4, 1997
1. INTRODUCTION
In order to provide high quality for specialized applications and to aid
in the development of the technology, many studies of electric arcs, especially
nontransferred arcs, have been made. Comparison between experimental
results and the models leads to refinement of the calculation by taking
increasingly sophisticated phenomena into account.
Many of these studies were only made on the plasma jet. We can men-
tion for example the measurements of Capetti and Pfender(1) and Dilawari
et al.(2) Other works by McKelliget et al.(3) and Chang and Ramshaw(4) have
treated numerical simulations of argon plasma jets flowing into cold air and
nitrogen respectively. For all these applications the behavior of the plasma
is essentially studied through turbulence like in the paper of Huang et al.(5)
concerning a two-fluid model of turbulence which can predict phenomena
that escape more conventional models, e.g., unmixing calculations. These
models of the free plasma jet require adjustable data for the temperature
1
Centre de Physique des Plasmas et de leurs Applications de Toulouse, E.S.A. No. 5002,
Universite Paul Sabatier, 118 route de Narbonne, 31062 Toulouse Cedex 4, France.
409
0272-4324/97/1200-0409$12.50/0 © 1997 Plenum Publishing Corporation
410 Bauchire, Gonzalez, and Gleizes
and the velocity profiles, such as upstream boundary conditions at the exit
of the torch.
In order to avoid having to make these assumptions some authors
included the arc region in their modeling. Westhoff and Szekely,(6) for
example, computed a nontransferred arc plasma torch and the resultant
plume in laminar flow, assuming that the cathode tip was flat rather than
pointed. More recently, Murphy and Kovitya(7) published a model consider-
ing gas mixing and the effects of turbulence by a k — e model, omitting from
the energy equation the term used to account for energy transfer due to
electron flow. Paik et al.(8) studied the determination of the arc-root position
in a DC plasma torch in laminar flow, but this model was not coupled with
modeling of the free plasma jet.
The present paper has two main objectives:
First, to calculate the influence of turbulence models on the theoretical
predictions. We present a study consisting of the application of three turbu-
lence models using two Prandtl's mixing length models and also a standard
k— e model. The work shows the turbulence effects and gives a better idea
of the utility of each model. It brings out the difficulty of using a turbulence
model, without adjustment, on experimental measurements.
Second, to study the influence of physical parameters on the tempera-
ture and the velocity profiles in laminar or turbulent flow. In particular, a
comparison is made between the properties of the plasma with argon and
air as plasma gases. These two points were pursued studying a simplified
geometry corresponding to our experimental configuration so as to compare,
in future works, predicted and measured values.
In order to validate our model and to show its level of predictivity, we
will present, in the last part, a comparison between our calculated results
and the experimental data previously obtained by Westhoff and Szekely.(6)
For this comparison numerous changes were made to our model in order
to exactly respect both the experimental geometry given by the authors and
their flow conditions.
2. MODELING
2.1. Assumptions and Governing Equations
The model adopted is based on the following assumptions:
The governing equations are written in an axisymmetric system of coor-
dinates and the operation of the torch is assumed to be in a steady state
with negligible gravity effects.
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 411
Axial momentum u ue
Radial momentum V ue
Energy h
field, and eN is the net emission coefficient of the plasma. The self-induced
magnetic field BT may be written as
AB v =0
BC
CD u=0 T= 500 K
DE u=0 v =0 T=1000K k =0 e =0
EF u=0 v=0 T=1000K k = kw £=£w
FG u = u(r) v =0 T= 300 K — —
GA u=0 v=0 T= 3000 K k=0 e =0
of the arc or the voltage between the electrodes. On line AA', the current
density is given by an exponential function after Hsu et al.(10) The computa-
tion domain used for potential resolution is AA'EE'. Knowing the current
density distribution on line AA' and the zero potential (on line EE'), we
can use an iterative scheme to obtain the potential field. The total voltage
calculated does not take into account the potential fall in front of the
cathode. This latter value, corresponding to the distance between line AA'
and the cathode tip, has to be added to calculate the total input power. This
point will be discussed in Section 5. For the inlet velocity profile u(r) three
initial profiles were used on line FG: a parabolic, a constant, and a Couette's
profile. Our calculations showed that the difference between the three profiles
diminishes as the power of the arc becomes greater. In all cases the maximum
difference for the velocity in the free plasma jet is about 5% over the range
50-400 A for the current intensity and 0.1-1.2 g/s for the inlet mass flow
rate. In the following results u(r) was chosen as a parabolic profile.
The operating and the ambient gases were both argon or both air.
Thermodynamic and transport properties for the argon plasma are taken
from Mostaghimi(11) and Bacri and Raffanel(12) for air. The values of the
net emission coefficient assume a homogeneous plasma column radius of
5 mm and are taken from the works of Gleizes et al.(13) The net emission
coefficient for air is assumed to be equal to that of pure nitrogen.
A preliminary study was made of the influence of the radial and axial
dimensions over the computational domain. For this comparison mesh grids
35 x 85 (5 cm) and 35 x 167 (10 cm) were used as the axial distance and grids
35 x 85 (1.5 cm) and 42 x 85 (3 cm) for the effect of the radius. The good
agreement between the values obtained with the different grid sizes led us
to use the 35 x 85 grid.
414 Bauchire, Gonzalez, and Gleizcs
where R is the pipe radius, y is the distance from the pipe wall as y =
R-r + rg, rg is the wall roughness equal to 0.01 mm; yc is obtained from
the continuity of lm and quantities with an asterisk denote values taken at
the wall. In the plume region, two cases are considered:
—Case I, the mixing length and the Prandtl's number expressions are
taken from the work of Gonzalez,(15) i.e., Eq. (6) for the mixing length and
Eq. (7) if r<R and Pr T =0.95 if r>R.
—Case II, a standard expression of the mixing length for a free jet is
used(16):
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 415
r1 and r2 are two radial positions chosen such that the velocity is equal to
0.15 and 0.85 times the centerline velocity respectively. C is a constant that
can be determined by experimental studies. Here C is made equal to 0.8 to
be in agreement with case I.
The turbulent kinetic energy and the dissipation rate are calculated via the
transport equations shown in Table I with
The values of the constants are given in Table III with modifications for
free jets as suggested by Rodi(20):
uw is the axial velocity at the first grid points away from the wall and the
values of kw and £w are deduced from the following expressions:
Fig. 2. Comparison of laminar and turbulence models in the plume: temperature profiles on
centerline, D0= 1.0 g/s and I=200 A.
The three turbulence models give very similar results for temperature
and axial velocity. We note that velocities are different in the downstream
part of the nozzle not far from the tip, although the flow is supposed to be
laminar in this zone. This is due to the elliptical character of the model.
Even if the flow has a main direction, local properties can be modified by
the downstream values. We can also note that the k — e model and case II
of the mixing-length model lead, in comparison to case I, to a faster decrease
of velocity and temperature in the last part of the jet (for x>3 cm). This
comes from the fact that for case I, the mixing length lm is not proportional
to the width of the mixing boundary layer.
The calculation times for the k- e model are longer than those for the
mixing length models: for the 35 x 85 mesh grid and 3500 iterations, the
calculation takes 1231 CPU seconds for a laminar flow, 1265 s for case I,
1266 s for case II, and 1565 s for k— s on a HP 750. So, these three models
are quite equivalent under these operating conditions although the k- E
model requires the solution of two additional conservation equations but
gives more information on turbulence in comparison with the mixing length
models. This detailed information may be necessary for other purposes. For
example, values of the local turbulence intensity or the kinetic energy are
418 Bauchire, Gonzalez, and Gleizes
Fig. 3. Comparison of laminar and turbulence models in the plume: axial velocity profiles on
centerline, D0= 1-0 g/s and I=200 A.
For the nozzle region, we used the mixing length model described in
Section 2.3.1 for case II in the plume region. We also made the calculation
with the k - s model over the whole computation domain with a wall func-
tion (cf. Section 2.3.2) to take into account the influence of the wall.
The temperature and axial velocity variations on the centerline are plot-
ted in Figs. 4 and 5. The boundary conditions on k and e are fixed on line
FG as kFG(r) = 0.003u2G(r) and eFG(r) = C u k 3/2 /0.03R.
We note a critical difference between the velocity profiles (Fig. 5), which
can be explained in two ways. First, the Prandtl's mixing length model was
proposed by Nicolet et al. for a high-enthalpy constricted arc heater with a
Reynolds number of 2 • 106. Turbulence is surely not as great in our case
(Re = 3000 at the nozzle exit), so the use of this model may strongly over-
estimate the turbulence intensity and thus underestimate the velocity.
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 419
Fig. 4. Influence of turbulence models in the torch and in the jet: temperature profiles on
centerline, D0= 1.0 g/s and I=200 A.
4. PARAMETRIC STUDY
The influence of the inlet mass flow rate and of the arc intensity on the
temperature and velocity profiles in the nozzle and in the free plasma jet
were studied considering laminar and turbulent (k— e model with the wall
function) flows.
Fig. 5. Influence of turbulence models in the torch and in the jet: axial velocity profiles on
centerline, D0= 1.0 g/s and I=200 A.
D0 = 1.0 g/s. The notation 2 -> 18 kK means that the minimum isotherm is
2000 K, the maximum isotherm is 18,000 K, and the interval is 2000 K. This
figure shows that an increase of the inlet mass flow rate leads to a constriction
of the plasma column and to an increase of the axial temperature in the
nozzle and in the upstream part of the jet. The properties of the plume are
determined by those of the arc, in particular by those at the exit of the torch.
Therefore, when D0 rises, energy losses increase (because of convection and
turbulence) and hence, the arc voltage rises: for I— 100 A and a turbulent
flow (case I), U= 16.78V when D0 = 0.5 g/s and U= 18.06V when D0 =
1.0 g/s. For a given section, the mean electric field increases with the losses.
So, the conductance G decreases leading to the contraction of the plasma
column and to the rise of the axis temperature. The downstream part of the
jet is under the influence of mixing with cold ambient gas. In this region
turbulence effects are strong and lead to a temperature decrease.
Fig. 6. Influence of the inlet mass flow rate on isotherms, /= 100 A and turbulent flow
(k— e model).
nozzle tip, a rise in the mass flow rate due to the cold gas drag. The quantita-
tive results are in good agreement with the works of Pateyron,(24) i.e., for
an axial distance of four times the diameter, the total mass flow rate D has
a value defined by the relationship:
where D0 is the inlet mass flow rate and k3 is a coefficient in the range 0.25-
0.45. We found the values k3 = 0.25 (laminar flow) and k3 = 0.33 (turbulent
422 Bauchire, Gonzalez, and Gleizes
Fig. 7. Mass flow rate vs. axial position for laminar and turbulent flow (k — e model).
flow) for a current intensity of 200 A. We can see that pumping of the
ambient gas rises with the turbulence and with the current intensity. Under
the effects of turbulence, the exchanges between the plasma jet and the
ambient gas rise, leading to a greater pumping of surrounding gas. The drag
also depends on the value of the exit velocity of the jet which rises with the
current intensity. So, the mass flow rate increases with the current intensity.
This behavior of the plasma with the current intensity was noted in experi-
mental works by Vardelle.(25)
In this paper we do not present the influence of geometric or mathemati-
cal parameters, but a preliminary study has shown that taking the position
of the open boundary CD (see Fig. 1) at R1 = 3.2 or R2 = 6.38 times the
nozzle diameter away from the axis gives rise to the same results. So, the
results are presented using R1 = 3.2. Other works on the modeling of plasma
jets(15,26) have shown that the boundary BC (see Fig. 1) should be located
at R1 equal to six times the nozzle diameter. In our case, for values of R1
of about 6 or 3 times the diameter, there are no significant differences in the
temperature profile. The main difference is encountered in the velocity pro-
files, with a difference of about 10% for the value on the centerline where
the mass density is very low.
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 423
Fig. 8. Influence of the length of the arc on the temperature field, D0 = 0.5 g/s and I=200 A.
exit of the torch) and for an arc length of l2 = 2l1 (zero potential at the exit
of the torch), the operating parameters being D0 = 0.5 g/s and I= 200 A with
a laminar flow. Although the current intensity is the same, the lengthening
of the arc leads to an increase of the arc voltage and hence of the power:
Pl1 = 2.5 kW and Pl2 = 3.6 kW. This results in the important difference
between the temperature fields seen in Fig. 8.
On the other hand, Fig. 9 shows the isotherms of the plasma for current
intensities of 100 and 200 A with arc lengths of respectively l2 and l1, consid-
ering the approximate relationship(8)
where xs is the arc root position and A a constant depending on the upstream
mass density and axial velocity. The differences between the temperature
fields in Fig. 9 are lower than those obtained in Fig. 8 because the powers
were P=1.6 kW for I=100 A (l 2 ) and P = 2.5 kW for I=200 A (l 1 ). Arc
restrike phenomena are complex and difficult to compute but we can deduce
from these results that plasma properties depend more on the power value
than on the current intensity.
Fig. 9. Influence of the product arc length x current intensity on isotherms, D0 = 0.5 g/s.
and argon respectively. It can also be noted that using air leads to a constric-
tion of the plasma jet and to an increase of the mass flow rate: k3 is equal
to 0.23 for argon and 0.25 for air because of the pumping of the surrounding
cold gas.
Figure 11 shows the temperature fields for the same current intensity
I= 100 A. Although powers are different, Pair = 2.3 kW and Pargon = 1.6 kW,
we find the same kind of results than those of Fig. 10 on the temperature
fields because of the strong difference between argon and air properties.
426 Bauchire, Gonzalez, and Gleizes
Fig. 10. Isotherms of argon and air plasmas, D0 = 0.5 g/s and P= 1.9 kW.
Fig. 11. Isotherms of argon and air plasmas, D0 = 0.5 g/s and I= 100 A.
their experimental results. Two kinds of changes have been made for this
comparison:
5.2. Comparison
For this given geometry, we compare the experimental axis temperature
at the exit of the torch with our model, for /= 250 A and an inlet mass flow
rate D0 = 0.29 g/s (0.59 scmh) of argon. Following the procedure described
at the end of Section 5.1, we obtain an arc length Larc= 12.25 mm, for which
the computed total torch power is approximately equal to the experimental
one (case No. B23). Figure 12 plots the temperature profile on the centerline
in the jet as obtained from Westhoff and Szekely's measurements and from
our calculations with the arc length Larc. The figure shows that the values
predicted by our model are higher than the experimental ones over a distance
of roughly 2 cm. This difference can be explained by our conditions at the
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 429
Fig. 12. Comparison between the measurements of Westhoff and Szekely (B23) and our calcula-
tion of the temperature profile on the centerline in the jet. x = 0 is the nozzle tip location.
fictitious anode which lead to a source term that is greater on the axis than
in Westhoff's conditions. In Fig. 13 we have plotted the radial temperature
profile at 37 mm from the nozzle tip obtained from Westhoff and Szekely's
measurements and from our calculations with the arc length Larc. The previ-
ous difference is seen to vanish after an axial position of the order of 2 cm
from the exit, and a good agreement occurs between the measurements and
our calculation. Near the exit of the nozzle, the overestimation of the axis
temperature leads to a slight overestimation of the axial velocity, as can be
seen in Fig. 14 which shows the axial velocity profiles at 1 mm from the
nozzle tip.
We tried to compare this temperature and the velocity profiles with
the mathematical formulation used as boundary conditions in the models
simulating only the free jet region.
More specifically, the radial temperature profile is often given by(2)
where Tmax is the maximum temperature, R0 the internal radius of the exit
430 Bauchire, Gonzalez, and Gleizes
Fig. 13. Comparison between the measurements of Westhoff and Szekely (B23) and our calcula-
tion of the radial temperature profile at 37 mm from the nozzle tip.
of the torch, and Tw the wall temperature. In our case, T max = 12,062 K,
T w =1000K, and R0 = 6.35 mm. There is a very good agreement with
2.1 <n<2.2 between Eq. (17) and the temperature profile obtained by the
model in laminar flow. The comparison is not so good for the velocity profile
using the following relationship:
with a maximum difference of about 30% (umax= 124 m/s). As can be seen
in Fig. 14, the radial velocity profile from Eq. (18) does not correspond to
our profiles or to the predictions of Westhoff et al. In conclusion, Eq. (17)
appears to be a good boundary condition approximation and is in agreement
with prediction models.
6. CONCLUSION
In this paper we present the study of a model of a DC plasma torch in
the nozzle and in the free jet regions, with turbulent and laminar flows.
Modeling of a DC Plasma Torch in Laminar and Turbulent Flow 431
Fig. 14. Axial velocity vs. radial position at 1 mm from the nozzle tip. Comparison between
the measurements of Westhoff and Szekely (B23) and our calculation.
We discuss the utility and the results obtained with three turbulence
models. The most important problems encountered in turbulence are deter-
mining the boundary conditions for the k — s model and adjusting the con-
stants for the Prandtl's mixing length model. Measurements are always
necessary for these calibrations. The study of these turbulence models does
not allow just one to be chosen but it leads to a better understanding of
their behavior and their advantages.
The study of physical and geometrical parameters shows that the
behavior of the plasma is coherent with the experimental and theoretical
results in the literature. The evolution of the mass flow rate is in agreement
with Pateyron's works(24) where turbulence effects or a current intensity
increase lead to greater pumping of the surrounding gas. We compared the
temperature fields for two gases, argon and air. A difference was noted in
the voltage drop and a decrease seen in temperature using air.
Using another geometry, we compared the model with experimental
measurements and theoretical results. Good agreement was encountered
with laminar flow. The temperature and velocity profiles were extracted at
the exit of the nozzle in order to compare them with the mathematical
formulation used as boundary conditions by the free jet model. The usual
432 Bauchire, Gonzalez, and Gleizes
temperature law presents a very good agreement with the model, but this is
not the case with the usual velocity law which presents a difference of about
30%.
This model will allow us to consider various perspectives including
mixing of dissimilar gases and the influence of particles injected in the jet.
We are in the process of adapting the model to other experimental conditions
now under study in our laboratory.
ACKNOWLEDGMENTS
The authors wish to thank the Conseil Regional Midi-Pyrenees for
partial support of this study under projet RECH/9300395.
REFERENCES
1. A. Capetti and E. Pfender, Plasma Chem. Plasma Process. 9, 329 (1989).
2. A. H. Dilawari, J. Szekely, J. Batdorf, R. Detering, and C. B. Shaw, Plasma Chem. Plasma
Process. 10, 321 (1990).
3. J. McKelliget, J. Szekely, M. Vardelle, and P. Fauchais, Plasma Chem. Plasma Process. 2,
31 (1982).
4. C. H. Chang and J. D. Ramshaw, Plasma Chem. Plasma Process. 13, 189 (1993).
5. P. C. Huang, J. Heberlein, and E. Pfender, Plasma Chem. Plasma Process. 15, 25 (1995).
6. R. Westhoff and J. Szekely, J. Appl. Phys. 70, 3455 (1991).
7. A. B. Murphy and P. Kovitya, J. Appl. Phys. 73, 4759 (1993).
8. S. Paik, P. C. Huang, J. Heberlein, and E. Pfender, Plasma Chem. Plasma Process. 13, 379
(1993).
9. S. V. Patankar, Numerical Heat Transfer and Fluid Flow, McGraw-Hill, New York (1980).
10. K. C. Hsu, K. Etemadi, and E. Pfender, J. Appl. Phys. 54, 3, 1293 (1983).
11. J. Mostaghimi, PhD Thesis, University of Minnesota (1982).
12. J. Bacri and S. Raffanel, Rep. 49277-86-1 and 40277-86-2, CPAT, Universite Paul Sabatier,
France (1986) and Plasma Chem. Plasma Process. 7, 59 (1987).
13. A. Gleizes, J. J. Gonzalez, B. Liani, and G. Raynal, J. Phys. D: Appl. Phys. 26, 1921
(1993).
14. W. E. Nicolet, C. E. Shepard, K. J. Clark, A. Balakrishnan, J. P. Kesselring, K. E.
Suchsland, and J. J. Reese, Acurex Corporation, AEDC.TR.75.47.
15. J. J. Gonzalez, PhD Thesis, Universite Paul Sabatier, France, No. 1100 (1992).
16. P. Proulx, PhD Thesis, Sherbrooke University, Canada (1987).
17. B. E. Launder and D. B. Spalding, Comput. Meth. Appl. Mech. Eng. 3, 269 (1974).
18. Y. P. Chyou and E. Pfender, Plasma Chem. Plasma Process. 9, 291 (1989).
19. D. A. Scott, P. Kovitya, and G. N. Haddad, J. Appl. Phys. 66, 5232 (1989).
20. W. Rodi, PhD Thesis, University of London (1972).
21. E. Pfender, J. Fincke, and R. Spores, Plasma Chem. Plasma Process. 11, 529 (1991).
22. W. M. Pun and D. B. Spalding, Rep. No. HTS/76/2, Imperial College, London (1976).
23. A. H. Dilawari and J. Szekely, Plasma Chem. Plasma Process. 7, 317 (1987).
24. B. Pateyron, PhD Thesis, No. 21-1987, Limoges University, France (1987).
25. M. Vardelle, PhD Thesis, No. 28-87, Limoges University, France, (1987).
26. J. J. Gonzalez, A. Gleizes, S. Vacquie, and P. Brunelot, Plasma Chem. Plasma Process.
13, 2, 237 (1993).
27. O. Betoule, PhD Thesis, No. 24-1994, Limoges University, France, (1994).