Download Experimental Technique - Metal Powder Industries Federation

Reprinted by permission from Advances in Powder Metallurgy & Particulate Materials—2012.
Copyright © 2012 Metal Powder Industries Federation. All rights reserved.
INVESTIGATION OF THE PULSATION PHENOMENON IN
CLOSE-COUPLED GAS ATOMIZATION
Andrew M. Mullis, Ian N. McCarthy and Robert F. Cochrane
Institute for Materials Research
University of Leeds
Leeds LS2-9JT, UK
Nicholas J. Adkins†
CERAM Research
Penkhull
Stoke-on-Trent ST4-7LQ, UK
ABSTRACT
High speed photography coupled with sophisticated image analysis has been used to study the low
frequency pulsation in the volume of melt being instantaneously delivered to the melt nozzle during
close-coupled gas atomization. We find that at low gas pressures the distribution of material at the melt
tip can be described by a log-normal distribution. At high gas pressure the distribution is better described
by two superimposed log-normal distributions, one with a high standard deviation when there is little melt
at the atomizer tip and a second with a lower standard deviation when there is more melt at the atomizer
tip. We associate this behavior with the transition between open- and closed-wake conditions in the gas.
We suggest that the methodology proposed represents a simple, non-invasive technique for characterising
the performance of gas atomizers.
INTRODUCTION
Close-Coupled Gas Atomization (CCGA) is the production technique of choice when fine (5-50 µm,
0.0002 - 0.002 in), highly spherical, metal powders are required. In principle CCGA is straightforward,
high pressure gas jets impinging upon a molten metal stream are used to disrupt the stream, breaking it up
into a spray of fine droplets. The gas may be delivered either from a single, annular orifice surrounding
the melt delivery nozzle, or by a series of discrete jets. In many commercial atomizers an annular gas
delivery mechanism is employed due to its simplicity, although the discrete jet design has been
demonstrated to be much more efficient in terms of gas consumption1 and has therefore been the focus of
much of the recent research into atomization. In both cases, liquid metal is delivered down the central
bore of the nozzle, wherein it wets the nozzle tip (a process termed pre-filming, which is itself dependent
upon the gas flow conditions) and is stripped off the circumferential edge of the nozzle by the gas.
However, the complex interaction between the high velocity gas and the metal results in a turbulent, and
often chaotic, flow with the result being that the details of the process are far from well understood.
Consequently, early work into gas atomization focused on empirical correlations between median particle
†
Now at: School of Metallurgy and Materials, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK.
1
size and process parameters2, 3, 4 such as gas pressure, gas flow rate and melt flow rate. The most widely
quoted of these empirical relationships is that due Lubanska,2 which correlates particle size with (1+G)-1/2,
where G is the gas:metal mass ratio.
For many potential powder metallurgy (PM) applications, both a small median particle size and a narrow
size distribution in the product are desirable. While considerable progress has been reported on improving
the efficiency with which the impinging gas jets disrupt the melt stream, and hence in reducing the
median particle size,1, 5 most atomizer designs produce a distribution of powder sizes which span one
order of magnitude or more.5 As pointed out by Nasr et al.,6 remelting of the out of specification product
can add substantially to the cost and energy usage of the overall process.
A number of factors contribute to this spread in the particle size distribution. In particular, common
foundry experience is that close-coupled gas atomizers pulse, that is experience variations in the amount
of metal being instantaneously delivered to the atomization tip, with this pulsation being in the frequency
range 5-50 Hz. This is commonly seen as a flickering of the luminosity of the atomization spray cone,
and has been quantitatively studied by applying Fourier analysis to high speed filming of the atomization
process.7 A significant scientific literature has built up attributing these fluctuations to the transition
between open- and closed-wake conditions.7, 8, 9 Clearly, such pulsing of the melt causes commensurate
variations in the instantaneous gas:metal ratio and hence presumably in the size of the resulting powders.
The wake region occurs due to the separation of the supersonic flow from the melt nozzle’s edge, and is
characterised by a recirculating subsonic gas flow. The far end of the wake region is determined by the
position of the ‘stagnation point’, a point of maximum pressure and minimum velocity where the majority
of the gas enters the wake region. Srivastava and Ojha10 state that gas entering the wake region does so via
a toroidal vortex. Gas exits the region from the periphery of the melt nozzle. In general, the higher the
stagnation pressure, the larger the amount of gas entering the region. Upon increasing the gas inlet
pressure the structure of the flow-field goes through a number of changes. The internal shocks caused by
the expansion of the issuing gas moves downstream. As they continue to move further downstream with
increasing inlet pressure the recompression shock that reflects from the internal sonic boundary moves far
enough to move past this boundary and crosses with the rest of the recompression to form a mach disk.
The formation of the disk cuts off the wake region from the surrounding flow, dramatically decreasing its
size and strength. It is this phenomenon that is termed ‘wake closure’. The magnitude of the inlet pressure
required to achieve wake closure is dependant on the gas inlet configuration; the gas inlet apex angle and
the gas jet exit area, as well as the size and geometry of the melt nozzle.8, 11
Wake closure has been an area of intense study as it has been postulated that operation within this region
may be beneficial in producing better powder products. Ting et al.8 reports that atomizing just above wake
closure led to an increase of 42% in fine particle yield when compared to operation in open-wake
conditions. It also led to a large decrease in the melt flow rate. However this is disputed by Mates and
Settles9, 12 who suggest that wake closure is of no real consequence in atomization as the closed-wake
structure is not preserved during the actual atomization process. The introduction of the dense melt
disturbs the structures, causing the disappearance of the mach disk, and the ‘opening’ of the wake.
The pulsatile model proposed by Ting et al.8 suggests that during atomization in the closed-wake
condition a dynamic balance exists between the momentum of the melt and the momentum of the
atomizing gas. In gas only conditions, equilibrium is reached and the familiar pattern of shockwaves and
flow features are established. The introduction of a liquid, be it molten metal or a low melting point fluid,
into the recirculation zone acts against the compressible gas, distorting and displacing the established
flow features. The changes in flow conditions caused by the introduction of the melt causes the melt flow
rate to slow or even stop, allowing the original flow features to re-establish. Once re-established melt is
again able to flow into the recirculation zone and the cycle is re-established.
2
EXPERIMENTAL TECHNIQUE
Imaging experiments were conducted on both an analogue (water) atomizer and a research scale metal
atomizer, both of which utilised an identical die of the discrete jet type with 18 cylindrical jets arranged
around a tapered melt delivery nozzle with an apex angle of 45°. The design, which is shown
schematically in Figure 1, is similar to the Ames HPGA-I1 and USAG13 designs. Although these designs
are known to be sub-optimal in their atomization performance, the cylindrical jets giving rise to choked
flow which limits the outlet gas velocity to Mach 1, we have used this geometry as it has been discussed
extensively in the literature. Moreover, unlike the convergent-divergent jets used to produce supersonic
gas flow, the simple cylindrical jet can be operated over a wide range of pressures allowing us to
investigate the characteristics of the atomization process as a function of pressure. Atomization
experiments were conducted over inlet gas pressures in the range 0.5 - 5.5 MPa (75 - 800 psi).
Figure 1. Schematic illustration of the geometry of the gas delivery die and melt nozzle used in the
atomization experiments.
Imaging of the melt spray cone was performed using a Kodak Ektapro 4540mx high speed digital motion
analyzer, operating at frame rates between 4,500 frames/s and 18,000 frames/s. The motion analyzer was
fitted with high magnification optics which allowed full frame images (covering a distance ~5 cm (2 in)
from the die) to be imaged at a working distance of 25 cm (10 in). Frames varied in size between
256  256 pixels at 4,500 frames/s to 256  64 pixels at 18,000 frames/s, with an eight-bit grey-scale
depth per pixel in all cases. Frames were stored separately as a high quality TIFF image without
interlacing in 1 Gb of fast memory, giving a total recording time of 3.64 s irrespective of frame rate
(corresponding to 16,384 frames at the lowest frame rate used to 65,536 frame at the highest frame rate
used).
For the metal atomization experiments Ni31.5Al68.5 (Ni-50 wt.% Al) metal was used, the melting point of
which is 1613 K (2444 F). The melt was superheated by 200 K (360 F), giving a melt temperature prior
to ejection of 1813 K (2804 F). At this temperature the melt stream is sufficiently bright that filming can
take place using the radiant light from the melt, even at high frame rates. Consequently, no external light
sources were used. Conversely, in the analogue atomization experiments imaging was performed in
reflected light provided by two high power halogen lamps mounted in the base of the atomizer. For both
the metal and analogue atomizers the melt was ejected from the guide tube under a constant overpressure
of 40 kPa (6 psi).
3
RESULTS OF THE IMAGING EXPERIMENTS
One feature that is apparent from the filming, particularly when a number of images are combined into a
movie, is that there are considerable fluctuations in the amount of material instantaneously being
delivered from the nozzle. This is illustrated in Figure 2, in which 2 frames taken from one imaging
sequence performed on the Ni-Al melt at a gas pressure of 3.5 MPa (500 psi) are shown. The frame
capture rate in this sequence was 4,500 frames/s.
Figure 2. Two frames from the high speed filming of the gas atomization of Ni31.5Al68.5 melt at 1813 K,
showing (a) copious melt at the atomization tip and (b) melt to the tip virtually choked off. The two
frames are separated by 0.842 s and the flow of melt resumed with 0.02 s of frame (b) being captured.
Here, because the imaging is performed using radiant light from the hot melt, the spray cone appears light
against a dark background. Pre-filming of the melt nozzle is evident, with the bright area to the top of the
frame in Figure 2a being the 5 mm (0.2 in) diameter circular base of the melt nozzle, not the outlet of the
inner guide tube. The frames shown are non-consecutive, and have been selected so as to illustrate the
extremes in the pulsation of metal delivered to the tip, with the frame number shown on each image
indicating from where in the sequence the image is taken (Note that due to the triggering mode used both
the time and frame number are counted backwards, consequently low frame numbers are images captured
later in real time). In particular, in Figure 2a it is evident that a significant quantity of melt is being
discharged from the melt nozzle, whereas in Figure 2b, recorded some 0.842 s later during the same
atomization run, the flow of melt to the nozzle has been virtually choked off and even the tip of the melt
delivery nozzle is barely distinguishable. We would stress here that neither the positioning nor settings on
the camera have been altered between these two frames being captured. Similar amounts of melt being
discharged from the melt nozzle would have given rise to images of similar brightness and the difference
between the images therefore reflects that amount of material instantaneously at the nozzle tip. Melt
returned to the nozzle shortly after frame 12448 (Fig. 2b) had been captured.
These fluctuations in the optical intensity apparent at the tip of the melt nozzle during filming have been
analysed quantitatively by taking a slice across each image. This slice is normal to the melt nozzle and 2
nozzle widths downstream of the nozzle outlet (see Figure 3). The lens aperture and camera gain are
adjusted such that during peak discharge no pixels are at saturation brightness. The mean grey level
intensity of the row of pixels within this slice is then calculated, giving a single statistic per frame which
may be used as a proxy for the optical intensity of the melt instantaneously at the tip of the melt delivery
nozzle, which may then be plotted as a time-sequence. As the grey level is recorded on an 8 bit scale this
statistic varies from zero (no discernible melt is being discharged from the melt nozzle) to 255 (the pixels
across the entire width of the frame in the slice are at saturation intensity).
4
QUANTITATIVE DATA ANALYSIS TECHNIQUES
In a series of recent papers14, 15 we have shown that if the Fourier transform of this average grey level time
sequence is taken two characteristic frequency regimes may be identified. A similar observation was
made by Ting et al.7 using a different methodology in which a 2-dimensional Fourier transform was
applied to the whole image. The low frequency component of this oscillation, for which the majority of
the spectral power is located at frequencies < 50 Hz, is thought to be related to the pulsation behavior7 of
the melt discharge which can be seen in Figure 2 and which is the subject of the investigation here.
Figure 3. Illustration of the sampling window used to measure the total volume of melt being discharged
from the melt nozzle (h = two nozzle diameters). Note that in this figure the image has been rotated by
90 counter-clockwise to save space, atomization occurs vertically downwards.
Conversely, we have postulated that the high frequency variation, which generally has a well defined
peak spectral power in the frequency range 300 - 1000 Hz depending upon the atomization parameters
and the thermophysical properties of the atomized fluid, is related to the motion of melt filaments within
the spray cone. This observation appears to be confirmed by recent observations using ultra high speed
(20 ns) pulsed laser imaging which has allowed the highly regular motion of such filaments to be
observed directly.16 However, as this high frequency variation in the optical intensity is related to the
motion of the melt at the nozzle tip, and not to the quantity of melt present, this high frequency
component of the signal is filtered out before we make any further analysis to understand the pulsation
behavior of the melt.
In order to remove the high frequency components of the variation a low band pass filter is applied to the
time-series data. For this we use a Butterworth IIR multi-section filter, as Butterworth filters are
characterized by a magnitude response that is maximally flat in the pass-band and monotonic overall. The
transfer function, H(z), for a filter of order n is given by;
H ( z) 
B( z ) b0  b1z 1  ...  bn z  n

A( z ) a0  a1z 1  ...  an z n
where a and b are length n+1 vectors containing the filter coefficients. Here we use a filter of order n = 7
in 4 sections. The pass-band, with gain of 1 (0 db) is defined as the region f  60 Hz, with an attenuation
rising smoothly to -80 dB at a frequency of 300 Hz. This filter design is considered appropriate for the
current application as those parts of the signal relating to melt pulsation (typically f  50 Hz, and with
negligible spectral power above 100 Hz) are preserved while components associated with motion of melt
filaments (which have negligible spectral power below 200 Hz) are very effectively removed. The
frequency transfer characteristics of the filter are shown in Figure 4.
5
DATA ANALYSIS RESULTS
The effects of applying such a filter to the grey level intensity data obtained from a typical atomization
experiment are shown in Figure 5. In this particular case the Ni-Al melt was atomized at a gas pressure of
3.5 MPa (500 psi). The time averaged value of G, the gas-to-metal ratio, was measured at 2.84 (kg gas to
kg melt) while the resulting powder had a mass-median diameter of 62.6 m (0.00246 in) and a
(geometric) standard deviation of 1.8. The atomization experiment was filmed at 18,000 frames/s over a
period of 3.64 s, giving a total of 65536 images. Each image was subsequently analyzed to give a single
value of the optical brightness of the melt in the analysis window as described above. Figure 5a shows the
unfiltered data, with the slow rise and fall in the mean grey level the result of melt pulsation and the
thickening of the trace depicting this variation being the result of the superimposed high frequency
motion of the melt filaments within the spray cone. Figure 5b shows the filtered data, from which it can
be seen that the high frequency oscillation around the mean level has been very effectively removed but
that the chaotic, low frequency signature of the melt pulsation is retained.
Figure 4. Frequency transfer characteristics of the Butterworth filter applied to the data. The filter has a
gain of 1 (0 dB) between 0 and 60 Hz dropping to -80 dB at 300 Hz.
Figure 5. Time sequence showing the variation of optical intensity at the melt delivery nozzle tip (a)
unfiltered and (b) after application of the Butterworth filter.
6
The median grey level in Figure 5b is 30.58, with a minimum of 5.73 and a maximum of 68.98. On the
assumption that (i) the gas flow rate is constant and (ii) that the measured and filtered grey level
corresponds directly to the volume of melt passing instantaneously through the measuring slice, this
would correspond to an equivalent variation in the instantaneous gas-to-metal ratio ranging from 1.26 (at
grey level 68.98, maximum metal available at the tip) to 15.16 (at grey level 5.73, minimum metal
available at the tip). Although no data is available relating the instantaneous gas-to-metal ratio to the size
of the particles produced, in light of the well studied statistical correlations between mean gas-to-metal
ratio and mass median particle size we would conjecture that such large variations in the instantaneous
ratio are likely to be a significant contributory factor to the observed spread in the particle size
distribution in gas atomized powders.
In order to investigate the variation in the mass flow of metal to the tip of the atomization nozzle further,
and hence attempt to elucidate a deeper understanding of the pulsation phenomena during gas
atomization, the mean grey level intensity has been grouped into bins (in the case of this data set of width
1.0), such that an intensity histogram can be constructed. This is shown in Figure 6. The resulting
distribution is very clearly bimodal, with intensity peaks at grey levels of 27 and 33. In general, such a
bimodal distribution would suggest that two separate physical processes are operating, or that conditions
within the process are switching between two possible sets of conditions. To explore this possibility
further cumulative frequency plots can be constructed. Figure 7a shows the (relative) cumulative
frequency plot for the example data set shown in Fig. 6. In Figure 7b we transform the cumulative
frequency plot by taking log10 of the intensity (grey) level and applying the inverse normal probability
function to the cumulative frequency. The result is that the data can indeed be modeled as comprising of
two segments, each of which is, to a good approximation, log-normally distributed. At low optical
intensities, corresponding to there being little metal at the atomization tip, the (geometric) standard
deviation for the resulting distribution is 2.0. Conversely, at higher optical intensities, corresponding to
their being more metal at the tip, the standard deviation is around 1.4.
Figure 6. Histogram of optical intensity measured during 3.64 s of close-coupled gas atomization at a
sampling rate of 18 kHz.
7
ANALOGUE ATOMIZER RESULTS
The analogue atomizer has exactly the same geometry as the metal atomizer described above, and except
where stated was run with the same operating parameters. Our previous high speed imaging studies14, 15, 16
have shown that the analogue atomizer is a very good model for the metal atomizer, displaying many
qualitatively similar phenomena including low frequency pulsation and high frequency precession of the
second (atomized) fluid. In the context of the investigation conducted here performing experiments on an
analogue atomizer has a number of advantages over performing metal atomization experiments, not least
that the analogue atomizer can be run over a much wider range of gas inlet pressures, which is useful in
elucidating the mechanism for the pulsation effect. Specifically, if the metal atomizer is run at too high a
pressure (> 5.0 MPa, 650 psi) freeze-off of the melt at the atomization tip is a significant risk whereas if
the metal atomizer is run at too low a pressure (< 2.0 MPa, 220 psi) incomplete atomization could result
in a stream of hot metal pooling in the bottom of the chamber. In contrast, there are no issues associated
with running the analogue atomizer in the pressure range 0.5 MPa  P  5.5 MPa (70 psi  P  800 psi).
Figure 7. (a) Relative cumulative frequency plot for the data shown in Figure 6 and (b) transformed by
taking the log10 of the intensity and applying the inverse normal function to the relative cumulative
frequency revealing the the data follows two discrete log-normal distributions.
However, it is worth noting one significant difference between the images obtained using the metal and
analogue atomizers. On the metal atomizer imaging experiments are performed in radiant light, whereas
those on the analogue atomizer are performed in reflected light. Consequently, in the case of the metal
atomizer, when no melt is present the field of view is completely black and the recorded mean grey level
intensity is zero. However, in the case of the analogue atomizer some stray reflection of the light from the
halogen lamps occurs from the walls of the atomization enclosure, so in the absence of melt some
background grey level is recorded. This is typically around 10 on the 0-255 scale used here. For each
experiment performed this background level is therefore determined prior to introducing the atomized
fluid into the chamber and subtracted from all subsequent measurements once the fluid is introduced.
The data analysis methodology follows that given above but for brevity we do not reproduce cumulative
frequency data for all of the experiments performed, instead we present here two representative examples
of the behavior observed. Figure 8 shows the inverse normal probability obtained from the cumulative
frequency data plotted against log10(optical intensity) for two pressures, 1.75 MPa (250 psi) and 3.5 MPa
(500 psi). What is interesting is that at the lower pressure we observe a single log-normal distribution
whereas at the higher pressure we observe two distinct distributions, as was the case for the metal
8
atomizer. The geometric standard deviations for these are 1.2 at 1.75 MPa (250 psi) and 1.2 and 1.6 for
the low and high standard deviation regimes at 3.5 MPa (500 psi) respectively. The pressure for this
transition in behavior appears to be around 3.0 MPa (440 psi), with atomization at low pressure always
giving a single regime of log-normal behavior and atomization at high pressure giving two regimes of
log-normal behavior, with the higher standard deviation occurring for low melt flow.
We also note that the values measured on the analogue atomizer appear to be significantly less than those
measured during the atomization of Ni-Al melt. Specifically, taking the case described above for
atomization at high gas pressure, the standard deviation observed during the atomization of Ni-Al melt
using radiant light from the melt at 1813 K (2804 F) was 2.0 in the low flow regime and 1.4 in the high
flow regime whereas the equivalent figures obtained using water as the analogue fluid in reflected light
were 1.6 and 1.2 respectively. As the pulsation phenomenon is intimately linked to the momentum
balance between the high pressure gas and the second fluid we suspect that the lower density of the
analogue fluid, 1000 kg m-3 (62.4 lb/cu ft) for water as opposed to 3750 kg m-3 (234 lb/cu ft) for
Ni31.5Al68.5 alloy at 1813 K (2804 F),17 is responsible for this. In deed, given the large difference in
density we are gratified to observe qualitatively similar behavior between the metal and analogue
systems.
Figure 8. Log-normal plots for the analogue (water) atomizer operating at a gas pressure of (a) 1.75 MPa
(250 psi) and (b) 3.5 MPa (500 psi). At low pressure the intensity appears to follow a single log-normal
distribution but at high pressure two distributions appear to be present, as was the case with the metal
atomizer.
9
DISCUSSION
The pulsation phenomenon in close-coupled gas atomizers is thought to arise due to the chaotic transition
between open- and closed-wake conditions down-stream of the tip.7, 8 However, it is also know that a
minimum gas pressure is required to produce wake-closure and that below this pressure the wake will be
permanently open. Ting et al.8 reported the wake closure pressure in the system they studied as being
around 4.0 - 4.5 MPa (580 - 650 psi). Aspiration pressure measurements16 suggest that for the system
considered here this transition occurs between 3.0 - 3.5 MPa (440 - 500 psi). It is therefore tempting to
speculate that the transition we observe here between the distribution of optical intensities being described
by a single log-normal distribution and being described by two superimposed distributions is a
consequence of the atomizer moving from being in the open-wake condition to being in the state where it
is alternating between the open- and closed-wake conditions. If this is the case we could then also
associate the two log-normal regimes observed at high gas pressure with the alternation between openand closed-wake conditions, with the high flow rate, low standard deviation regime corresponding to the
open-wake condition and the low flow rate high standard deviation regime corresponding to the closedwake condition.
For the metal atomizer the crossover point between the two distributions occurs at an optical intensity of
23.5 (intersection of the two trend lines in Figure 8b), which corresponds to the atomizer being on the
high standard deviation distribution around 25% of the time. However, because there is less metal at the
tip during these times, this corresponds to significantly less than 25% of the product being produced
during these periods. In fact, the fraction of product produced in this regime can be obtained by
estimating the integral


23.5
0
255
0
I  f dI
I  f dI
where I is the intensity grey level and f the frequency density. Using this procedure we estimate around
14% of the total metal volume was discharged while the atomizer was in the high standard deviation state.
The corresponding figures for the analogue atomizer are that it was in the high standard deviation state
21% of the time and that 16% of the total fluid volume was discharged during this time. Moreover, if we
accept that we can associate the open- and closed-wake conditions with the periods of high and low
standard deviation then the estimates above for the duration these periods can be applied directly to the
open- and closed-wake, i.e. the metal atomizer was in the closed-wake condition 25% of the time and
produced 14% of the final product through a closed-wake.
Further, if the observed transition between high and low standard deviation behavior is related to the
open- to closed-wake transition, the behavior observed here has significant implications for our
understanding of the pulsation phenomenon in close-coupled gas atomization. In particular, it would
appear that to claim that pulsation occurs due to the transition between open- and closed-wake conditions
is overly simplistic. Rather, we observe that the atomizer transitions from an open-wake regime where it
pulses with relatively low intensity to a closed-wake regime in which it pulses with much greater
intensity. However, it is still the case that the closed-wake can be associated with low flow of material
and that in all likelihood the ultimate cause of the pulsation is short timescale fluctuation in the aspiration
conditions due to the chaotic interaction of the supersonic gas.
10
SUMMARY AND CONCLUSIONS
We have shown elsewhere,14, 15, 16 using high speed photography, that the amount of material
instantaneously at the melt delivery tip in a close-coupled gas atomizer varies on time scales of 0.05 –
0.01 s. In this paper we have shown that these fluctuations can be described statistically using the lognormal distribution. At low gas pressures these fluctuations seem to follow a single distribution whilst at
higher gas pressures two distributions with quite different standard deviations appear to be present, with a
high standard deviation being characteristic of low flow rates and a significantly lower standard deviation
being observed at high flow rates. For the atomization geometry considered here this transition between
following a single to following a dual distribution appears to occur around 3.0 MPa (440 psi). At high gas
pressure the atomizer appears to spend 20-25% of its time in the high standard deviation state and to
switch between the two states on time-scales of 0.05 – 0.01 s. We conjecture that the low standard
deviation state can be associated with the atomizer being in an open-wake condition and, conversely, that
the high standard deviation state can be associated with the atomizer being in a closed-wake condition.
We propose that the above methodology represents a simple, non-invasive technique for characterising
the performance of gas atomizers and would be appropriate to off-line and potentially on-line monitoring.
Off-line, the technique could be used to evaluate the effects of changes in atomizer configuration while,
with appropriate automation, on-line monitoring would provide an additional tool for ensuring correct
process operation and ultimately quality assurance of the product.
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
I.E. Anderson, R.S. Figliola and H. Morton H., “Flow mechanisms in high pressure gas
atomization”, Mater. Sci. Eng. A, 1991 vol. 148, pp. 101-114.
H. Lubanska, ”Correlation of spray ring data for gas atomization of liquid metals”, J. Metals, 1970,
vol. 22, pp. 45-49.
E. Klov and W.M. Schfer, “High Pressure Gas Atomization of Metal”, 1972, Syracuse University
Press, Syracuse, pp. 57.
D. Bradley, “On the atomization of a liquid by high velocity gases: Parts 1 and 2.”, J. Phys. D: Appl.
Phys., 1973, vol. 6, pp. 1724-1736 and pp. 2267-2272.
I.E. Anderson and R.L. Terpstra, “Progress Toward Gas Atomization Processing with Increased
Uniformity and Control”, Mater. Sci. Eng. A, 2002, vol. 326, pp. 101–109.
G.G. Nasr, A.J. Yule, and L. Bendig, “Industrial sprays and atomization design: Analysis and
applications”, 2002, Springer Publishing pp. 494.
J. Ting, J. Connor and S. Ridder, “High-speed cinematography of gas-metal atomization”, Mater.
Sci. Eng. A, 2005, vol. 390, pp. 452-460.
J. Ting, M. Peretti and W.B. Eisen, “The effect of wake-closure phenomenon on gas atomization
performance”, Mater. Sci. Eng. A, 2002, vol. 326, pp. 110-121.
S.P. Mates and G. S. Settles, “A study of Liquid Metal Atomization Using Close-coupled Nozzles,
Part 2: Atomization Behavior” Atomization and Sprays, 2005, vol. 15 pp. 41-59.
V.C. Srivastava and S.N. Ojha, “Effect of aspiration and gas-melt configuration in close coupled
nozzle on powder productivity”, Powder Metal., 2006, vol. 49, pp. 213-218.
I. E. Anderson, R.L. Terpstra, J.A. Cornin and R.S. Figliola, “Verification of Melt Property and
Closed Wake Effects On Controlled Close-Coupled Atomization Processes”, Adv. Powder Metall.
Particulate Mater., 2006, vol. 1, pp. 1-16.
S.P. Mates and G.S. Settles, “A study of liquid Metal Atomization Using Close-Coupled Nozzles,
Part 1: Gas Dynamic Behavior” Atomization and Sprays, 2005, vol. 15, pp. 19-40.
V. Anand, A.J. Kaufman, and N.J. Grant, “Rapid solidification of a modified 7075 aluminium alloy
by ultrasonic gas atomization”, Rapid Solidification Processing, Principles & Technologies II,
compiled by R. Mehrabian, B.H. Kear and M. Cohen, Claitor, Baton Rouge, LA, 1978, pp. 273-286.
11
14. A.M. Mullis, N.J. Adkins, Z. Aslam, I.N. McCarthy and R.F. Cochrane, “High frame rate analysis of
the spray cone geometry during close-coupled gas atomization” Int. J. Powder Metall., 2008, vol.
44, pp. 55-64.
15. I.N. McCarthy, Z. Aslam, N.J. Adkins, A.M. Mullis, and R.F. Cochrane, “High speed imaging and
Fourier analysis of the melt plume during close-coupled gas atomization” Powder Metall., 2009, vol.
52, pp. 205-212.
16 A.M. Mullis, I.N. McCarthy and R.F. Cochrane, “High Speed Imaging of the Flow During CloseCoupled Gas Atomization: Effect of Melt Delivery Nozzle Geometry” J. Mater. Process Tech.,
2011, vol. 211, pp. 1471-1477.
17 Intermetallic Materials Processing in Relation to Earth and Space Solidification (IMPRESS) Project
(NMP3-CT-2004-500635) Deliverable Report D7.3.
12