You are on page 1of 8

A Mathematical Model of Gas Tungsten Arc Welding

Considering the Cathode and the Free Surface


of the Weld Pool
W.-H. KIM, H.G. FAN, and S.-J. NA
A two-dimensional axisymmetric numerical model, including the influence of the cathode and the
free surface of the weld pool, is developed to describe the heat transfer and fluid flow in gas tungsten
arc (GTA) welding. In the model, a boundary-fitted coordinate system is adopted to precisely describe
the cathode shape and deformed weld-pool surface. The current continuity equation has been solved
with the combined arc plasmacathode system, independent of the assumption of current density
distribution on the cathode surface, which was essential in the previous studies of arc plasma. It has
been shown that the temperature profile, the current, and the heat flux to the anode show good
agreement with the experimental data. Moreover, the current and the heat-flux distributions may be
affected by the shape of the cathode and the free surface of the weld pool.

I.

INTRODUCTION

THE gas tungsten arc (GTA) welding process has been


widely used and has produced spectacular results. The studies of heat transfer and fluid flow in the weld pool, especially in the GTA weld pool, has been an area of active
research in recent years. For the effective usage of these
models with the weld pool, accurate information about the
welding arc influencing the molten pool is a prerequisite,
and so the study of heat transfer and fluid flow in the welding arc is equally important.
Modeling heat transfer and fluid flow in the arc plasma
has been well documented[14] These studies all dealt with
an arc plasma between a tungsten electrode (cathode) and
a water-cooled copper plate (anode). The anode was represented as a flat surface having a constant temperature.
Figure 1 shows a schematic sketch of a GTA welding operation with a deformed weld pool. It has been observed
that the surface of the weld pool becomes markedly depressed at high current levels, and the assumption of a flat
surface is no longer valid.[5] Although Choo et al.[6] presented a model of high-current arcs with a deformed anode
surface, the specified weld-pool shapes have been approximated as stepwise, and the cathode tip shape was limited
to being flat-ended. Experimentally, Lin and Eager[7] measured the arc pressure with different electrode shapes, and
showed that the sharp cathode produces a higher arc pressure than the blunt one. Therefore, the cathode shape is an
important factor influencing the welding arc characteristics
and the transferring phenomena on the base plate.
It should also be emphasized that in most previous studies on heat transfer and fluid flow in the arc plasma, a
current density profile has to be assumed over the surface
plane of the cathode,[6,8] although it has been found that the
theoretical predictions are sensitive to the current density
W.-H. KIM, Senior Researcher, is with the Agency for Defense
Development, Taejon, 305-600, Korea. H.G. FAN, Postdoctoral Fellow,
and S.-J. NA, Professor, are with the Department of Mechanical
Engineering, KAIST, Taejon, 305-701, Korea.
Manuscript submitted August 27, 1996.
METALLURGICAL AND MATERIALS TRANSACTIONS B

at the cathode.[3] A model without any assumption of the


current density at the cathode surface was presented until
recently,[9,10,11] and Fan et al.[12] used a similar model to
describe the heat and mass transfer in pulsed GTA welding.
In the present article, the model addresses a pointed-tip
cathode that fits with the actual situation. The distribution
of current density, which is determined primarily by the
welding current and the cathode shape, is calculated with
the combined arc plasmacathode system,[3] independent of
the assumption of current density distribution in the cathode. The welding arc is modeled with a predeformed weldpool surface, which is deduced from previous experimental
observations,[13] and the current and the heat flux to the
anode calculated for various cathode shapes are compared.
Because the cathode shape and the free surface at the anode
are curved, it is difficult to represent them by rectangular
grids. Thus, a boundary-fitted coordinate system is adopted,
to precisely describe the cathode shape and deformed weldpool surface profile.[8] The aforementioned work lays a
foundation for considering the interaction between the
welding arc and the molten pool. The ultimate objective is
to develop a unified model, including the cathode, the arc
plasma, and the molten pool.

II.

MODELING OF ARC

A. Governing Equations
In GTA welding, the process usually uses a direct current
straight polarity. In this model, a tungsten electrode with a
diameter of 3.2 mm is used. The assumptions adopted for
modeling are summarized as follows.
(1) The arc is assumed to be pure argon in local thermodynamic equilibrium (LTE) and is assumed to be optically thin to radiation.
(2) The arc is steady and radially symmetric. For simplicity
of the analysis, the temperatures of the cathode and
anode are assumed to be constant.
(3) The arc plasma can be assumed to be a laminar flow.
Buoyancy and the heating effect of viscous dissipation
are negligible.
VOLUME 28B, AUGUST 1997679

Fig. 1Schematic of GTAW operation with molten pool.

The free-burning arc should be analyzed with a combined model of fluid mechanics and electromagnetics, and
satisfy a group of magneto-hydro-dynamics (MHD) equations. With the aforementioned assumptions, the conservation equations are expressed in terms of cylindrical
coordinates, as follows.
Equation of mass continuity:
](ru)
1 ](rrv)
1
50
r ]r
]z

Fig. 2Geometric configuration for calculating radiative heat flux from


arc plasma to base plate.

[1]

Conservation of radial momentum:


1 ]
]v
]
]v
]P
(rrvv 2 mr ) 1
(ruv 2 m ) 5 2
r ]r
]r
]z
]z
]r
1 ]
]v
v
]
]u
1
(mr ) 2 2m 2 1
(m ) 2 jzBu
r ]r
]r
r
]z
]r

[2]

Conservation of axial momentum:

1 ]
]u
]
]u
rruv 2 mr
1
ruu 2 m
r ]r
]r
]z
]z
]P
1 ]
]v
]
]u
52 1
(mr ) 1
(m ) 1 jr Bu
]z
r ]r
]z
]z
]z

[3]
Fig. 3Computational domain for welding arc.

Conversation of energy:

1 ]
k ]h
]
k ]h
rrhv 2 r
1
ruh 2
r ]r
Cp ]r
]z
Cp ]z
5

Under the axisymmetric cylindrical coordinate, the current continuity can be written in terms of the potential, as
follows:

j r2 1 j z2
5 Kb jz ]h
j ]h
2 SR 1
(
1 r
)
s
2 e CP ]z
CP ]r

[4]

However, there are Lorentz force terms in the momentum


equations, and the energy equation contains the Joule-heating term, and an additional term that represents the transport of electron enthalpy due to the drift of electrons.
Therefore, it is necessary to solve Maxwells equations for
the electromagnetic field.
According to Ohms law, the following is true:
jr 5 2s

]f
,
]r

680VOLUME 28B, AUGUST 1997

jz 5 2s

]f
]z

[5]

]
]f
1 ]
]f
(s r ) 1
(s ) 5 0
r ]r
]r
]z
]z

[6]

At last, the self-induced magnetic field Bu is calculated


by Amperes law, as follows:
Bu 5

m0
r

* j rdr
r

[7]

The previous partial differential equations, Eqs. [1]


through [7], are solved together to obtain the distributions
of arc-plasma temperature, velocity, current density, etc.
The plasma properties for this analysis, including the denMETALLURGICAL AND MATERIALS TRANSACTIONS B

Table I.

Boundary Conditions for Welding Arc

AG or BG

GF

FE

ED

DA or BCD

u
v
h

]u/]r 5 0
v50
]h/]r 5 0

]f/]r 5 0

u50
v50
h 5 hconst
(T 5 1000 K)
f50

]u/]r 5 0
](rrv)/]r 5 0
h 5 hconst
(T 5 1000 K)
]f/]r 5 0

u 5 ugiven
v50
h 5 hconst
(T 5 1000 K)
]f/]z 5 0

u50
v50
h 5 hconst
(T 5 3000 K)
2s]f/]z 5 j0

Qr 5

p/2

* * * 4p{(r 2 r )

4Sr rl
du drdl
1 l2 1 4rr0 sin2 u}3/2

[10]
where H and R are the height and radius of the welding
arc, respectively. The geometric configuration for calculating Eq. [10] is shown in Figure 2. The heat transfer due to
the electron flow was calculated as follows.
Qc 5 ja (2.76 1 VW)

[11]

where ja is the current density into the base plate, the constant 2.76 is the approximating factor for the kinetic energy
of electrons assumed to be at 10,000 K, and Vw is the work
function of the base plate.
B. Calculation Domains and Boundary Conditions

Fig. 4Mesh generation with boundary-fitted coordinates.

sity, heat capacity, viscosity, and electrical and thermal


conductivity, are treated as temperature dependent and are
taken from the published data.[2]
Between the arc column and the electrode surface, there
is a thin transition layer in which steep gradients of arc
parameters occur. It was assumed that the cathode and anode boundary layer would have a thickness of 0.1 mm,[4]
but the calculated results were not sensitive to this assumption for the high-current arcs considered in this study. The
energy used to ionize the plasma in the cathode boundary
layer Qioniz is expressed as follows:[6]
Qioniz 5 |Jc|Vc

[8]

Here, Qioniz is a positive energy source supplied to the arc


column at the cathode boundary layer, and Vc is the cathode
fall voltage, which can be approximated to be 4.3 V.[6]
The heat is transferred to the anode plate by a complex
mechanism of convection, radiation, and electron flow.
Each term was calculated using a formula taken from works
by McKelliget and Szekely[4] and Choo et al.[6] The convective contribution is written as follows.
Qc 5

v
0.515 mabrab 0.11
(
) (mara )0.5 (hab 2 ha)
Pra
mara
r

[9]

where the subscripts a and ab denote the anode plate and


the edge of anode boundary layer, respectively, and Pra is
the Prandtl number at the anode plate. The radiative heat
flux at a point r0 of the base plate was calculated numerically by the following formula.[8]
METALLURGICAL AND MATERIALS TRANSACTIONS B

The calculation domains and boundary conditions used


are shown in Figure 3 and Table I. A large calculation
domain of AGFEA, including the cathode, is used for solving the electrical potential equation. At the surface of the
anode (GF), the electrical potential is assumed to be constant. The equation ]f/]r 5 0 has been set on EF to represent the condition that no current flow crosses this
boundary. Over the cathode cross section (AD), a uniform
current density of j0 5 2s]f/]z is assumed, defined as the
input current divided by the area of cathode cross section
(j0 5 I/(pR21)). Zero gradients are assumed for the electric
potential along the boundary of DE and the centerline of
AG.
A mesh generation for a 200-A arc with 1-mm depression at the anode is shown is Figure 4, in which 50 nodes
in the r direction and 50 nodes in the z direction are located
in the entire domain. The boundary-fitted nodes are generated in order to describe the cathode shape and the free
surface at the anode. The finer mesh is used near the cathode and the anode, because a steeper gradient of dependent
variables is expected along these surfaces.
To solve the momentum and energy equation, a smaller
domain of BCDEFGB, excluding the cathode, is used. Basically, a no-slip condition at the solid surface and a symmetric condition along the centerline are adopted. A
temperature of 1000 K is assumed on the anode surface of
GF. In the process of welding, the apex of a sharpened
electrode usually tends to experience some melting and
rounding up, and the temperature at the cathode cone surface BCD is assumed to be 3000 K. Across the boundary
EF, a constant mass flow rate and a 1000 K temperature
condition are used. Along the boundary DE, the radial-velocity component is neglected, and the axial-velocity component is determined from the formula of pipe flow, as
follows.[8]
VOLUME 28B, AUGUST 1997681

Fig. 5Geometric transformation and grid structure with notation.

1 ]
]F
rrvF 2 rGF
r ]r
]r

]
]F
ruF 2 GF
5 SF
]z
]z

[13]

where GF is the general diffusion coefficient and SF is the


corresponding source term. Then, Eq. [13] was transformed
to a generalized curvilinear coordinate system (Figure 5)
and resulted in the following equations.
Fig. 6Temperature profile compared with experimental results for 200-A
arc and 5-mm arc length.

u52

Q
pr

ln (r/R2)
R 2 r 1 (R 2 R )
ln (R2 /R1)
2
2

2
2

R42 2 R41 1

2
1

(R22 2 R21)2
ln (R2 /R1)

[12]

where Q is the inflow rate of shielding gas, R1 is the radius


of the electrode, and R2 is the internal radius of the shielding nozzle. The temperature of the inflow shielding gas,
with a flow rate of 101/min, was assumed to be 1000 K.

III.

NUMERICAL PROCEDURE

A. Transformation
To solve the governing equations with the finite-difference method, the physical domain represented by a boundary-fitted coordinate system must be transformed into a
rectangular domain. Before applying the transformation, the
governing equations can be expressed in a generalized formula using a general dependent variable F, as follows.

682VOLUME 28B, AUGUST 1997

]F ]z ]j
1 ]
]F ]j
rrvF 2 rGF (
1
r ]j
]j ]r
]z ]r ]r
1 ]
]F ]j
]F ]z ]z
1
rrvF 2 rGF (
1
)
r ]z
]j ]r
]z ]r ]r
]
]F ]j
]F ]z ]j
1
ruF 2 GF (
1
)
]j
]j ]z
]z ]z ]z
]
]F ]j
]F ]z ]z
1
ruF 2 GF (
1
)
5 SF
]z
]j ]z
]z ]z ]z

~
~

[14]

!
!

where
]j
1 ]z
5
,
]r
J ]z

]z
1 ]z ]j
1 ]r
52
,
52
,
]r
J ]j ]z
J ]z
1 ]r
]r ]z
]r
5
, J5
2
J ]j
]j ]z
]z

]z
]z
]z
]j

[15]

The transformation coefficients were calculated numerically, using the second-order central difference method. In
the transformed domain, the grid size was set to be unity
to simplify the calculation.
B. Technique of Solution
The calculation domain for u, v, and h excludes the solid
cathode. This is done by rendering the control volume of

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 7Anode current density and heat flux for 200-A arc with 6.3-mm arc length (voltage 5 16.4 V).

the solid cathode inactive, so that the remaining active control volume forms the desired irregular domain
(BCDEFGB).
To solve the potential equation, the field is specified by
employing the true electrical conductivity of the solid cathode and the plasma fluid in their respective regions. Therefore, the problem is solved as a conduction problem
throughout the entire calculation domain.
The transformed Eq. [14] is discretized using the control
volume approach[14] in the transformed domain (Figure 3),
which results in an equation of the following form.
aPFP 5 aEFE 1 aWFW 1 aTFT 1 aBFB 1 S

[16]

where the general dependent variable F can be any of u,


v, h, P, or f. The values of ap, etc., are the coefficients that
result from Eqs. [1] through [7], and S is the source term
from Eqs. [1] through [7]. Variables of the equations are
solved iteratively. The numerical iteration procedure is as
follows.
(1) Read the welding conditions, generate the grid, and calculate the transformation coefficient.
(2) Initial arbitrary values of P, u, v, T, and f are selected.
(3) Material properties of plasma are evaluated for each
grid point by linear interpolation of the known material
properties as a function of temperature.
(4) Equation [6] is solved for the potential distribution f
(r, z), and the values of jr, jz and the Lorentz force are
obtained using Eqs. [5] and [7].
(5) Updated values of u, v, and P are obtained from Eqs.
[1] through [3] using the SIMPLEC method.
(6) Updated values of h are obtained from Eq. [4].
(7) Return to step (3), and this loop is repeated until the
convergence is achieved.
Convergence is declared when the following condition is
satisfied.

F
L

i51 k51

m
i,k

2 Fm21
i,k

Fmax

, F,

F 5 0.01

METALLURGICAL AND MATERIALS TRANSACTIONS B

[17]

where L and N are the total number of nodes in the r and


z directions, respectively, and m is the number of iterations.
In addition, the current density integrated over the anode
surface yields a total current within 1 pct error of the welding current.

IV.

RESULTS AND DISCUSSION

The temperature contours predicted for a 5-mm-long arc


at 200 A with a 60 deg cathode vertex angle are shown in
Figure 6, where the temperature distribution measured by
Haddad and Farmer[15] is also given for comparison. As
shown, the typical bell shape of the arc periphery expressed
by the isotherm of 10,000 K is clearly observed. Comparison between the calculated and the measured temperature
distributions shows a fairly good agreement. Figure 7
shows a comparison of the calculated current density and
the heat-flux distribution at the anode with experimental
measurements of Nestor.[16] The good agreement shown
provides a measure of reliability for the numerical model.
Figure 8 shows the current density distribution for the
200-A arc formed on the surface of a depressed pool, previously determined according to Lin and Eagars
experiment.[13] As shown, the welding current leaves the
anode surface and enters the cathode tip through the arc
space. The current density achieves its greatest value at the
tip of the cathode. In order to illustrate the current density
distribution in the cathode clearly, a large standard scale
(150 A/mm2) is used. As a result, the current density distribution is not clear near the anode, where the value of
current density is very low. In fact, the current flow near
the deformed weld pool appears to be perpendicular to its
surface.
Figure 9 shows the temperature contours calculated for
the depressed weld pool. As shown, the maximum temperature occurs near the tip of the cathode, because the current
density achieves its biggest value there, while the cathode
must be maintained at 3000 K. The temperature contours
appear to be parallel to the deformed weld-pool surface in

VOLUME 28B, AUGUST 1997683

Fig. 8Calculated current density for 200-A arc with 1-mm depression.

Fig. 9Calculated temperature contours (10,000 to 22,000 K) for a 200-A


arc with 1-mm depression.

the arc center, because the anode is assumed to maintain a


constant temperature of 1000 K.
The corresponding velocity vectors of plasma flow are
shown in Figure 10. It is shown that the fluid flows inward
and downward along the slant of the cathode at first, and
then downward along the axis of symmetry. Due to the
stagnation effect at the anode, the fluid is then deflected,
and flows radially outward along the free surface of the
molten pool.
Figure 11 shows the current density and heat flux for a
200-A arc with 607 (solid line) and 120 deg (dashed line)
684VOLUME 28B, AUGUST 1997

Fig. 10Calculated velocities for a 200-A arc with 1-mm depression.

tip angles. Because of the lack of experimental data, and in


order to investigate the varying tendency of current density
and heat flux with different cathode tip angles, the same
weld-pool shape was predefined. Because the surface depression increases the arc length in the arc center, and the
hump area around the pool decreases it, a bimodal distribution (only half is seen, because of symmetry) is observed
for the current density. In addition, it can be seen that the
smaller the cathode vertex angle, the bigger the peak current density. The reason for this is that a smaller cathode
tip angle results in a reduced radius of the cathode tip conductive section, which causes an increase in the cathode
current density. Therefore, with a decrease in the cathode
vertex angle, the temperature near the cathode increases
because of the intensive Joule heating, which causes an
increase in the degree of ionization, and consequently the
increase in current density.
The variation of the heat flux entering into the weld pool
is also shown in Figure 11. Because the heat flux into the
weld pool is mainly determined by the electron flow, the
heat flux also shows a bimodal distribution similar to the
corresponding current density. The ratio of the maximum
to the minimum is not as large as that of the current density,
because the deformed pool affects the convective contribution much less than it does the electronic contribution.
As the cathode vertex angle decreases, the gas jet velocity
increases,[8] which causes the increase of convective heat
transfer. Since convection continues to add more heat to
the center of the molten pool with a sharper cathode tip
angle, the difference between heat fluxes with different
cathode vertex angles is larger than that of the current density in the middle region of the weld pool.
The simulation results for a 260-A arc with a 1.2-mm
surface depression are similar to those of 200-A arc (Figure
12). Since the temperature distribution of the arc plasma
influences its electrical conductivity, which determines the
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 11Anode current density and heat flux for a 200-A arc with 1-mm depression (voltage 5 16.4 V).

Fig. 12Anode current density and heat flux for a 200-A arc with 1.2-mm depression (voltage 5 17.6 V).

current flow, the temperature near the anode has a great


influence on the current flux distribution along the anode
surface. It is well known that the increasing current results
in the increase in temperature, and since the electrical conductivity of pure argon gas is nearly saturated above the
temperature of 15,000 K, the electrical conductivity of arc
plasma is not markedly affected by the cathode shape at
high current levels. Therefore, compared with the situation
of lower current, the current and the heat flux to the anode
are not markedly affected by the cathode shape at higher
current levels.
V.

CONCLUSIONS

A two-dimensional axisymmetric model, including the


influence of the cathode and the free surface of the weld
pool, has been developed to describe the heat transfer and
fluid flow in GTA welding. The distribution of current denMETALLURGICAL AND MATERIALS TRANSACTIONS B

sity, which is determined primarily by the welding current


and the cathode shape, is calculated with the combined arc
plasmacathode system.
Using a boundary-fitted coordinate system, the rounded
and sharply pointed cathode tip and the free surface of the
weld pool could be described. Through the geometric transformation from a nonrectangular physical domain to a rectangular, uniformly spaced computational domain, the
nonlinear conservation equations were solved by the finite-difference method in the transformed domain.
As a result of simulations for a 6.3-mm-long arc at 200
A, the temperature profile and the current and the heat flux
to the anode show good agreement with the experimental
data.
Calculations performed for deformed anode surfaces
have shown that the current and the heat-flux distribution
may be in a bimodal shape because of the depression of
the weld-pool surface. The current and the heat flux to the
VOLUME 28B, AUGUST 1997685

anode increase with the decreasing cathode vertex angle,


but are not markedly affected by the cathode shape at high
current levels.

Bu
Cp
e
h
H
I
ja
jc
jz, jr
k
Kb
l
L, N
P
Pr
Q
Qc
Qc
Qioniz
Qr
r, u, z
r0
R
R1
R2
SR
T
u, v
Vc
VW

SYMBOLS
azimuthal magnetic field, Wb/m2
heat capacity, J/kg
electronic charge, C
plasma enthalpy, J/kg
height of welding arc, mm
arc current, A
anode current density, A/m2
cathode current density, A/m2
axial, radial current density, A/mm2
thermal conductivity, W/(m z K)
Boltzmans constant, J/K
arc length, m
total number of nodes in j and z directions of
grid
pressure, Pa
Prandtl number 5 Cpm/k
inflow rate of shielding gas, m3/s
convective contribution to anode heat flux,
W/m2
electron contribution to anode heat flux, W/m2
heat source from cathode fall to ionize plasma,
W/m2
radiative contribution to anode heat flux, W/m2
cylindrical coordinate system
radial coordinate of a point on base plate, m
radius of welding arc
cathode radius, m
internal radius of shielding nozzle, m
radiative heat loss term, W/m3
temperature, K
axial, radial velocities, m/s
cathode fall voltage, V
work function of anode material

Greek symbols
GF
general diffusion coefficient

error
l
parameter for calculating Qr, m

686VOLUME 28B, AUGUST 1997

m
m0
j, z
r
s
f
F

viscosity, kg/(m z s)
permitivity of free space, H/m
transformed coordinate system
density, kg/m3
electrical conductivity, 1/(ohms z m)
potential, V
general dependent variable

Subscripts
a
ab
cal
exp
i, k

anode plate
anode boundary layer
calculated
experimental
j and z direction node numbers

REFERENCES
1. K.C. Hsu, K. Etemadi, and E. Pfender: J. Appl. Phys., 1983, vol. 54,
pp. 1293-1301.
2. R.T.C. Choo, J. Szekely, and R.C. Westhoff: Metall. Trans. B, 1992,
vol. 23B, pp. 357-69.
3. J.J. Lowke, P. Kovitya, and H.P. Schmidt: J. Phys. D: Appl. Phys.,
1992, vol. 25, pp. 1601-06.
4. J. Mckelliget and J. Szekely: Metall. Trans. A, 1986, vol. 17A, pp.
1139-48.
5. M.L. Lin and T.W. Eagar: Weld. J., 1985, vol. 64 (6), pp. 163s-169s.
6. R.T.C. Choo, J. Szekely, and R.C. Westhoff: Weld. J., 1990, vol. 69
(9), pp. 346s-361s.
7. M.L. Lin and T.W. Eagar: Metall. Trans. B, 1986, vol. 17B, pp. 60107.
8. S.Y. Lee and S.J. Na: Proc. Inst. Mech. Eng. Part B: J. Eng.
Manufacture, 1995, vol. 209, pp. 153-64.
9. C. Delalondre and O. Simonin: J. Phys. Coll., 1990, vol. 51, pp.
C5-C199.
10. P.Y. Zhu, J.J. Lowke, and R. Morrow: J. Phys. D: Appl. Phys., 1992,
vol. 25, pp. 1221-30.
11. H.G. Fan and Y.W. Shi: J. Mater. Processing Technol., 1996, vol.
61, pp. 302-08.
12. H.G. Fan, S.J. Na, and Y.W. Shi: J. Phys. D: Appl. Phys., 1997, vol.
30, pp. 1-9.
13. M.L. Lin and T.W. Eagar: Transport Phenomena in Materials
Processing, PED-Vol. 10, HTD-Vol. 29, ASME, New York, NY,
1983, pp. 63-69.
14. S.V. Patankar: Numerical Heat Transfer and Fluid Flow, Hemisphere,
Washington DC, 1980.
15. G.N. Haddad and A.J.D. Farmer: Weld. J., 1985, vol. 64 (12), pp.
339s-342s.
16. O.H. Nestor: J. Appl. Phys., 1992, vol. 33, pp. 1638-48.

METALLURGICAL AND MATERIALS TRANSACTIONS B

You might also like