Вы находитесь на странице: 1из 8

An Electrode Extension Model for

Gas Metal Arc Welding

Short circuiting transfer can be predicted within 9%


for a given electrode feed speed

BY T. P. Q U I N N , R. B. M A D I G A N A N D T. A. SIEWERT

ABSTRACT. The electrode extension dur- Introduction contact with the workpiece) is often used
ing gas metal arc welding is predicted for thin or heat-sensitive materials (Ref.
using a one-dimensional model of the This paper presents a model for elec- 1). It is desirable to be able to predict the
melting electrode. Joule heating in the trode extension during gas metal arc transfer mode from the process variables
electrode, heat directly applied to the welding (GMAW). In GMAW, an arc is both for automatic control and for weld
end of the electrode from the condensing created between the continuously fed, schedule design.
electrons, and heat transferred from the consumable electrode and the work- Halmoy (Refs. 2-4) neglected thermal
droplet, together with conduction along piece. The melted electrode acts as filler conduction and used an energy balance
the electrode are considered. The ther- metal in the weld. The length of the solid at the tip of the electrode to calculate the
mal conductivity, the thermal diffusivity, electrode from the point of current trans- steady state (Ref. 2) and dynamic melting
and the electrical resistivity of the elec- fer to where the electrode is melted, the rates (Ref. 4), as well as the EE. He as-
trode material are allowed to vary with electrode extension (EE) determines the sumed a fixed EE and heat generation
temperature. The steady-state electrode length of the arc and, therefore, the from the arc and Joule heating. Waszink
extension is predicted to an accuracy of amount of power used in the welding and van den Heuvel (Ref. 5) assumed a
1.9 mm (0.074 in.). The onset of short- process. Transfer mode (globular, spray, fixed EE and included the effects of con-
circuiting as the current is decreased for etc.), arc stability, and deposition rate de- duction, the Thomson effect, radiation,
a given electrode feed speed is predicted pend on the EE (Ref. 1). and Joule heating. They found the losses
within 9%. Dynamic analysis shows that from radiative heat transfer and the
The way in which the metal is trans-
the gas metal arc welding process acts as Thomson effect to be about two orders of
ferred from the electrode to the weld pool
a low-pass filter for electrode extension magnitude less than the contributions of
affects the final shape and quality of the
with respect to the square of the current Joule heating and heat from the arc. Con-
weld (Ref. 1). Spray transfer (when the di-
(proportional to power) and with respect ductive heat transfer was larger than or of
ameters of the detached droplets are
to electrode feed speed. As the mean the same order of magnitude as Joule
equal to or smaller than the electrode di-
welding current is increased, the elec- heating in about the first 25% of the elec-
ameter) is the highest production transfer
trode extension (or arc length if the con- trode from the contact point. The heat
mode. Short circuit transfer (when the
tact-tube-to-work distance is constant) conducted from the liquid at the elec-
electrode comes into direct electrical
has a smaller response to perturbations in trode tip was 33% of the total Joule heat-
the current or electrode feed speed. The ing at 137 A and 1 1 % of the total Joule
quasi-linear transfer functions between heating at 238 A for the 1.2-mm (0.045-
electrode extension and current squared in.) steel electrode they studied. Data for
and between electrode extension and KEY WORDS the EE, current and electrode feed speed
electrode feed speed can be described by or wire feed speed were experimentally
one zero, two pole parametric fits. The Electrode Extension determined.
transfer functions are linear in the ampli- GMAW
tude of the excitation up to 10% of the Kim, et al. (Ref. 6), concluded that an
One-dimension Model energy balance approach was "inade-
mean excitation. The model transfer Temperature Profile
functions were verified with experi- quate to explain the observed melting
Describing Function phenomena." Using a steady state, two-
ments. Spray Transfer Mode dimensional model with conduction, he
Welding Current predicted the temperature distribution in
Electrode Feed Speed the electrode. Again, EE, current and
Transfer Functions electrode feed speed were determined
75 P. QUINN, R. B. MADICAN and T. A. SIE- Stefan Problem experimentally. He modeled the heat
WERT are with National Institute of Standards transferred to the melting tip of the solid
and Technology Boulder, Colo.

W E L D I N G RESEARCH SUPPLEMENT I 241-s


heating, heat from electrons that con- tion heat is applied directly to the solid
r^H dense directly on the solid electrode, and electrode at z = £ giving
Contact heat transferred from the liquid drop at
Tube
the electrode tip was developed for the QA = YVCI
spray transfer mode. The resulting set of (5)
UJ
S
Ohmic
Heating
I2r(T)
T< y
T partial differential equations
one-phase Stefan problem) was solved
(a where V c is the apparent condensation
voltage and y is the fraction of electrons
a. A
-S
0
1 for the transient and steady-state cases. that condense on the solid electrode. As
O vz
Quasi-linear transfer functions (using the shown in the appendix, the rate of heat
_i Heat 0

UJ from ii method of describing functions (Ref. 9) transferred from the melt can be approx-
Condensing Ul
> Electrons between EE and the square of the current imated by
Ui
°A ^ and EE and electrode feed speed were de-
a ^~~~--* veloped. The transfer functions describe Vc(\-y)l
X
o v^m
Q /' V r
HGat from
H
TD
the dynamic relationship between the
input, current squared or electrode feed
Qn
2 (6)
QC Liquid
Melt speed, and the output, EE. y must be determined experimentally. As
<
Ui seen in the appendix, the equations are
W) first nondimensionalized and then
Ul F/g. 1 — The melting electrode model. Model
solved using a finite difference method.
cc The equations governing steady-state
Neglecting heat losses due to the
electrode as coming from two sources: Thomson effect and radiation, the one- heat transfer and their method of solution
heat transferred from the liquid, and heat dimensional governing equation for the are also shown in the appendix.
directly delivered to the solid electrode rate of heat transfer in the electrode is There is a lack of data across the tem-
by the arc. perature range of interest (room temper-
There has been no attempt known to _d_ dT),l2(t)r(T) ature to melting point) for the thermal dif-
the authors to predict EE from the process K(Ty fusivity a(T), electrical resistivity r(T) and
variables (current, electrode feed speed, dzKy'dzj A* thermal conductivity K(T) for the low-
electrode composition, and electrode di- carbon steel electrode (E100S-1) used in
ameter) using a model that includes con- dZ dt ) the experiments; y is also unknown. To
duction. approximate the material properties, the
(1)
Dynamic models of G M A W are data for pure Fe (Refs. 13-15) are used
needed to facilitate control of the (see Fig. 1). Here, T is temperature; K is and corrected for the material used in the
process. Wu and Richardson (Ref. 7) used experiments. (For pure Fe, the data at the
thermal conductivity; I is current; r is re-
Halmoy's model of the electrode (Ref. 4) melting point have an estimated uncer-
sistivity; A is the cross-sectional area; v z
in an overall dynamic model of the entire tainty of ±20%, so a similar procedure
is electrode feed speed, c is specific heat
GMAW process. Kannatey-Asibu (Ref. 8) would be required even if a pure Fe elec-
and p is the density. The boundary con- trode were used.) The data for Fe are mul-
used an experimentally identified first-
ditions are tiplied by a constant over the temperature
order equation to model the dynamics of
the arc length where EE plus arc length is range. To find the multiplication con-
equal to the contact-tube-to-work dis- T ( Z = 0) = Tc, stants for the material properties and to
tance (CTWD). find y, the error between the experimen-
T(Z = £(t)) = Tm
The purposes of this research were: 1) tal steady-state EE £ss and the model 4s
(2) minimized as the constants are varied.
develop a dynamic and steady-state where Tc is the temperature of the con-
model that predicts EE for GMAW from Grid searches ranging ±15% around the
tact tube and T m is the melting tempera- nominal values of Fe with 5% steps are
the other process variables; 2) determine
ture. An additional condition is neces- used to find a(T m ), r(Tm), K(Tm). y was var-
the steady-state process variables at
which short circuit transfer takes place; sary because the boundary moves with ied between 0.05 and 0.25 in 0.005 steps.
and 3) quasi-linearize the model to a set time: To understand the dynamics of the
of transfer functions for use in automatic system around the steady-state solutions,
, ,9T QA(t) Qjt) the method of describing functions (Ref.
control. A thermal conduction model
I '' JdZ
7 A A
that includes heat sources from Joule 9) is used. Approximate transfer func-
p(T)Lz{t tions between the dimensionless EE s(t) =
t/£0 and the dimensionless, volumetric
(3)
Joule heat generation
(the energy balance at the tip of the solid
Table 1 — Optimal Material Constants and y electrode). Here, Q A is the rate of heat
for E100S-1 and Nominal Material Constants
for Fe (Refs. 12-15). The Nominal Values for Fe transfer to the solid electrode from con- MO yK{Tm ){Tn
Were Used Except Where Noted densing electrons (z=i ). Q m is the rate of
and between s and the Peclet number
heat transfer from the liquid melt. This set
fe E100S-1 of equations with the addition of the ini-
a(T m ), m 2 /s 6.30 X 1CT6 6.93 X 1 0 - 6 tial conditions, W it =
a(Tm )
K(Tm), W / ( m . K ) 34.6 31.1
r(Tm), /jQ-rn 1.32 1.52 *(f = 0) = y are calculated. W and q.- are varied about
2.72 X 103 —
L, J/kg T(Z,t- oW„ their steady-state values W ss and q i s s :
T m , °C 1510 — (4)
c(T m ),J/(kg.K) 751 — is a one-phase Stefan problem (Ref. 10). -a .sin(2n:(or)
4 y'ss ' aqJq •
Vc, V 6.00 — Following the developments of Kim
7 — 0.08 W- • W„ +a,A,sin(2nan)
(Ref. 11), a percentage of the condensa-
(7)

242-s I OCTOBER 1 9 9 4
where aqj and a w are the amplitudes of
the oscillations.
16.0- • ,''
ta(Tm) jk
14.0- y'' A

to B
~ 12.0-
± y A.

. * • *
t
A
is the Fourier number, and co is the di- H 10.0- y A.
TJ
mensionless frequency of oscillation.
The response of s in x is then calculated 1 8 "°-
•H
A y

using the temperature distribution of the s. 6 -°- y' 4k

steady-state solution as the initial condi- Fig. 2 — Model


4.0- y ^
• r electrode extension
tion. To calculate the magnitude and \ Mndpl = ^xr*1 rim ent
calculated using the
2.0-
phase of the transfer function at co, a and "" parameters in Table 1
\\t are found in '0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 1f ,0
and the experimental
M o d e l ft (mm) electrode extension.
s =a sin(tm+ yr)
(8)
model. The I 2 or v z was varied sinu- phases Z£-Z\2 or Z(-Zvz.
through a least squares fit. The magni-
soidally about a mean value at selected
tude of the transfer function at co is then
frequencies; v z or I, respectively, was Result
predicted to be a/aq:. The phase of the
maintained constant. The data recording
transfer function is predicted by \|f. Once rates were reduced because of the rela-
transfer function data at discrete fre- Using the optimal material constants
tively long data recording period (~30 s). and the optimal y (Table 1), the standard
quencies are determined, parametric, ra- Here, the voltage, v z ; and current were deviation between the steady-state EE
tional curves are fitted to the transfer sampled at 210 samples/s after being measured in the experiments and the
functions according to the procedure of low-pass filtered with a cutoff frequency steady-state EE predicted by the model is
Lin and Wu (Ref. 16). of 100 Hz. The video data were recorded 1.9 mm for the E1 00S-1 electrode — Fig.
at 250 frames/s. 2. The steady-state model EE is derived
Experiments The end of the solid electrode was de- from the relationship between qj ss and
fined as the point midway between Wss-Fig.3.
Bead-on-plate welds were made with where the electrode starts to neck (spray Figure 3 can also be used to determine
a 1.14 mm (0.045 in.) E100S-1 electrode, transfer) and where droplets formed. The a lower bound on the current at which
1 5-22 mm (0.6-0.9 in.) CTWD, and 95% point of current transfer was assumed to short circuiting will occur. By assigning
Ar-5% C 0 2 gas. A low-noise regulator be at the exit of the contact tube. The EE i0 = CTWD and using the optimal mate-
was used as described in Ref. 1 7 to main- was determined from the video data by rial properties in Table 1, Fig. 3 can be
tain constant current. To obtain EE mea- counting the number of pixels between transformed to a relationship between
surements, a 10 mW He-Ne laser and the contact tube and electrode end frame current and v z . Short circuiting will actu-
632-nm bandpass filter were used to cre- by frame. For the dynamic experiments ally occur at currents higher than those
ate a shadowgraph of the electrode and the EE was measured automatically to predicted because droplets will bridge
base plate (Ref. 17). The shadowgraph within about 0.9 mm by capturing the between the electrode tip and the weld
images were recorded on a high-speed video data frame by frame and detecting pool; the current predicted here is that
video system. The contact tube, the elec- the outline of the electrode using an which would cause the electrode to enter
trode, and the workpiece were imaged. image processor. the weld pool. Using the data in Ref. 1 7,
The current was measured with a Hall The phase and amplitude of the cur- the predicted short circuiting is com-
effect transducer with an absolute error rent squared or electrode feed speed and pared to the actual onset of short circuit-
of 1 % . The voltage between the torch the EE in the dynamic experiments were ing — Fig. 4. The largest difference be-
and the baseplate was measured within determined by least-squares fitting the tween the predicted and measured
0.5% absolute error. The axial velocity of data with a sine wave. The EE to current currents at the onset of short circuiting is
the electrode or feed speed (vz) was mea- or EE to electrode feed speed transfer 9% at 187 mm/s (442 in./min).
sured as the electrode entered the wire function were then constructed by taking For large W such as are produced
feeder. A pinch roller (16 mm diameter) the ratio of the magnitudes IIJ?/I 2 II or when v z or £0 is large, the model predicts
was attached to an optical encoder (5000 ll<?/vzll and taking the difference in the a steady-state temperature distribution
pulses/rev) and the resulting pulse train
was frequency converted to give a volt-
age signal proportional to v z . The rms un- 500i
certainty from calibration tests for the 450-
transducer was 2 mm/s. Model
A"7
400- A A /
Experiments were undertaken to de- Experiment A/
350-
velop and verify the model for steady- h6
300
state conditions. The welding current, / A A

welding voltage and v2 signals were low- 2s 250-

pass filtered with a cutoff frequency of 200-

3000 Hz and recorded on a laboratory 150


computer at 8928 samples/s. The video 100
data were recorded at 3000 frames/s with t?
50 Fig. 3 — Model qy ss
0.2 mm/pixel resolution. The data were calculated using the
0
recorded for about 2 s during welding. 3 50 100 150 200 250 300 3 50
parameters of Table 1
Experiments were also conducted "s,Vz
and the experimental
Wss
under dynamic conditions to verify the «(TJ results.

WFI n i N O RFSFARGH SUPPI FMFNT I 243-s


500-
450- 1 + a, cr + a2 CT"
Model
400- where a is the Laplace variable. For s/qj
Experiment [17] at W = 75 and for s/W at mean W < 1 75,
350-
the best fitting rational f u n c t i o n , ac-
300- cording to the criteria of Lin and Wu
2 250- (Ref. 16), is a one pole, no zero model
(bi = 0, a 2 = 0). For s/qj and s/W, a-, and
Fig. 4 — The model- 200-
a 2 decrease with increasing mean W —
predicted onset of short
150- Figs. 10, 11. For s/qj, b 0 and b, increase
circuiting transfer as
the welding current is 100- with increasing W (Fig. 12), but for s/W,
decreased, calculated b 0 and b, decrease w i t h increasing
50-
using the parameters of mean W — Fig. 1 3.
Table I and the experimen- o- The transfer functions magnitude and
tal results. The CTWD 3 50 100 150 200 250 300 350
phase for s/qj and sAA/ were nearly inde-
was 13 mm. Vz(mm/s)
pendent of tne excitation amplitude for
a q j <0.10q j s s and a w < 0 . 1 0 W s . For s/qj,
example, for W = 75, lls/qll = 0.16 until the identified parametric coefficients
T-Tc
u= — co = 0.35 where it begins to roll off; for W with aqj = 0.075qj ss were within (4 x
Tm-Tc = 425, lls/qll = 0.20 until W=20 where it 1 0~4, 5 x 10- 6 , 8 x 1 0- 4 , 3 x 1 0-6) of the
which is nearly linear in x = 7./£0 until a begins to roll off. For s/qj (or sAV), the (a,, a 2 , b 0 , b,) with aq: = 0 . 1 0 q j s s — Figs.
phase starts near 1 80 deg (0 deg for s/W) 10, 12. For s/W, the identified parametric
point near the melting boundary is coefficients with a w = 0.075W SS were
for co « 1, passes through 90 deg (-90
reached where there is a sharp rise in u within (6 x 10- 3 , 2 x 1 0-f>, 2 x 10- 2 , 3 x
deg for sAV), and returns above 90 deg
— Fig. 5. For example, when W = 300.5, 10- 6 ) of the (a 1; a 2 , b 0 , b,) with a w =
(-90 deg for s/W) as co increases.
u = 0.238 x for 0 < x < 0.98 with a cor- 0.10WSS — Figs. 11, 13. Since
The model prediction for <f/l2was
relation coefficient of 0.998; 75% of the

<2m
compared to experiment for v z = 0.12
increase in u occurs in the last 2% of the ry
m/s and sinusoidal oscillations in I 2 of
electrode. For small W, u rises more grad-
amplitude 7 to 10% of the mean I 2 = 9/ 2
ually; when W = 0.5, 75% of the increase A K(Tm)(Tm-Tr
62500 A 2 (I = 250 A) — Fig. 8. The rms
in u occurs in the last 23% of the elec-
differences between the model predic- a 10% perturbation in qj corresponds to
trode — Fig. 5.
tion and the experiment were 0.6 x 10 - 7 a 32% perturbation in I.
The dynamic response of EE to sinu- m/A2 in magnitude and 11 deg in phase;
soidal excitation in I 2 and v z shows a sin- the model prediction is reasonably close Discussion
gle frequency response at the frequency to the experimental results. The model
of the excitation (nonlinear systems can prediction for £/vz was compared to ex-
show response at frequencies other than The value of y found here, 0.08, is
periment for I = 250 A and sinusoidal os- near that predicted by Kim, et al. (Ref. 6),
the excitation frequency). The response cillations in v z of 10% of the mean 0.12
of the EE is characterized by the transfer of 0.1 < y < 0.25 for spray transfer. The
m/s. The model again closely predicts the optimal values for cc(Tm), r(T m ), and K(Tm)
functions (Figs. 6, 7), which show that the experimental results; the rms differences
process acts as a low-pass filter for EE lie within 1 5% of the nominal values for
between the model prediction and the pure Fe, as would be expected for the
with respect to I 2 and v z . The magnitude experiment were 0.04 s_1 in magnitude
of the transfer functions in s/q; and s/W low-alloy steel E100S-1 used in the ex-
and 8 deg in phase. periments. There is a 20% uncertainty in
are essentially flat with increasing co until
The least squares parametric fits of the these material constants for Fe itself. Even
the cutoff frequency is reached, after
transfer functions s/q.- and s/W (Figs. 6, 7) ifa(T m ), r(Tm), and K(Tm) for the electrode
w h i c h the magnitude decreases one
were two pole, one zero models of the were k n o w n , the steady-state experi-
order of magnitude for a one-decade in-
form ments would need to be conducted to de-
crease in co. The magnitude decreases
with increasing W but the cutoff fre- termine y. The prediction of the current
quency increases with increasing W. For at the onset of short circuiting within 9%
offers an independent confirmation of
the model as does the data for the trans-
fer function seen in Figs. 8 and 9. The
(small) differences between the model

If III
transfer functions and those found exper-
1.0 - i 1
imentally (rms differences of 0.6 x 10~7
jjj i l l
m/A2 in magnitude and 11 cleg in phase
for i/\2 and 0.04 s - 1 in magnitude and 8

Fig. 5 — Model
u 0.5 —
lp!If deg in phase for U\7 indicate that the
model can be used to design and imple-
ment a control system for EE.
temperature Dynamic analysis of the EE shows that
distribution
the EE, and hence arc length, becomes
where
1.00
more stable as the mean current (qj) in-
150^ 0.75 creases — the EE is less affected by per-
..Li, w{t)JM, ""a so turbations in the current or wire feed
T„,-Tc w
a(Tm) lp •^25
300-* 0 X speed.
and x = Z /10.

244-s I OCTOBER 1994


Conclusions
W Model Parametric Fit
75 a
1) A dynamic and steady-state model 225 o
that predicts electrode extension in the 425 « - - -
spray transfer mode of gas metal arc
welding has been developed. The model !
predicted the steady-state electrode ex- - .
tension within 1.9 mm and the current at 0.1
=
the onset of short circuiting within 9% for S
an E100S-1 electrode. ).01 ©- -- -€> -© & officaee D— O-O Geosear©- G - / * " ^ .
2) Dynamic analysis using the method — x -* x >exx»st— — -*< — x —x x-5?§Ji&»,
of describing functions shows that as the 0.001 - 1 l l l l l I i i ii i i i i iiiii i ii nm

welding current is increased, the elec-


180 Fig. 6 — The
trode extension (or arc length if the con-
transfer functions
tact-tube-to-work distance is constant)
of s/qj where
becomes more stable. Gas metal arc ~ 135
N& "O x
welding acts as a low-pass filter for elec- 10
trode extension with respect to the 5 90- ^*****^_VJ \t\/?K(Tm ){Tm-Tc)
mag = ~-?— -,
square of the current (proportional to OJ

power) and with respect to the electrode i I I i Mil 1 1 1 lllll 1 1 I 1 lllll 1 1 1 1 1 III phase - Zl - Zl2, and a> =
45 «(Tm)
feed speed. 0.01 0.1 1 10 100
3) The quasi-linear transfer functions f is the frequency . W -
0) *{Tm
between electrode extension and current
squared and between electrode exten-
sion and electrode feed speed can be de-
scribed by one zero, two pole paramet- W Model Parametric Fit
75 a
ric fits.
225 o
4) The transfer functions between the 425 x - - -
electrode extension and the current 1
squared are linear in the amplitude of the
current squared excitation up to 10% of
the mean current squared. The transfer 0.1
tn
functions between the electrode exten- ro
2
sion and the electrode feed speed are 0.01
also linear in the amplitude of the elec-
trode feed speed excitation up to 1 0% of 0.001
the mean electrode feed speed.
5) The dynamic response of the model 180
was verified by experiments: at a mean
Fig. 7 — The
current of 250 A and a mean electrode 135 transfer functions
feed speed of 0.12 m/s, the model trans- CO of s/W where
fer function between electrode extension CB

and current squared was within 0.6 x


10~7 m/A 2 in magnitude and 11 deg in
45 _i ni i I M
phase of the experimental results be- 0.01 0.1 1 50 100
and(0—7^-^rf; fis the frequency

tween 0.1 and 2 Hz. The model transfer «(U


co
function between electrode extension c[T„)
and electrode feed speed was within
0.04 s_1 in magnitude and 8 deg in phase
of the experimental results.
l.Op
References

1. American Welding Society, 1 9 9 1 . 0.1


Welding Handbook, Vol. 2, pp. 109-113, -
Miami, Fla. S
2. Halmoy, E. May 1979. Wire melting
rate, droplet temperature, and effective anode 0.01
melting potential. Int. Conf. Arc Physics and
Weld Pool Behavior, The Welding Institute, 0
London, England.
3. Halmoy, E. 1986. Electrode wire heat- -45
ing in terms of w e l d i n g parameters. The Fig. 8 — Model
Physics of Welding, ed. J. F. Lancaster, pp. -90 a Experiment transfer function t/P
330-336, Oxford, Pergamon Press. Si Model Prediction calculated using the
4. Halmoy, E„ and Broten, H. 1979. Dy- -135
values in Table 7 for
namic response of the wire melting rate. Pub- I I Mini _l I _J I vz = 0.12 m/s and
lic Session Bratislava, pp. 5 4 - 6 1 , Interna- -180
0.01 0.1 1.0 10.0 the mean I = 250 A
tional Institute of Welding. and the experimental
Frequency (Hz)
5. Waszink, J. H., and van den Heuvel, G. results.
J. 1979. Measurements and calculations of

WELDING RESEARCH SUPPLEMENT I 245-s


o Amplitude = 7.5% of the itean 0 Amplitude = 7,5% of the mean
" Amplitude = 10.0% of the
x Amplitude = 10.0% 0 c rhe mean
la) 0.20 . . 2 •

5 6
0.16 . if.

0.12
.:.
0.08 0.08

0.04 s

«
n * 1 «
- O 1 n r

O) 2.4

2.0 -
„ l.(
9
1"
0.4 ®

150 250 35
K

Fig. 9 — Model transfer function calculated Fig. 10 — The coefficients in the denominator Fig. 11 — The coefficients in the denominator
using the values in Table I for a mean vz = of parametric fits to the model transfer func- of parametric fits to the model transfer func-
0.12 m/s and I = 250 A and the experimental tion tion
results.
, _t A2K(T„)(Tm-Tc) s _ l a(Tm)
as a function of
, as a lunction W vz
ir(Tm)
the resistance of wire extension in arc weld- JQvz
ing. Int. Conf. Arc Physics and Weld Pool Be- w = o\T„,
-
havior, pp. 227-239, The Welding Institute,
<*(ry
London, England.
dustry, Vol. 109, pp. 1 72-1 76. ture Solid Materials, 1, pp. 7 - 5 8 1 , MacMillan
6. Kim, Y.-S., McEligot, D. M., and Eagar,
9. Leigh, J. R. 1983. Essentials of Nonlin- Co., New York, N.Y.
T. W. 1991. Analyses of electrode heat trans-
ear Control Theory, Topics in Control, pp. 13. Touloukian, Y. S., Powell, R. W., Ho,
fer in gas metal arc welding. Welding Journal,
10-22, IEEE. C. Y , and Nicolaou, M. C. 1973. Thermal dif-
70(1): 20-s to 31-s.
10. Crank, ). 1984, Free and Moving fusivity. Thermophysical Properties of High
7. W u , G.-D., and Richardson, R. W. 1 989.
Boundary Problems, Oxford University Press, Temperature Solid Materials, 10, pp. 2 - 8 2 ,
The dynamic response of self-regulation of the
Oxford, England. IFI/Plenum, N.Y.
welding arc. Recent Trends in Welding Sci-
11. Kim, Y.-S. 1982. Metal Transfer in Gas 14. Touloukian, Y. S., and Buyco, E. H.
ence and Technology, pp. 9 2 9 - 9 3 1 , ASM In-
Metal Arc Welding, Ph.D. Thesis, Massachu- 1970. Specific Heat. Thermophysical proper-
ternational, Materials Park, Ohio.
setts Institute of Technology. ties of high temperature solid materials, 4, pp.
8. Kannatey-Asibu, E., Jr. 1987. Analysis of
12. Touloukian, Y. S. 1967 The elements. 1-102, IFI/Plenum, N.Y.
the G M A W process for microprocessor con-
Thermophysical Properties of High Tempera- 15. Touloukian, Y. S„ Powell, R. W., Ho,
trol of arc length. Journal of Engineering for In-
C. Y., and Klemens, P. G. 1970. Thermal con-
ductivity. Thermophysical Properties of High
Temperature Solid Materials, 1, pp. 9-1 09,
0 Amplitude = 7.5% of the mean 0 Amplitude = 7.5% of the rrean IFI/Plenum, N.Y.
* Amplitude = 10.0% of the mean * Amplitude = 10.0% of the
< al 0.0Q
• ® * '"
,a
> 0.00
e u 1 6. Lin, P. L., and Wu, Y. C. 1982. Identifi-
s
« e cation of multi-input multi-output linear sys-
-0.05 -0.05 tems from frequency response data. Journal of
Dynamic Systems, Measurement, and Con-
-^ -0.10 - •° -0.10
trol, Vol. 104, pp. 5 8 - 6 4 .
17. Heald, P. R., Madigan, R. B., Siewert,
-0.15
-0.15
-„ a T. A., and Liu, S. 1991. Droplet transfer modes
1
for a MIL 100S-1 G M A W electrode. NISTIR
3976, Gaithersburg, M d . , NIST.
1 8. Press, W. H., Flannery, B. P., Teukolsky,
0.0 8 X o o O 0.0 0 0 S. A., and Vetterling, W. T. 1988. Numerical
a
• •- ° Recipes in C, pp. 6 0 1 - 6 2 1 , Cambridge Uni-
"2 "0.5 2 -0-5
8 0 versity Press, Cambridge, England.
" 19. Forsythe, G. E., and Wasow, W. R.
* -1.0
- • -»•«

1960. Finite-Difference Methods for Partial


-1.5 - -1.5 " Differential Equations, pp. 6 8 - 7 1 , John Wiley
and Sons, New York, N.Y.
5j 15C 250 350 450 0 15C 250 350

Nomenclature

Fig. 12 — The coefficients in the numerator of Fig. 13 — The coefficients in the numerator of a A m p l i t u d e of
parametric fits to the model transfer function parametric fits to the model transfer function response in s
a1,a2,b0,bi Coefficients in fit of
s _ t A2K(Tm)(Tm-Tc) 5 a[Tm _
, as a function •, as a function of transfer f u n c t i o n s
1) I' ~W~ v2 '
Wm) a„;, a w A m p l i t u d e of sinusoidal e x c i -
ofW = w=-^y t a t i o n in qj or W .
or I C a[Tm

246-s I OCTOBER 1994


A Cross-sectional area
of the e l e c t r o d e u =
c(T) Specific heat T-Tr a Um
1
Dimensionless temperature Dimensionless
C(a) = v = excitation frequency

i/a(u[rm-rc]+rc) u
Reciprocal of JK(C[Tm-Tc] + Tc)d^ Appendix
thermal diffusivity
Transformed A s s u m i n g t h a t t h e d r o p has a u n i f o r m
Electron charge temperature temperature distribution, and neglecting
v
m V at u = 1 (melting) other losses (radiation), t h e rate of heat
W i r e feed speed transfer to the d r o p c a n be w r i t t e n (Ref.
Va A n o d e fall 11)
y +1 Apparent condensation
V, -Qm+(l-y)VcJ +
2 voltage
K{rm)r{Tm){Tm-T( V o l u m e of the d r o p
jrdJ2DdV-^(cpVD(TD-Tm)) =0
M u l t i p l y i n g factor
v
used for c o n v e n i e n c e " (A1)
Kt) W e l d i n g current W(t) = where
JD C u r r e n t density
in t h e d r o p 3 kT„
vc = <t>+vA +
k a{Tm) 2 e
B o l t z m a n n ' s constant Peclet N u m b e r (A2)
K(T) Thermal conductivity x = Z/£0 Dimensionless If the rate of d r o p l e t f o r m a t i o n is as-
Electrode extension coordinate s u m e d to be steady-state a n d the average
-Fig. 1 Z Spatial c o o r d i n a t e d r o p l e t g e o m e t r y is u s e d , K i m s h o w e d
Latent heat -Fig. 1 (Ref. 1 1 , p. 155) that

Subscripts A Vj g, tD
^•(o=
l\t)tlr{Tm 0 Refers to the v a r i a b l e
Vc{i-r)i
A2K(l att = 0 2
Dimensionless, volu- (A3)
ss Refers to the v a r i a b l e
m e t r i c , heat g e n e r a t i o n For a d r o p l e t radius o n the order of the
at t h e steady-state s o l u t i o n
Rate of heat transfer e l e c t r o d e d i a m e t e r (spray transfer), t h e
QA(D
to s o l i d e l e c t r o d e Greek Joule h e a t i n g is several orders of m a g n i -
from condensing t u d e less t h a n the c o n t r i b u t i o n f r o m e l e c -
e l e c t r o n s (z = £) a(T) Thermal diffusivity = pc t r o n c o n d e n s a t i o n . Q m is f i n a l l y a p p r o x -
Rate of heat transfer Fraction of electrons i m a t e d by
QmO") T
to s o l i d e l e c t r o d e that c o n d e n s e o n the
f r o m t h e m e l t (z = i) solid e l e c t r o d e (0 < y < 1 )
r(T) Resistivity of Qr 2
electrode (A4)
N o t e that the m o d e l w i l l be most ef-
fective for spray transfer. Equation 3 c a n
r Resistivity of t h e S e c o n d Kutateladze n o w be w r i t t e n as
D
l i q u i d in the d r o p number

-K(T)^ + KL(t)[ 111


i"[y-rc]+Tc) 5: s(t)
2
dl
Dimensionless pm
Transformed c o o r d i n a t e
D e n s i t y of e l e c t r o d e
pirn m dt ) (A5)
resistivity Equation 1 c a n be m a d e d i m e n s i o n -
p' D e n s i t y of t h e l i q u i d
s(t) t(t)Ao Dimensionless less in t h e f o r m o f
electrode extension in the d r o p
t Time
d2v(u)
+ q, T r u
Period of d r o p l e t d e - dx2 ""
tachment
Temperature of e l e c t r o d e ; s u b -
4 Fourier n u m b e r 1 ; l
' dx dr
scripts c a n d m refer to (A6)
T at z = 0 ( c o n t a c t tip) a n d a Laplace v a r i a b l e w i t h dimensionless boundary conditions
z = £ ( m e l t i n g ) . Subscripts e 0 Work function of
a n d D refer to the t e m p e r a t u r e electrode
of the c o n d e n s i n g electrons a n d Phase of the response
the t e m p e r a t u r e of the d r o p . in s

W F I D I N C R F S F A R C H SI JPPI F M F M T I 9 d 7 - c
T h e steady-state Equations A 9 to A11 c u r r e n t I, the e l e c t r o d e feed speed v z , a n d
v{x=0,t) = 0, are treated like an eigen v a l u e p r o b l e m qj(W).
(Ref. 18) w i t h Equation A 9 r e w r i t t e n as
v x = slT , r =v„ T h e transient p r o b l e m , Equations A 6
t w o first o r d e r e q u a t i o n s by letting
to A 8 , is s o l v e d by using t h e c o o r d i n a t e
transformation
dx +N(*)) dv
vr dx
X\ W ( T ) y2=v ^=W)
(A7) Yi = Oj (A16)
(A12)
a n d d i m e n s i o n l e s s initial c o n d i t i o n s to fix the moving boundary according to
w h i c h gives
(Ref. 10). The resulting nonlinear partial
S ( T = 0 ) = 1,
dy, differential equation is solved by using
y3r*(u) + C(u)WH
V(X,T = o)= vnU(x) dx central differences in t, for unequally
1
' (A8)
dy2 spaced nodes (Ref. 19). The nodes are
T h e g o v e r n i n g e q u a t i o n for steady- dx spaced in a geometric progression with
state c a n be by f o u n d by setting the t i m e
dy3 more nodes near the % = 1 boundary. A
derivatives in Equations A 6 to A 8 to z e r o =0
dx backward differentiation method for stiff
a n d c h a n g i n g t h e partial derivatives w i t h (A13)
equations is used to solve the resulting set
respect to x to o r d i n a r y d e r i v a t i v e s : The boundary conditions for Equation
of ordinary differential equations.
A13 are
To calculate v, K(T) is first smoothed
-±± + qjr*{u)- y,(0) = 0 from room temperature to the melting
dx point using a moving average to avoid
dv(u) Yl^)=vm step changes in K(T) when the material
C(u)W
dx 1 changes from one solid phase to another;
(A9) -y^=-{Fy3)2+XW ( A U ) a piecewise cubic spline is then fit to the
H e r e t h e b o u n d a r y is f i x e d at s 1, giv-
A r e l a x a t i o n m e t h o d is used to solve for tabulated K(T) and integrated. Tabulated
ing the b o u n d a r y c o n d i t i o n s
y-i, y 2 , a n d y 3 (Ref. 18). T h e steady-state values of oc(T) and r(T) are then written as
= 0, m o d e l predicts q. g i v e n W ,
functions of v, smoothed and fitted with
:1) =
piecewise cubic splines.
(A10)
T h e p r o b l e m is subject to the a u x i l i a r y Q7 = q,(ty) = q , ( ^
c o n d i t i o n t h a t e n o u g h e n e r g y m u s t be
s u p p l i e d at x = 1 that t h e e l e c t r o d e u n - T h e steady-state EE, is g i v e n f r o m
dergoes a phase c h a n g e :

dv_ A2K(Tm)(Tm-Tr
' dx +H |2 =XW
(AID
Equation A11 determines the qj (the , | ^ l
welding current) for which the solutions (A15)
of Equations A9 and A10 are valid. given the material constants, the welding

248-s I JUNE 1994

Вам также может понравиться