WELDING RESEARCH

SUPPLEMENT TO THE W E L D I N G JOURNAL, MAY I g95

Sponsored by the American Welding Society and the Welding Research Council

A Study on the Effect of Contact

Tube-to-Workpiece Distance on Weld Pool
Shape in Gas Metal Arc Welding

Experiments establish contact tube-to-workpiece distance as

an important variable in shaping weld geometry

BY J. -W. KIM A N D S. -J. NA

ABSTRACT. Computer simulations of the with those of experiments, and both assigned to the process variables which
three-dimensional heat transfer and fluid these results demonstrate that the contact control the formation of the weld by in-
flow in gas metal arc (GMA) welding of tube-to-workpiece distance exerts a fluencing the depth of penetration, the
have been studied for analyzing the ef- considerable influence on the formation bead width, and the bead height. Varia-
fect of contact tube-to-workpiece dis- of the weld pool and the resulting weld tion of the welding power and effective
tance on the weld pool shape by consid- shape by affecting the arc length and radii of the welding heat flux and current
ering the driving forces for weld pool welding current. path distribution due to the change of the
convection, the electromagnetic force, contact tube-to-workpiece distance can
the buoyancy force and the surface ten- Introduction affect the weld pool forrnation and even-
sion force at the weld pool surface, and tually the weld geometry. And because of
also the effect of molten electrode The strength of the weld joint depends the large amount of the heat supplied at
droplets. In the numerical simulation, upon the size of welds, when the strength the weld site over a short period of time,
difficulties associated with the irregular of the welct metal is given. In order to get there are rnany problems in and around a
shape of the weld bead have been suc- the proper weld size, specific values are welded joint such as generation of distor-
cessfully overcome by adopting a bound- tion, residual stress, and reduced strength
ary-fitted coordinate system that elimi- (Ref. 1). Accurate predictions of the weld
nates the analytical complexity at the size and the above problems require a
weld pool and bead surface boundary. precise analysis of the weld thermal
The method used in this paper has the ca- KEY WORDS cycle. The importance of a good model
pacity to determine the weld bead and for the weld pool convection in the analy-
penetration profile by solving the surface Weld Pool Shape sis of the thermal cycle has been empha-
equation and convection equations si- (]MAW sized by a number of investigators.
multaneously. Fluid Flow A number of researchers have shown
The experiments are performed to Heat Transfer that the convection in the weld pool can
show the variation of the weld bead 3-Dimensional Model strongly affect the weld pool geometry
geometry due to the change of the con- Boundary-Fit. Coord. and consequently the quality of the re-
tact tube-to-workpiece distance. The cal- Surface Deformation sultant welds including the joint penetra-
culated weld shapes correspond well Arc Length Distance tion, undercutting, and porosity (Refs.
Velocity Field 2-11). In recent years, considerable
Filler Metal Effect progress has been made, mainly for gas
J. -~" KIM is an Assistant Professor at Yeung-
nam University, and S. -J. NA is a Prol~,ssor at tungsten arc (GTA) welding processes, in
Korea Advanced Institute of Science and Te~ h- modeling the fluid flow and heat transfer
nolo qy, 7~e/on, Korea. condition of weld pools.

WELDING RESEARCH SUPPLEMENT[ 141-s

 consequently, the volved. Conversely, reducing the arc
weld bead surface. length makes the bead narrower and in-
Boundary-fitted co- creases the height of the reinforcement.
ordinate systems, In GMA welding systems, the power
employed in con- source has a flat or constant voltage char-
junction with finite acteristic and the electrode wire is fed
difference tech- into the arc at a fixed speed. The major
niques, remove reason for selecting the constant voltage
some of the difficul- source is the self-regulating character in-
ties encountered in herent in this system. The constant volt-
representing the age system compensates for the variation
complex geome- in the contact tube-to-workpiece dis-
tries with the Carte- tance, which readily occurs during weld-
sian or cylindrical ing, by automatically supplying the in-
Fig. 1 - - Schematic sketch of GMA welding system.
polar coordinate creased or decreased welding current to
system. attain an equilibrium. The power source
The work presented here considers provides the proper current so that the
The early studies of the fluid flow in
the three-dimensional quasi-steady heat melting rate is equal to the wire feed rate.
weld pools (Refs. 2-4) described the
and fluid flow analysis for the moving As the contact tube-to-workpiece dis-
steady-state and electromagnetically dri-
heat source of the GMA welding process tance increases, the arc length increases
ven flows in predetermined hemispheri-
in which the driving forces for the weld and the welding current decreases mo-
cal regions. In an important study on the
pool convection, the electromagnetic mentarily, as the characteristic of the
fluid flow and heat transfer conditions in
force, the buoyancy force and the surface constant voltage power source predicts.
a stationary GTA weld pool, Oreper, et al. However, this also decreases the elec-
(Ref. 5), considered the effect of the sur- tension force at the weld pool surface,
trode melting rate. With high-conductiv-
face tension gradient on the weld pool and also the effect of molten electrode
ity electrode materials such as aluminum
convection for the first time. The surface droplets are investigated. The relation of
or copper alloys, the resistance and the
tension effect on the weld pool geometry the contact tube-to-workpiece distance
voltage drop in the electrode extension
has been experimentally investigated by to the arc length is discussed for analyz-
can be considered as negligible. There-
Heiple, et al. (Ref. 6). Kou, et al. (Ref. 7), ing the effect of the contact tube-to-work-
fore, the decreased melting rate and the
presented the results of a quasi-steady, piece distance on the weld pool forma-
constant wire feed speed bring the weld-
three-dimensional computer model for tion. A series of experiments has been
ing current and the arc length back to the
the flow and heat transfer conditions in performed to show the variation of the
normal condition (Ref. 13).
the weld pool of the GTA welding weld geometry due to the change of the For the ferrous electrode, however,
process. An inherent limitation in all the contact tube-to-workpiece distance. Nu- the change of the contact tube-to-work-
aforementioned modeling efforts of the merical simulations have been per- piece distance eventually influences the
GTA welding process is the assumption formed for analyzing the behavior of a welding current. Increasing the contact
that the weld pool surface is nonde- quasi-steady GMA weld pool so as to de- tube-to-workpiece distance reduces the
formable or flat. A study by Tsao, et al. fine the principal operating parameters welding current due to the increased re-
(Ref. 8), presented a two-dimensional governing the fluid flow and heat trans- sistance in the circuit. The resistance in
stationary weld pool convection model fer phenomena and to investigate the ef- the electrode extension of the steel wire
for the GMA welding process, which also fect of the contact tube-to-workpiece plays an important role in the generation
adopted the assumption of the flat weld distance on the weld pool shape. of the Joule heat and then affects the
pool surface. More recently, Kim, et al. melting phenomena. It has been reported
(Ref. 9), studied the phase change prob- Contact Tube-to-Workpiece that the temperature of molten droplets is
lem at the liquid-solid interface and the Distance in G M A Welding lower in the case of the long electrode
deformation problem at the weld pool extension than that of the short electrode
surface in the stationary GTA welding Arc Behaviorand Weld Shape extension (Ref. 14). This indicates that
process by adopting a boundary-fitted the high temperature of the electrode ex-
coordinate system. The three-dimen- Among the welding variables, the arc tension promotes the detachment of
sional computer model based on the de- length, which is related with the welding droplets from the electrode end. The in-
formable surface condition was formu- voltage and contact tube-to-workpiece creased contact tube-to-workpiece dis-
lated first by Zacharia, et al. (Refs. 10, distance, has a close relationship with tance consequently brings on a reestab-
11), for the fluid flow and heat transfer the weld bead width. Generally, the bead lished equilibrium state of a somewhat
analysis in the weld pool of the GTA width increases, as the arc length is in- extended contact tube-to-workpiece
welding process. But the deformed pool creased. This can be explained by con- (Ref. 13).
surface was approximated by small steps sidering the characteristics of the weld- Thus, it can be expected that the con-
of the finite difference grids closest to the ing arc. The welding arc has a tact tube-to-workpiece distance affects
pool surface. point-to-plane relationship and is thus the weld bead geometry as well as the arc
Molten surface deformation, particu- bell-like in shape with the point at the length. Moreover, the variation of the
larly in the case of GMA welding, plays end of the electrode and the wide portion welding power due to the change of the
a significant role in the actual weld bead at the surface of the weld (Ref. 12). In- contact tube-to-workpiece distance can
size and should be considered in order to creasing the arc length makes the bead affect the weld pool formation and even-
evaluate the weld pool convection accu- wider due to the widened arc area at the tually the weld pool shape. A series of ex-
rately. In this study, a boundary-fitted co- weld surface, and consequently reduces periments were performed to study the
ordinate was chosen to handle the the reinforcement height because the effect of the contact tube-to-workpiece
largely deformed weld pool surface and, same volume of the filler metal is in- distance on the weld pool shape.

142-s I MAY 1995

Experiment where and JI are the Bessel functions of the first
kind and of the zero and first order, re-
In order to show the variation of the
weld bead geometry due to the change of
the contact tube-to-workpiece distance,
I puu - flu pvu - p u
/
spectively. The other components of J
and B are zero.

a series of experiments was performed for

various contact tube-to-workpiece dis- I uw-,w. I BoundaryConditions
tances. The welding power source used pcuT-kT; [pc T-kqJ The boundary conditions for the fluid
was a standard commercial unit and the
welding gun was mounted directly on a
computer-controlled single-axis motion
table. Welding was done in the flat posi-
pwu-Mu
G=lpwv-,uvZ
/ i0 / _p~ + fx
l andS=t-py+fY I
flow and temperature distributions are as
follows:

Top surface
tion, with the vertical gun moving at a
prespecified constant speed along the IPww - liwz I [-p~ + fz i
stationary workpiece. [pcwT - k d J [o ~ J V.n=O (3a)
The following welding conditions
were selected for experiments: liV(V.t,).n= ( (t(7 )YT
There are three distinct driving forces -t dT ) .t,
Tip-to-workpiece distance: 15, 20,
for the weld pool convection, i.e., the (3b)
and 25 mm; wire feed speed: 150 mm/s;
electromagnetic force, the buoyancy
welding speed: 7 mm/s; welding voltage:
force and the surface tension gradient at
31.5 V; wire diameter: 1.2 mm; shielding
the weld pool surface. The first driving
gas: 80% Ar - 20% CO2. in the weld pool (3c)
The electrode wire of the type of AWS force is caused by the interaction be-
ER70S-3 was used for the mild steel tween the divergent current path in the
weld pool and the magnetic field it gen- u = - U , v = w = 0 in the solid region
workpieces. The weld geometries were (3d)
determined by sectioning the resulting erates. The second driving force is
weldment. These cut specimens were caused by the temperature gradient
polished and etched, and then a macro- within the weld pool, and the third by the - k V T . n = h ( T - T~) for r >_a
picture of the weld was taken to measure temperature gradient at the weld pool
the weld size. surface.
In the case of the current flux falling w m+~ = w m + vd at the weld pool surface
Formulation on the flat surface, the analytical solution (3e)
of the electromagnetic force can be suc- The normal and tangential unit vectors in
Governing Equations cessfully obtained by using the Bessel Equations 3a-3c are determined on the
functions (Ref. 7). According to the com- surface node and the equations 3b and
Figure 1 shows the schematic diagram 3c represent the boundary conditions for
of the GMA welding process, in which where jr = I "~ . ' ' surface tension gradient at the pool sur-
the heat source is moving at a constant face. Equation 3e is for considering the
speed U, and a constant supply of the effect of molten electrode droplets,
coshTinh _~__~2~(H-z)dZ, JZ= j--~ ~ J o ( * r ) • where m represents the internal iteration
consumable electrode is maintained
through the center of the welding nozzle. number of the calculation loop.
The coordinate system (x, y, z) moves ( ~t2b2 ~ sinh lirn I
with the heat source at the same speed, Exp c qT i) sinhZ ((~.H)
H - z) d,, :-- - k V T . n = %(z.n) for r < a
(4a)
and its origin coincides with the center of
the electrode.
( sinh (4b)
~M,(Zr).Exp ---~-) sinh(ZU) dZ where z is the unit vector of the z direc-
The molten pool convection was con- tion and h, a combined heat transfer co-
sidered as laminar and incompressible efficient for the radiative and convective
(Refs. 7-11). And the following addi- putational analysis of Tsai, et al. (Ref. 12),
boundary conditions, which can be cal-
tional assumptions were adopted for the for the deformed surface, the electro-
culated from the following relationship
further simplification: magnetic force at a point 2.5 mm (0.1 in.) (Ref. 1):
1) Physical properties are constant ex- apart from the origin is less than a tenth
cept for the thermal conductivity, the of the maximum value for the welding h = 24"1x 10-4 ~"TI'6I(W/m2°C) (5)
specific heat and the density in the buoy- current of 150 A in GTA welding. By the
ancy term (Boussinesq approximation). result of repeated calculations in this where ~ is the emissivity of the surface of
2) Heat, current and pressure distrib- study, the surface deformation within the the body. A value of O.9 was assumed for
utions of the arc have Gaussian charac- 3-mm (0.12-in.) radius around the origin E, as recommended for the hot rolled
teristics. was revealed not to be severe. Therefore, steel (Ref. 16).
The governing equations, describing the body force in Equation 1 can be ex- Bottom surface
the continuity equation, the momentum pressed by using the approximated elec-
u=-U, v=w=O
equation and the energy equation for tromagnetic force and buoyancy force as (6)
steady-state velocity and temperature follows:
- k V T . n = h(T-T~)
fields in the workpiece, may then be writ- (7)
ten as follows: f = Z x B - p / J g ( T - T, ) Center plane
(2)
E,+Fy+G z =S au 8w
where r is the radius, i.e., ~/" ÷ y~, ~ the v=O, --=--=0
(1) ~y ay
depth from the weld pool surface, and J0 (8)

WELDING RESEARCH SUPPLEMENT I 143-s

 ±l±(F+,c)}+ a
ax Laz
+ Xc)I-±(F + Xc):o
8Zy ~ j 8z
(18)
N ii IIIII![II ! Substitution of the values previously
NllgNllIIIIIIIIIIIIIIII defined into the Euler equation gives a
somewhat complex form of the equation,
because the surface tension ~ is position
111
iii1111111111111 dependent in the distributed temperature
field. But the derivative term of the sur-
face tension is as small as about a factor
Fig. 2 I An example of grid used (center plane). of 100 less than that of the other terms.
Ignoring the derivative term of the sur-
pl led to the solid region. Tile pool bound- face tension, the following equation can
c~T
m = 0 ary was then determined by tracing the be derived.
c?y (9) locations where the temperature corre-
Side surface
u=-U, u=w=O
(10)
sponds to the liquidus temperature of the
workpiece. of(/ )
l+z~ z~-2z~zyz~+ 1+z2, zyy

(1+ z; + z[,)
)
- k v r . n : h ( r - r~) {11) Surface Deformation
= pgz - Par{ - Z (19)
Front surface boundary where the surface tension r~ at the weld
The weld pool surface under the arc
u=-U, v=w=O (12) pressure forms a shape which satisfies the pool surface was assumed to vary with
equilibrium condition of the surface in the temperature as follows (Ref. 19):
T=T~ (13) the gravitational field. Hence, it is usually
convenient to use the condition of equi- ~,(T) = < - r , ) + ( d ~ , / d T )
Rear surface boundary librium by directly solving the variational (T-T~) N / m (20)
problem subject to the constraint that the
The actual boundary condition of the volume of the weld pool is constant. For The radial distribution of the arc pressure
temperature imposed on the rear bound- the precise calculation of the surface de- Parc employed was adopted from the ref-
ary surface can be defined only in the formation, the convection in the liquid erence (Ref. 20). In the computational
case of the infinite calculation domain, pool, the electromagnetic force itself, procedure, the iterative method of the fi-
where the boundary temperature is equal and the effect of metal transfer should be nite difference has been used to solve the
to the ambient temperature (T~). How- considered. In this study, however, those nonlinear differential Equation 19. After
were assumed to be negligible for sim- these calculations, the computational re-
ever, the following numerical boundary plifying the problem. In a previous study,
condition was applied because a finite sult is applied to the constraint pre-
the effect of convection in the liquid pool scribed. If it does not satisfy the con-
domain must be used. was found to be less than that of the arc
straining equation, the computation
u=-U, b=w=0 pressure (Ref. 15). Thus, the total energy
(14) should be repeated after modification of
to be minimized includes the surface en-
ergy connected with the change in the Lagrange multiplier kin Equation 19. The
h~ = - \ - £ £ D ax area of the weld surface, the potential en- value of k is modified by using the golden
(15) ergy in the gravitational field, and the section search technique (Ref. 21 ).
I where T b is the temperature of the work performed by the arc pressure dis-
I boundary and the subscript i means the placing the weld pool surface. The total Numerical Procedure
internal node of the boundary. It was as- energy is therefore given by the following
sumed in Equation 15 that the tempera- equation (Ref. 18): Transformation
ture gradient is conserved at the rear cal-
culation boundary, which moves with r r I G~I~Zx+Z ~ q- Boundary-fitted coordinate systems,
the welding speed (Ref. 17). In calcula- employed in conjunction with the finite
tions, however, the solutions were found difference technique, remove the diffi-
not to be very sensitive to this boundary culties encountered in representing the
condition. (16) complex geometries with the Cartesian
The solid/liquid interface was treated where s means the surface of the work- or cylindrical polar coordinate system
as the free boundary problem involving piece. (Ref. 22). In the boundary-fitted coordi-
the heat and fluid flow. This is due to the Since the volume of the fed wire is ad- nate system, the finite difference grid
fact the solid/liquid interface does not ditionally supplied in GMA welding (Vw), lines are aligned with the surface bound-
have to be considered as a boundary. The aries so that the interpolation of flow vari-
the constraining equation is as follows:
effective-viscosity method was used for ables and the representation of normal
the equation of motion, in conjunction derivatives at boundaries can be per-
with a viscosity-temperature relationship Sf-zdxdy=-S!Gdxdy=Vw formed accurately. If a regular finite dif-
- - (17)
covering both the solid and liquid phase. The necessary condition for the solu- ference grid is used to represent the com-
Since no fluid flow exists in the solid tion follows from the application of the plex geometries, the boundaries intersect
phase, the very large viscosity was ap- appropriate Euler equation. the grid lines, which makes the computer

144-s I MAY 1995

programming complex and the evalua- coefficients. ET
tion of derivatives inaccurate. The governing Equation 21 can be
In order to gain a better accuracy, a represented by the following transport
grid mesh with variable spacing was em- equation in which 0 denotes all the de- / I
ployed. The grid spacings are fine near pendent variables and F the diffusion co-
the location of the heat source and coars- efficient:
ened with distance away from it. An ex-
ample of such a grid mesh is shown in
Fig. 2 for the center (y = 0 mm) plane. In
a boundary-fitted coordinate method,
the physical geometry is transformed into
a rectangular, uniformly spaced finite dif- TM

ference mesh. ~B
Equation 1 is transformed to a general
curvilinear coordinate system (~, q, ~),
Fig. 3 - - Grid structure and labelin~ around a
which results in Equation 21. ~rid node p.
E ~ + E17qx+ E¢~ + ~ y + F~Tq~,+
F~y +G~ z + G,~r/,,+GcG = S
diffusion and the source terms. For the
(21)
convection terms, the hybrid differenc-
wn,~,-~ K~ : ',/r,r'c-Y¢',T), ing scheme (Ref. 23) is enlployed. The fi-
{0w0- + 0,,,,. + ,,. +
nite difference equation is arranged by
{own- 0,,,,, + c: ,,
(22)
collecting the terms according to the grid
nodes around a control volume as shown
Numerical Scheme in Fig. 3. The final expression is given by
Equation 23 in which A represents the
Discretization of Equation 22 is per- link coefficients between grid nodes:
formed by using the finite difference ap-
At,@t, - At@L + A~,¢w + ANgN
proximation in the transformed domain.
The second order central differencing + A s ¢ 5 + AT~/ +AI~@B +S"
technique is used for approximating the (23)

Start ~)
Jt = 1 / xqLy~z ¢ - y C z ~ j - ~ i
/ Input Parameters/
I
Calculation of Temperature
(Flat surface)
and S1 = I -L O ~ y + t)17fly + P¢~-Y) + t y I
~C-alculation of Pool Surface
[,, Mesh Generation I
I
Calculation of Transformation
Coefficients In=n+ I I
I
Calculation of Velocity
The transformation coefficients, {×, {y, {z, and Temperature
I
qx, qy, qz, ~×, ~y and ~, are determined Output of Velocity/
numerically using the second order cen- and Temperature /
tral differencing technique on the grid [
generated to fit the irregular boundaries. Estimate of Fusion Boundary
In the transformed domain the grid sizes [
(i.e., A{, Aq and A~) are set to be unity to Calculation 9~ Pool Surface
simplify the calculation of transformation Z~'(ij)

Table 1 - - Parameters U s e d in
Compulalions

p = 7200 kg/m s Too = 2 0 ° C /' Output of Bead Profile1/

# 0.006 I~g/m s Ti = 1 5 2 0 ° C [
/3 = 10 -4 K -1 Iv = 2500°C ( End )
#m = 0.26 X 10 -6 H / m H = 12 m m
d ~ / d T = 10 - s N / m ° C U = 7 mm/s Fig. 4 - - Flow chart of calculations.

WELDING RESEARCH SUPPLEMENT I 145-s

 where
80 5* = 51 + ANE ~NE + ANW ~NW + ANT ~NT
+ ANB~)NR + AsECSE + Asw~sw + AsT~ST
~l~C + AsB ~SB + AETF#ET+ AEB ~EB + AWT YggT
>.
\ 1200
+~w~~B
Ap = AE + Aw + AN + As + Ar + AB

/I c
° ~

O
Thus, the nonlinear equations of the
t- transportation are approximated by a sys-
"0
C t
o tem of linear algebraic equations.
40 el_ A velocity-pressure correction algo-
°_
rithm SIMPLE-C (Ref. 24) was used for the
U
E I) solution procedure of the discretized
I. t./,//'I 13.
U') equations. This kind of procedure re-
r" k quires a grid staggering between the ve-
I--
locity or temperature nodes and the lo-
cations where the pressure is estimated
for the numerical stability in calcula-
tions. The unevenly spaced grids 73 x 31
x 17 were used for the 45.5 x 20 x 12-
500 1000 1500 °C 2000 mm (length x width x thickness (H)) cal-
Te~f~re culation domain. The overall solution
procedure is shown in Fig. 4, and the
welding parameters and the physical
Fig. 5 - - Temperature-dependent thermal conductivity and specific heat of AISI 1010 (Ref. 16).
properties used in the calculations are
shown in Table 1 and Fig. 5. The iterative
procedure for the calculation of the ve-
locity and temperature distribution was
continued until the following conver-
gence criterion was satisfied:
Table 2 -- Measured Weld Sizes and Welding Currents According to Various Contact
Tip-to-Workpiece Distances (L)
M a x . I ~ - @ ° d l <0.001
L (mm) BeadWidth (mm) BeadHeight (mm) JointPenetration(ram) WeldingCurrent (A)
15 11.32 3.12 5.05 300 And the following convergence criterion
20 12.60 2.82 3.95 280 was used for the overall calculation;
25 13.20 2.79 2.84 260
Iz n - zn-~l < 0.01 mm

The typical computer time was about

1600 minutes CPU time on the CRAY2S
computer.
Table 3 - - Welding Parameters Relevant to Contact Tube-to-Workpiece Distance Variation (L)
Results a n d Discussion
Tube-to-Workpiece a b c e Pmax Q
Distance (mm) (mm) (mm) (mm) (mm) (kN/m 2) (W) Experimental Results and Discussion

15 18.0 5.2 5.4 1.6 0.85 6615 The results of the experiment were il-
20 10.0 6.5 6.0 2.0 0.80 6170
lustrated in Table 2 and Fig. 6, in which
25 12.0 7.8 7.2 2.4 0.75 5733
the measured weld sizes and pho-

A B C

Fig. 6 - - Photographs o f weld geometry according to change o f contact tube-to-workpiece distance (L). A - - L = 15 ram; B - - L = 20 m m ; C - - L =
25 mm.

146-s I M A Y 1995
tographs are represented respectively. A
typical finger-like penetration, frequently
produced in GMA welding with Ar or its
mixture shielding gas, can be seen in all
the figures.
The welding current was decreased
according to the increasing contact tube-
to-workpiece distance as shown in Table
2. The welding current determines the
welding power because the welding volt-
age was kept at a specified value in this
experiment. But the largest bead width
was acquired in the case of the longest
contact tube-to-workpiece distance in
spite of the low welding power as shown
in the figure. This is probably due to the
increased arc length by the increase of
the contact tube-to-workpiece distance. Fig. 7 - - An example o f calculated weld surface.
Consequently, increasing the arc length
makes the bead wider and the reinforce-
ment lower, as mentioned earlier.
It can be also seen in the figure that the
depth of joint penetration distinctly in- X-AX[S
creases with the decrease of the contact -15.0 -10.0 -5.0 0.0 5.0 [mm]
tube-to-workpiece distance. This is con- I I I I I

sidered to be due to the constriction of the

distributed heat and current flux by the
decreased arc length, which results in a
large electromagnetic force, which pro- 0.0
motes the convection effect in the weld

[
pool.
The experiments demonstrate that the
contact tube-to-workpiece distance ex-
erts a considerable influence on the for- N 00
200[mm/sec]
mation of the weld pool and the resulting
weld shape by affecting the arc length (mm)
and welding current. The contact tube-
to-workpiece distance is thus considered
Fig. 8 - - Velocity distribution in weld p o o l (center plane, y = 0 mm).
an important variable, which can control
the formation of welds.

Calculation Results and Discussion calculated weld surface is illustrated in loop in the middle part of the weld pool
Fig. 7, where the weld pool deformation is mainly due to the electromagnetic
At certain combinations of welding due to the gravitation and arc force can force. By the interaction between the di-
parameters, the GMA welding process be seen. The weld pool surface under the vergent current path in the weld pool and
results in a finger-like penetration such as arc center was depressed below the un- the magnetic field it generates, a down-
illustrated in Fig. 6. Previous models, melted weld surface. ward electromagnetic force is produced
which used the conduction model only, Figure 8 shows the side view of the near the z-axis and this force pushes the
were unable to explain the phenomena fluid flow pattern in the weld pool due to high-temperature liquid metal in that re-
of this finger-like penetration in the GMA the combined effect of the buoyancy gion downward to the weld pool bottom.
weld. Thus, it is expected that weld pool force, the electromagnetic force and the The flow pattern induced by the electro-
convection, which causes more heat to surface tension gradient. As can be seen magnetic force allows the welding heat
be transferred from the heat source to the from the figure, the weld pool has a tail to be delivered efficiently from the heat
weld root, plays an important role in the about 15 mm (0.6 in.) long, due to the source to the weld root, and conse-
formation of the weld penetration. travel of the heat source. At extremely quently results in a deep weld penetra-
high welding speeds, the arc may be un- tion. The second circulation loop in the
Effect of the Three Driving Forces symmetric. But the symmetric arc current rear part of the weld pool is mainly due
and heat distributions were still used for to the surface tension gradient at the pool
Calculations were performed by con- the calculation in this study, because the surface and the effect of the primary cir-
sidering the three driving forces, the elec- speeds of the electron movement and culation. The temperature gradient of the
tromagnetic force, the buoyancy force plasma flow are much higher than the surface tension is generally negative, that
and the surface tension force at the weld normal welding speed. is, the surface tension is lower near the
pool. The values of the relevant parame- There are two circulation loops in the center of the weld pool surface, where
ters employed are those from the condi- convection pattern in Fig. 8, one in the the temperature is higher, and higher
tion of 20 mm (0.78 in.) contact tube-to- middle part and the other in the rear part near the pool boundary, where the tem-
workpiece distance listed in Table 3. A of the weld pool. The primary circulation perature is lower. Therefore, the liquid

WELDING RESEARCH SUPPLEMENT I 147-s

 lation procedure.
Figure 11 shows the fluid flow pattern
Y-AXIS in the weld pool due to the combined ef-
-8.0 -4.0 0.0 4.0 8.0 [mm]
I I I @ I fect of the three driving forces and the
droplets of the molten electrode. As can
be seen from the side view in Fig. 11A, the
0.0.
weld pool length in the weld line direc-
tion is longer than that in the case of con-
sidering the three driving forces only due
to the promoted fluid motion. The mag-
nitude and direction of the fluid veloci-
4.0 ties, especially under the center of the arc,
are considerably different from those in
: Convectlon
Fig. 5. It was thus presumed that the
...... : Pure conduction molten electrode droplets have an influ-
8.0 ence on the formation of the weld pool by
[mm]
promoting the fluid motion and then on
the weld shape. Figure 11 B shows the top
view of the convection pattern on the
Fig+ 9 - - Comparison of weld bead shapes for two different calculation methods. weld surface. As shown, the flow at the
free surface is outward, from the center to
the pool boundary. The temperature gra-
metal near the center of the weld pool common mechanism of the free-flight dient at the weld pool surface and conse-
surface is driven toward the weld pool metal transfer, the molten filler metal is quently the shear stressdue to the surface
boundary. delivered across the arc as fine droplets. tension gradient are larger in the front part
The detail descriptions for the effect of Thus, the transferred molten droplets of the arc than that in the rear part. There-
the three driving forces have been de- provide an additional heat flow in the fore, high velocities are seen near the
scribed in the authors' previous study weld pool by promoting the fluid motion front pool boundary, while low velocities
(Ref. 25). in the weld pool, compared with the case are near the rear pool boundary.
In calculation results of the heat trans- of considering the three driving forces The effect of molten electrode
fer without considering any convective only. In order to include the effect of droplets on the weld size and shape was
molten droplets into the numerical represented in Fig. 12 by comparing the
motion in the weld pool, there can be
model in this study, it was assumed that calculated weld bead with that in the
seen no deep penetration shape-- Fig. 9.
the molten electrode is transferred to the case of considering the three driving
As shown in the figure, it is evident that weld pool surface with the distributed ve-
the liquid metal flow plays a very impor- forces only. By considering the molten
locities of the Gaussian characteristics as
tant role in the formation of the finger-like electrode droplets, the depth of penetra-
shown in Fig. 10, where V r is the volu-
penetration in GMA welds, and the elec- tion was considerably increased. How-
metric feed rate of the electrode. The ef-
tromagnetic force is the dominant driving fective radius of the velocity distribution ever, the variation of the velocity distrib-
force for the liquid metal flow. was assumed also to be affected by the ution radius of the molten electrode
contact tube-to-workpiece distance as shows only a small change of the weld
Effect of Molten Electrode Droplets shown in Table 3. The distributed veloc- pool shape as shown in Fig. 13 for the
ities were then added to the converged employed conditions.
In GMA welding processes with the velocity components of the fluid at the
spray transfer mode, which is the most weld pool surface in the iterative calcu- Effect of Contact Tube-to-Workpiece
Distance

For investigating the effect of the con-

tact tube-to-workpiece distance on the
weld pool formation, the welding para-
3v, meters relevant to the contact tube-to-
workpiece distance variation are selected
as shown in Table 3 by considering the re-
sults of the arc behavior analysis. From
the analysis, it was considered that the in-
crease of the arc length due to the in-
creased contact tube-to-workpiece dis-
tance causes the distribution of the heat,
current flux and arc pressure to be ex-
panded (Refs. 12, 20, 26).
In the experiment for the contact tube-
to-workpiece distance variation, it was
revealed that the effect of the contact
tube-to-workpiece distance on the weld
shape is considerable. Figures 14 and 15
show the fluid flow pattern at the center
Fig+ 10 Velocity distribution of molten electrode transferred. plane (y = 0 mm) in the weld pool for the

148-s I MAY 1995

 X-AXIS
-15.0 -10,0 -5.0 0.0 5.0 [mm]

I I I ' I I

! ..!!! ! !ilii
X
I
N

5.0 200 [ mm/sec ]

[mm]

X-AXIS
-15,0 -10.0 -5.0 0,0 5.0 [mm]
I I ! I I

0,0
\ \ \ \ ~ •

1~dllf1111dd//

. n
5.0

10.0 200[mm/sec]
[mm]

Fi,~4. 11 Velocity distribution in weld pool by additional effect o f molten droplets. A - - ( enter plane (y = 0 ram); B - - projected top surface.

contact tube-to-workpiece distances of 15 15-mm contact tube-to-workpiece dis- droplets could make the calculation more
and 25 mm, respectively. As the contact tance in comparison with the measured accurate.
tube-to-workpiece distance decreases,the weld shape, but slightly overestimated in Figure 17 illustrates the effect of the
radii of the welding current flux and dis- the condition of 25-mm contact tube-to- contact tube-to-workpiece distance on
tributed velocity of the molten electrode workpiece distance, which is considered the calculated weld shape and shows the
were considered to decrease and conse- to be due to the rough model for trans- considerable variations between the
quently the maximum velocity of the fluid ferred molten droplets. In spite of many as- weld shapes as could be also seen in the
motion is increased. Figure 16A-C shows sumptions and simplifications, however, it experimental ones.
the comparisons between the calculated is revealed that the calculated weld
weld shape and experimental one. The shapes correspond well with those of the Conclusions
calculated penetration shape was some- experiments. It can be thus expected that
what underestimated in the condition of a better model of molten electrode A three-dimensional convection

WELDING RESEARCH SUPPLEMENT[ 149-s

 model has been developed to calculate
Y-AXIS the bead shape, velocity field and tem-
-8.0 -4,0 0.0 4.0 e.O [mm] perature field in the GMA weld pool, and
I I I I 1 then to investigate the effect of the con-
tact tube-to-workpiece distance on the
weld shape. In computer modeling, three
distinct driving forces for the weld pool
0.0
convection were considered: the buoy-
ancy force, the electromagnetic force
and the surface tension at the weld pool
surface. In addition to these, the effect of
molten electrode droplets was also con-
4.0
sidered for the accurate analysis of the
N
process. A boundary-fitted coordinate
wlth d r o p l e t eFFect system was used to handle the deformed
weld pool and weld bead surface.
. . . . . . : without d r o p l e t eFFect The result of computations has shown
e.O that the weld pool has a long tail behind
[mm] the arc due to the travel of the heat
source. The electromagnetic force may
Fig. 12 - - Effect o f molten droplets on weld shape. have a strong effect on the weld pool
convection and consequently on the
shape and size of the GMA welds, while
the effect of the buoyancy force and the
Y-AXIS surface tension force at the weld pool
-8.0 -4.0 0.0 4,0 8.0 [mm] surface is relatively small. The primary
t I circulation of the fluid flow was shown in
the middle part of the weld pool due to
the electromagnetic force, and the out-
ward flow near the pool boundary of the
0,0 free surface due to the surface tension
gradient. The narrow and deep joint pen-
etration could be acquired in the result of
calculations by considering the weld
pool convection, which cannot be seen
4.0 in the calculation result of the heat trans-
fer without considering any convective
_ _ : d = 2 . 0 m m motion. As the effective radius of the cur-
rent distribution decreases, an increase
...... :d=3.0mm in the weld pool penetration can be also
8.0 expected. By considering the molten
[mm] electrode droplets, which promotes the
fluid motion in the weld pool, the depth
of penetration was considerably in-
Fig,. 13 - - Effect o f velocity distribution radius o f molten droplets on weld shape. creased in comparison with that in the
case of considering the .three driving
forces only. In spite of many assumptions
and simplifications, it was seen that the
X-AXIS
calculated weld shapes correspond well
-15.0 -10.0 -5.0 0.0 5.0 [mm] with those of experiments.
I I I I I From the experiment and calculation
results, it was revealed that the contact
=~-=~!~=-!::=,=,- 7-, tube-to-workpiece distance exerts a con-
siderable influence on the formation of
' ° , I I : ¢ ~
the weld pool and then the weld bead
0.0
shape by affecting the arc length and
welding current. The contact tube-to-
workpiece distance is thus considered as
one of the important variables that con-
trols the formation of the GMA welds.
5.0 2 0 0 [m m / s e c ]
I
[rnm] References

1. Goldak, J., Bibby, M., Moore, l., House,

Fig. 14 - - Velocity distribution at center plane (y = 0 mm) in weld p o o l (L = 15 mm). R., and Patel, B. 1986. Computer modeling of
heat flow in welds. Metall. Trans. 17B(9):

150-s I MAY 1995

587 - 600. r
2. Sozou, C., anc] Pickering, W. M. 1976.
Magnetohydrodynamic flow due to the dis-
X-AX I S
charge of an electri( (urrent in a hemispheri-
cal container. J. Fluid Mechanics 73: -15.0 -10.0 -5.0 0.0 5,0 [mm]
641-650. I I I I I
3. Andrews, J. G., and Craine, R. E. 1978.
Fluid flow in a hemisphere induced by a dis-
tributed source of current. J. Fluid Mechani(s
84: 281-290.
4. Atthey, D. R. 1980. A mathematical
model for fluid flow in a weld pool at high cur-
rents. J, Fluid Mechanics 98: 787-801.
5. Oreper, G. M., and Szekely, J. 1984.
Heat- and fluid-flow phenomena in weM
pools. J. Fluid Mechanics 147: 53-79. |~
6. Heiple, C. R., and Roper, J. R. 1982. 5.0 200 [ mm/~ec ]
Mechanism for minor element effect on GTA [mm]
fusion zone geometry. Welding Journal 61 (4):
97-s to 1()2-s.
7. Kou, S., and Wang, Y. t4. 1986. Com- Fig. 15 - - Velocity distribution at center plane (y = 0 ram) in weld p o o l (L = 25 ram).
puter simulation of ~onvection in moving arc
weld pools. Metal/. Trans. 17A(12):
2271-2277.
10. Zacharia, T., Eraslan, A. H., and Aidun, 13. Ando, K. 1967. Control method of au-
8. Ksao, K. C., and Wu, C. S. ]988. Fluid
D. K. 1988. Modeling of autogenous welding. tomatic arc welding. J. Japan Weldin~ Society
flow and heat transfer in GMA weld pools.
Welding Journal 67(3): 53-s to 62-s. 36(5): 78 88.
Welding Journal 67(3): 70-s to 75-s.
11. Zacharia, T., Eraslan, A. H., and Aidun, 14. Ando, K., Nishiguchi, K., and Fukuda,
9. Kim, S. D., and Na, S. I. 1989. A study
D. K. 1988. Modeling of nonautogenous K. 1967. Temperature characteristics of the
on heat and mass flow in stationary gas tung-
welding. Welding Journal 67(1 ): 18-s to 27-s. droplet detaching from the wire tip in MIG
sten arc welding using the numerical mapping
12. Cary, H. B. 1979. M o d e r n Welding welding - - effect of Joule's heating. J. Japan
method. Proc. Inst. Mech. Engrs. Part B - - J.
Technology. Prentice-Hall, Inc., Englewood Welding Society 36(10): 59-66.
En#ineerin,~ Manufa( ture 203:23 ~-242.
Cliffs, N.J., pp. I69-178. 15. rsai, M. C., and Kou, S. 1990. Ele~.tro-

Y-AXIS Y-AXIS
-8.0 -4.0 0.0 4,0 B.0 [mm] -B.O -4.0 0,0 4.0 B.O [mini
I i t

i !~']
0.0 0.0~

Ii:~ii!!il

4,04-
~N
~ 4'0

- - : Celculetlon _ _ : C~]cul~tlon

...... : Experiment ..... : Exper [ment

8.0 8.0-L
[mm] [mm]
A B

Fig. 16 - - Comparisons o f calculated weld shapes

Y-AXIS with experimental ones. A - - L = 15 m m ; B L=
8,0 [mm]
-8.0 -4.0 0.0 4.0 20 m m ; C L = 25mm. F!i!~il
i i i i I

0.0
H

cz~
N~ 4 . 3

_ _ : Celculetlon

. . . . . . : Experiment
8.0
[mm)
C

WELDING RESEARCH SUPPLEMENT I 151-s

 n normal unit vector of weld pool
Y-AXIS
surface
-8.0 -4.0 0.0 4.0 8.0 [mml max
maximum value of arc pressure
! I ! I I distribution
Q effective heat input
qa heat density distribution of arc

0.0
=3Q/Tea2 exp [-3(r/a) 2]
r,z,O coordinate system of moving
heat source
tl tangential unit vector parallel
to x-z plane
4.0
J t2 tangential unit vector parallel
to y-z plane
...... : L. ~ 15.0 mm W, liquidus temperature of
weld metal
_ _ : L -- 2 0 . 0 mm
T~ maximum temperature of weld
8.0
...... : L = 25.0 mm pool surface
(mm]
T~ ambient temperature
U welding speed
Fig. 17 - - Effect of contact tube-to-workpiece distance on calculated weld shape. V velocity vector
u,1)~w x-, y-, z-component of velocity
vector V
magnetic force induced convection in weld on the three-dimensional analysis of heat and molten electrode droplet
Vd
pools with a free surface. Welding Journal fluid flow in gas metal arc welding using
69(6): 241-s to 246-s. boundary-fitted coordinates. Proc. of ASME velocity at the weld pool
16. Touloukian, Y. S. 1967. Thermophysi- Symposium on Welding and Joining surface
cal Properties of High TemperatureSolid Ma- Processes, 1991 ASME Winter Annual Meet- x,y,z coordinate system of moving
terials. Vol. 3. The Macmillan Co., New York, ing, PED-Vol. 51, Atlanta, Georgia, pp. heat source
N.Y. 159-173.
17. Na, S. J., and Lee, S. Y. 1987. A study 26. Tsai, N. S., and Eagar,T. W. 1985. Dis- Greek Symbols
on the three-dimensional analysis of the tran- tribution of the heat and current fluxes in gas
sient temperature distribution in gas tungsten tungsten arcs. Metall. Trans. 16B(12):
thermal expansion coefficient
arc welding. Proc. Inst. Mech. Engrs. Part B - - 841-846.
J. Engineering Manufacture 201:149-156. F diffusion coefficient
18. Landau, L. D., and Lifshitz, E. M. 1959. Appendix viscosity
Fluid Mechanics. Pergamon Press, Oxford, IJm magnetic permeability
U.K. Nomenclature ~,n,~ transformed coordinates
19. Lin, M. L., and Eagar, 1-. W. 1985. In-
fluence of arc pressure on weld pool geome- P density
effective radius of heat density
try. Welding Journal 64(6): 163-s to 169-s. a do/dT temperature gradient surface
20. Lin, M. L., and Eagar,T. W. 1986. Pres- distribution of arc
tension
sures produced by gas tungsten arcs. Metal. b effective radius of current
V gradient operator
Trans. 17B(9): 601-607. density
21. Haug, E. J. and Arora, J. S. 1979. Ap- distribution of arc
plied Optimal Design. John Wiley & Sons, B magnetic flux vector Superscripts
New York, N.Y. c effective radius of pressure
22. Vanka, S. P., Chen, B. C.-J., and Sha, W. distribution of arc r,z,0 r-, z-, 0 - c o m p o n e n t of depen-
T. 1980. A semi-implicit calculation proce- dent vectors
dure for flows described in boundary fitted co- c
specific heat
effective radius of velocity x,y,z x-, y-, z-component of depen-
ordinate systems. Numerical Heat Transfer3: e
1-19. distribution of molten dent vectors
23. Patankar, S. V. 1980. Numerical Heat electrode
Transfer and Fluid Flow. McGraw-Hill, New f body force vector Subscripts
York, N.Y. g gravitational acceleration
24. Jang, D. S., Jetli, R., and Acharya, S. H depth of solution domain x,y,z partial derivative with respect
1986. Comparison of the PISO, SIMPLER, and / welding current to x, y, z
SIMPLEC algorithms for the treatment of the ~,q,~ partial derivative with respect
pressure-velocity coupling steady flow prob- J
current density vector
lems. Numerical Heat Transfer10: 209-228. k thermal conductivity to%n,~
25. Kim, J. W., and Na, S. J. 1991. A study L tube-to-workpiece distance

1 5 2 - s l M A Y 1995