Professional Documents
Culture Documents
on Argon-Shielded
Gas Metal Arc Welding Processes
ABSTRACT. It has been observed experi- aluminum and stainless steel to include velopment of the first model representing
mentally that small additions of oxygen carbon steels (Ref. 1). Today, oxygen ad- the gas metal arc.
to the argon shielding gas affect the gen- ditions to the shielding gas are widely The main purpose of this article is to
eral operation of GMAW processes. By used in many applications of the GMAW present the results of our study of the ef-
theoretically modeling the arc column, it process. It is commonly believed that fects of oxygen addition on the charac-
is shown that the addition of 2 to 5% oxy- oxygen reduces the surface tension of the teristics of an argon arc for GMAW of iron
gen to argon has an insignificant effect on metal and improves the arc stability and and aluminum. We shall also relate these
the arc characteristics. This corresponds arc root behavior. So far, no information findings to previous studies of the impor-
to the minor changes in the thermophys- has been published in the open literature tance of oxygen in the anode and cath-
ical transport and thermodynamic prop- regarding the effects of oxygen additions ode region, in order to give a more com-
erties caused by the oxygen addition. on the characteristics of the arc column. plete description of the influence of
Therefore, it is concluded that the addi- The main reasons for this have been the oxygen additions on the performance of
tion of oxygen to the argon shielding gas difficulty in obtaining high-temperature the GMAW process.
mainly affects the anode and the cathode data for thermophysical properties for an One section of the article describes
regions. From the literature, it was found argon-oxygen mixture, and the lack of a the GMAW process, and another section
that the formation of oxides initiates arc- gas metal arc model. However, recent summarizes previous work on the influ-
ing at the cathode and decreases the developments at CSIRO have increased ence of oxygen addition on arc welding
movement of the cathode spots. These the reliability of calculations of the ther- processes. We then describe the method
oxides can also improve the wetting mophysical properties for a large number of calculation of the thermophysical
conditions at the workpiece and the elec- of gas mixtures. Also, some recent work properties and the calculation of the arc
trode. Finally, oxygen is found to affect in our laboratory at MIT has led to the de- characteristics. Finally, by examining our
the surface tension gradient and thereby calculated results in conjunction with
the convective flow of liquid metal in the previously reported work, we discuss the
weld pool. influence of oxygen additions on argon-
KEY WORDS shielded GMAW processes.
Introduction
Oxygen Additions Background
In the early fifties, it was found that the Argon Shielding Gas
addition of small amounts of oxygen to GMAW Figure 1 shows the main components
the argon shielding gas in gas metal arc Arc Characteristics of the GMAW system, which is seen to
welding (GMAW) allowed the range of Gas Properties consist of the consumable electrode
weldable materials to be extended from Arc Column (anode), the anode-fall region, the arc
Theoretical Model column, the cathode-fall region and the
P. G. JONSSON is with MEFOS, Lule~, Swe- workpiece (cathode). The figure also in-
Computational Model
den. A. B. MURPHY is with CSIRO, Division cludes the gas-shielding nozzle, through
of Applied Physics, Lindfield, Australia. J.
Oxide Formation
which the shielding gas is supplied to the
SZEKELY is with Department of Materials Sci- Surface Tension
arc.
ence and Engineering, Massachusetts Institute
When an arc is struck between the
of Technology, Cambridge, Mass.
consumable
;"'-- electrode (anode)
anode f a l l
arc column
Q metal
droplet
~ f
anode and the cathode, a current flows point of iron is 3343 K (Ref. 5)). Based on gas with 2 to 4% oxygen and the com-
through the electric discharge between these findings, iron will be treated as a position of the welding wire, are critical
the electrodes. The arc current is spread nonthermionic electron emitter in this in achieving a high fracture toughness.
laterally and a jet is formed which gives study. According to the authors, the fracture
rise to a flow in the direction of the cath- toughness is determined by the amount
ode (workpiece). The gas impinges on Previous Work of acicular ferrite formed, which is de-
the workpiece and is spread in a direc- pendent on the number and size of oxide
tion parallel to the workpiece. Also, the Initially, some general observations of inclusions.
current distribution at the anode gives the influence of oxygen on the GMAW
rise to a high heat generation in the near process will be presented. Thereafter, the Effects Due to Oxide Formation
anode area, which results in a rapid melt- review of the importance of oxygen in
ing of the consumable electrode. GMAW operations will be separated into It is commonly believed that non-
Droplets are formed at the melted tip, de- two main parts: 1) the effects due to oxide thermionic cathodes operate by the con-
tach from the electrode, and are trans- formation; and 2) the effects due to the in- tinuous formation and decay of several
ferred to the workpiece under the influ- fluence on surface tension gradients of small electron-emitting sites (cathode
ence of arc forces. The melted droplets liquid metals. In doing this, we choose to spots, Refs. 3, 11). The site lifetime for
solidify at the workpiece and form the include results from other arc welding thick oxide films on mild steel has been
welded joint. processes for the sake of completeness. shown to be as low as 1 ns (Ref. 12). The
Of the two workpiece materials dis- In general, experimental observations oxides, formed by reactions between
cussed in this article, aluminum is known have shown that oxygen additions to a oxygen and workpiece elements, are be-
to be a nonthermionic cathode charac- shielding gas have a major impact on the lieved to act as a source of electrons, be-
terized by a mobile cathode spot (Ref. 2). weld pool, leading to an increased length, cause they usually have a lower work
It is not clear, on the other hand, whether width, volume and section (Ref. 6). How- function than the metal (Ref. 13). The ox-
thermionic or nonthermionic emission of ever, use of an oxidizing gas can cause ides are also thought to be charged by in-
electrons occurs from iron cathodes (Ref. some ferrous metals to become porous, cident positive ions and to emit electrons
3). However, the following observations and therefore reducing elements are usu- by a tunneling mechanism and a switch-
indicate that iron shows more of a non- ally added to the electrode to compensate ing mechanism for oxide thicknesses
thermionic than a thermionic cathode for this effect (Ref. 7). Also, it has been smaller than 5 nm and larger than 10 nm,
behavior. First, Essersand Walter (Ref. 4) found that addition of too high a propor- respectively (Ref. 14). For welding arcs
found that the calculated value of the tion of oxygen, 7 to 8% or more, to the acting on an oxidized steel workpiece,
current density, as a result of thermionic argon shielding gas increases slag forma- each emitting site has been found to carry
emission, is very low. Second, the mobile tion and leads to firmer attachment of the up to 1 A of current (Ref. 15). The emit-
cathode spot behavior usually observed slag to the surface (Ref. 8). ting sites are also said to move outward
for nonthermionic cathodes is also ob- It has also been found that an im- toward fresh oxide and to consume it at
served in GMAW of steel (Ref. 3). Third, provement in the toughness of ferritic a certain rate (Ref. 16). The area over
calculated values of the current density, steels may be obtained by increasing the which the oxide is removed increases
as a result of thermionic emission, have oxidizing potential through addition of with an increase in the change in the
shown good agreement with experimen- up to 2% oxygen to the argon shielding cathode voltage drop (Ref. 17).
tal data on materials with boiling points gas (Ref. 9). Moreover, Francis, etal. (Ref. The mobility of nonthermionic cath-
of at least 4000 K (Ref. 3) (the boiling 10), found that the use of an argon-based ode spots is undesirable in GMAW, be-
~y/~T < 0
z,w 1 inflow
(a)
-F
anode
B C
plasma
column
outflow
8y~T > 0
I G
H
(b) cathode
Fig. 3 - - Convection due to thermocapillary forces at: A - - a Fig. 4 - - Outline of integration domain.
negative surface tension gradient (pure metal); B - - a positive
surface tension gradient (metal containing oxygen).
Ar
1.5 98%Ar-2%0z -
107
95%Ar-5%Oz
I
E 1.0
10 g
v
0.5
105 Ar r-
98%Ar-2%0z
95%Ar-5%02 ~ [ . . . I ...I .... I ....
10 4
,,,[ .... I .... I .... I .... I .... 0.0
5000 10000 15000 20000 25000 30000 5000 10000 15000 20000 25~0 O~O
Fig. 5 - - The enthalpy as a function temperature. Fig. 6 - - The density as a function o f temperature.
demonstrated that the base metal com- mixtures (Refs. 45, 46), which has been where x, is the mole fraction and Mj is the
position is indeed critical in controlling modified by Kovitya (Refs. 47-49) to molecular weight of the jth species. The
the weld bead shape. allow for treatment of ionized gases. enthalpy and specific heat of the mixture
All calculations are performed assum- are calculated using the expressions
Theoretical Modeling ing local thermodynamic equilibrium. given by Kovitya (Ref. 49), which in-
The initial step in the determination of the cludes terms for the Debye-H0ckel cor-
Calculation of Thermophysical thermophysical properties is the calcula- rection.
Gas Properties tion of the equilibrium composition of the The transport coefficients; viscosity,
gas or gas mixture, which is done using thermal conductivity, and electrical con-
The model of the arc column that we the principle of minimization of the Gibbs ductivity; are calculated using the Chap-
use requires values of the density, en- free energy. The species considered are man-Enskog method (Refs. 52-54). De-
thalpy, specific heat, viscosity, thermal Ar, Ar +, Ar ++, Ar +++, 02, 02 +, O-, 0% tails of the calculations are given by
conductivity, electrical conductivity, and O+% 0++% and the electron. The pres- Murphy (Ref. 55). The thermal conduc-
radiative emission coefficients as a func- ence of metallic species is neglected, tivity is calculated as the sum of the con-
tion of temperature for pure argon and since the focus is on studying the effect of tributions due to the translational, inter-
mixtures of argon and oxygen. The radi- oxygen additions on the argon plasma. nal, and reaction components. The
ation loss terms for both argon and the The required partition functions for the translational component is further bro-
argon-oxygen mixtures are taken from considered species are calculated from ken down into contributions due to
experimental data for argon of Evans and the data tabulated by Moore (Ref. 50) for heavy particle and electron motion; the
Tankin (Ref. 43). The radiation loss terms neutral and positively charged electron component is calculated to a
for the argon-oxygen mixtures are as- monoatomic species, and from data given third level of approximation, while the
sumed to be the same as for argon, be- in the JANAF tables (Ref. 51) for other heavy particle component is calculated
cause the total radiation of oxygen is of species. The Debye-HCickel correction is to a second level. Note that the first level
the same order of magnitude as for argon applied, as described by Kovitya (Ref. 49). of approximation for the thermal con-
(Ref. 44) and only 2 to 5% oxygen is The density of the mixture at temper- ductivity is identically zero. The electri-
added. The other quantities are calcu- ature T and pressure P is calculated using cal conductivity is calculated to a third
lated using a code designed to calculate level of approximation, neglecting the in-
the equilibrium compositions and ther- P~.~jxjMi fluence of ion transport. The viscosity is
p- - -
mophysical properties of nonionized gas RT (2) calculated to a first level of approxima-
12500
15000 .... l .... I .... I .... I .... I'''
'''I .... r .... I .... I .... I .... -
-- -- Ar
12500 - - 98%Ar-2%Oz
i
E 10000 -- 98%Ar-2%Oz
95%Ar-5%02 O3
I
10000
>~ 7500
7500
,~ 5O00
5000 0
o
2500
2500
5 o .... I~',, , I . . . . I . . . . I . . . . I . . . .
¢ 0 -r~,,, ' r ~ " ~ , , , I . . . . I .... I .... I,,, -~ ~ 0 5000 10000 15000 20000 25000 30000
5000 I0000 15000 20000 25000 30000
O3
Temperature (K) Temperature (K)
Fig. 7 - - The specific heat as a function o f temperature. Fig. 8 - - The electrical conductivity as a function o f temperature.
I
i-
I I _ _ _ 98%Ar-2%Oz
E 0.20 ~ ~ o -- • -
v
v
F.
c)
0
f/]
0
c~
0 0000 10000 15000 20000 25000 30000 0 5000 10000 15000 20000 25000 30000
E CJ
Fig. 9 - - The thermal conductivity as a function o f temperature. Fig. 10 - - The molecular viscosity as a function o f temperature.
tion. Devoto (Refs. 56, 57) has verified The collision integral tabulations of tion. However, the data here are ex-
that the levels of approximation used give Levin, e t a l . (Ref. 62), and Stallcop, e t a l . pected to be considerably more reliable
results accurate to within 1% for argon, (Ref. 63), are used directly for the O-O than less-recent values (Ref. 55). This,
except for the electrical conductivity at and O+-O interactions respectively. Fi- and the small concentrations of oxygen
very low levels of ionization. Since the nally, the collision integrals for electron- considered in the calculations, means
electrical conductivity is very low under neutral interactions are treated by nu- that our transport properties of argon-
these conditions, the absolute error is al- merical integration of the oxygen mixtures will not be significantly
ways small (less than 200 Sm-1). momentum-transfer cross-section data less accurate than our results for pure
The collision integrals, integrals over presented by Itikawa (Refs. 64, 65), sup- argon.
a Maxwellian distribution of the collision plemented at low and high energies by The calculated thermophysical prop-
cross-sections for interactions between further data from Itikawa's sources. erties for the shielding gases considered
the various species, are required to cal- Our values for the thermodynamic in this paper are shown, and discussed
culate the transport coefficients. Most of properties of argon and argon-oxygen later.
the integrals are calculated from the in- mixtures are expected to be accurate to
termolecular potentials given by Aubre- within a few percent. The main source of Description of Arc Model
ton, et al. (Ref. 58). However, in a num- uncertainty in the calculation of transport
ber of cases improved data are used. The properties is in the collision integrals. A two-dimensional steady-state math-
collision integrals for the Ar-Ar and Ar-O 2 These are generally well defined in the ematical model has been developed to
interactions are respectively calculated case of interactions between argon predict properties such as temperature,
from the HFDTCS2 potential given by species, leading to good agreement, velocity, and voltage in the gas metal arc.
Aziz and Slaman (Ref. 59) and the ESMSV within a few percent, between recent cal- In an earlier study, the model was applied
potential given by Pirani and Vecchio- culations by different authors of the trans- to GMAW of aluminum in an argon at-
cattivi (Ref. 60). The charge-exchange port properties of argon (Ref. 55). There mosphere (Ref. 66). Calculated values for
cross-section required in the calculation is, in contrast, considerable variation be- temperatures, at an axial location
of the collision integrals for the O-O in- tween published collision integral data halfway between the electrode and the
teraction is taken from the experimental for some interactions between oxygen workpiece, were found to differ by 0 to
results of Rutherford and Vroom (Ref. 61 ). species, particularly the O+-O interac- 6.1% and 0 to 3.8% in comparison with
I0 . ,
' i . . . . I . . . . I ' 6000
. . . . I . . . . I . . . . I '
Ar
98%Ar-2%0 z C -- - - - Ar
y
5000
8 95%Ar-5%02 98%Ar-2%0~
I I o - o
o2~ 4000
I
6 v
3000
0 2000
4
E
=,
r~
o~
tO00 ~-
J
, I .... I .... I,
100 200 300 400 E 0 --' .... I .... I .... I ,
0 t00 200 gO0 400
Current (A)
Current (A)
Fig. I I - - The m a x i m u m value o f the mass flow plotted as a func- Fig. 12 - - The m a x i m u m value o f the m o m e n t u m flux as a function o f the cur-
tion o f the current. rent.
18
v
///
400
so i--
16
Z bS~ _e
300
14
v77 // v
E
200
12'
D
o
o >
I . . . . I . . . . I ,
10 ' ' ' lO0
100 200 300 400 o 2 4 B 8 to
C u r r e n t (A) Axial d i s t a n c e ( m m )
spectroscopically measured tempera- detailed description of the mathematical and the self-induced magnetic field, re-
tures (at different radial locations, Ref. model can be found in an earlier publi- spectively.
67), for arc currents 150 and 250 A, re- cation (Ref. 66). The boundary conditions for the
spectively. For this investigation, the welding arc are listed in Table 1 and the
model is applied to both GMAW of alu- Mathematical Formulation variables used in Table 1 are defined in
minum and iron. the Appendix. In the anode region (BC,
The primary difference to the earlier The boundaries of the computational CD in Fig. 4, a no-slip condition is used
model application to GMAW of alu- domain for the welding arc are shown in for the momentum boundary conditions.
minum (Ref. 66) is that the arc column is Fig. 4. The system is assumed to be ax- The enthalpy at the anode ha is set to the
modeled under the assumption of local isymmetric, steady-state and laminar. value corresponding to the melting tem-
thermal equilibrium (LTE). In the earlier The flasma is assumed to be at atmos- perature of pure iron (1810 K) or pure
model application, non-LTE conditions pheric pressure, in local thermodynamic aluminum (933 K). The equation for the
in the cathode and anode fall regions equilibrium, optically thin to radiation conservation of charge continuity is the
were accounted for in a simplified way and the influence of metal droplets is ne- only equation to be solved within the
(through source terms in the plasma re- glected. Also, the consumable electrode electrode region, and therefore, the elec-
gion close to the anode and the cathode). is assumed to be cylindrical, and the tip tric potential is set to be constant in re-
However, in this investigation the focus of the electrode and the workpiece sur- gion DA - - Fig. 4. In the region DE, a
is on the majority of the plasma region faces are assumed to be flat. zero mass-flow gradient dpw/dz was
where LTE exists and not on the small In applying the above assumptions, postulated, together with an inlet en-
non-LTE boundary layer regions (< 0.1 the equations, written in cylindrical co- thalpy hi corresponding to a temperature
mm (Ref. 68)). Therefore, we neglect the ordinates, that need to be solved are the of 300 K. Since it is not clear where in-
simplified approach to account for non- conservation of mass, radial and axial flow (EF) and outflow (FG) will take place
LTE boundary regions that was used in momentum, thermal energy, and charge in the fringes of the arc column, zero ra-
the earlier study (Ref. 66). continuity (expressed in the form of the dial mass flow d(pu)/dr and electric po-
The paragraphs below summarize the electrical potential). The definitions of tential gradients were specified. The en-
key assumptions, equations, and bound- the electric potential and Ampere's law thalpy is assumed to correspond to a
ary conditions used in the model. A more are used to calculate the current densities temperature of 300 K for the gas en-
oo •" a?
15000 ddZ ~ •~ ej
e
10000 "
e~ . . . . Ar
. ------ Ar
- - 98%Ar-2%0~
5000 r-"
[- ~ ~ 95%Ar-5%02
- I.... I .... L .... I .... I,, ,
o ~. . . . . . t,,,,I .... I .... I .... [.~ -5 u ,
2 4 6 8 10
o 2 4 6 8 lO
Axial d i s t a n c e ( m m )
Axial distance (mm)
trained into the system and the enthalpy r>& the electrode diameter were 150 to 400
~=0
gradient dh/dr is assumed to be zero for (4) A, 10.0 and 1.2 mm, respectively.
mass flowing out of the system.
In the cathode region (GHI), a no-slip Method of Solution Thermophysical Gas Properties
condition is used for the momentum
equations. The enthalpies at the cathode The solution of the governing equa- As mentioned earlier, the thermody-
surface within he,i and outside he,o the tions and boundary conditions is ob- namic and transport properties of the
weld pool (cathode spot) region are taken tained by using a modified version of shielding gaseswere calculated using the
at temperature values of 1810 and 1000 2/E/FIX, a two-dimensional steady-state program developed at CSIRO (Ref. 47).
K for iron and 933 and 600 K for alu- code based on a finite volume scheme Figures 5 to 7 show the enthalpy, density
minum, respectively. The radius of the (Ref. 69). During a calculation, the dif- and specific heat as a function of tem-
cathode spot Re, is defined as an average ference equations are solved by iteration perature. It is obvious from the figures
value representing the movement of the until the residuals are less than 1% of the that the addition of oxygen has an in-
cathode spot. Theoretical calculations of magnitude of the respective variables. significant effect on the density, and only
the weld pool profiles showed that the The current balance is satisfied within small effect on the enthalpy and the spe-
weld pool radius in argon-shielded sys- 1% for all calculations. cific heat. The largest enthalpy differ-
tems is 3.2 to 3.5 mm (0.125 and 0.138 Theoretical inputs in the form of ther- ences occur in the temperature range of
in.) for welding currents of 150 to 220 A mophysical gas properties are calculated 2500 to 12000 K, for which oxygen ad-
(Refs. 27, 28). Based on these values of as described earlier. The values of the ditions tend to increase the value of the
the weld pool radius, a sensitivity calcu- electrical conductivity of iron and alu- enthalpy slightly. The most noticeable ef-
lation was done to study the effect of the minum at their respective melting points fect of oxygen additions on the specific
cathode spot radius on the calculated arc are taken from Touloukian (Ref. 70). heat is found at a temperature of 3000 K,
characteristics. It was found that the cal- where a peak value is found, corre-
culated maximum velocity varies less Results sponding to the dissociation of oxygen
than 1.7% and the maximum tempera- molecules.
ture varies less than 0.1% for a 2.7 to 4.2 Our calculations were done for The transport properties; electrical
mm (0.11-0.16 in.) range of the cathode GMAW of aluminum and iron using the conductivity, thermal conductivity and
spot radius. Therefore, Rc was chosen to following shielding gases: 1) pure argon; molecular viscosity; are plotted vs. tem-
be 2.7 mm. 2) 98% argon - 2% oxygen; and 3) 95% perature in Figs. 8-10, respectively. The
It is assumed that a single value of the argon - 5% oxygen by weight. The stud- electrical conductivity is found to in-
current density Jc is valid within the cath- ied range of currents, the arc length and crease slightly with increased oxygen
ode spot (weld pool) region and that the
current density is zero outside the cath-
ode spot region. This assumption is based
on the strong dependence of the current Table 1 - - Boundary Conditions for the Arc
density on surface temperature; the tem-
peratures in the weld pool region are sub- Fig. 4 u w h
stantially higher than in the rest of the BC, CD, AD 0 0 ha 0 (AD)
workpiece. Therefore, the current den- DE 0 apw hi a~
------0 --=0
sity conditions at the cathode are given az 8z
by: EG apu aw hi (inflow) a~
--=0 --=0 --=0
I #r 8r ah ar
- - = 0 (outflow)
Jc=-~c r <-Rc ar
(3) IH 0 0 hc.i Given by
Equations 3 and 4
HG 0 0 hc.o. 0
Sixth International Conference on Aluminum Weldments (INALCO '95), April 3-5, 1995, Cleveland,
Ohio. This conference will be conducted concurrentlywith the AWS Annual Convention and Exposition. Topicswill
focus on design,fabrication, inspectionand qualitycontrol of aluminum weldments, with special emphasis on design
of weldments, tools for design, application examples, non-welding joining processes, behavior improvement, cast
and extruded aluminum, inspectiontechniques and codes and standards.
1995 International Conference on Microbially Influenced Corrosion, May 8-10, 1995, New Orleans,
Louisiana. Three broad areas are envisioned in the scope of this conference: biological aspects, materials
including weldments, and corrosion. Topics include: theoretical, materials and welding considerations, case
studies; monitoring technologies; biocides and treatment; economics; and regulatory considerations
mechanisms. A mini-exposition will be held.
Aluminum Welding Seminar, October 5-6, 1995, Pittsburgh, Pennsylvania. Presentations include an
introduction to aluminum joining; selection of aluminum alloys and characteristics; gas tungsten arc welding of
aluminum; gas metal arc welding of aluminum; aluminum welding metallurgy; metal preparation for welding; weld
discontinuities--causes and cures; plasma cutting and gouging; performance of aluminum welds; welding
exposure studies; and laser cutting and welding. Leading manufacturers will exhibit.
Eleventh North American Welding Research Conference: Joining for the Automotive
Industry, October 17-18, 1995, Southfield, Michigan. Tentative sessions titles include: design for welding;
tailored blank manufacturing; alternative materials (plastics, adhesives, aluminum, magnesium); process
control/NDT; resistance welding; alternative processes-weld bonding, power beams; welding automation;
TQJ.
Watch for the following conferences to be scheduled for the fall of 1995:
• Welding of Stainless Steels and Nickel Alloys .Structural Steel Applications
For further information, contact: Conferences, American Welding Society, 550 N.W. LeJeune Road,
Miami, FL 33126, 800-443-9353, Ext. 278, or 305-443-9353, Ext. 278, Fax: 305-443-6445.