You are on page 1of 10

ELSEVIER

Materials Science and Engineering A208 (1996) 20 29

A numerical simulation of gas flow field effects on high pressure


gas atomization due to operating pressure variation
J. Mi a, R.S. Figliola a, I.E. Anderson b
~Department of Mechanical Engineering, Clemson University, Clemson, SC 29634, USA
bMetals and Ceramics Division, Ames Laboratory, Ames, 1,4 50011, USA

Received I1 May 1995: in revised form 17 August 1995

Abstract

High pressure gas atomization is emerging as an efficient method of producing high yields of ultrafine metal and alloy powders.
The results of a numerical study of the gas-only flow field of a high pressure, confined feed, annular jet gas-metal atomizer are
compared with experimental data and are reported in this paper. The axisymmetric, turbulent, compressible Navier-Stokes
equations are solved for the gas-only flow in the vicinity of the melt tip. A parametric variation of gas atomizing pressure and
its effect on gas flow field is examined. The numerical results demonstrate the existence of a strong recirculation region, mixing
layer, and shock structures downstream of the melt tip in good agreement with experimental observations and past studies. The
paper focuses on the characterization of the flow field, including a comparison of numerical and experimental data and how the
results support a droplet formation mechanism over a range of atomizing pressures.
Keywords: Gas atomization; Numerical simulation

1. Introduction

Fine metal and alloy powder is essential to the


effective development of a number of innovative powder
processing technologies, such as reactive sintering, explosive compaction, metal injection molding, thermal
spraying, rapid solidification, arc plasma coating and
spray deposition processing. Products manufactured
using fine powders (dso< 2 0 / l m ) can exhibit unique behavior and properties which are unachievable by conventional manufacturing techniques. Microstructural
refinement, chemical homogeneity, efficiency of near
final shape production and competitive handling costs
are the benefits of this technology [1-3]. High pressure
gas atomization (HPGA) is an effective method of
producing high yields of ultrafine metal and alloy powders. The powders produced by H P G A have much lower
mean particle diameters and much more uniformly
distributed size range than other commercially available
processes. Currently, many commercial metal and alloy

powders have been produced by HPGA, including high


alloy tool steels, nickel-based superalloys, aluminum
alloy etc. [4-8].
Although H P G A techniques have received increasing
attention, the fundamental momentum and solidification
processes associated with the technology remain far
from being completely understood. Moreover, extension
of the physics [9-11] used to describe the droplet
formation mechanism for low velocity gas atomization
to the high pressure regime is open to question. Previous
studies [12-15] suggest that the specific flow behavior of
both the gas and metallic melt in the H P G A gas-melt
region controls the momentum and energy exchange
necessary for droplet formation and that this flow
behavior can be modified by adjusting atomizer operating conditions. It has been also observed that a good
correlation exists between the gas-only flow and actual
atomization process [12,16]. As a result, a numerical
study of the high pressure regime gas-only flow under
appropriate operating conditions would be useful.
0921-5093/96./$15.00 Cc31996 -- Elsevier Science S.A. All rights reserved
S S D I 0921-5093(95)10046-6

J. Mi et al..' Materials Science and Engineering A208 (1996) 20-29

The objective of this paper is to report on a numerical study to characterize the gas flow behavior of the
HPGA process in an effort to understand the contribution of the high velocity compressible gas flow to the
molten metal droplet formation mechanism, and to give
insight into the correlation between atomization conditions, such as gas atomizing pressure, and the gas flow
patterns in the near-tip region. By understanding the
process dynamics and with a knowledge of this flow
behavior, the information may be useful for the intelligent control of HPGA through prediction of some of
the most important parameters of the atomization processes and real-time modification of the atomization
parameters.

2. Formulation of the problem


The geometry of the HPGA nozzle discussed below
and used for the parametric variation of operating
pressure is selected to match that used in the previous
experimental study [13], as shown in Fig. 1. The melt
feed tube tip design has a 45 taper angle, with a tip
extension length of 1.93 mm. The surface of the melt tip
frustum portion meets exactly the innermost edge of the
gas jet ring. The gas nozzle consists of an annular gas
jet which is fed from an annular gas manifold and
arranged in 10 mm diameter ring. The gas jet is aimed
at an angle of 22.5 , with respect to the axis of the
symmetry of the nozzle. The jet has a straight 0.813 mm
diameter exit path. A 4.76 mm diameter melt feed tube
is inserted into an 8.43 mm diameter melt tip base to

INOZZLE CENTERLINE

i"
i

1.93 mm

21

deliver a molten metal stream during actual gas atomization process, which is blocked in this study.
Based on the assumptions of steady state, axisymmetric, compressible, and turbulent flow, the conservation
equations for the present problem are presented in
terms of the continuity, momentum, and energy equations:
continuity equation,

~(pu,)
c?x,

-0

(1)

momentum equation,

~7,,
8s
energy equation,

a(puh )

t~.,.,

c:.,.

a(
,,,

a/

,,,

(2)

ah \

where xi is the coordinate, u, is the mass-averaged


velocity component, p is the time-averaged pressure, h
is the mass-averaged static enthalpy, p is the time-averaged density. Here, ll~fr is the effective viscosity which
includes molecular viscosity #1 and turbulent viscosity
#,. The turbulent viscosity #t is determined from the
high Reynolds number form of the two-equation k - e
turbulence model. In this model, the turbulent viscosity
is computed from the local values of the kinetic energy
of the turbulence k, and its dissipation rate E. The
turbulence model coefficients are the standard constants
recommended by Launder and Spalding [17]. For the
present study, the modifications to the standard k--~
turbulence model with regards to the compressibility
effect and shear layer effect are not taken into consideration. Numerical experiments showed these effects to
have little impact on the converged solutions but increased computational time. K,er is the effective conductivity, which also includes laminar and turbulent effects.
To make direct comparison among the computational
results, the foregoing conservation equations and following boundary conditions are normalized by the stagnation values assumed to exist at the gas manifold. The
thermodynamic properties of the gas are related
through the equation of state for a perfect gas.
Eqs. (1)-(3) can be transformed to the form of a
generalized transport equation [18]:

Gt

i
l

0'

'i,d 4 5 . ,

'

" I "'
I
Fig. I. The H P G A nozzle geometry.

div(pui) = div(F grad ) + Se

(4)

where & is any dependent variable under consideration


and F is the exchange coefficient. The term on the
left-hand side of the above equation represents the flux
due to convection. The first term on the right-hand side
represents the flux due to diffusion. S~ is any other
source term for which cannot be included in these
two terms.
A general procedure based on the "SIMPLEST"
algorithm [18] is applied to solve all the governing

22

J. Mi et al. ; Materials Science and Engineering A208 (1996) 20- 29


CONSTANT PRESSURE BOUNDARY

O
Z

GAS JET INLET

HALF CENTRAL SECTION


OF MELT TIP

CENTERLINE OF ATOMIZATION NO77~ E


Fig. 2. Computational domain.

equations. The differential equations represented by Eq.


(4) are formulated using the "control volume" approximations [18]. This involves subdivision of the computational domain into a number of control volumes, each
associated with a grid point, and integration of the
governing equations over the computational domain
using a fully implicit finite difference scheme.
The computational domain of the problem is shown
in Fig. 2. Only half of the actual nozzle flow field is
solved with symmetry about the nozzle center-line assumed. The lower hatched region represents half of the
central section of the flow-blocked melt tip. At the
inflow boundary of the constant width gas jet, the
enthalpy, pressure and velocity of the gas are specified
together with turbulent kinetic energy and its dissipation rate. The assigned gas property values are based on
an isentropic expansion of the stagnation properties
assumed to exist in the gas manifold to those at the gas
jet inlet [19]. With the high gas atomizing pressures
typically used, the flow in the jet is "choked". That is,
the gas stream at the jet inlet is at Mach number of
unity and subsequently expands. The inlet turbulent
kinetic energy is specified by assuming the turbulent intensity to be 1%. Therefore, the inlet values of turbulent
kinetic energy k and its dissipation rate E are given as

(0.01 Win) 2
kin --

'

fin =

t.'~33.';421./~"
, ~i,,,
L

(5)

where Wi, is the gas inlet velocity, L is a typical length


scale taken to be 0.001 of the nozzle diameter, and c;, is
a constant. At the outflow boundary, a static pressure is
specified and held constant at a value equal to the
ambient pressure. This requires that the boundary be
located at a distance which is sufficiently far downstream. Across the center-line of the gas atomization
nozzle no mass, energy or momentum flux is permitted.
Along solid surfaces, the boundary conditions are no
slip and adiabatic. A zero-flux condition is applied for
the static enthalpy for the adiabatic condition. The
standard k - ~ model is applicable only to flow regions
with high turbulent Reynolds numbers and cannot be
applied near a wall, where viscous effects become dominant. Hence, a wall function approach is used to
accommodate the viscous effect [20].

,.<

3. Results and discussion

The atomizing gas chosen for the study is argon. The


only operational variant is the stagnation gas manifold
pressure, and consequently the gas inlet pressure.
Twelve cases are reported for stagnation gas manifold
pressure varying from 0.25 MPa to 5 MPa.
Before proceeding to the computation of the HPGA
gas-only flow, a sonic underexpanded axisymmetric gas
jet was taken as a test case for code validation. This
case was heavily studied and is well documented
[21,22]. The numerical code was used to predict the
shock structures within this flow field for one nozzle
exit condition (stagnation pressure/ambient pressure =
6.76). The calculated results were found to preserve the
salient features of the flow field. Based on the validation study, the gas accelerates to a Mach number of 3.8
on the center-line just upstream of a normal shock or
Mach disk. Then the flow within the subsonic region
decelerates to around 0.51 downstream of the disk
before accelerating again to supersonic. These values
are in reasonable agreement with the respective approximate values of 3.5 and 0.45 accepted in the literature
[21]. The Mach disk can be clearly seen around 1.6 jet
diameters downstream of the nozzle for the calculation,
compared with experimentally reported 1.58 jet diameters [22]. The radius of the Mach disk is calculated to
be around 0.24 nozzle diameters, compared with the
reported experimental value of 0.28 nozzle diameters.
With the code thus validated, we turned attention towards the HPGA model.
The calculation domain shown in Fig. 2 is defined by
a discretized flow field which extends 6 melt tip base
diameters downstream with 400 non-uniform cells in the
main gas flow direction and 1 diameter with 75 non-uniform cells in the lateral direction. This provides suflicient accuracy in the overall calculation of all variables,
as well as permitting a reasonable computer execution
time. Use of non-uniform grid spacing in both directions allows for a finer grid in the regions of steeper
gradients, such as boundary layer, shock wave, and
recirculataion cell, and a coarser grid in the far field,
where the flow is more smooth. The solution is judged

J. Mi et al... Materials Science and Engineering A208 (1996) 20 29

23

/ t-

...........\

-\

RECIRCULATION REALIGNMENT SUPERSONIC NORMAL SUBSONIC SUPERSONIC NOT"-~.~.


REGION
POINT
FLOW
SHOCK FLOW
FLOW
CENTERLINE
Fig. 3. HPGA gas flow velocitydistribution at the stagnation pressure or 4 MPa.
as being grid independent in all flow regions based on
numerical experimentation. For the calculation, adequate convergence of the solution is usually achieved
after 4000 sweeps with 10 iterations on the pressure
correction equation. After this number of sweeps and
iterations, the absolute whole field sums of residual
errors in the finite volume equations for the mass,
momentum and energy are less than 1% of their inlet
reference values. About 20 s of CPU time is necessary
to make one sweep on the SUN SPARC station 2 used
for the numerical simulations.
As an example of a typical flow field, Fig. 3 shows
the H P G A gas flow velocity distribution near the melt
tip at the stagnation pressure of 4 MPa (only some of
the velocity profiles are shown). The gas exhausts from
the H P G A gas manifold and leaves the exit path at a
pressure higher than the ambient condition, such that a
supersonic underexpanded jet is formed along the melt
tip boundary. The flow gives rise to a boundary layer
along the melt tip which separates as it attempts to turn
the melt tip base corner. Outside this very thin
boundary layer, the supersonic flow continues to expand as it moves towards the center-line. The gas
between the realignment point, where the center-line
velocity vanishes, and melt tip base forms a recirculating flow. Gas flow in the recirculation region is towards
the melt tip along the center-line, and then turns radially outwards along the melt tip base. The recirculation
region is surrounded by the supersonic gas flow which
attempts to entrain the gas from the region as a direct
consequence of the shear action along the interface
between two flows. There is a large velocity gradient
present thus producing a strong mixing layer. The
mixing layer approaches the nozzle axis and encounters
a recompression. The flow rapidly accelerates after the
realignment point bringing about a shock wave normal
to the center-line at higher stagnation gas pressures
studied. Flow immediately downstream of the normal
shock is subsonic. The general flow pattern is in global
agreement with a previous experimental study in which
several gas flow visualization techniques were used [12].
Fig. 4 illustrates the predicted H P G A gas flow maximum Mach number with gas manifold stagnation pressure. For the lowest stagnation pressure of 0.25 MPa

used in this study, the maximum Mach number is just


above sonic, at 1.29, while, for the highest stagnation
pressure of 5 MPa, the maximum Mach number can
exceed 6. However, correction for rarefaction effects
above 5 have not been included in the numerical model.
The location of the maximum Mach number moves
downstream with increasing stagnation pressure. At
lower stagnation pressures, the maximum Mach numbers appear above the melt tip frustum portion surface,
while at higher stagnation pressures the maximum
Mach numbers move beyond the melt tip base plane
inward to the center-line. On the contrary, with increasing stagnation pressure, the maximum Mach number
within the recirculation region increases at lower stagnation pressure, then decreases at median stagnation
pressure, finally increasing again at higher stagnation
pressures. The predicted results are shown in Fig. 5.
The maximum gas flow Mach number within the recirculation region varies from 0.5 to 0.9, although at very
high stagnation pressures recirculating flow along the
center-line can approach a near sonic condition. At
high stagnation pressures, the intensity of recirculation
near the melt tip base is enchanced, which can be
responsible for more violent mixing in the region and

e~

<

os

1.5

2s

3.5

STAGNATIONPRESSURE(MPA)
Fig. 4. HPGA gas flow maximum Math number.

4,s

24

J. Mi et al. ; Materials Science and Engineering A208 (1996) 20 29


1
O,g

0.8
0.7
~

0.6

05

<

04

~- 0.3

< 0.2
0.1

05

,.5

2'5

3'5

,5

STAGNATION PRESSURE (MPA)


Fig. 5. Maximum gas flow Mach number within the recirculation
region.

more effective energy transfer between atomizing gas


and molten metal in the actual atomization process.
The highly underexpanded gas jets can induce the
enhanced entrainment of surrounding gas along the
exterior of the recirculation region. With increasing
stagnation pressure, the realignment point moves
downstream from the melt tip base along the center-line
causing a longer recirculation length. However, by increasing stagnation pressure further, the extremely
strong, high velocity gas flow compresses the recirculation flow, causing the realignment point to move back
towards melt tip with a subsequent decrease in recirculation length. The recirculation length with gas stagnation pressure is shown in Fig. 6. The recirculation
length, or the distance of realignment point from the
melt tip base along the center-line, is normalized by
melt tip base diameter. These results are in general
2
1

8t

1.6
[,(~

1.4

.d 1.2

~ 0.8
~ 11.6
e~

0.4

0.2

O0

0'.5

,15

215

315

STAGNATION PRESSURE (MPA)


Fig. 6. Recirculation length.

415

agreement with experimental measurements [15].


The turbulent kinetic energy is a valid indication of
the violence of mixing in the turbulent flow field. Contours of non-dimensional turbulent kinetic energy at
the stagnation pressure of 4 MPa are presented in Fig.
7. It is clear that turbulent kinetic energy is substantial
in the mixing layer, recirculating flow and subsequent
normal shock region, but drops off in other flow fields.
This suggests that in the mixing layer more violent
mixing occurs than in other flow regions.
For all cases, at all stagnation pressures, the underexpanded flow experiences the formation of an expansion
wave structure on exiting from the exit path as seen in
Fig. 8. The bright region in the figure indicates a high
Mach number gas flow, while the dark region represents a low Mach number gas flow. In the center of the
expansion wave (the brightest) the flow achieves its
maximum Mach number. By increasing stagnation
pressure to about 3 MPa, the expansion waves interact
downstream of the melt tip to form a Mach disk. Fig.
9 presents the gas flow Mach number along the nozzle
center-line at the stagnation pressure of 4 MPa. In Fig.
9 absolute Mach number values are used; therefore, we
note that the first peak in the figure represents a reverse
recirculating flow. The axial length along the center-line
has been normalized by the melt tip base diameter.
Around 2 melt tip base diameters downstream of the
melt tip base, the Mach number drops abruptly from
2.4 to 0.2, indicating that a normal shock occurs at this
location.
The strength of these Mach disks for some high
stagnation pressures can be expressed in terms of the
pressure rise across the shock [19], i.e. the ratio between
the after-the-shock static pressure and the before-theshock static pressure as shown in Fig. 10. The figure
shows that with increasing stagnation pressure, the
strength of the shock increases. All these results are in
reasonable agreement with experimental observations
by Anderson and Figliola [12].
With increasing stagnation pressure, the calculated
location of the Mach disk moves downstream from the
melt tip, as shown in Fig. I I. In the figure, the position
of the shock is normalized by the melt tip base diameter. These numerical values are also compared with
experimental results [15] in the figure. Clearly, the
shock location gas pressure curves for the numerical
and experimental data have the same slope but are
displaced from each other. In this regard, all of the
experimental results use the inlet pressure measured at
the controlling regulator valve for reference. This is not
the same reference value as the jet entry-manifold value
used here. There can be a significant stagnation pressure drop due to tubing, coupling and the valve body of
the gas delivery system and internal manifold shocks.
We do not attempt to account for these losses in this
study.

J. Mi et al../ Materials Science and Engineering ,4208 (1996) 20-29

25

6,".

0.04
O. 06

0.08
O. 20
O. 22
0.

~.4

Fig. 7. H P G A gas flow non-dimensional turbulent kinetic energy contours at the stagnation pressure of 4 MPa.

!-!"<!"ki

1.3
!7
2.1
2.5
2.9

3.a
4
.

4.b

Fig. 8. H P G A gas flow Mach number contours at thc stagnation pressure of 4 MPa.

Because of this, the experimental pressure value measured at the regulator will be higher than the stagnation
pressure within the gas manifold and will depend on the
plumbing and delivery system used. Hence, displacement of the data is expected. Therefore, and primarily
because of this uncertainty, it is difficult to establish an
exact comparison based on the stagnation pressures.
Also, a contributor to any discrepancy between the
numerical results and the experimental data to which
they are being compared can be the increased mass flow
rate that resulted from using the width of the gas

annulus jet as the same dimension as the diameter of


an experimental discrete jet. However, it can be
seen that the predictions show the correct trend, with
the Mach disk moving away from the melt tip and
increasing in strength as the gas stagnation pressure
increases.
The base pressures developed across the melt tip are
of practical interest because a subambient pressure
developed there may have beneficial effects on atomizing stability and possibly powder size refinement and
melt tip base may be a suitable site for pressure sensors

J. Mi et al. / Materials Science and Engineering A208 (1996) 2 0 29

26
3.0

MELT TIP BASE PLANE

N O R M A L SHOCK~

n
t.t.I 2.0
Z
"r
L)
<

1.0

0.0
0.0

2.0

4.0

6.0

NONDIMENSIONAL LENGTH ALONG THE NOTZI .E CENTERLINE


Fig. 9. Mach number along the nozzle center-line at the stagnation pressure of 4 MPa (compare directly with Fig. 8).

for an in situ controller [4-8,12]. Calculated base pressures for three stagnation pressures, shown in Fig. 12,
are consistent with limited measurement results of 12
MPa [15]. Here, the radial distance from the center
along the melt tip base is normalized by the melt tip
base radius. The base pressure is a maximum near the
center where the flow divides and turns radially outward along the base. Moving outwards from the center,
the base pressure decreases. At a lower stagnation
pressure of 0.5 MPa, near the center the base pressure
is above atmospheric pressure, while away from the
center the pressure becomes subambient. For a median
stagnation pressure of 1.67 MPa, the whole melt tip
base experiences a subambient pressure. At a higheer
stagnation pressure 5 MPa, the pressure at the center
portion of the melt tip base increases to above atmospheric conditions, while a larger outer portion of the
base develops subambient pressures. At all operating
pressures, a radial pressure gradient is set up that will

tend to drive the flow radially outwards from the base


center towards the base edge. At higher stagnation
pressures, this gradient increases considerably. The
overall pressure distribution along the melt tip base
supports the experimental observations that melt is
driven radially along the melt tip base by pressure
forces and entrainment [12]. The melt discharging from
the melt feed tube would be forced radially outwards
towards the corner of the melt tip where it would
experience direct contact with the still expanding high
kinetic energy gas flow. In fact, a radial pumping would
direct the melt into the region of the highest gas
velocity, providing a ready source of energy required to
produce fine droplets.
In the previous studies of HPGA [4-8,12], it was
suggested to use orifice pressure to determine the range
and magnitude of stable melt aspiration in advance of
atomization trials. From these numerical results and
some earlier experimental results I151, it is clear that the
2.5

1-

v -

NUMERICAL

18~
16:-

14~
~1,5

,-1

~1o

05[

00

L
2

i
4

i
8

i
8

I0
1

12

STAGNATION PRESSURE (MPA)


STAGNATION PRESSURE (MPA)
Fig. 10. Strength of the normal shock.

Fig. I I. Comparison of shock location between numerical predictions


and experimental results.

J. Mi et al. : Materials Science and Engineering A208 (1996) 20 29

I
-----

5[ .

27

X - - 0.5 MPA
[
+ - - 1.67 MPA
[
* - - 5 MPA
]
O-- MEASUREDI
t

(12 MPA)

~,

:<
t.~
,.v
~+++++'~+*++,++++
,

,.

0.5

+++~++++,++.++,
I

EXPERIMENTAL

o;0

10

STAGNATION PRESSURE (IVIPA)


Fig. 14. C a l c u l a t e d a n d m e a s u r c d melt tip base c e n t e r pressures.
0~

02

03

0.4

0.5

0.6

0.7

0.8

09

D I M E N S I O N L E S S R A D I A L DISTANCE
Fig. 12. C a l c u l a t e d a n d m e a s u r e d b a s e p r e s s u r e a l o n g the melt tip
base.

pressure distribution along the melt tip base is not


uniform. While at some higher stagnation pressures a
fraction of base pressure could be above atmospheric
pressure, desirable atomization conditions could still
exist as has been our experience. Based on this study,
caution is advised when using orifice pressure as a
diagnosis, i.e. the monitoring pressure port should not be
at the center of the base but rather at least 0.45 radii from
the center.
Fig. 13 shows the variation in the calculated base
center pressure ratio vs. overall pressure ratio. The base
center pressure ratio is defined as the ratio between base
center pressure and stagnation pressure. The base center
pressure ratio decreases with decreased overall pressure
0.5

045
O

0.4
0.35

7~

0.3
0.25
0.2

Z
~

0.15

~l:

0.1

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

O V E R A L L P R E S S U R E RATIO
Fig. 13. C a l c u l a t e d b a s e c e n t e r p r e s s u r e r a t i o vs. o v e r a l l p r e s s u r e
ratio for HPGA.

ratio until the overall pressure ratio of 0.03. Further


decreases in overall pressure ratio produces no significant
change in the base center pressure ratio. This phenomena
has been well investigated and documented in rocket plug
nozzle studies [23,24]. For an overall pressure ratio above
0.03 the wake region remains " o p e n " but then becomes
"closed" as pressure ratio is lowered below 0.03. This
behavior is very similar to that observed in an axisymmetric plug nozzle proplusive jet with cylindrical shrouds
and conical truncated plug. At low values of operating
pressure to ambient pressure ratio the wake region is
"open", which is sensitive to ambient conditions and
unsteady in nature. During the open wake operation the
base pressure is essentially equal to the ambient pressure.
When the operating pressure to ambient pressure ratio
has increased sufficiently, the wake "closes" and the
separated base region is no longer sensitive to ambient
conditions. During the coised wake operation the base
pressure ratios remain relatively constant. It is noted that
the operation of the H P G A nozzle at high pressure ratios,
i.e. at low stagnation pressures, is comparable with the
open wake condition for a plug nozzle. The relatively
high base pressures and unsteady flow behavior at these
operating conditions are not so attractive for stable
atomization. The operation of the H P G A nozzle at high
stagnation pressures corresponds to the closed wake
operation for plug nozzle and brings subambient base
pressures and relatively stable flow conditions.
Fig. 14 illustrates the calculated melt tip base center
pressures as a function of different gas manifold stagnation pressures. The curve demonstrates that melt orifice
pressure decreases with increasing stagnation pressure
to reach a minimum and then increases with further
increasing stagnation pressure. It is inferred that increasing the stagnation pressure does not progressively
lower the orifice pressure, thereby increasing the aspiration effect. The result compared favorable with experimental results of a similar atomizer, in which inlet
pressures at the regulator valve were used [12].

28

J. Mi et al. / Materials Science and Engineering ,4208 (1996) 20-29

-~

~.~5mm stream diameter


I ~,~8.5mm tip edge diameter

Atomized
Spray

iihdi

1
_L
6.Tmm

6.4ram --~

~--

5.

6.

7.

8.

231111 frame interval, g2/.le exposure

Fig. 15. High speed cinematograph of actual gas atomization process. Reprinted from Anderson and Figliola [12].

Overall, there is good qualitative agreement between


the numerical flow solutions and various gas flow visualization results, including the Schlieren images, water
table shadowgraphs, and white smoke images of the gas
flow in the atomization process [12-15]. The predominant flow discussed is apparent in both numerical and
experimental results. There is also good quantitative
agreement between the calculated base pressure results
and static pressure measurements.
However, there are discrepancies between the actual
atomizer nozzle configuration and the physical characteristics of the numerical model and these would affect
exact comparisons. The flow studied in the experiments
is produced by either 18 or 20 discrete gas jets. Therefore, the process is absolutely three dimensional with
discrete jet interactions. Conversely, the numerical simulation models an annular slit nozzle, as a consequence
of its axisymmetric assumption.
Based on a numerical gas flow simulation with supporting experimental evidence [12], the present study
had documented the flow structure and provided supporting evidence for the dominant mechanisms responsible for melt break-up in the HPGA process. A low
pressure outside the mixing layer together with relatively higher pressures within the recirculation region
bring on a lateral pumping of the melt into the mixing
layer. The pressure gradient along the melt tip base
forces melt to flow radially outward to the boundaries
of the recirculation zone to make contact with the high
speed gas flow. Along the mixing layer, atomization
occurs with vigor. The liquid film can behave much as
a wave breaking down into droplets owing to the

rapidly growing instabilities brought on by the interaction from the high speed gas.
To illustrate how the gas-only flow field plays a
relevant role in the actual process, consider the high
speed film frames taken from the atomization of 2 kg
charge of Cu-2wt.%Zr using the HPGA nozzle described [12]. The sequence, as shown in Fig. 15, was
recorded on 16 mm film at a nominal rate of 10 000
frames s ~ using 2000 W incandescent diffuse backlighting. In the sequence shown, the atomizer is aligned
with the gravity vector so the main flow is downwards.
The melt enters the established flow field and immediately flows radially outwards along the melt tip base. At
the edge of the melt tip the melt turns downwards and
flows along the mixing layer. Soon the recirculation
region is outlined. Break-up can be observed. In the
final frame a high void fraction within the central core
is apparent, indicating that the bulk of the metal mass
flow is along the mixing layer. Even at subsequent
frames well into the atomization run this overall flow
field persists. While shock structures will not show up
with this method, the overall structures of the flow field
are precisely consistent with the numerical predictions
of the gas-only flow field.

4. Conclusion

This study has revealed the gas flow characteristics of


an HPGA process by using a numerical simulation. The
effect of gas atomizing pressure on the gas flow field is
cxamined. The main gas flow features characterized in

J. Mi et al.

MateriaLs Science and Engineering ,4208 (1996) 20-29

the numerical solutions include a strong recirculating


region, a mixing shear layer at all gas stagnation pressures and normal shock structures at some higher stagnation pressures. The intensity of the recirculation flow
and mixing layer is found to increase with increasing
stagnation pressure. At the elevated pressure, a Mach
disk forms at the center-line. The location of the Mach
disk moves downstream with increasing stagnation
pressure and, meanwhile, its strength also increases. All
these numerical results agree with experimental results,
at least qualitatively. Calculated melt tip base pressures
reach a maximum at the center of the melt tip base and
a decrease to a minimum when moving away from the
center. This pressure gradient derives the radial flow of
the melt observed during experiments.
Ultimately, the present numerical study provides support to the previous experimental evidence and verifies
the melt break-up hypothesis proposed in the earlier
studies. Recirculation flow and adjacent strong mixing
flow are responsible for the primary melt break-up. It
also demonstrates that the atomizer performance in the
high pressure regime can be affected and controlled by
the interaction between molten metal and gas flow, in
particular the gas flow behavior near the melt tip base.

References
[l]
[2]
[3]
[4]

J.D. Ayers and I.E. Anderson, J. Met., 37 (1985) 16.


E.J. Savage and F H . Froes, J. Met., 36 (19841 20.
N.J. Garant. J. Met., 35 (1983) 2(1.
I.E. Anderson and B.B. Rath, in S.K. Das, B.H. Kear and C. M.
Adam (eds.). Proc. Syrup. on Rapidly Solidified CO'stalline AIhO',~, TMS AIME, Warrendale, PA, 1985, p. 219.

29

[5] W.O. Busch and A.O. Sepulveda, A~h'ances m Powder Metallurg.v, Vol. 2, Prac. 1989 Powder Metallurgy ('onf and Exhihitwn, MPIF, Princeton, N J, 1989, p. 97.
[6] C.S. Lin. G.J. Jih. C.S. Chang, and S.Y. Chiou, Adram'es m
Powder Metallurgy, Vol. 2, Proc. 1989 Pou~h,r metallurgy ConL
and E~hihition, MPIF, Princeton, NJ. 1989, p. 657.
[7] R.A. Ricks and T.W. Clyne, J. Mater. Sei. Lett.. 4 II9,~51 814.
[8] SD. Ridder'and F.S. Biancaniello, Mater. Sci. Eng.. 98 (19881
47.
[9] N. Dombrowski and W R . Johns, ('hem. l:)tg. Sci.. 18 (1963)
2O3.
[10] D. Bradley, J. Phys. D, 6 (19731 1724, 2267.
[1 I] J.B. See and G . H Johnson, Powder Technol., 21 (1978) 119.
[12] I.E. Anderson and R.S. Figliola, in P.t;. Gummeson and D.A.
Gustafson (eds.L Modern Derelopment in Powder Metallurgy,
APMI, Princeton, N J, 1988, p. 205.
[13] I.E. Anderson. R.S. Figliola and H. Morton, :'~hm,r. Sci. Eng.,
148 (1991) 101.
[14] RS. Figliola, I.E. Anderson and tl. Morton, in ,S~vnthesis and
,4na[ysis in Materials Proc.: ('haraeterization and Dmgnostics ~[
Ceramic attd Metal Particulate Processing. TMS, Warrendale,
PA, 1989, p. 37.
[15] H. Morton, Master lhesis, Clemson University, 1990.
[16] A. Unal. Metall. Trans. B, 20(1989)833.
[17] B.E. l,aunder and D.B. Spalding, Comput..~lethod Appl. Mech.
Eng., 3 (1974) 269.
[18] S.V. Patankar, Numerical Heat Tran.~/kr attd Fluid Flow.
McGraw-Hill Hemisphere, New York, 198(I.
[19] M.A. Saad. Compressihh, Fluid Flow, Prentice-Hall. Englewood
Cliffs, N J, 1985.
[20] W. Rodi, Turbulence Mo~&l,~ arid Their Application in Itydraulic~ A State ~/ the Art Revim. International Association
[br Hydraulic Research, Delft, 1980.
[21] ('.D. Donaldson and R.S. Snedckcr, J Fluid Mech., 45 (1971)
281.
[22] G.S. Love, ('.E. (;rigsby, L.P. Lee and M . J Woodling, NACA
Tech. RT. R-6, 1959.
[23] C.R. Hall and T J . Mueller, J. Spacecr., 9 (1972) 337.
[24] W.P. Suit and T.J. Mueller, J. Spacecr.. I0 (1973) 689.

You might also like