (c) Isotherms at different times in the cross-sectional
plane parallel to the welding seam (i.e., in the x-o-z plane) when the effective
power of the arc is (a) qpl.eff 5 3940 W ('55.2 pct of the power used
earlier) at y 5 3.4 mm, (b) qpl.eff 5 4280 W ('60 pct) at y 5 3.4 mm,
and (c) qpl.eff 5 4280 W at y 5 2 mm.
the groove surface is lower than the melting point. So, '70
pct of the joint length of the weld (in the range of x 5
225.4 to 28.2 mm) is not welded properly, which makes
the welding joint weaker. This means that the power considered is too low.
Figure 15(b) shows the isotherms of the MP temperature
in the x-z plane at y 5 3.4 mm, when the heat liberation
rate of the heat source is qpl 5 4280 J/s ('60 pct of the
power used in the example considered). It shows that, with
a slight increase in power, the temperature rise is higher
than the melting point for most of the length (.85 pct) of
the weld joint at the groove surface, and the bottom surface
is still a solid (in the range of x 5 225.4 to 24.2 mm).
Consequently, most of the molten filler will be retained in
the groove. Only ,15 pct of the weld-joint length (at the
right-hand side of Figure 15(b) at t 5 5 s in the range of x
5 24.2 to 0 mm) is not perfectly welded, due to the loss
of material from drainage. This makes the welding joint
much better than the case discussed earlier.
Figure 15(c) shows the isotherms of the MP temperature
in the x-z plane at y 5 2 mm, when the heat liberation rate
of the heat source is qpl 5 4280 J/s. It provides a check on
the use of 4280 J/s as the heat liberation rate of the heat
source at another part of the groove surface. It shows that,
only in the range of x 5 219.2 to 27.2 mm of the weld
joint, the temperature rise at the groove surface is higher
than the melting point and that, at the bottom surface, the
material is still a solid. This means that only '47 pct of the
length of the weld joint at this part of the groove is perfect.
In the range of x 5 225.4 to 219.2 mm, the temperature
rise is not sufficiently high for a perfect weld joint. As in
the range of x 5 27.2 to 0 mm (at the right-hand side of
the figure at t 5 5 s,), it is also not a perfect weld, due to
the loss of material from drainage.
It can be seen from Figures 15(a) through (c) that, for the
case where the length of the weld is rather short, one cannot
expect to obtain a perfect weld joint by merely selecting a
constant optimal value of the welding power. The reasons
for this are the following. (1) The welding process is totally
transient and, hence, hardly quasi-steady state. So, the MP
size, including its depth, cannot be constant. It increases
gradually and continuously from the beginning of the welding process until the establishment of the quasi-steady-state
conditions. If the selected power is sufficient for the initial
stages, it most certainly will be too high for the later part,
or, it it is adequate for the later part, it will most certainly
be too low for the initial stages. (2) The boundary effect
magnifies the size of the MP, especially during the last part
of the welding cycle. Refer to Figure 3, where the image
heat source relative to the boundary BD is moving,
approaching the primary heat source with the same velocity
but in the opposite direction. Usually, the temperature rise
in front of the moving heat source is lower and drops rather
significantly with respect to the distance. In front of a moving
heat source, the farther the distance, the lower the temperature. Thus, when the distance between this pair of heat
sources is large (at t 5 1, 2, 3, or 4 s), the effect of the
image heat source can be rather small or insignificant. But,
when the primary heat source is approaching and nearing
the boundary, so does the image heat source. The distance
between them becomes very short, and the effect of the
image source becomes significant. When the centers of each
arrive at the boundary, the effect of the image heat source
will be at a maximum. In fact, it is so significant, its effect
would be as if the power of the primary heat source had
doubled. (3) When the isotherm of the MP approaches the
bottom-surface boundary, its effect becomes stronger and
stronger and changes the direction of the isotherm to be
perpendicular to the bottom surface. Thus, the shape and
size of the MP changes significantly during the last part of
the welding cycle, as shown in Figures 15(a) through (c).
As for the image heat source relative to the boundary
AC, it is moving away from the primary one. At the very
beginning (it coincides with the primary heat source), it
exerts the maximum influence. With increasing distance
between them, the effect of the image heat source decreases
gradually. But, for a moving heat source, the temperatures
at those points behind it usually decrease rather slowly. At
the same time, in this transient stage, the temperatures of
those points around the moving heat source are gradually
increasing with respect to time. The combination of these
two factors makes the size and shape of the MP rather close
at time t 5 1, 2, 3, or 4 s, as shown in Figures 15(a)
through (c).
For optimization of the welding process, two approaches can
be considered: (1) the use of a variable effective arc power,
and (2) the use of a variable welding speed. For a variable
arc power (and, consequently, the variable heat liberation
Table II. qpl, Peff, P, and I at Different Times*
ti , s
qpl
Peff (5qpl 1 371.7), W
P (5Peff/0.67), W
I (5P/V)
4280
4652
7315
261.2
4280
4652
7315
261.2
4271
4643
6929
247.5
4213
4585
6843
244.4
4000
4372
6525
233.0
3424
3796
5665
202.3
*Using Eq. [19] as the function of the variable heat liberation

rate of the moving disc heat source.
The length of the weld seam is L 5 25.4 mm and the welding
speed is y 5 5.08 mm/s; thus, the time of the welding process is
5 seconds.
rate of the heat source), the functional relationship of the heat

liberation rate of the heat source, for this case, is given by:
qpl 5 qplmax (1 2 0.2(vti /L)5)
[19]
Using Eq. [19], the heat liberation rate of the heat source
at different times is different, as shown in Table II (in this
case, qplmax 5 4280 J/s). It can also be seen that the heat
liberation rates of the heat source at time t 5 1, 2, or 3 s
are very close to that used in the example discussed previously (the difference is less than 0.25 pct). So, the size of
the MP will be very close to that given in that example.
Only when t 5 5 s does the heat liberation rate of the heat
source drops by '20 pct, so that the size and the depth of
the MP can be significantly reduced. Based on Eq. [19], the
relevant values of the effective arc power (Peff ), apparent
arc power (P), and welding current (I ) at different times of
ti are also given in Table II. Substituting Eq. [19] into Eq.
[18], an equation for the moving disc heat source with a
variable heat liberation rate, for this example, can be
obtained as
uM 5
r5t
r50
3.1576qplv
4lap 3/2r 2o
ri5ro
ri50
e23(ri/ro) ? ri dri ? on50 e2Xnv/2a
dt
vt 2 2v/V
1 2 0.2
t 3/2
L
? I0
1 !1
r iV 2
2v
22
2v
Xn 1
V
? exp 2v 2
u2n
4v
2 1y2
2
[20]
It can be seen that by using Eq. [20], the results of the

calculation are very similar to those obtained for Figures
15(b) and (c), except that the size of the MP at time t 5 5
s is somewhat small, as shown by the dotted lines in those
figures for the variable power. The percentage of a perfectly
welded part is, thus, significantly increased. It can be seen
that only by considering variable arc power can a good
welding joint can be obtained. Similar results can be obtained
by considering the use of a variable welding speed. The
variable power and/or variable speed can easily be realized
these days using numerically controlled machine tools.
VII. CONCLUSIONS
1. The analytical model developed here correctly predicts
the development of the weld pool under moving arc conditions, with melting at the leading edge of the pool and
weld solidification at the trailing edge.
2. General solutions (both transient and quasi-steady state)

for the temperature-rise distribution in arc welding are
developed by considering a moving disk heat source with
a pseudo-Gaussian heat-intensity distribution that is close
to practice. The model can yield accurate results, even
close to or at the heat source. The transient analysis
developed here complements the quasi-steady-state analysis originally developed by Rosenthal for point and line
moving heat sources and subsequently modified by other
researchers for plane heat sources under quasi-steadystate conditions. It also complements the numerical techniques, such as finite-difference method and FEM analyses. The analysis is exact, and the solution can be obtained
quickly and inexpensively, compared to the FEM.
3. The analytical method developed here can be applied for
both the transient and quasi-steady states. It can also be
used to determine the time required for reaching the quasisteady state and to solve the equation for the temperature
distribution, be it transient or quasi-steady states. Thus,
it can be used for welding short-length workpieces to
predict the thermal response in terms of various input
parameters, such as welding power, arc velocity, and weld
geometry. It can be used also to calculate the temperature
on the surface as well as with respect to the depth at all
locations, including points very close to or at the heat
source. It, thus, provides a clearer physical understanding
of the process.
4. The general solutions developed here were applied to long
workpieces where quasi-steady-state conditions would
prevail. The results of the temperature isotherms on the
surface as well as on the top surface along the direction
of the welding at various distances show good agreement
with the original work of Rosenthal.
5. While there is some agreement between the results of
the analytical model and the FEM model,[8,9] there are
significant differences. This is attributed to the differences in the boundary conditions used and the role of
image heat sources in welding short, thin workpieces.
6. The analytical method developed here facilitates optimization of the process parameters for good welding practice. It was shown that, for the case of welding short
lengths, it is not possible to optimize the welding process
by merely selecting a constant value of welding power
or arc velocity. Optimization for a good-quality weld and
the lowest cost can be realized only by considering an
appropriate variable welding power and/or arc velocity,
the methodology for which was outlined here.
7. In this investigation, the values of thermo-physical properties used correspond to a temperature of ' 4008C. This
is because, for this material, the thermal conductivity,
l is found to be a constant (0.60 J/cm. sec.8C) in the
temperature range from 17008C30008C and is nearly
the average (slightly higher) in the range of 08C15008C.
It is also the maximum temperature where specific heat
increases with temperature or minimum temperature
above which the specific heat is constant. Since the maximum widths and depths of the MP and HAZ are nearly
the same from ' 4008C to 13008C [38], we have taken
the lower temperature, namely, ' 4008C, as the temperature at which the thermo-physical properties are to be
taken for the analysis. This is also the temperature at
which experimental results reported by Tekriwal and
Majumdar[89] seem to agree well with the analytical

values.
8. Currently, the ability to incorporate varying thermal properties with temperature is a limitation of the analytical technique, in view of the complexity involved in solving the
mathematical equations. In the same vein, there are some
limitations with the numerical techniques, such as the determination of the boundary conditions used and the need to
take into account the image heat sources. Also, it would
not be easy to determine the optimum conditions for good
welding practice by the numerical techniques. Even if this
can be done, it would be time-consuming.
ACKNOWLEDGMENTS
This project was initiated by a grant from the NSF U.S.China Cooperative Research Project on the Thermal Aspects
of Manufacturing. One of the authors (RK) thanks Dr. Alice
Hogen, NSF, for facilitating this activity and for her interest
in this project. The authors are indebted to the NSF for their
continuing support to one of the authors (RK) at OSU on
the various aspects of the manufacturing processes. Thanks
are due, in particular, to Drs. L. Martin-Vega, B.M. Kramer,
K. Rajurkar, and Delci Durham, Division of Design, Manufacturing, and Industrial Innovation, and to Dr. Jorn Larsen
Basse, Tribology and Surface Engineering program. The
authors also thanks the MOST Chair (most eminent scholars
program) for Intelligent Manufacturing for enabling the
preparation of this article. The authors thank the reviewers
for their valuable comments and suggestions that has greatly
enhanced the quality of the manuscript.
NOMENCLATURE
a
Ln
Qrg
qpt
qrg
qpl
q0
R0
R0, R1, R2, R3,

and R4
X (or X0)
X1, X2, X3,

and X4
thermal diffusivity of the medium (cm2/s)

latent heat or heat of fusion
heat liberated by the instantaneous ring
heat source (J)
heat liberation rate of a point heat source
(J/s)
heat liberation rate of a ring heat source
(J/s)
heat liberation rate of a moving disc heat
source (J/s)
heat-liberation intensity of a moving disc
heat source (J/cm2?s)
distance between the center of the moving
disc heat source and the point where the
temperature rise at time t is concerned
(cm)
distances between the center of the relevant-image moving disc heat sources
and the point where the temperature rise
at time t is concerned (cm)
projection of the distance on the X-axis
between the center of the moving disc
heat source and the point where the temperature rise at time t is concerned (cm)
projection of the distances on the X-axis
between the center of the relevantimage moving disc heat sources and the
point where the temperature rise at time
t is concerned (cm)
t
ts
X, y, and z
r0
ri
l
uM
r
I0( p)
time of observation or the time after the

initiation of a moving disc heat
source (s)
the time when the moving disc heat source
is shut off after completion of the welding process (s)
coordinates of any point M in a moving
coordinate system, where the temperature rise is concerned
the radius of the moving disc heat source
or moving ring heat source (cm)
the radius of a segmental ring heat
source (cm)
thermal conductivity of the medium (J/cm
? s ? 8C)
temperature rise at any point M at any time
t (8C)
density of the medium (g/cm3)
modified Bessel function of the first kind,
order zero, (5 1/2p
2p
epcosada)
For ease of computation, the modified Bessel function (I0( p))

is approximated as follows (refer to Reference 41 for details):
when 0 # p , 0.2
I0( p) 5 1
when 0.2 # p , 1.6 I0( p) 5 0.935 exp (0.352p)

when 1.6 # p , 3
I0( p) 5 0.529 exp (0.735p)
when p $ 3
I0( p) 5
1
exp ( p)
!2pp
REFERENCES
1. J.C. Jaeger: Proc. R. Soc. NSW, 1942, vol. 76, pp. 203-24.
2. H.S. Carslaw and J.C. Jaeger: Conduction of Heat in Solids, 2nd ed.,
Oxford University Press, Oxford, United Kingdom, 1959.
3. D. Rosenthal: 2nd Congr. Nat. Sci., Paris, France, 1935, pp. 1277-92.
4. D. Rosenthal: Welding Res. Suppl., 1941, pp. 220-s-234-s.
5. D. Rosenthal: Trans. ASME, 1946, vol. 68, pp. 849-66.
6. D. Rosenthal and R. Schmerber: Welding Res. Suppl., 1938, pp. 2-8.
7. J. Goldak, A. Chakravarti, and M. Bibby: Metall. Trans. B, 1984, vol.
15B, pp. 299-305.
8. P. Tekriwal and J. Mazumder: ASM Int. Conf. on Trends in Welding
Research, Gatlinburg, TN, 1986, ASM INTERNATIONAL, Metals
Park, OH, 1986.
9. P. Tekriwal and J. Mazumder: Welding Res. Suppl., 1988, pp. 150-s156-s.
10. N. Christensen, V. de L. Davies, and K. Gjermundsen: Br. Welding

J., 1965, vol. 12, pp. 54-75.
11. A.A. Wells: Welding Res. Suppl., 1952, pp. 263-s-267-s.
12. R.L. Apps and D.R. Milner: Br. Welding J., 1955, pp. 475-85.
13. J.M. Barry, Z. Paley, and C.M. Adams, Jr: Welding Res. Suppl., 1963,
vol. 42, pp. 97-s-104-s.
14. Z. Paley, J.N. Lunch, and C.M. Adams, Jr: Welding Res. Suppl., 1964,
vol. 43, pp. 71-s-79-s.
15. C.L. Tsai: Am. Soc. Met., Proc. of The Conference on Trends in Welding
Research in the United States, New Orleans, LA, ASM International,
Metals Park, OH, 1982, pp. 91-108.
16. A.C. Nunes, Jr: Welding J., vol. 62 (6), pp. 1652s-1170s.
17. T.W. Eager and N.S. Tsai: Welding J., 1983, vol. 62 (12), pp. 346s-355s.
18. T. Zacharia, A.H. Eraslan, and D.K. Aidun: Welding Res. Suppl., 1988,
pp. 18s-27s.
19. T. Zacharia, A.H. Eraslan, and D.K. Aidun: Welding Res. Suppl., 1988,
pp. 53s-62s.
20. K.S. Boo and H.S. Cho: Proc. Inst. Mech. Eng., 1990, vol. 204, pp.
175-83.
21. C.L. Tsai, S.G. Lee, Y.L. Shim, J.J. Jaeger, and C.P. Chasten: ASME
Heat Transfer in Materials Processing, (JHTD), 1992, vol. 224, pp.
9-17.
22. T. Kasuya and N. Yurika: Welding Res. Suppl., 1993, pp. 107s-115s.
23. S.K. Jeong and H.S. Cho: Welding Res. Suppl., 1997, pp. 223s-232s.
24. N.T. Nguyen, A. Ohta, K. Matsuoka, N. Suzuki, and Y. Maeda: Welding
Res. Suppl., 1999, pp. 265-s-274-s.
25. V. Pavelic, R. Tanbakuchi, O.A. Uyehara, and P.S. Myers: Welding
Res. Suppl., 1969, vol. 43, pp. 295-s-304-s.
26. Z. Paley and P.D. Hibbert: Welding Res. Suppl., 1975, pp. 385-s-392-s.
27. E. Friedman and S.S. Glickstein: Welding Res. Suppl., 1976, pp. 408s-420-s.
28. E.L. Wilson and R.E. Nickell: Nucl. Eng. Design, 1966, vol. 4, pp.
277-86.
29. G.W. Krutz and L.J. Segerlind: Welding Res. Suppl., 1978, pp. 211s-216-s.
30. E. Friedman: Trans. ASME J. Pressure Vessel Technol., 1975, pp.
206-13.
31. S.-J. Na and S.-Y. Lee: Proc. Inst. Mech. Eng. (London), 1987, vol.
201, pp. 149-56.
32. N. Silva Prasad and T.K. Sankara Narayanan: Welding Res. Suppl.,
1996, pp. 123-s-128-s.
33. R.J. Grosh, E.A. Trabant, and Hawkins: Q. Appl. Math., 1955, vol.
13, pp. 161-67.
34. R.J. Grosh and E.A. Trabant: Welding Res. Suppl., 1956, pp. 396s-400s.
35. J. Isenberg and S. Malkin: ASME Paper No. 74-WA/Prod-5, ASME,
Fairfield, NJ, 1974.
36. S. Kou, D.K. Sun, and Y.P. Le: Metall. Trans. A, 1983, vol. 14A, pp.
643-53.
37. G.H. Little and A.G. Kamtekar: Computers Struct., 1998, vol. 68,
157-65.
38. R. Komanduri and Z.B. Hou: unpublished research (2000).
39. in Metals Handbook, 9th ed., vol. 6, Welding, ASM INTERNATIONAL, Metals Park, OH, 1989, pp. 31-34.
40. M.R. Spiegel: Mathematical Handbook of Formulas and Tables,
Schaums Outline Series, McGraw-Hill, New York, NY, 1993.
41. Z.B. Hou and R. Komanduri: Trans. ASME, J. Tribology, 1998, vol.
120, pp. 645-51.

Thermal Analysis of The Arc Welding Process Part I. General Solutions

Uploaded by

Copyright:

Available Formats

Thermal Analysis of The Arc Welding Process Part I. General Solutions

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Thermal Analysis of The Arc Welding Process Part I. General Solutions

Uploaded by

Copyright:

Available Formats

Thermal Analysis of the Arc Welding Process: Part I.

ARC welding is one of the most common manufacturing

Equation [2], even for a unidirectional heat flow, e.g., along

In order to solve this problem, Rosenthal incorporated an

By experimentally observing the shape of the weld pool on

The first part of Eq. [4], namely, e2vX/2a, is an asymmetric

II. REVIEW OF LITERATURE

Goldak et al.[7] developed a mathematical model for weld

results. This, however, does not restrict the applicability of

Table I. Data for the Arc Welding Process

a similar reason, the welding current (I ) is also increased

from 66 to 69 pct on mild steel. The GTA welding, however,

and Majumdar, is 67 pct, which closely compares with the

to determine the differential temperature rise at any point

P 5 V ? I ? h 5 28 ? 400 ? 0.67 5 7504 W

In the following text, the solution for a moving ring source

Figure 5 is a schematic of the moving ring heat source,

Integrating Eq. [6] from ti 5 0 to ti 5 t, the solution for

Here, I0( p) is the modified Bessel function of the first kind,

Thus, the effective heat liberation rate is given by

For changing Eq. [7] into a moving coordinate system,

The following substitutions are made to transform the

The solution of a moving ring heat source in an infinite

Figure 6 shows the variation of f(v) with v.[41] Consider

limit of integration a very small but a finite value, perhaps

The coefficient C can be determined by integrating the

Here, 2ri dri 5 d(r 2i ) 5 r 2o d(r 2i )/r 2o 5 r 2od (r 2i /r 2o) 5r 2o dz

Substituting for C in Eq. [13], we get

The differential temperature rise at any point M caused

The temperature rise at any point M caused by the entire

METALLURGICAL AND MATERIALS TRANSACTIONS B

C. Solution for a Moving Disc Heat Source with a

are considered (Figure 3). The temperature rise at any point

R1 5 !X 2 1 y2 1 (2H 2 z)2, the projection of R1 on

R4 5 !(2(L 2 vt) 2 X )2 1 y2, 1 z2,

u3 5 V !r 2i 1 R23; u4 5 V !r 2i 1 R24; u5 5 V !r 2i 1 R25

e23(ri/ro) ? ri dri ? on50

IV. RESULTS OF THE ANALYSIS AND

MP is at all times larger than the thickness of the steel plate

boundary. The effect of the bottom-surface boundary (nearly

effect of the boundary (or the image heat source) would

temperature distribution at that boundary should be made a

for this example by Tekriwal and Mazumder[8,9] using the

The boundary AC at that time is rather far away from the

of the MP at t 5 9 s is the same as that at t 5 11 s, 13 s,

Table II. qpl, Peff, P, and I at Different Times*

*Using Eq. [19] as the function of the variable heat liberation

rate of the heat source), the functional relationship of the heat

e23(ri/ro) ? ri dri ? on50 e2Xnv/2a

It can be seen that by using Eq. [20], the results of the

2. General solutions (both transient and quasi-steady state)

Majumdar[89] seem to agree well with the analytical

R0, R1, R2, R3,

X1, X2, X3,

thermal diffusivity of the medium (cm2/s)

time of observation or the time after the

For ease of computation, the modified Bessel function (I0( p))

when 0.2 # p , 1.6 I0( p) 5 0.935 exp (0.352p)