Prevention of Nozzle Wear in Abrasive Water Suspension Jets

Prevention of Nozzle Wear in
Abrasive Water Suspension Jets
(AWSJ) Using Porous Lubricated
Nozzles
Umang Anand
e-mail: umang@jhu.edu
Joseph Katz
katz@titan.me.jhu.edu
The Johns Hopkins University,
Department of Mechanical Engineering,
Baltimore, MD 21218
1
This paper introduces a novel method for preventing nozzle wear in abrasive water jets. It
consists of using a porous nozzle, surrounded by a reservoir containing high-viscosity
lubricant, which is exposed to the same driving pressure as the flow in the nozzle. The
pressure difference across the porous medium, generated due to the high-speed flow in the
nozzle, continuously forces lubricant through it. The resulting thin oil film forming on the
walls of the nozzle protects the walls from the impact and shear caused by the abrasive
particles. The porous nozzles were manufactured using Electric Discharge Machining and
examined with Scanning Electron Microscopy. Two test facilities were used for evaluating
the porous lubricated nozzles. The first was a two-dimensional facility, supporting a 145
␮m wide nozzle with windows on both sides, which enabled visualization of the oil film
and measurements of the liquid and abrasive-particle velocities using Particle Image
Velocimetry. The measured slip velocities were also compared to computed values from a
simple numerical model involving one-way coupling. The second facility used a 200 ␮m
axisymmetric nozzle to determine the extent of nozzle wear under different conditions. We
found that the presence of an oil film substantially reduced the extent of nozzle wear, from
111 percent of the diameter, when the nozzle was not lubricated, to 4 percent, when the oil
viscosity was 1800 mm2/s and its flow rate was 2.4 percent of the water flow (over the
same period). The wear increased as the lubricant flow rate and viscosity decreased. The
presence of the oil film also improved the coherence of the jet. 关DOI: 10.1115/1.1491977兴
Introduction
Abrasive water jets, namely water jets containing abrasive particles, have a considerable niche in the material processing industry. Like laser cutting instruments they are accurate, easily managed and cause very little loss of material. However, abrasive jet
cutting does not involve high temperatures, which is characteristic
to laser cutting, and as a result they are suitable for practically any
material. Furthermore, the instrumentation required for high-speed
jets is simpler and much cheaper. Consequently, jet cutting can be
implemented in a broad range of industries, ranging from small
machine shops and quarries, to large sheet metal, composites or
ceramic processing in the car and aircraft industries.
The most troublesome difficulty associated with high-speed
slurry jets, which presently limits their usefulness, is wear of the
nozzle walls 共Conn 关1兴, Dubensky et al. 关2兴兲. Since the jet speed
ranges between 100–500 m/sec, and the particle size can be as
high as 40 percent of the nozzle diameter, it does not take long to
destroy a nozzle. Consequently, in current systems nozzles must
be replaced frequently, sometimes every 10–20 minutes 共Dubensky et al. 关2兴, Kovacevic and Evizi 关3兴, Mort 关4兴兲. The wear of the
nozzle walls also leads to the jet becoming incoherent, which
causes an increase in the kerf width on the workpiece, deterioration of surface quality and loss of cutting accuracy. Hence, wear
of the nozzle requires constant maintenance and inspection, which
leads to machine down-time and increases the process costs.
Present attempts to solve this problem include: 共a兲 Pure water is
injected through an orifice and the abrasive particles are then fed
at low pressure through a side tube. The water entrains the particles as both travel through a mixing tube 共or ‘‘focusing tube’’兲
whose diameter is typically three times larger than that of the
Contributed by the Tribology Division for publication in the ASME JOURNAL OF
TRIBOLOGY. Manuscript received by the Tribology Division October 2, 2001 revised
manuscript received April 16, 2002. Associate Editor: L. San Andrés.
168 Õ Vol. 125, JANUARY 2003
orifice 共Hashish et al. 关5兴, Hashish 关6兴, Momber and Kovacevic
关7兴兲. This approach is typically referred to as Abrasive Water Jets
共AWJ兲, as opposed to abrasive water suspension jets 共AWSJ兲 that
involve injection of premixed slurry through the nozzle. Because
of the wear problem, essentially all the commercial jet cutting
systems are based on this principle; 共b兲 Use of nozzles made of
very hard materials, such as diamond and boron carbide 共Dubensky et al. 关2兴, Hollinger and Mannheimer 关8兴, Miller 关9兴兲; 共c兲
Keeping the particles softer than the nozzle walls 共Dubensky et al.
关2兴, Mort 关4兴兲; 共d兲 Attempts to modify the flow structure in order
to keep the particles away from the wall 共Horii et al. 关10兴, Okita
et al. 关11兴兲. All the presently available means have major deficiencies. In AWJ 共seeding downstream of the jet兲 the particles are not
accelerated to levels that are close to that of the liquid velocity,
and, hence, requiring substantially higher pressures to achieve the
same cutting effect 共Hashish et al. 关5兴, Hashish 关6兴, Momber and
Kovacevic 关7兴兲. The process also causes considerable expansion,
scattering and unsteadiness 共Hollinger and Mannheimer 关8兴兲. Furthermore, even in AWJ systems, wear of the mixing tube is also a
serious problem 共Hashish 关12兴, Nanduri et al. 关13兴兲. Modification
to the jet flow structure by introducing secondary swirling flows
near the nozzle walls is useful only in relatively slow flows and
small particles. It also causes jet expansion and secondary flow
phenomena that limit the capability to control the process. Diamond nozzles are expensive and difficult to form into desirable
shapes. Using particles softer than the nozzle walls reduces their
cutting effectiveness.
Thus, a solution to the wear problem must still be found. It may
enable us to increase the jet speed, and reduce its diameter even
further 共present sizes range between 100–500 ␮m兲, allowing
much higher precision, deeper cutting, and wider implementation
in problematic materials including ceramics. The present paper
introduces such a solution.
Copyright © 2003 by ASME
Transactions of the ASME
Fig. 1 A Sketch illustrating the principles of the method for
preventing nozzle wear
2
The Lubricated Porous Nozzle
The proposed solution to solve the wear problem is sketched in
Fig. 1 共Katz 关14兴兲. The nozzle is made of a porous material and is
surrounded with a reservoir containing a high viscosity lubricant
that is exposed to the same pressure that drives the flow in the
nozzle. The lubricant is forced continuously through the porous
medium as a result of the pressure difference created due to the
high-speed flow in the nozzle. The lubricant injection rate, which
is controlled by the pressure difference, the nozzle geometry
共thickness兲, permeability of the porous medium and lubricant viscosity, is designed to create a thin layer 共film兲, with a typical
thickness of 5 ␮m, on the walls of the nozzle. This film of high
viscosity fluid protects the walls of the nozzle from the shear and
impact of the abrasive particles. Since the lubricant is constantly
replenished, sites where particles ‘‘gouge’’ the film are repaired,
preventing damage to the solid walls. Provided that the proper
lubricant 共viscosity兲, film thickness and nozzle geometry 共flow
rate through the porous medium兲 are selected, this approach provides a reliable but yet very simple method to prevent nozzle
wear. Due to the differences in viscosity between the lubricant and
the water 共can be as high as 4000:1兲, the oil consumption is minimal, typically about 1 percent of the water flow rate.
The idea of using a porous nozzle has been introduced before
by Tan and Davidson 关15兴 and Tan 关16 –18兴. They used a fluidized
sand bed as a source of the abrasive slurry as well as a source of
the water forced through the porous nozzles to lubricate them, i.e.,
the nozzle was exposed to the same slurry on both sides. Their
experiments were performed at low pressures of 1–2 MPa, i.e., at
low velocities, and as a result did not address the wear problem
under relevant conditions. As demonstrated in this paper, water
does not have sufficiently high viscosity to prevent wear. In fact,
lubricants with viscosities that were three orders of magnitude
higher than that of water were essential. Furthermore, the forcing
of liquid containing particles through the porous medium would
quickly clog it due to the high-pressure difference across the
nozzle.
3
Experimental Setup
The experiments were performed using the supply system of
lubricant and abrasive particles illustrated in Fig. 2. The filtered
共1-micron兲 tap water was pressurized using a 7.5 kW, positive
displacement pump with maximum pressure of 69 MPa and maximum flow rate of 9.5⫻10⫺5 m3 /s. We typically operated at pressures of up to 34.5 MPa. A regulating valve at the exit from the
pump was used to control the flow rate, which was measured with
Journal of Tribology
Fig. 2 Schematic of the system supplying slurry particles and
lubricant to the test chamber
a turbine flow meter 共Hoffer Flow Controls Model HO兲. Based on
the manufacturer’s specifications and calibration, the measurement uncertainty was less than 1 percent. A pressure gauge 共PSITRONIX Model PG5000兲 attached in the main line monitored the
pressure upstream of the nozzle. Based on the manufacturer’s
specifications and calibration, the measurement uncertainty was
less than 1 percent.
The water also pressurized the chambers containing the lubricant and abrasive slurry. The slurry chamber contained a concentrated mixture of slurry particles and water. During the experiments part of the water was injected into this chamber from
below, which entrained some of the particles, then flowed out
from the top of the chamber, and mixed with the main stream.
Injection from below was necessary since the particles were
heavier than water and tended to settle. A perforated hemispherical cap was placed at the inlet of the slurry chamber. The resulting
small jets mixed the slurry in the chamber, and prevented blockage of the inlet when the chamber was loaded with particles.
A small loosely fitted piston separated the water from the lubricant in the oil chamber. This piston ensured that the lubricant and
the water would not mix and form an emulsion. There is substantial evidence that the permeability of porous media is reduced due
to the transport of emulsions through them 共McAuliffe 关19兴, Soo
and Radke 关20,21兴兲. The oil line upstream of the nozzle contained
a 2-micron filter to remove any dirt from the lubricant. The presence of this element was critical. Experiments performed without
this filter resulted in the clogging of the porous medium. Figure
3共a兲 shows Scanning Electron Microscope 共SEM兲 images of the
clogged porous surface resulting from experiments with no filter
and the same surface with embedded dirt after the surface coating
was removed. Figure 3共b兲 shows a SEM image of the porous
surface recorded after the experiments with the filter installed. As
is evident, the nozzle surface remained free from blockage.
We constructed two types of nozzles and the housings to support them: A two-dimensional nozzle made of porous stainless
JANUARY 2003, Vol. 125 Õ 169
Fig. 3 SEM images of „a… the clogged porous surface when a
filter was not used, showing a coating of dirt on the surface and
dirt inside the pores, and „b… the porous surface recorded after
the experiments with a filter installed, free from blockage
3.1 The Two-Dimensional Nozzle Housing. The interior of
the two-dimensional nozzle housing is shown in Fig. 4 and the
nozzle is illustrated in Fig. 5共a兲 and 共b兲. The nozzle consisted of
two 1.57 mm thick, symmetric 共mirror image兲 inserts/sections
made of porous, 316-stainless steel. The slurry flowed in the narrow gap between these two sections and the grooves served as oil
reservoirs. The two porous sections were inserted in a housing,
with a matched slot and openings for oil and slurry, as shown in
Fig. 5共a兲. The nozzle geometry consisted of a 4.83 mm long quarter ellipse followed by a 1.52 mm long straight portion with
smooth transition between them. The length of the straight section
was chosen to accelerate the abrasive particles to nearly the liquid
velocity near the nozzle exit 共analysis follows兲. The porous sections were manufactured in our laboratory using Electric Discharge Machining 共EDM兲. The shape of the oil reservoir and the
nozzle geometry determined the thickness of the porous layer
separating the nozzle from the oil reservoir, which in turn influenced the oil injection rate into the nozzle.
As shown in Fig. 4, the oil entered the oil reservoirs through the
oil ports 共Fig. 5共a兲兲 and flowed into the nozzle through the porous
layer. The slurry entered through the upper port. The porous inserts were covered with metal shims of matched shapes on both
sides, to prevent leakage of oil and water between the surfaces,
and were then inserted inside the main body. They were pressed
on both sides by sapphire windows. The gap between the porous
inserts varied for each nozzle set due to their compression by the
windows. The actual gap 共nozzle width兲 for the present nozzle
was measured to be 0.145 mm. One window covered the entire
porous section and the other covered only the converging and
high-speed sections of the nozzle. This arrangement allowed direct observations of the flow, film layer and particle trajectories
inside the nozzle.
steel with windows on both sides was used to visualize the flow
and the oil film and to measure the liquid and abrasive-particle
velocities in the nozzle. An axisymmetric nozzle was used to determine the extent of nozzle wear and investigate the effect of
lubricant properties and flow rate.
3.2 The Axisymmetric Nozzle Housing. Figure 6共a兲 shows
the components of the axisymmetric nozzle setup and Fig. 6共b兲
shows a cross-section of the porous nozzle. The nozzle consisted
of a short converging section of quarter-circular shape followed
by a straight section. We were forced to use a shorter nozzle due
to limitations in the ability to manufacture the nozzle precisely
without smearing the porosity of the interior walls. This design
Fig. 4 Components of the two-dimensional nozzle housing
170 Õ Vol. 125, JANUARY 2003
Transactions of the ASME
Fig. 5 „a… The shape of the porous inserts and the slot in which they were inserted in the
two-dimensional housing, and „b… geometry of the two-dimensional nozzle. All dimensions are
in mm.
also allowed us to vary the outside diameter of the nozzles, which
determined the thickness of the porous medium separating the
nozzle from the oil reservoir. Hence, by changing the thickness we
could vary the oil flow rate 共details follow兲. The oil entered
through the oil port and collected in the reservoir surrounding the
porous nozzle 共Fig. 6共a兲兲. It then flowed through the porous me-
dium due to the pressure difference, to create a thin film on the
nozzle walls. The abrasive slurry entered through the upper port.
Two sets of copper shims, on both sides of the nozzle were used
as seals.
The nozzles were made of porous, 316 stainless steel and machined using EDM. The EDM machining parameters required to
Fig. 6 „a… Components of the axisymmetric nozzle housing, and „b… cross-section of the porous axisymmetric nozzle. All
dimensions are in mm.
Journal of Tribology
JANUARY 2003, Vol. 125 Õ 171
Fig. 7 SEM micrograph of the top of the axisymmetric nozzle
maintain the porosity have been established and verified by observations using SEM. Figures 7 and 8 show SEM images of the top
and cut section of the porous nozzle, respectively. The quality of
the surface varied substantially, depending on the method used for
manufacturing the porous material, and the EDM machining parameters, such as energy level, spark frequency and cutting speed.
We experimented with different materials and machining parameters in order to maintain a uniform pore distribution and prevent
the ‘‘smearing’’ of the pores on the surface of the nozzle. As
illustrated in Fig. 9, decreasing the cutting speed and energy level
Fig. 9 Effect of EDM machining parameters on the porous surface. SEM image: Top to bottom—improvements by decreasing
the cutting speed and energy level.
improved the surface quality. The machining parameters were also
adjusted for preventing oil flow in the undesired regions of the
nozzle.
4
Fig. 8 SEM micrograph of the cut section of the axisymmetric
nozzle
172 Õ Vol. 125, JANUARY 2003
Velocity Measurements
A schematic of the data acquisition setup is shown in Fig. 10.
Due to the high velocities in the nozzle 共⬎150 m/s兲 the exposure
time had to be very short, in the order of nanoseconds, to avoid
blurred images. Consequently, we used a dual head Nd-YAG laser
with pulse duration of 5 ns as a light source. However, due to the
laser coherence, direct illumination of the nozzle generated undesired interference patterns that obscured the image. As a result,
when we needed uniform background illumination 共for PIV applications兲, a glass container with an emulsion of oil and water containing fluorescent dye-Pyromethene 597 dissolved in the oil, was
inserted between the laser and the window. The laser excited the
dye and caused bright fluorescence at a broad range of wavelengths. This method effectively created a 5 ns flash of uniform
illumination. The two independent laser heads enabled us to generate 15-pulse pairs/sec with an essentially unlimited in-pair delay, as low as 100 ns. Particle Image Velocimetry 共PIV兲 was used
for measuring the velocity of the particles and the liquid inside the
nozzle. The images were recorded using a 12 bit, 4 frames per
second, 2048⫻2048 pixels digital camera 共SMD-4M4兲. All the
observations were performed using a long working distance 共50.8
mm兲 microscope objective with a resolution of 1.75 ␮m, manufactured by Infinity Photo-Optical Company. The magnification
was 0.695 ␮m per pixel. We recorded silhouette photographs, i.e.,
the camera faced the light source and as a result the oil film and
the particles appeared as shadows. The camera and the laser were
synchronized. The time separation between pulses, varying between 150 ns at the end of the nozzle to 1.5 ␮s at the top of the
nozzle, was too short for relying on the laser electronics. Consequently, the exact timing was measured using a photodiode and a
500 MHz, 1GS/s oscilloscope, and then recorded by the computer
using a GPIB board.
Sample images of the straight section of the nozzle recorded
using the fluorescent bulb with and without oil injection are presented in Fig. 11. In Fig. 11共b兲 the oil layer creates dark patterns
with a bright background. Figure 11共c兲, obtained by subtracting
Fig. 11共b兲 from Fig. 11共a兲, clearly shows the formation of the oil
layers on the two walls. The oil used in these experiments had a
Transactions of the ASME
Fig. 10 Schematic of the setup for data acquisition
Fig. 11 „a… The nozzle with water only, „b… with water and oil,
and „c… the oil layers on the walls of the nozzle. The flow was
from top to bottom. The section shown is 4.42Ë x Ë5.84 mm
and 0.145 mm wide. For definition of x , see Fig. 5„b….
Journal of Tribology
viscosity of 1800 mm2/s 共at 25°C兲. In spite of this high viscosity,
the high shear rates in the nozzle caused considerable entrainment,
as can be observed from the protrusions and eddy-like structures
in Fig. 11共b兲 and 共c兲. However, the typical flow rate of lubricant in
the two-dimensional facility was still very low, below 1 percent of
the flow rate of water. Note that the characteristic Reynolds number of this flow was 22,000, based on the nozzle exit diameter.
Without oil, transition to a turbulent boundary layer would have
been triggered immediately due to the wall roughness, and even
with the oil, the eddy-like structures indicate that the flow was at
least transitional.
The same type of ‘‘fluorescent bulb’’ illumination was used
while observing the motion of the abrasive 共slurry兲 particles and
the 共almost兲 neutrally buoyant tracer particles used for measuring
the liquid velocity. As mentioned before, velocity measurements
were performed using PIV 共Adrian 关22兴兲 using software and procedures developed in our laboratory 共e.g., Roth and Katz 关23兴,
Sinha and Katz 关24兴兲. This method is based on seeding the flow
with microscopic tracer particles and recording double exposure
images on the same or on separate frames. The measured particle
displacements and the known time delay between exposures are
used for determining the velocity. In the present experiments, the
tracers were 4 ␮m diameter 共Std. dev.–1.5 ␮m兲 spherical nylon
particles that had a specific gravity of 1.14. As shown in Fig.
12共a兲 and 共b兲 and Fig. 13, the double-exposure images were recorded on the same frame due to the short delay between exposures 共⬍1 ␮s兲. The images were enhanced using an in-house enhancement program based on the histogram equalization
algorithm, and the velocity was calculated using an autocorrelation code 共Roth and Katz 关23兴兲.
In order to protect the sapphire windows from being damaged
due to the impact of the abrasive particles during the visualization
experiments in the two-dimensional nozzle, we used 20– 45 ␮m
Celestite 共Strontium Sulphate, a naturally occurring mineral兲 as
slurry particles instead of the typical industry standard of Garnet.
Celestite has a Mohs Hardness of 3–3.5, much lower than the
Garnet’s 7–7.5, but both have almost the same specific gravity,
3.95 versus 4.0, respectively. The Celestite particles also had the
same characteristic shape as the Garnet particles, and, hence, they
had similar hydrodynamic behavior. Due to the large difference in
size between the liquid tracers and the slurry particles, they could
easily be distinguished. The large particles were removed from the
image before calculating the liquid velocity. The velocity of the
slurry particles was measured separately, also using autocorrelation analysis, but also subtraction of enhanced edges of the
particles. Sample images of the slurry and tracer particles in the
nozzle are presented in Fig. 14共a兲–共c兲.
Figure 15 shows the measured centerline velocity of the liquid
JANUARY 2003, Vol. 125 Õ 173
Fig. 13 The converging section of the nozzle with enhanced
double exposure image of the tracers. The flow was from top to
bottom. The section shown is 0.84Ë x Ë2.26 mm. For definition
of x , see Fig. 5„b….
the straight section of the nozzle. Near the nozzle exit, the relative
velocity decreased to negligible levels, for example to 1.88 m/s,
i.e., 1.2 percent of the local liquid velocity, at x⫽5.94 mm.
Fig. 12 „a… The nozzle with water and tracer particles, and „b…
enhanced double exposure image of the tracers. The flow was
from top to bottom. For location, see Fig. 11.
and its standard deviation for a pressure upstream of the nozzle of
14.48 MPa, with and without lubrication. As can be observed,
injection of oil caused virtually no change in the centerline liquid
velocity. Note that the centerline liquid velocity estimated using
the Bernoulli equation would be 170.2 m/s. For each section of
the nozzle, we used 24 instantaneous realizations to calculate the
average velocity. From Dong et al. 关25兴 and Roth et al. 关26兴 the
sub-pixel accuracy in velocity measurement using auto-correlation
analysis was about 0.3 pixels 共the standard deviation between
measured and exact results was 0.2 pixels兲, and depended mostly
on the number of particles per window. Since the typical particle
displacement for the PIV images was 30 pixels, the uncertainty in
liquid velocity measurements was about 1 percent. This number
was reduced further after averaging. The uncertainty in the displacement of individual large particles could be maintained at a
similar level provided they were larger 共⭓20 pixels, according to
Sridhar and Katz 关27兴兲. These conditions were satisfied in the
present measurements.
Using a sample of 103 slurry particles, Fig. 16 shows the measured slip velocity of slurry particles, i.e., vជ l ⫺ vជ p 共the average
liquid velocity minus the slurry particle velocity兲, in the last 1.93
mm of the nozzle and illustrates the decrease in slip velocity along
174 Õ Vol. 125, JANUARY 2003
5
Numerical Analysis of Particle Slip
In order to compare the measured slip velocity to expected
levels, we performed a simple numerical analysis of the velocity
of spherical particles using Eq. 共1兲 that accounted for inertia, virtual mass, pressure gradients and drag forces 共Maxey and Riley
关28兴, Sridhar and Katz 关27,29兴兲. Since the pressure gradients associated with the nozzle geometry were five orders of magnitude
higher than the buoyancy forces, we neglected the latter. We also
neglected the lift forces, as we were interested in the streamwise
motion. Based on the assumption of spherical particles we used a
virtual mass coefficient of 0.5.
冋
D ␳P
d ␯ជ p
dt
⫹
1
2
␳l
冉
d¯␯ p
d ṫ
⫺
d ␯ជ l
dt
冊册
⫽⫺
3
4
␳ l 兩 ␯ជ p ⫺ ␯ជ l 兩 共 ␯ជ p ⫺ ␯ជ l 兲 C d
⫺ⵜpD
(1)
here, v is the velocity, C d is the drag coefficient, ␳ and ␮ are the
density and viscosity, respectively, p is the pressure, D is the diameter of the particle and the subscripts p and 1 refer to the
particle and liquid, respectively. The drag coefficient, presented in
Eq. 共2兲, was based on an empirical expression available in Clift
et al. 关30兴 for Re⬍3⫻105 共the present range was less than 103 兲
C d⫽
24
0.42
关 1⫹0.15 Re0.687兴 ⫹
Re
共 1⫹4.25⫻104 Re⫺1.16兲
(2)
Transactions of the ASME
Fig. 15 The centerline liquid velocity in the two-dimensional
nozzle measured using PIV, with and without injection of oil.
The velocities were obtained using 135Ã24 measurements for
each case. Every 5th data point, representing an average of 24
measurements, is shown for clarity. The error bars represent
the standard deviation values at these points. For dimensions,
see Fig. 5„b….
Fig. 14 „a – c… show three samples of double exposure images
of the nozzle with water and oil along with slurry „large dark
objects… and tracer particles „small dark objects…. The flow was
from top to bottom. For location, see Fig. 11.
where Re⫽
␳ l 兩 ␯ជ l ⫺ ␯ជ P 兩 D
␮
We also assumed a steady liquid flow, i.e.,
being spherical, i.e., they had larger drag coefficients and their
virtual mass coefficient was different than 0.5. Also, the assumed
one-way coupling was doubtful due to the size of the particles
compared to the width of the nozzle. Analysis 共results not shown兲
performed to identify the required drag that would match the computed slip velocities to the experimental values indicated that the
C d would have to be tripled. Such drag coefficients are consistent
with the published data on non-spherical particles 共Clift et al.
关30兴, Haider and Levenspiel 关31兴兲.
Most of the slurry particles appeared to be moving in the center
of the nozzle as the sample images in Fig. 14 show. However, as
shown at the left side of Fig. 18, in some cases the slurry particles
⳵␯ជ l
⳵␯ជ l
1 ⳵p
⫽ ␯ជ l
⫽⫺
⳵t
⳵x
␳l ⳵x
The numerical analysis was performed for several particle diameters assuming one-way coupling, i.e., neglecting the effect of
the particle motion on the liquid flow 共although the particle size
compared to the nozzle diameter made the validity of this assumption doubtful兲. We also assumed that the particle traveled along
the centerline of the nozzle. A 10th order polynomial was fitted to
the measured centerline liquid velocity with oil injection and this
curve was used as an input to the numerical calculations. An
Adams–Bashforth, third order scheme was used to march forward
in time and we used time steps of 0.05 ␮s. By varying the time
steps and repeating the calculations we verified that a step of 0.05
␮s was short enough not to have an effect on the results.
Figure 17共a兲 compares the 10th order curve fit to the measured
liquid velocity with the computed velocity of a 25 ␮m slurry
particle. Figure 17共b兲 shows the computed slip velocities for several particle diameters. At the entrance to the straight section the
computed slip velocity of a 25 ␮m slurry particle 共⬃23 m/s兲, was
higher by 50 percent than the measured value 共⬃15 m/s兲. Near the
exit the measured slip velocities 共⬃2– 4 m/s兲 were substantially
lower than the computed values 共⬃13 m/s兲. However, if the comparison had been performed using the results for a 35 ␮m slurry
particle, the discrepancy would have increased to 90 percent and
400 percent at the entrance and exit from the straight section,
respectively. This discrepancy could have been a result of several
factors, such as the fact that the real slurry particles were far from
Journal of Tribology
Fig. 16 Measured slip velocity of the slurry particles with
nominal diameter of 20–45 ␮m close to the exit from the nozzle
„4.42Ë x Ë6.35 mm… and the linear least squares fit to the data.
The standard deviation for all the values is 5.2 mÕs. For definition of x , see Fig. 5„b….
JANUARY 2003, Vol. 125 Õ 175
Fig. 17 „a… The 10th order curve fit to the measured liquid velocity and the computed velocity of a 25 ␮m spherical slurry
particle based on Eq. „1…; and „b… the computed slip velocities
at the centerline of the nozzle for several diameters of slurry
particles. For dimensions, see Fig. 5„b….
gouged the oil layer. We found that in such cases, the oil layer
quickly 共subsequent image兲 replenished itself and maintained its
integrity.
6
Fig. 18 Slurry particles impinging the nozzle walls. The flow
was from top to bottom. The section shown is 4.42 mmË x
Ë5.03 mm. For definition of x , see Fig. 5„b….
Wear Tests Using the Axisymmetric Nozzle
Garnet particles with nominal size of 25 ␮m were used as abrasive for all the wear experiments. The slurry concentration inside
the slurry chamber was 4.44⫻10⫺3 g/cm3 . In all the present
cases, the upstream pressure was 14.48 MPa 共same liquid velocity兲 and the run-time was 1 hour and 45 min. Oils with three
different viscosities, 460 mm2/s, 1800 mm2/s, and 4000 mm2/s 共at
25°C兲 were used as lubricants. Figure 19 shows the time required
to empty the 125 cm3 oil reservoir as a function of viscosity. Also
shown is the flow rate of oil relative to that of water, R 共oil flow
rate ratio兲. As is evident, the time required to empty the chamber
varied linearly with viscosity, in agreement with the Darcy’s Law
176 Õ Vol. 125, JANUARY 2003
共Eq. 共3兲兲. Accordingly, the relative oil flow rate curve exhibited a
hyperbolic behavior. Note that these results included slight leakage of oil around the porous nozzle.
According to Darcy’s law
␯ជ o ⫽⫺
1
¯P
K̄•ⵜ
␮o
(3)
Here, v̄o and ␮ o are the oil velocity and viscosity, respectively, K̄ is
¯ P is the
the permeability tensor of the porous medium, and ⵜ
pressure gradient within the porous medium.
Transactions of the ASME
Fig. 19 Time taken to empty the 125 cm3 reservoir for three
different oils
Integrating Eq. 共3兲 for a cylindrical annular flow in a medium
with uniform properties, the oil flow rate, Q o , is given by
Q o⫽
2 ␲ KL⌬ P
ro
␮ o ln
ri
冉冊
(4)
where ⌬ P is the pressure drop across the nozzle, L is the height of
the nozzle, and r o and r i are the external and internal radii of the
axisymmetric nozzle, respectively. To examine the effect of viscosity on the wear at similar oil flow rates, we used nozzles of
three different external diameters, 5.08 mm, 3.81 mm, and 2.54
mm, keeping the internal diameter the same 共about 200 ␮m兲.
However, being inversely proportional to ln(ro /ri), the flow rate
increased only by 27 percent when the external diameter was
halved.
A series of images of the nozzle exit, prior to and after twentyone runs 共five minutes each兲 are presented in Fig. 20–21. As a
reference case, Fig. 20 shows SEM images of the nozzle exit
taken before and after a test with a non-lubricated nozzle (R
⫽0). The nozzle exit diameter increased from an initial size of
202 ␮m to 426 ␮m based on the diameter of a circle with an
equivalent area, i.e., a change of 111 percent. Figure 21 shows
sample images of the nozzle exit taken before and after experiments with oil injection but at different conditions. For the nozzle
in Fig. 21共a兲, ␮ o ⫽1800 mm2 /s and R⫽0.024. The nozzle exit
diameter changed from an initial size of 202 ␮m to 210 ␮m, i.e.
an increase of 4 percent. For the nozzle shown in Fig. 21共b兲, the
oil flow rate ratio was reduced to R⫽0.014, and consequently, the
diameter changed from an initial size of 210 ␮m to 232 ␮m, i.e.
an increase of 10.5 percent. Figure 21共c兲 shows sample images for
Fig. 20 The reference case: SEM images of the nozzle exit
recorded „a… before and „b… after 105 minutes of test using abrasive slurry but without any oil injection
Journal of Tribology
Fig. 21 SEM images of the nozzle exit recorded before and
after 105 minutes of test with „a… oil viscosity ␮ o
Ä1800 mm2 Õs, and flow rate ratio R Ä0.024; „b… ␮ o
Ä1800 mm2 Õs, R Ä0.014; and „c… ␮ o Ä460 mm2 Õs, R Ä0.041.
␮ o ⫽460 mm2 /s and R⫽0.041 共for a lower viscosity oil兲. The
nozzle exit diameter changed from 208 ␮m to 248 ␮m, i.e., an
increase of 19.5 percent. The uncertainty in the measurement of
wear results was less than 1 percent.
Figures 22共a兲 and 共b兲 show SEM images of the cut sections,
after the experiment, of the nozzles shown in Fig. 20 共111 percent
wear兲 and Fig. 21共a兲 共4 percent wear兲, respectively. Figures 23共a兲
and 共b兲 show the top of the same nozzles after the experiment,
respectively. Comparing them with Figs. 7 and 8, it is evident that
besides being larger, the nozzle with 111 percent wear had substantial wear grooves and ridges, and the interior porous surface
was smeared. The wear of this nozzle was also quite asymmetric,
both in the interior walls and on the top 共Fig. 23共a兲兲. Conversely,
the images of the nozzle with 4 percent wear demonstrate that the
wear was essentially symmetric and very small. The porous surface on the internal wall of this nozzle remained quite similar to
that of a new nozzle, although some wear could be seen close to
the entrance of the nozzle.
Table 1 summarizes the different experiments conducted with
axisymmetric nozzles. The results are also presented graphically
as a function of the oil flow rate ratio and viscosity in Fig. 24.
Each point represents the overall wear after twenty-one identical
runs, each lasting five minutes, and each performed with a fresh
load of slurry and lubricant. The slurry load was prepared by
JANUARY 2003, Vol. 125 Õ 177
Table 1 A listing of the different experiments conducted on
the axisymmetric nozzles
Fig. 22 SEM images of the cut section of the nozzles recorded
after 105 minutes of test: „a… using abrasive slurry but without
oil injection „nozzle of Fig. 20…, and „b… using abrasive slurry
and injection of oil with ␮ o Ä1800 mm2 Õs and R Ä0.024 „nozzle
of Fig. 21„a…….
Fig. 23 SEM images of the top of the nozzles recorded after
105 minutes of test „a… without any oil injection „nozzle of Fig.
20…; and „b… ␮ o Ä1800 mm2 Õs and R Ä0.024 „nozzle of Fig.
21„a…….
178 Õ Vol. 125, JANUARY 2003
measuring a fixed quantity of abrasive particles using a high precision balance and then mixing them with a measured volume of
water. During the experiments the small water jets at the bottom
of the chamber helped in maintaining a well-mixed slurry. The
average particle concentration in the nozzle was 4.44
⫻10⫺3 g/cm3 and most likely decreased slightly during each run.
It is evident that the presence of an oil film on the nozzle walls
had a substantial impact on the extent of nozzle wear. Both the
lubricant flow rate and the viscosity of the oil were important
parameters affecting the extent of the wear. For the same oil viscosity, the wear increased as the lubricant injection rate decreased.
However, as the data for ␮ o ⫽460 mm2 /s suggested, the wear
seemed to reach a plateau that depended on the viscosity. This
trend may be a result of reaching a phase where adding more oil,
instead of increasing the film thickness just increased the amount
of oil being entrained into the stream. This issue will be verified in
future observations using the two-dimensional nozzle. Lowering
the oil viscosity for the same flow rate ratio caused a substantial
increase in the nozzle wear. Unfortunately, the present combination of nozzle size, permeability and pressures prevented us from
Fig. 24 The effect of oil viscosity and flow rate on the increase
in effective nozzle diameter due to wear. Without oil the wear
was 111 percent. Each point represents the overall wear after
twenty-one identical runs, each lasting five minutes, and each
performed with a fresh load of slurry and lubricant.
Transactions of the ASME
increasing the flow rate ratio of the ␮ o ⫽4000 mm2 /s oil beyond 1
percent. The trends, however, suggest that a 4 percent wear could
be achieved at R⬃1.5 percent, and that at R⫽2.5 percent the wear
may be reduced to below 2 percent. These statements are at this
stage only speculations.
We had also observed that as long as oil injection occurred in
the nozzle, the jet stream from the exit of the nozzle was coherent
and well defined. Once the oil injection stopped, the jet broke into
droplets and spread immediately at the exit of the nozzle. This
effect could be attributed to the smoothening of the jet walls by
the presence of the oil layer.
7
Conclusions
This paper introduces a novel solution for preventing nozzle
wear in Abrasive Water Suspension Jets 共AWSJ兲 used for jet cutting. The nozzle was made of porous material and was surrounded
by a reservoir containing a high viscosity lubricant. The lubricant
reservoir was exposed to the same pressure that drove the flow in
the nozzle. The pressure difference created due to the high-speed
flow in the nozzle, continuously forced the lubricant through the
porous medium, resulting in the formation of a thin film of high
viscosity fluid on the interior walls of the nozzle. This lubricant
film protected the walls of the nozzle from the abrasive particles,
and substantially reduced the extent of nozzle wear.
A facility with a two-dimensional nozzle with windows on both
sides was used for observations of the lubricant film and for measuring the velocity of liquid and slurry particles in the nozzle. In
spite of the high oil viscosity, the high shear rates in the nozzle
caused considerable entrainment. However, the typical flow rate
ratio of lubricant in the two-dimensional nozzle was still below 1
percent. When the particles gouged the oil layer, it immediately
replenished itself and maintained its integrity. The velocity measurements showed that the centerline liquid velocity was not affected significantly by the injection of oil. The measured slip velocity decreased along the straight section of the nozzle. In fact,
near the exit the measured slip velocity decreased to less than 2
percent of the local liquid velocity, i.e., to a negligible level. The
measured velocities of the slurry particles relative to that of the
liquid, i.e. the slip velocities, were also compared to the computed
values from a simple numerical model that assumed one-way coupling and spherical particles. The measured slip velocities and the
computed values showed discrepancies, which may be attributed
to the difference between the assumed and actual drag and virtual
mass coefficients 共the slurry particles were far from being spherical兲. Assuming one-way coupling when the particle size was
14 –31 percent of the nozzle diameter was also questionable at
best.
Tests were also conducted using axisymmetric nozzles to determine the extent of nozzle wear and investigate the effects of lubricant viscosity and flow rate. It was found that the presence of
oil substantially reduced the wear of the nozzle walls, from 111
percent of the diameter to 4 percent, our best result to-date, over
the same period. For this case the oil flow rate ratio was only 2.4
percent. The wear increased as the lubricant flow rate and viscosity decreased. However, the tests indicated that increasing the oil
flow rate beyond a certain level had a diminishing effect on the
wear. Thus, increasing the viscosity promised to be a better approach for future improvements. The presence of the oil film also
improved the coherence of the jet.
This paper clearly demonstrates that the porous lubricated
nozzles can substantially reduce the extent of nozzle wear of abrasive water suspension jets. Once several issues associated with
commercializing this technology are resolved, it may expand the
use and applications of high-speed abrasive waterjet cutters. Being able to accelerate the particles to nearly the liquid velocity
with minimal damage to the nozzle, even when the nozzle is made
of plain stainless steel, is a substantial improvement over other
presently used techniques. Compared to the present commercial
abrasive water jet 共AWJ兲 cutters, the smaller jet diameter and the
Journal of Tribology
lower pressure required to achieve the same cutting effect, may
result in cost savings, higher cutting efficiency and more precise
cutting. A more durable nozzle may also enable further reduction
in nozzle diameter, hence, even greater cutting precision, and
higher particle speeds that may lead to deeper cutting.
Acknowledgments
We are grateful to Andy Conn, first for introducing us to the
problem of nozzle wear, and then for his continued advice during
the course of this project. Yury Ronzhes and Steve King provided
engineering support. Seed funding that enabled us to buy the
pump and some of the equipment was provided by National Science Foundation under Grant No. 9320153. We would like to
thank Jet Edge Corp., USA for providing graduate student support
for one year.
References
关1兴 Conn, A. F., 1991, ‘‘A Review of the 10th International Symposium of Jet
Cutting Technology,’’ International Journal of Water Jet Technology, 1共3兲, pp.
135–149.
关2兴 Dubensky, E., Groves, K., Gulau, A., Howard, K., and Mort, G., 1992, ‘‘Hard
Ceramics for Long Life Abrasive Water Jet Nozzles,’’ Proceedings of the 11th
International Conference on Jet Cutting Technology, St. Andrews, Scotland,
Sept. 8 –10.
关3兴 Kovacevic, R., and Evizi, M., 1990, ‘‘Nozzle Wear Detection in Abrasive
Waterjet Cutting Systems,’’ Mater. Eval., 48, pp. 348 –353.
关4兴 Mort, G. A., 1991, ‘‘Long Life Abrasive Water Jet Nozzles and Their Effect on
AWJ Cutting,’’ Proceedings of the 6th American Water Jet Conference, Houston, Texas, August 24 –27, pp. 315–344.
关5兴 Hashish, M. A., Kirby, M. J., and Pao Y. H., 1987, ‘‘Method and Apparatus for
Forming a High Velocity Liquid Abrasive Jet,’’ United States Patent No.
4,648,215.
关6兴 Hashish, M., 1990, ‘‘Abrasive-Fluidjet Machinery Systems: Entrainment Versus Direct Pumping,’’ Proceedings of the 10th International Symposium on Jet
Cutting Technology, Amsterdam, The Netherlands, Oct. 31–Nov. 2, pp. 99–
114.
关7兴 Momber, A. W., and Kovacevic, R., 1998, Principles of Abrasive Water Jet
Machining, Springer-Verlag, New York.
关8兴 Hollinger, R. H., and Mannheimer, R. J., 1991, ‘‘Rheological Investigation of
the Abrasive Suspension Jet,’’ Proceedings of the 6th American Water Jet
Conference, Houston, Texas, August 24 –27, pp. 515–528.
关9兴 Miller, D. S., 2001, ‘‘Micro Abrasive Waterjet Cutting,’’ Proceedings of the
11th American Water Jet Conference, Minneapolis, Minnesota, August 18 –21,
pp. 59– 69.
关10兴 Horii, K., Matsumae, Y., Cheng, X., Takei, M., Hashimoto, B., and Kim, T. J.,
1990, ‘‘Developments of a New Mixing Nozzle Assembly for High Pressure
Abrasive Water Jet Applications,’’ Proceedings of the 10th International Symposium on Jet Cutting Technology, Amsterdam, The Netherlands, Oct. 31–
Nov. 2, pp. 193–206.
关11兴 Okita, Y., Nakamura, K., Nishimoto, K., and Yanasaka, N., 2000, ‘‘Performance of Abrasive Water Suspension Jet Nozzles With a Single Annular Conduit,’’ Proceedings of the 6th Pacific Rim International Conference on Water
Jet Technology, Sydney, Australia, October 9–11, pp. 111–114.
关12兴 Hashish, M., 1994, ‘‘Observations of Wear of Abrasive-Waterjet Nozzle Materials,’’ ASME J. Tribol., 116, pp. 439– 444.
关13兴 Nanduri, M., Taggart, D. G., Kim, T. J., Ness, E., Haney, C. N., and Bartkowiak, C., 1996, ‘‘Wear Patterns in Abrasive Waterjet Nozzles,’’ Proceedings of
the 13th International Symposium on Jetting Technology, Sardinia, Italy, October 29–31, pp. 27– 45.
关14兴 Katz, J., 1999, ‘‘Lubricated High Speed Fluid Cutting Jet,’’ United States
Patent No. 5,921,846.
关15兴 Tan, R. B. H., and Davidson, J. F., 1990, ‘‘Cutting by Sand-Water Jets From a
Fluidised Bed,’’ Proceedings of the 10th International Symposium on Jet Cutting Technology, Amsterdam, The Netherlands, Oct. 31–Nov. 2, pp. 235–251.
关16兴 Tan, R. B. H., 1991, ‘‘Fluidised Abrasive Jets,’’ Proceedings of the First Asian
Conference on Recent Advances in Jetting Technology, Singapore, May 7– 8,
pp. 100–107.
关17兴 Tan, R. B. H., 1995, ‘‘A Model for Liquid-Particle Jet Flow in a Porous
Re-Entrant Nozzle,’’ International Journal of Water Jet Technology, 2共2兲, pp.
97–102.
关18兴 Tan, R. B. H., 1998, ‘‘Discharge of particle-liquid jets through porous converging nozzles,’’ International Journal of Water Jet Technology, 3共1兲, pp. 13–19.
关19兴 McAuliffe, C. D., 1973, ‘‘Oil-in-Water Emulsions and Their Flow Properties in
Porous Media,’’ J. Pet. Technol., 25, pp. 727–733.
关20兴 Soo, H., and Radke, C. J., 1984, ‘‘The Flow Mechanism of Dilute, Stable
Emulsions in Porous Media,’’ Ind. Eng. Chem. Fundam., 23, pp. 342–347.
关21兴 Soo, H., and Radke, C. J., 1986, ‘‘A Filtration Model for the Flow of Dilute
Stable, Emulsions in Porous Media—I. Theory,’’ Chem. Eng. Sci., 41共2兲, pp.
263–272.
JANUARY 2003, Vol. 125 Õ 179
关22兴 Adrian, R. J., 1991, ‘‘Particle-Imaging Techniques for Experimental Fluid Mechanics,’’ Annu. Rev. Fluid Mech., 23, pp. 261–304.
关23兴 Roth, G. I., and Katz, J., 2001, ‘‘Five Techniques for Increasing the Speed and
Accuracy of PIV Interrogation,’’ Meas. Sci. Technol., 12共3兲, pp. 238 –245.
关24兴 Sinha, M., and Katz, J., 2000, ‘‘Quantitative Visualization of The Flow in A
Centrifugal Pump with Diffuser Vanes, Part A: On Flow Structure and Turbulence; Part B: Addressing Passage Averaged and LES Modeling Issues in Turbomachinery Flows,’’ ASME J. Fluids Eng., 122, pp. 97–116.
关25兴 Dong, R., Chu, S., and Katz, J., 1992, ‘‘Quantitative Visualization of the Flow
Within the Volute of a Centrifugal Pump, Part A: Technique,’’ ASME J. Fluids
Eng., 114, pp. 390–395.
关26兴 Roth, G., Mascenik, D. T., and Katz, J., 1999, ‘‘Measurements of The Flow
180 Õ Vol. 125, JANUARY 2003
Structure And Turbulence Within A Ship Bow Wave,’’ Phys. Fluids, 11共11兲, pp.
3512–3523.
关27兴 Sridhar, G., and Katz, J., 1995, ‘‘Drag and Lift Forces on Microscopic Bubbles
Entrained by a Vortex,’’ Phys. Fluids, 7共2兲, pp. 389–399.
关28兴 Maxey, M. R., and Riley, J. J., 1983, ‘‘Equation of Motion for a Small Rigid
Sphere in a Nonuniform Flow,’’ Phys. Fluids, 26共4兲, pp. 883– 889.
关29兴 Sridhar, G., and Katz, J., 1999, ‘‘Effect of Entrained Bubbles on the Structure
of Vortex Rings,’’ J. Fluid Mech., 397, pp. 171–202.
关30兴 Clift, R., Grace, J. R., and Weber, M. E., 1978, Bubbles, Drops and Particles,
Academic Press, New York.
关31兴 Haider, A., and Levenspiel, O., 1989, ‘‘Drag Coefficient and Terminal Velocity
of Spherical and Nonspherical Particles,’’ Powder Technol., 58共1兲, pp. 63–70.
Transactions of the ASME