ASTM - STP 408 - Erosion by Cavitation or Impingement

EROSION BY CAVITATION
OR IMPINGEMENT
A symposium
presented at the
Sixty-ninth Annual Meeting
AMERICAN SOCIETY FOR
TESTING AND MATERIALS
Atlantic City, N. J., June 26-July 1,1966
ASTM SPECIAL TECHNICAL PUBLICATION NO. 408
List price $20.00; 30 per cent discount to members
published by the
AMERICAN SOCIETY FOR TESTING AND MATERIALS
1916 Race Street, Philadelphia, Pa. 19103
© BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1967
Library of Congress Catalog Card Number: 67-12411
NOTE
The Society is not responsible, as a body,

for the statements and opinions
advanced in this publication.
Printed in Baltimore, Md.

March, 1967
Foreword
The papers in the Symposium on Erosion by Cavitation or Impringe-

ment were presented at the Sixty-ninth Annual Meeting of ASTM held at
Atlantic City, N. J., June 26-July 1, 1966.
The symposium consists of ten papers and eight discussions, and
focuses mainly on the damage mechanism and the material response to
it. The symposium was sponsored by Committee G-2 on Erosion by
Cavitation or Impingement under the chairmanship of F. J. Heymann,
Westinghouse Electric Corp.
Related
ASTM Publications
Erosion and Captation, STP 307 (1961), $4.00

Contents
Introduction 1
A Test Rig for Studying Impingement and Cavitation Damage—
J. F. RIPKEN 3
Discussion 18
The Concept of Erosion Strength—A. THIRUVENGADAM 22
Discussion 36
Material Destruction Due to Liquid Impact—G. HOFF, G. LANGBEIN,
AND H. RIEGER 42
On the Time Dependence of the Rate of Erosion Due to Impingement
or Cavitation—F. j. HEYMANN 70
Discussion 100
Water Jet Impact Damage in a Cobalt-Chromium-Tungsten Alloy—
D. J. BECKWITH AND J. B. MARRIOTT Ill
Erosion of Steam Turbine Blade Shield Materials—ALLEN SMITH, R. P.
KENT, AND R. L. ARMSTRONG 125
Discussion 152
Experience With a 20-kc Cavitation Erosion Test—j. M. HOBBS 159
Discussion 180
Accelerated Cavitation Damage of Steels and Superalloys in Sodium
and Mercury—s. G. YOUNG AND j. R. JOHNSTON 186
Discussion 213
Scale-Effect Investigation of Cavitation Erosion Using the Energy
Parameter—K. K. SHALNEV, j. j. VARGA, AND G. SEBESTYEN 220
Discussion 236
Correlation of Cavitation Damage with Other Material and Fluid
Properties—R. GARCIA, F. G. HAMMITH, AND R. E. NYSTROM 239
Discussion 280
This page intentionally left blank
EROSION BY CAVITATION OR IMPINGEMENT
(With special emphasis on the interrelationships between
material properties and erosion damage)
Introduction
Five years ago, during the 1961 Annual Meeting, ASTM held its
first Symposium on Erosion and Cavitation, the proceedings of which
were published as STP 307. As a direct result of the interest stimulated
by that Symposium, Technical Committee G-2 was established, and the
1966 Symposium was the first to be held under its sponsorship. In 1961
six papers were presented; the 1966 Symposium heard eighteen presen-
tations, including six from abroad. Of these ten are contained in this
volume; some of the others will eventually appear in other ASTM pub-
lications.
There have been many symposia dealing with cavitation as a fluid
flow phenomenon as well as a damage-producing phenomenon. Curi-
ously though, there had been a notable lack of communication in this
country (though not in Europe) between those concerned with cavita-
tion damage and those concerned with liquid impingement damage.
The two ASTM Symposia are, to the best of my knowledge, the first in
this country to focus on the damage mechanism and the material response
to it, and thus on the common aspect of cavitation and impingement
attack—for it is now widely (though not universally) accepted that the
principal direct cause of damage in both instances is the mechanical
stressing due to the high-speed impact of a liquid surface upon a solid
surface, though corrosion can, certainly, enter the picture under ap-
propriate conditions and mechanical and chemical effects can reinforce
each other. The mechanical stress theory was proposed by Cook in 1928,
but over the years there have been many who doubted the possibility of
sufficiently high mechanical stresses and postulated principally chemical
mechanisms or a variety of other more fanciful mechanisms. I believe
that the days of wild conjecture are over and that investigators today
at least talk the same language and agree on the kind of quesions to ask.
This is not to say that all stimulating disputes have vanished: such still
exist, as for instance concerning the exact nature of the interaction be-
tween mechanical and corrosive effects, and the exact meaning of the
various phases observed in erosion rate-time histories.
i
2 EROSION BY CAVITATION OR IMPINGEMENT
ASTM is interested in the properties of materials and the ways of

defining, testing, prescribing, and controlling these properties. Thus,
naturally, we are interested not only in the mechanism of erosion, but
in how the resistance of materials to erosion can be objectively defined,
how it can best be tested, and how test results should be interpreted, and
whether this property can be correlated with other known material prop-
erties. Also of interest is the influence of the physical as well as chemical
properties of the impinging liquid. All of these points are still in need of
further enlightenment, even though the literature of the past forty years is
replete with comparative erosion test data, some of which are valuable but
few of which can be quantitatively compared with one another. Here,
again, the right questions are finally being asked, and some of these
questions and proposed answers are discussed in these papers.
The range of practical situations in which this type of damage pre-
sents serious problems seems to be continually growing. To the "classi-
cal" examples of ship propellers, hydraulic turbines, and steam turbine
blades operating hi wet steam, there had been added diesel cylinder
liners, pumps, valves and orifices, and condenser tubes; and even more
recently aircraft and missile surfaces subject to ram erosion, hydro-
foils, and pumps and turbines in liquid-metal Rankine cycle space
power plants. Moreover, there is some evidence that this type of mecha-
nism may occur hi bearings, gear teeth, and in instances where damage
has previously been attributed purely to fretting corrosion.
The first generation of cavitation and impingement erosion research
was characterized by much fundamental dispute, general uncohesive-
ness, and the frequent "rediscovery" of the same findings without added
elucidation. In the past dozen years or so, this research has come of age
and has acquired a more solid foundation and a more definite direction.
It is hoped that these papers will help to document this advance and to
light the way more clearly for the future course.
Frank J. Heymann
Senior Engineer, Westinghouse Electric Corp.
Lester, Pa., symposium chairman.
/. F. Ripken1
A Test Rig for Studying Impingement and

Cavitation Damage
REFERENCE: J. F. Ripken, "A Test Rig for Studying Impingement and

Cavitation Damage," Erosion by Cavitation or Impingement, ASTM
STP 408, Am. Soc. Testing Mats., 1967, p. 3.
ABSTRACT: A new type of test facility for simulating impingement and
cavitation damage is described. The facility consists of a rotor with a
material specimen attached at the periphery in such a manner that there
is repeated impact with a column of liquid drops during high-speed rota-
tion of the specimen in vacuum. Impact speeds up to 1250 ft/sec were
employed. Preliminary tests indicate that erosive weight loss from the
specimen is similar in character to that produced by other cavitation
damage facilities. In some instances, weight loss appeared to be directly
associated with the fatigue failure properties of the material. A refined
ability to control impact conditions permits detailed study of the failure
mechanics. While primarily intended to simulate a postulated cavitation
damage mechanism, it may prove useful in simulating other impact
damage applications.
KEY WORDS: accelerated test, cavitation, impact, erosion, fatigue
For nearly forty years various investigators have attempted to develop

and routinely employ test procedures which would realistically evaluate
the service resistance of fabricated materials exposed to erosion by water
impact. These evaluations were necessary for the design selection of
materials for steam turbine blades, for hydraulic machinery, and more
recently for underwater ship appendages and for aircraft windshields.
Field testing, which is an expensive and very slow method of accumu-
lating data, has largely given way to simplified and accelerated testing
in the laboratory. In the case of steam turbines, this has been accom-
plished with whirling blades impacting high-speed jets. Studies of air-
craft rain erosion have also used whirling blades in a spray and, for
higher speeds, a projecting of fluid slugs at stationary solids or firing
1
Professor, St. Anthony Falls Hydraulic Laboratory, University of Minnesota,
Minneapolis, Minn.
3
solids at stationary drops of liquid. For hydraulic devices and ship

members exposed to cavitation, the accelerated test apparatus has varied
widely, but three basic types have found considerable use. These are
the venturi throat or recirculating tunnel, the vibratory apparatus, and
the submerged rotating disk with cavitating perforations.
Cavitation damage studies in these three types of devices show a
general similarity of findings, but a number of significant differences
continue to appear in the quantitative values derived from the various
test programs.
A study of the findings from these devices indicates a present ability
to make suitable material selections for many design problems and
serves to indicate something of the basic qualitative nature of the cavi-
tation damage mechanism involved. It is, however, evident that these
devices are inherently incapable of providing a reasonably complete
quantitative portrayal of the damage mechanism. The prime difficulty
with the existing test systems for accelerated cavitation damage lies in
their inability to provide adequate experimental control over the many
variables that are concurrently involved in the cavitation damage phe-
nomenon. This suggests that the test control problems might be greatly
simplified by eliminating the complex and obscuring parts of the phe-
nomenon that have to do with the creation of the cavity, and by concen-
trating instead on treating the erosion solely as a consequence of the
mechanical action of cavity collapse. That this is a realistic approach is
supported by the opinion of many investigators as summarized by
Eisenberg et al [I].2 While gross fluid jet impact studies were used many
years ago by Rheingans [2] to simulate cavitation damage for materials
selection, the method was abandoned in favor of the seemingly simpler
vibratory test. Abandonment was probably to a considerable extent due
to a lack of evidence showing that cavitation damage was basically the
product of a fluid jet impact erosion mechanism. However, the concept
of jet impact as a cavitation damage mechanism appears to have some
substance as a result of the work of Ellis, Naude, Plesset, and co-workers
[3-5] at the California Institute of Technology and later studies of
microjets by others. The Cal Tech work has, by high-speed photography,
served to show that cavities collapsing near boundaries may collapse
unsymmetrically with the formation of a reentrant jet or interface front
which moves through the cavity to impinge on the solid boundary. The
physical observations by Ellis have shown that the jet velocities are quite
high and that the shape of the jet tip is probably significant to the pres-
sure developed on the boundary. Later studies by Shutler and Mesler
[6] confirm the presence of the jet but question the damage mechanism
proposed by Ellis. The jet mechanism of cavitation damage is therefore
a
The italic numbers in brackets refer to the list of references appended to this
paper.
RIPKEN ON TEST RIG FOR STUDYING DAMAGE 5
unproven at this time but has been tentatively accepted for the needs of
this study.
In light of these recent findings regarding the jet mechanism in cavity
collapse, the test facility described in this paper was designed to strip the
cavitation phenomenon to its bare essentials by examining the erosion
action of an impacting simple fluid element on a boundary solid. In
consequence, the design bears a considerable resemblance to other
impingement devices which have been developed for steam turbine and
aircraft windshield tests.
Currently, studies of damage on steam turbine blades are made with
sample pins or blocks attached to the periphery of a disk rotating in
air at high speed. A recent version of this type of facility has been em-
ployed also for material evaluations for cavitating hydraulic machinery
[7]. In this type of test, fluid impact is achieved by continuous ejection
of a high-speed water jet positioned with its axis parallel to the disk
shaft and passing between the rotating specimens. By this mechanism, a
fluid element of substantial mass impacts on the specimen at high fre-
quency (several hundred per second) and with high velocity (up to 1000
ft/sec). The resulting damage rate is very high. The device has produced
interesting results but is not considered suitable to fundamental studies
because of an inherent inability to independently control the frequency
and velocity of impact and to control the geometry of the jet front which
initially impacts on the solid. The high damage rates in these tests quite
probably relate to grossly abnormal thermal, chemical, or electrical ef-
fects resulting from the high stressing and the high rate of change of
stresses. These secondary effects might be quite different for elastomers
or plastics than they are for metals and should be subject to rate control
as well as stress control in any fundamental study.
The windshield erosion studies have even further simplified the fluid
impact test mechanism by either firing a solid specimen at a stationary
droplet [8] or by projecting (up to 4000 ft/sec) a fluid slug at a stationary
solid [9]. The first method has permitted some control of the shape and
size of the fluid interface and of the impacting velocity, but only in single
impacts. It does not permit measurement of impact pressure transients.
The second method permits measuring transient pressures in the solid
but does not permit refined control of the jet size or shape because of
the inherent instabilities of an interface under highly dynamic conditions.
More important perhaps is the failure of these systems to produce the
repetitive impacts or rate control which are basic to the fatigue failures
which are believed fairly common with cavitation damage. These two
methods together, however, have served many of the needs of rain
erosion studies in that these studies are generally concerned with single-
impact failure for droplets impinging on thin-plate structures at super-
sonic speeds.
The foregoing methods were not considered directly applicable to

simulating cavitation damage, but they did serve to point the way to a
modification which appears to be a workable compromise. This modifica-
tion consisted of designing an apparatus in which a drop of slow-moving
water would impact on a small target of test material moving at a high
velocity. The mass of the impacting liquid and the velocity of impact
were to be rather readily and accurately controlled, and controlled rates
of impact repetition were to be provided.
FIG. 1—The rotor assembly.
FIG. 2—A mounted test specimen at the rotor tip.
The resulting experimental equipment described in this paper produces

many drops of water of a selected uniform size and introduces them into
the path of a target of test material mounted on a rotating arm. The
introduction of the drops into this path is controlled with precision to
subject the same point on the target to repetitive blows at a selected
impact speed.
Experimenal Apparatus
The basic facility consists of a rotor with a material specimen attached
at a periphery in such a manner that there is impact with a column of
liquid drops during rotation of the specimen.
The aluminum-alloy rotor as shown in Fig. 1 consists of two central

disks supporting two projecting arms, the tips of which contain mounting
sockets for the test specimen as shown in Fig. 2. The specimen rotates
in a circle of 23.68 in. diameter at a present maximum of 12,000 rpm,
giving a tangential speed of 1250 ft/sec. (It is anticipated that future
changes will permit increasing the speed to 1500 ft/sec or more.) Speed
FIG. 3—The test rig—housing chamber closed.
is infinitely adjustable down to the minimum values of interest of about

400 ft/sec. The specimen has a target face of 1A in. diameter as shown
in Fig. 2. A small target volume of about l/s cm3 favors sensitive weight-
loss determinations. The tapered target is mounted in a tapered recess
in the rotor arm and is drawn snug with a draw screw tapped into the
rear of the target.
The rotor is spun within a protective chamber by a variable-speed,
directly connected, electric motor of 1 hp. The general assembly is shown
in Fig. 3 in the closed condition used for testing and in Fig. 4 in the
open position used for changing test specimens.
The drive system is patterned after systems successfully employed
with ultracentrifuges [10]. The drive shaft is a l/s in. diameter stainless
steel tube gripped by collets mounted on the motor shaft and on the
rotor. The tube provides flexible coupling which permits the rotor to
find its own center of rotation without elaborate dynamic balance proce-
dures. This also permits a very smooth high-speed operation even after
loss of target material. This stability is quite important for photographic
FIG. 4—The test rig—housing chamber open.
purposes. (It has been found possible to superimpose nearly 4000 re-
peated occurrences on a single film without loss in desirable sharpness
of the target or drops.)
Starting is a problem in this flexible system for speeds up to 300 rpm.
A Teflon guide bearing is provided to limit the undesired motions of the
rotor in this speed range. As soon as the first critical is reached, the rotor
spins smoothly and no longer touches the guide bearing.
The chamber pressure around the rotor is reduced to 0.01 atmos by
continuous vacuum pumping. This is necessary to reduce the aerody-
namic drag of the rotor and to reduce the wind disturbance on the drops
which are introduced into the target path. The value of 0.01 atmos
represents the limit of the available vacuum system. A lower value is

considered desirable.
The vacuum seal for the drive shaft is a close-fitting babbitt sleeve
approximately 1 in. long which is positioned over a hole in the chamber
cover. It presses against an O-ring at the bottom to provide a vacuum
seal. Oil is fed continuously into a cup at the top of the sleeve to provide
a liquid seal for the shaft. An oil slinger and catch cup on the shaft
below the sleeve collects the oil to prevent scattering throughout the
chamber. The rate of oil feed is only a few cubic centimeters per hour.
The small drops of water (degassed, distilled water) needed for target
impacting are generated in the evacuated bell jar above the main housing
chamber shown in Figs. 3 and 4. The drops are produced by attaching a
fine glass capillary nozzle in axial alignment with the vibratory dome of
a speaker element. A number of devices of this type are reported in de-
tail in the literature [11,12]. The test liquid flows through the capillary
nozzle from a reservoir and is valve controlled. Because of the forced
vibrations, the liquid discharges in a discrete series of drops directed
vertically downward through a small hole hi the top of the main housing
chamber. Flow in the capillary is induced by the pressure difference
which exists between the reservoir at atmospheric pressure and the
capillary nozzle which is in the evacuated bell jar. A manually operated
shutter deflects the drops away from the impact area or allows them to
strike the target as desired. Current tests have been conducted with drops
of 0.047 in. diameter, but substitution of other nozzles will permit other
size selections. The system readily provides electronic count of the num-
ber of test impacts.
The signal to drive the vibrating capillary is derived from a photo-
electric pickup and a slotted wheel mounted on the rotor. The best drop
production seems to occur between 600 and 1000 cps. This range is
determined by flow rate and jet diameter and is given in Refs 11 and 12
as la < X < 14a, where a is the jet diameter and A. is a wavelength
based on jet velocity and vibrator frequency. The number of slots on the
"pickup wheel" must be determined by the wavelength criterion and
by the desired rotational speed. Some flexibility was introduced by feed-
ing the output of the photoelectric cell into a General Radio tone burst
generator. This instrument counts a preset number of pulses and switches
from one stable state to the other. In this way a square wave is generated
which is an accurate submultiple of the output frequency of the photo-
electric pickup.
The output of the tone burst generator is then fed to an audio am-
plifier and then to the speaker element of the drop generator. The par-
ticular wave form driving the generator seems to have little influence on
the drop production. This system presently provides one drop impact
per target revolution with only a small fraction of the total drops pro-
duced striking the target.
The location of the drop as it impacts the target is controlled by the
phase of the electrical signal with respect to the rotor position and the
rate of flow of the liquid to the capillary. The differential pressure be-
FIG. 5—Volume loss for various materials exposed to various speeds and cycles
of impact.
tween the atmosphere and the vacuum hi the chamber is the force mov-
ing the liquid through the capillary. Flow rate is controlled by a pinch
clamp on the supply tube. While the electrical phase control appears
adequate for drift-free long-term tests, the simple flow control requires
continuous monitoring and is in need of further refinement for stability
of drop impact location.
Two small windows in the protective chamber permit viewing the
moving target in either full face or profile. Illumination is provided by a
(a) Velocity = 500 ft/sec, N about 145,000 cycles.

(a) Velocity = 750 ft/sec, N about 145,000 cycles.
(c) Velocity = 1000 ft/sec, N about 58,000 cycles.
General Radio strobotac which is synchronized with the motion of the

rotor with a second photoelectric pickup and slotted wheel.
Vaporization of the spent water drops produces fogging of the view-
ing windows and has necessitated the addition of cooled surfaces in the
chamber for accelerating removal of the fog.
The present system permits investigation of a large number of impacts.
Additional refinement is needed, however, to permit selection of the
impact rate and positioning of any number of drops in the impact area
down to a single event. This will require a mechanical deflection or
shuttering system for this purpose.
Experimental Procedure
The application of the device to date has been limited to a brief series
of tests intended to show the capabilities.
A representative number of metals were selected and machined into
the specimen form as shown in Fig. 2.
The equipment was adjusted so that the center of the specimen face
would impact a liquid drop on each revolution. Four test speeds were
arbitrarily selected at 500, 750, 1000, and 1250 ft/sec.
A simple measure of weight loss as affected by running time was made
by stopping the apparatus periodically and removing and weighing the
specimen. It was possible to watch the progress of the erosion visually
with a small telescope. This was of great help in establishing the incre-
ment of exposure time hi a test run.
Preliminary Findings
The four materials which were exposed to weight loss-damage tests
have physical properties briefly summarized as follows:
aluminum type 1100 F annealed, ultimate strength 13,000 psi,
BHN 23;
cast iron, physical properties unknown;
Type 430 stainless steel, annealed, ultimate strength 70,500 psi,
BHN 152;
Type 304 stainless steel, cold drawn, annealed, ultimate strength
90,000 to 125,000 psi.
A graphical summary of the test data for these four materials is
shown in Fig. 5. Many plotting parameters might be employed for such
data. In this case, a plotting based on volume of material removed
versus number of impacts sustained appeared to be a meaningful repre-
sentation.
The general character of the damage inflicted on a test specimen is
shown in photographs taken at the completion of a test series. Figure 6
shows for the aluminum alloy in part (a) the results at 500 ft/sec. Part
(b) shows the results for 750 ft/sec and part (c) for 1000 ft/sec. The
photographs demonstrate plastic flow with considerable cratering or up-
lift deformation at the edge of the impact region. The plastic uplift
undoubtedly leads to occasional loss of material in fairly large pieces
and contributes to the somewhat erratic losses evidenced for aluminum
in Fig. 5. The less deformable materials appear to experience losses in
a smoother and more gradual progression. Part a of Fig. 6 shows im-
pact positioning fairly well confined, whereas part b shows some wander-
ing of impact around a deep central hole. Part c again shows a condition
of some wandering about the deep central hole and additionally shows
a large area of secondary erosion by spray following the initial im-
pacting. Visual studies of the impacting drops indicate that with a
smooth surface the drops spread slightly radially on the target face and
then rebound in a spray which moves radially outward and away from
the face. This spray evidently moves fast enough so that it clears the
target sweep path without again striking the target. However, as shown
in Fig. 6c, the spray rebounding from a highly roughened surface moving
at 1000 ft/sec does make a second damaging impact.
FIG. 7—Impact damage, Type 430 stainless steel annealed.
A test on the Type 430 stainless steel failed to show any volume loss
(based on weight loss) with an impact velocity of 500 ft/sec when test
exposure was terminated at about 14.5 X 104 impact cycles, but yielded
the data of Fig. 5 at 750 and 1000 ft/sec. Figure 7 shows a very slight
evidence of plastic deformation. It is noteworthy that the last points on
the curve of Fig. 5 for a velocity of 1000 ft/sec represent a punching
through of the target specimen which was approximately % 6 in. thick
at the impact point. It is interesting to note that the diameter of the
larger outer end of this hole as shown in Fig. Ib is approximately the

diameter of the impacting drops (0.047 in.) even after approximately one
quarter million impact cycles.
The impact tests on the Type 304 stainless steel were run at 1000 and
1250 ft/sec with results as shown in Fig. 5. Photographs of these speci-
FIG. 8—Damage-rate values for 304 and 430 stainless steels.
mens are not included but have a considerable resemblance to those of

the 430 stainless steel shown hi Fig. 7. The major difference between
the 304 and 430 alloys hi Fig. 5 is the substantial "incubation" or delay
time before loss occurs with the 304 alloy.
Eisenberg et al [1] outlined four zones of cavitation damage, based
on vibratory tests, which also seem applicable to impact erosion damage.
These zones which are evaluated hi terms of loss per unit of time are
described as:
Zone 1. An incubation or no-loss zone. In cavitation tests with a

vibratory apparatus this was shown to depend on amplitude for a given
frequency.
Zone 2. Accumulation zone. A zone in which the energy-absorption
rate increases with time, resulting in increasing loss of material with
increasing test duration.
Zone 3. Attenuation zone. The rate of material loss reaches a peak
value and begins to decrease. This zone is reported to be characterized
by the formation of isolated deep craters on the surface of the test mate-
rial, indicating that the attenuation of energy absorption is associated with
the influence of the craters on the bubble-collapse process.
Zone 4. Steady-state zone. The rate of material loss reaches a con-
stant value.
These four zones are not directly identified in the accumulated mate-
rial loss type of plotting used in Fig. 5 but can be partially distinguished
in an alternate plotting using rate of material loss versus number of im-
pact cycles. An alternate plot of this type for data relating to the 304
and 430 stainless steels is shown in Fig. 8. Zone 1 appears to be negligi-
ble for the 430 steel but appreciable for the 304 steel. The plotting of
Fig. 8 serves to show the same general form as material-loss data from
other types of facilities as summarized by Eisenberg et al [1],
Analysis
There is some indication in the limited test data of Fig. 5 that mean-
ingful damage evaluations can be derived from Zone 1 for an impact
type of facility. This is based on the fact that the facility can produce
controlled impact conditions which are subject to a fairly rational analy-
sis of the resulting loading and that controlled numbers of load cycling
can be applied. In short, this concept is one of fatigue failure in which
some combination of stress value and cycles of application determines
the failure.
The concept is not well defined with most of the materials shown in
Fig. 5 but is given some substance if the data for the 304 stainless steel
are converted to values of stress and cycles of stress. The cycles of
stress to failure are arbitrarily evaluated by assuming that failure occurs
when significant loss of material begins or where the horizontal line
respresenting Zone 1 intersects the sloping line representing Zone 2. For
the 304 stainless steel with an impact velocity of 1000 ft/sec the inter-
cept occurs at a time equivalent to 1.85 X 105 cycles and for a velocity
of 1250 ft/sec at 3.4 X 104 cycles.
The value of the peak impact pressure stress may be roughly approxi-
mated by the expression p = kpcv employed many years ago by Ackeret
and deHaller and given more recent consideration by Engel [13]. In
this expression, k is a constant which approximates unity when the im-
pacting faces of both the liquid and the solid are plane but, according to
Engel approximates one half when the liquid face is spherical; p is the
water density or 1.94 slug/ft3; c is the acoustic velocity which is as-
sumed as 4800 ft/sec; and v is the relative velocity of impact. With this
the v test value of 1000-ft/sec yields p = 32.3 x 103 psi, and the v
test value of 1250 ft/sec yields p = 40.5 X 103 psi. The equivalent value
is p = 24.2 X 103 psi for the test which failed to yield measurable dam-
age for 2.2 X 105 cycles at v = 750 ft/sec.
The foregoing stress-cycle values are plotted in Fig. 9 together with
longitudinal fatigue failure test values for an annealed 304 stainless steel
as given in Ref 14. The relative agreement of these different types of
FIG. 9—A comparison of impact failure with failure by a standard fatigue

method for 304 stainless steel.
test data may perhaps be fortuitous but is nevertheless encouraging in

a preliminary experiment.
Conclusions
The equipment described in this paper has shown a capability for
eroding several typical structural materials by repeated impact of small
liquid drops. It has a capability for repetitive impact on a small area of
sample material with close control of the mass and velocity of the drop.
The pattern of weight-loss damage in tests with the impact facility
shows a distinct similarity to weight-loss values obtained from cavitation-
type test facilities. The similarity indicates that this type of facility may be
useful not only for evaluating materials for applications exposed to
cavitation but also for other types of impact erosion evaluations.
Limited tests with ductile materials show marked deformation prior
to loss of weight. More detailed tests with this apparatus can serve to
more clearly define the transition from plastic flow to actual loss of
material and the extent to which plastic flow may occur in the "incuba-
tion" period of harder materials.
The incubation period as defined in weight-loss cavitation damage
studies is also evident in these impact damage studies with a self-hard-
ening material. Under specific laboratory conditions, the incubation
period has been shown to be subject to refined study with this type of
apparatus. The period is a particularly important one, for its limit serves
to define the conditions under which a desirable type of material begins
to fail. Preliminary findings indicate that failure represented by the limit
of incubation may be rather directly associated with the better known
fatigue failure properties of the material.
Further studies are to be undertaken relative to wider control of drop
size, drop numbers, and rate of impacting. The influence of these varia-
bles together with variations of the fluid properties such as pc surface ten-
sion, corrosiveness, and so forth, will be studied relative to the damage
resistance of selected metals.
A cknowledgment
The author gratefully acknowledges the support of the Office of Naval
Research and the David Taylor Model Basin of the U.S. Department of
the Navy and expresses sincere appreciation for the many contributions
made by his colleague J. M. Killen, assisted by S. D. Crist and R. M.
Kuha of the staff of the St. Anthony Falls Hydraulic Laboratory.
References
[1] P. Eisenberg, H. S. Preiser, and A. Thiruvengadam, "On the Mechanism of
Cavitation Damage and Methods of Protection," Meeting Paper No. 6, Soc.
Naval Architects and Marine Engrs., November, 1965.
[2] W. J. Rheingans, "Prevention and Reduction of Cavitation and Pitting in
Hydraulic Turbines," Engineering Bulletin No. 11, Allis Chalmers, 1949.
[3] M. S. Plesset and T. P. Mitchell, "On the Stability of the Spherical Shape
of a Vapor Cavity in a Liquid," Quarterly of Applied Mathematics, Vol 13,
1956.
/4] C. F. Naude and A. T. Ellis, "On the Mechanism of Cavitation Damage by
Nonhemispherical Cavities Collapsing in Contact with a Solid Boundary,"
Transactions, Am. Soc. Mechanical Engrs., Vol 83, Series D, December,
1961.
[5] A. T. Ellis, M. E. Slater, and M. E. Fourney, "Some Flow Approaches to
the Study of Cavitation," Symposium on Cavitation and Hydraulic Machinery,
IAHR, Sendai, Japan, September, 1962.
[6] H. D. Shutler and R. B. Mesler, "A Photographic Study of the Dynamics
and Damage Capabilities of Bubbles Collapsing Near Solid Boundaries,"
Transactions, Journal of Basic Engineering, Am. Soc. Mechanical Engrs.,
Vol 87, June, 1965.
[7] J. M. Hobbs, "Problems of Predicting Cavitation Erosion from Accelerated
Tests," ASMS Paper 61-HYD-19, Am. Soc. Mechanical Engrs., 1961.
[8] D. C. Jenkins, "Erosion of Surfaces by Liquid Drops," Nature, Vol 176, Au-
gust 13, 1955.
I8 EROSION BY CAPTATION OR IMPINGEMENT
[9] J. A. Brunton, "Deformation of Solids by Impact of Liquids at High Speeds,"

Symposium on Erosion and Cavitation, Am. Soc. Testing Mats., June, 1961.
[10] A. Weissburger, Physical Methods of Organic Chemistry, Interscience Pub-
lishers, Inc., New York, 1949.
[11] J. M. Schneider and C. D. Hendricks, "Source of Uniform-Sized Liquid
Droplets," Review of Scientific Instruments, Vol 35, No. 10, October, 1964.
[12] R. G. Sweet, "High-Frequency Oscillography with Electrostatically Deflected
Ink Jets," Review of Scientific Instruments, February, 1965.
[13] O. G. Engel, "Waterdrop Collisions with Solid Surfaces," Journal of Research
for the National Bureau of Standards, Vol 54, May, 1955.
[14] Anonymous, Metals Handbook, Am. Society Metals, 8th edition, 1961, p.
419.
DISCUSSION
W. S. Owen1 (written discussion)—It is clear that the strain rate im-

posed upon the material of a turbine blade by an impacting water drop
is much faster than the strain rate used in conventional tension testing.
Thus, it seems more reasonable to attempt a correlation of erosion loss
with yield stress, work-hardening coefficient, etc., measured at the ap-
propriate strain rate. In the past few years, much new data from uni-
axial tension tests carried out under fast impact conditions have become
available, and I wonder if any of the authors have found it possible to
make use of these data in their correlation studies.
/. F. Ripken (author)—Dr. Owen raises a question as to whether
correlations of the erosion loss have been attempted using the tensile
characteristics of the material taken at high strain rates. This is a rela-
tively new concept which has been given related consideration in tension
fatigue tests by Thiruvengadam.2 His limited data for five metals show
that low rate tensile fatigue data yield failure values about 15 per cent
higher than data taken at rates comparable to his cavitation erosion stud-
ies. Similar data are not available for the 304 stainless steel shown in
Fig. 9, but fatigue data values taken at rates comparable to the impact
rates would appear to be a more meaningful way in which to plot Fig. 9.
/. M. Hobbss (written discussion)—This new apparatus appears to
1
Department of Materials Science and Engineering, Cornell University, Ithaca,
N. Y.
2
A. Thiruvengadam, "High Frequency Fatigue of Metals and Their Cavitation
Damage Resistance," Technical Report 233-6, Hydronautics, Inc., December,
1964.
'Properties of Fluids Div., National Engineering Laboratory, East Kilbride,
Glasgow, Scotland.
DISCUSSION ON TEST RIG FOR STUDYING DAMAGE 19
be a very useful research facility, but it will need further refinement if it

is not to suffer some of the shortcomings of the equipment upon which it
is intended to improve.
Professor Ripken criticizes the water jet impact machine on the
grounds that the frequency and velocity of impact are interdependent.
He also attributes the high damage rates in these tests to "grossly abnor-
mal thermal, chemical, or electrical effects resulting from the high stress-
ing and the high rate of change of stresses," but has not stated his rea-
sons for doing so. The discusser finds it difficult to see why these effects
should be attributed to only one particular type of test equipment and
not to the phenomenon in general.
An examination of the results given in the paper indicates that the
damage rates obtained, when rationalized, are almost as high as those
obtained in the water jet impact machine used at the National Engineer-
ing Laboratory. The results for Type 304 steel at 1000 ft/sec as pre-
sented in Fig. 8 of the author's paper give a maximum rate of volume
loss of about 7 mm3 per 106 cycles. The volume of a 0.047 in.-diameter
spherical drop is approximately 0.89 mm3. Using the rationalized ero-
sion rate suggested by Heymann,4 this gives:
ER = 7/(0.89 x 106) = 7.87 x 1Q-6
The nearest austenitic steel that we have tested is Type 321 (tensile
strength, 95,000 psi). This eroded at a rate of 7 mm3 in 24 X 103 im-
pacts at 660 ft/sec with the side of a 0.635 in.-diameter water jet. The
volume of water colliding with the test piece per impact is in this case
12.9mm3. Hence,
EH = 7/(12.9 x 24 x 103) = 2.26 x 1Q-5.
In spite of the lower impact velocity, this is nearly three times greater
than ER. This may be the result of the relatively higher stresses pro-
duced by the collision between the test piece and the side of a continuous
jet of circular cross section instead of with a spherical drop. However,
the difference is hardly sufficient to dissociate the new equipment from
the criticisms leveled at the old.
As described in the paper, the system provides one drop impact per
target revolution; that is, frequency and velocity are therefore interde-
pendent. If, at the present, the test drop production occurs between 600
and 1000 per second, considerable improvement appears to be necessary
in order to achieve independent control of impact frequency.
The progress of the cumulative erosion with time is very similar to
4
F. J. Heymann, "Second Quarterly Progress Report, October 1965 Through
January 1966; Basic Investigation of Turbine Erosion Phenomena," Contract NPS
7-390, WANL-PR(DD)-007, Westinghouse Astronuclear Laboratory, Pittsburgh,
Pa., 1966.
that obtained using resonant-cavity erosion test devices. In these tests,

it was found that the relative merits of different materials were best
assessed from the times required for pits to reach a certain depth.5 In
the paper under discussion, measurement of the incubation period is
fairly straightforward, but for many purposes a knowledge of erosion
rate is also required. As the variation of erosion rate with time may not
be similar to its variation in practice, where the distribution of droplets
over the surface is random, the interpretation of erosion rate-time data
may require considerable skill.
Mr. Ripken—The writer agrees with Dr. Hobbs' observation that
certain of the stress values and rate of change of stress values are in-
herent to cavitation phenomena in general rather than to a particular
type of test equipment. This agreement, however, relates to the values
pertinent to any one cycle of impact on a given element of area. Evi-
dence from various sources indicates that in most prototype cavitation
problems, the rate of cycling on a given area is probably quite low, and
some relaxation or normalizing presumably has an opportunity to occur
between cycles. In contrast, accelerated test environments markedly in-
crease this rate of cycling and may permit accumulations of secondary
effects which are abnormal and detrimental. The water jet impact
machine is believed to provide these high rate cycling conditions over
most of the erosion area of the test specimen, and it is quite correct that
the same high rates will prevail in the repeating drop impact machine
if the machine is set for one drop impact per target revolution as it has
been for the data shown. It is, therefore, not surprising that Dr. Hobbs
found substantial agreement when applying Heymann's analysis to his
jet impact data and the data shown herein. The prime difference in the
machines is that the drop machine has been designed to permit inde-
pendent control of the effective drop count to less than one per revolu-
tion using a suitable drop deflection mechanism. Interruption of the jet
stream also is conceivable but would appear to offer some difficulties.
Independence of velocity and drop rate could also be achieved in the
drop facility by randomizing the impact position over a specified area
with a drop deflection mechanism. While Thiruvengadam's data2 indi-
cate that high repetition rates do not drastically alter the damage proper-
ties of a metal, there is evidence that elastomers or plastics may be more
susceptible to high rate values. The lower thermal conductivity of these
materials could lead to thermal buildup and consequent damage.
Olive G. Engel® (written discussion)—In using the apparatus de-
scribed, the chamber pressure around the rotor is dropped to 0.01 atmos
5
D. J. Godfrey, "Investigations of Cavitation Damage. Part II—An Acoustic
Method of Producing Cavitation Damage," AML Report A/3(c), Admiralty
Materials
9
Laboratory, Poole, Dorset, England, 1959.
Chemical physicist, Space Power and Propulsion Section, General Electric Co.,
Evendale, Ohio.
DISCUSSION ON TEST RIG FOR STUDYING DAMAGE 21
to reduce aerodynamic drag and wind disturbance of the drops that are
introduced. This pressure is equivalent to 7.6 mm Hg which is lower
than the room-temperature vapor pressure of water (roughly 20 to 30
mm Hg). Consequently, it can be expected that bubbles of water vapor are
present in the drops which the test specimens intercept. This constitutes
a reduction of the density of the drop liquid and should affect the extent
of damage produced.
The point could be clarified in the following way: compare test results
on weight loss obtained at a chamber pressure of x mm above and x mm
below the vapor pressure of water at the test temperature with the other
test variables held constant, and let x increase in magnitude until the
point of difference is found.
If the test facility is to be used in determining only comparative
damage ratings, then, of course, the density of the drop liquid is of no
importance so long as it remains constant for all the materials being
rated.
Mr. Ripken—Dr. Engel's concern with regard to possible density
variations due to vaporization is well taken, and her suggested test for
verification of density stability should prove useful. Some confirmation
of her concern has been experienced in that generation of drops proved
difficult until it was learned that prior and severe degassing of the water
was essential for stable drop generation. This treatment presumably
reduced the nucleate centers for interfacial vaporization within the drop.
The room-temperature vapor pressure cited by Dr. Engel is perhaps
excessive in that the chamber temperature is maintained far below freez-
ing to reduce chamber fogging. While the exposure of the water to this
reduced chamber temperature is very fleeting before impact, it is proba-
ble that the drop temperature is closer to freezing values than to room
values.
A. Thiruvengadam1
The Concept of Erosion Strength
REFERENCE: A. Thiruvengadam, "The Concept of Erosion Strength,"

Erosion by Cavitation or Impingement, ASTM STP 408, Am. Soc.
Testing Mats., 1967, p. 22.
ABSTRACT: In general, the problem of erosion of materials can be
divided into two categories. One is the understanding of the threshold
for each material wherein the impact stresses reach a limiting value just
sufficient to initiate detectable erosion either at the first blow or after
repetitive blows. Evidence is presented to show that the dynamic yield
strength of a material controls the threshold for the single impact,
whereas thejendurance limit is the important property representing the
threshold for multiple impacts.
The second problem is the prediction of the amount of damage if the
erosive forces are above the threshold for the material. The designer
needs some numerical value of a property that governs the volume of
erosion of a material. As of now, there is no single property that can be
used for this purpose, just as we use various properties of materials to
represent their response to static, fatigue, and creep loadings. A recent
suggestion to use the strain enerjy^^^fjhj^malejial^-as-givenr-by the area
of the stress-^&în^diagranrfrom a simple tension test, has a few limita-
tions suciuas strain-rate effects, environmental effects (for example, tem-
perature and corrosion), and the scarcity of stress-strain data under these
.conditions'. In order to overcome these limitations, a new concept known
as erosion strength is introduced, and it is defined as the energy ab-
sorbed per unit volume of material up to fracture under the action of the
erosive-forces Jn various environments. The methods to determine the
erosion strength from an erosion test are outlined.
If^the concept is accepted by the engineering profession, erosion
strength would take its place among the other mechanical properties of
materials such as yield strength, ultimate strength, fatigue strength, creep
strength, hardness, and corrosion fatigue.
KEY WORDS: erosion, cavitation, impingement, impact, erosion strength,
strain energy, fatigue (materials)
Nomenclature
Ae = Area of erosion
Cj = Velocity of sound in liquid
Cm = Velocity of sound in material
Ea = Energy absorbed by material
1
Senior research scientist, Hydronautics, Inc., Laurel, Md.
22
THIRUVENGADAM ON CONCEPT OF EROSION STRENGTH 23
Em = Young's modulus of material

/ = Average depth of erosion
/ = Intensity of eriosion
KI = Bulk modulus of liquid
p = Water hammer pressure
Se = Erosion strength
t = Time of erosion
UT = Threshold impact velocity
YD = Dynamic yield strength of material
zt = Acoustic impedance of liquid
FIG. 1—Definition sketch for the material response to erosive forces.
zm = Acoustic impedance of material

a = A factor tending to unity at high velocities
ft = ^(Em/Ki)112 = A factor used in Engel's theory
AF = Volume eroded
pi = Liquid density
pm = Material density
<re = Endurance limit of material
The word "erosion" from the Greek "erodere"—to gnaw away—

generally means the surface destruction and removal of material by ex-
ternal mechanical forces. In the majority of practical cases, these forces
are in the form of multiple impacts produced by the dynamic impinge-
ment of a liquid on a solid surface or of a solid on another solid surface.

Numerous examples may be cited where such phenomena are important,
such as:
1. cavitation damage of hydrodynamic systems such as hydraulic
turbines, ship propellers, and valves;
2. erosion of steam turbine blades due to the impact of condensed
liquid droplets;
3. erosion of aircraft structures as a result of collisions wkh rain
drops;
4. sand erosion of water tubrines and other control devices; and
5. erosion of space vehicles as a result of impacts with meteorites.
In all of these erosion problems, one of the important aspects is the
prediction of the material response to a given set of input conditions such
as the collapse of a bubble on or near the surface, the impingement of a
liquid drop or jet, the impact of a solid particle. In general, the impact
energy of the input system will produce on the material one of the fol-
lowing effects depending upon the intensity of impact and the number
of repetitions (see Fig. 1):
1. there may not be any permanent deformation;
2. the material may deform after a certain number of repetitive im-
pacts;
3. a permanent deformation may develop at the onset of the first
blow; and
4. it may plastically flow on the first blow itself or after a certain
number of repetitions as a result of high strain rates.
Based on these arguments, one can arrive at two types of problems.
The first one is the understanding of the threshold conditions wherein
the impact stresses reach a limiting value just sufficient to initiate de-
tectable erosion either at the first blow or after repetitive blows. The
second case is the prediction of the amount of damage if the erosive
forces are above the threshold for the material. It is the aim of most of
the investigations to provide the designer with numerically expressed
properties to represent the behavior of the materials under the afore-
mentioned two cases. The property (or properties) that controls the
threshold conditions, where in the criterion is the observable yielding
of the material, need not be the same as the property that represents
the volume of erosion of a material wherein the material is being frac-
tured and removed from the surface. It is the purpose of this paper to
discuss these two cases in the light of some of the results obtained in our
laboratory in connection with cavitation erosion, as well as those pub-
lished in the literature.
It is found that the dynamic yield strength and the endurance limit
are the two properties controlling the threshold conditions due to single
impact and multiple impacts, respectively. However, the volume of ero-
sion when the erosive forces are above this threshold is controlled by
some property of the material that represents the energy-absorbing ca-
pacity of the material. Recent attempts [1,2]2 to correlate the strain en-
ergy of the material from the simple tension test are handicapped by
a few limitations such as strain-rate effects, environmental effects, and
the scarcity of the stress-strain data under these conditions. To over-
come these difficulties, the concept of erosion strength is introduced with
methods of determining this strength from an erosion test. In many prac-
tical cases, the erosion forces may be assisted by environmental effects
such as corrosion, embrittlement, and temperature. The methods to de-
termine the erosion strength under these conditions are also outlined.
Threshold Criteria
It is becoming increasingly common to observe a threshold parameter
such as the threshold intensity of cavitation damage, the threshold am-
plitude of oscillation, the threshold velocity of flow, the threshold impact
velocity, and so forth, in erosion problems such as cavitation damage,
turbine blade erosion, rain erosion, jet or drop-impact erosion. When a
cylindrical column of liquid impinges on the surface of a material, the
maximum pressure (generally known as the "water hammer" pressure)
developed by the impact is given by de Haller as
where [3]:
Ur = impact velocity,
Pi = density of liquid,
pm = density of material,
Ci = velocity of sound in liquid, and
Cm = velocity of sound in the material.
For most practical cases of liquids and materials involved, the ratio of
PiCi/pmCm is small, and this term may be neglected. Then the water-
hammer pressure becomes
Such an estimate for a spherical liquid drop colliding with a solid surface
has been given by Engel [4] as
where a is of the order of unity at high velocities; the factor Y% is due to

the spherical shape of the drop.
2
paper.
TABLE 1—Threshold criteria.
Water-Hammer theory EngePs theory
Em = Young's modulus of material

KI = Bulk modulus of liquid
Threshold Density Velocity Dynamic Yield Modulus of Endurance i
Impact
a
"e <r>
No. Material Liquid Velocity. of Liquid, of Sound, Strength, Elasticity, Limit Ref
pi , slugs/ft' Ci , ft/sec YD, psi Em , psi <re , psi PiCiUr UipiWEm1'*
Ui , ft/sec
1 . . . . Copper water 99 (30 value of piCiUi = 5750 21 500 (15 16 X 108 12 000 1.85 0.25 0.26 10
m/sec) psi (= 4 kg/mm2) kg/mm2)
2 . . . . Steel water 295 (90 PiCiUi= 17 100 psi (= 12 72 000 (50 30 X 10« 36000 2.1 0.18 0.20 10
m/sec) kg/mm2) kg/mm2)
3 . . . . Chromium water 400 1.94 4800 30 X 10" 50 000 1.9 0.19 0.20 11
Steel
4 . . . . Stellite 6 water 600 1.94 4800 36 X 106 80 000 2 0.20 0.20 11
5 . . . . Stellite water 1100 1.94 4800 36 X 106 80 000 1 0.10 0.18 7 (P.
4-4)
6 . . . . Chromium water, 800 to 900 1.94 4800 100 000 30 X 10" 2 0.4 0.20 12
Steel single
impact
7 . . . . Stellite 6 water, 1100 to 1.94 4800 150000 36 X 106 2 0.2 0.18 12
single 1200
impact
Experiments by Brunton [6,8] show that the measured values of the

pressures closely agree with the values predicted by the water-hammer
equations.
The threshold criterion for the case when a drop makes a single im-
pact on a metal would be that the water-hammer pressure should exceed
the dynamic yield strength of material producing a detectable permanent
deformation [9]. It may be mathematically stated as
where YD is the dynamic yielcfWength of the material. However, in

practice, the material deformation startsâifer repeated multiple impacts.
In such a case, if we assume that the significant material property would
be the endurance limit of the material, then the threshold criterion is
modified as
where <je is the endurance limit.

There are a few published data for multiple impact of water drops
on metals [5,7], and the results are summarized in Table 1. These data
seem to check well with the criterion given by Eq 5 which states that the
endurance limit of the metal is the property controlling the threshold.
Even in cavitation damage, the threshold is characterized by the en-
durance limit of the metals [1,13], Earlier experiments [1] with a two-
dimensional water tunnel and with a rotating-disk apparatus showed
that the threshold velocity of flow depended upon the endurance limit
of metals. Recent experiments with a magnetostriction oscillator pro-
ducing cavitation on an oscillating piston also showed that the endurance
limit of metals controls the threshold intensity of cavitation damage [13].
While the foregoing discussion tends to establish the endurance limit
as the sole material property controlling the threshold conditions in a
multiple-impact test, there is a differing point of view wherein the
modulus of elasticity also would become important. Engel [14] gives
the following relationship for the threshold velocity of a liquid drop to
produce a dent in the metal:
where:
zi = piCi of liquid, and
zm = pmCm of material.
The foregoing equation was verified by the results of two sets of ex-
periments: one involved collisions of mercury drops with such metals as
28 EROSION BY CAPTATION OR IMPINGEMENT
copper, aluminum, lead, and steel; the other, collisions of water drops
with metals such as aluminum, copper, and lead [75]. Engel [14] also
reported that factor 19 in Eq 6 becomes 1 in the case of impact of rigid
metal spheres against metal targets.
Now, Eq 6 can be rewritten in the form
Since (from [75])
if we assume the Poisson's ratio, v, as nearly Vs for most metals. Term

Em is the Young's modulus of elasticity of the material, and /3 is a fac-
tor to be evaluated.
For multiple impacts of drops, YD can be replaced by the endurance
limit (as before), giving
The threshold criterion would become
From Eqs 5 and 9,
moreover,
where KI is the bulk modulus of the liquid.

Hence
For water, ^ = 300,000 psi, and for steel, Em = 30 X 106 psi.

Hence, from Eq 10, the value of l/£ can be predicted as 0.2. The ex-
perimental data given in Table 1 verify this value of 0.2. It seems that
the water-hammer theory and Engel's theory closely agree.
Necessity for the Concept of Erosion Strength

So far, the discussion was confined to the threshold conditions at
which the materials start being damaged and their relationship to dy-
namic yield strength and endurance limit (including modulus of elas-
ticity). The easiest design approach would be to determine the threshold
values and use the proper materials. However, it is not always possible
to limit the designs to threshold conditions. Practical and economic con-
siderations may warrant a design in which we may have to live with
some erosion hi a given material and predict the life of the system. In
this case, it is essential to know the property of the material that con-
trols the- volume eroded in a given time.
There have been several attempts to correlate the resistance of a ma-
terial to erosion with any one of its known mechanical properties such as
yield strength, ultimate strength, hardness, and so forth, without much
success. The evidence for this statement may be found in Ref 16 for
cavitation damage and Refs 5,7 for drop-impact erosion. This situation
led to a premise (to quote from Peterson [77]) that "it is not possible
to provide the designer with numerically expressed 'properties' which can
be fed into formulas for proportioning parts, as one can do for static,
fatigue and creep loadings." Qualitative and comparative screening tests
led to a duplication of efforts in addition to confusing the problems in-
volved.
Recent analyses [1,2] in connection with cavitation damage pointed
out that the property correlated should represent the energy-absorbing
capacity of the material up to fracture. Available experimental results
showed that the strain energy of a material up to fracture as given by
the area of the stress-strain diagram represents the erosion resistance as
far as cavitation damage is concerned. However, there are a few genuine
limitations to this approach, namely:
1. The strain-rate effects. The strain energy values used for these cor-
relations were obtained at relatively low strain rates, whereas the erosion
phenomena generally takes place at high strain rates. This effect could
become very important for strain-rate-sensitive materials.
2. The mode of stressing in erosion is radically different from the
simple tension test.
3. The environmental effects such as corrosion, high temperature,
vacuum, low temperature, and embrittlement cannot be quantitatively
reproduced in the strain energy measured from a simple test.
4. Above all, the availability of the stress-strain data up to fracture
under these conditions itself is a great limitation. This is mainly because
the strain energy itself is not a commonly used material property. How-
ever, its importance is being increasingly felt as the understanding of
fracture mechanics progresses.
In order to overcome these limitations, the concept of erosion strength
is introduced. Just as yield strength, ultimate strength, fatigue strength,

all define a certain physical state and behavior of the materials, erosion
strength is specifically defined to represent the erosion resistance of ma-
terials. It is hoped that this direct approach will lead to a more general-
ized understanding of the phenomenon of erosion as a whole.
Definition of Erosion Strength
During the process of erosion, a certain volume of material is frac-
tured from the surface of the parent material as a result of the work'
FIG. 2—Correlation between strain energy and reciprocal of rate of volume

loss (Ref. 2).
done by the external forces. The energy absorbed by the volume of the
material fractured is given by
where:
Se = erosion strength which represents the energy-absorbing capacity
of the material per unit volume under the action of the erosive
forces,
AF = volume of material eroded, and
Ea = energy absorbed by the material eroded.
The measurement of A Fin a laboratory experiment is not very difficult.
If we can devise a method by which we can accurately determine the

energy absorbed by the material under the action of erosive forces, then
the erosion strength as defined here can easily be determined.
Determination of Erosion Strength

The logic behind the method of determination of erosion strength will
be developed by illustrating the specific case of cavitation damage. In
this case, the eroding forces are caused by the collapse of cavitation
FIG. 3—Effect of NaCl concentration on the amplitude versus damage rate re-
lationship for aluminum 1100 F in steady-state zone (Ref 21).
bubbles near the material surface. Recent experiments show that the rate
of volume loss is inversely proportional to the strain energy of the ma-
terial for a group of materials [2] as shown in Fig. 2. For five metals in
this group, the strain-rate effects are shown to be not very important in
the experimental investigations on the high-frequency fatigue of these
metals [18]. These experiments Were carried out in a magnetostriction
apparatus at 14 kc as described in Ref 19. It is known that in both cavi-
tation damage tests and liquid-impact tests [7,19,20], the rate of damage
is time dependent in the initial "zones of damage," and it finally reaches
a steady state. Successful correlations have been obtained only in the
steady-state zone.
If one defines the intensity of erosion ^s the power absorbed by the
material per unit area, then the intensity / is given by
where:
Ae = area of erosion,
t = test duration, and
i = average depth of erosion.
In our cavitation erosion test, the relationship between the rate of ero-
sion and the amplitude of vibration is obtained as shown in Fig. 3 [21].
The intensity of erosion was calculated by assuming that the erosion
strength is identically the same as the strain energy for the group of ma-
terials shown in Fig. 2. The correlation shown in this figure provides the
justification for the foregoing assumption. Furthermore, the high-fre-
quency fatigue tests mentioned earlier show that materials such as 316
stainless steel, Monel, tobin bronze, 2024 aluminum, and 1100 alumi-
num do not exhibit significant effects of strain rates [18]. The same
group of materials (or one among this group) may be used to determine
the intensity of erosion of a given test device under a set of test condi-
tions. We may call this procedure the calibration of the test device
wherein we obtain the numerical value of the intensity of erosion of the
test device. Once such a calibration is accomplished, the erosion strength
of any material may be experimentally determined by measuring the rate
of depth of erosion with this calibrated test device from Eq 12.
This procedure is feasible with any type of erosion whether it is cavi-
tation, liquid-impact, or solid-impact erosion. It may even be extended
to wear of materials due to friction.
Environmetal Effects on Erosion Strength
It is realized that the erosion strength will be affected by various en-
vironments such as corrosion, temperature, and vacuum just as other
strengths like fatigue, creep, yield, and ultimate strengths are affected by
these environments. The main problem in such cases is the determina-
tion of the intensity of erosion caused by the erosive forces from the
purely mechanical point of view. Once this value is known, then the
measurement of the rate of depth of erosion will give the erosion strength
in that environment, taking into account the environmental effects.
In the following example, the procedure adopted to determine the ero-
sion strength of materials under cavitation in salt (NaCl) solutions is
illustrated. Figure 3 shows the results of erosion tests in a magnetostric-
tion oscillator using different concentrations of NaCl including distilled
water (zero concentration) [21]. From these data one can infer that the
intensity of cavitation damage is not affected by the concentration of
NaCl up to 9 per cent, and it is the same as in distilled water. If we use

a steel specimen instead of aluminum in the same experiment using 3
per cent NaCl solution, the erosion increases threefold as compared to
the erosion in distilled water (Table 2 [27]). The erosion strength is re-
duced to one third of that in distilled water. It is interesting to compare
TABLE 2—Erosion strength of 1020 steel in NaCl solution.°

1 Intensity of erosion at the test conditions using
1100 F aluminum (see Fig. 4)... / = 0.3 w/m
2 Average depth of erosion of 1020 steel in distilled
water î/jt = ~2.5jnjyestt
3 Average depth of erosion of 1020 steel in 3 per cent
NaCl solution (__//_/ = ~0.7 inVyâjL
4 Erosion strength of 1020 steel = 5e = ////
In 3 per cent NaCl solution. .. ^^j'^'P^
In distilled water ^S7^J20J300j>s
0
Data from Ref 21 are used; nomogram of Ref 13 is also used.
FIG. 4—High-frequency corrosion fatigue of SAE 1020 steel (Ref 18).
this result with the corrosion fatigue data obtained at the same frequency
in 3 per cent NaCl solution as shown in Fig. 4 [18].
The same approach may be extended to other environments and
liquids. Another interesting case is the determination of the cavitation
erosion strength of materials in high-temperature liquid sodium [22,23].
The strain-energy value of Type 316 stainless steel at the test tempera-
ture was used to estimate the intensity of erosion. The erosion strength
of other metals, for which even the simple tensile properties are scarce
in these environments, may easily be determined by this method.
Conclusions
While the threshold conditions of erosion may be correlated to exist-
ing properties such as dynamic yield strength and endurance limit, there
is no common property that can be used for predicting the volume of the
material eroded due to fracture of material from its surface. Recent
suggestions to use the strain energy of the material have a few limitations
such as strain-rate effects, environmental effects, and availability of
strain-energy values. The concept of erosion strength is introduced spe-
cifically to overcome these limitations and to provide the designer with
some numerically expressed property for designing erosion-resistant
structures. The methods to determine the erosion strength in various en-
vironments are outlined. Although the examples used for the discussion
pertain mostly to cavitation erosion, it is believed that these methods can
be equally applied to any type of erosion tests involving liquid- and
solid-impact phenomena.
Acknowledgments
This investigation was supported by the Office of Naval Research, De-
partment of the Navy, under Contract Nonr-3755(00) (FBM), NR 062-
293.
References
[1] A. Thiruvengadam, "A Unified Theory of Cavitation Damage," Journal of
Basic Engineering, Transactions, Am. Soc. Mechanical Engrs., Vol 85, Series
D, No. 3, September, 1963, pp. 365-376.
[2] A. Thiruvengadam and S. Waring, "Mechanical Properties of Metals and
Their Cavitation Damage Resistance," Technical Report 233-5, Hydronautics,
Inc., June, 1964.
\3] F. P. Bowden and J. H. Brunton, 'The Deformation of Solids by Liquid Im-
pact at Supersonic Speeds," Proceedings of the Royal Soc., London, A, Vol
263, October, 1961, pp. 433^50.
[4] O. G. Engel, "Water Drop Collisions with Solid Surfaces," Journal of Re-
search, Nat. Bureau of Standards, Vol 54, 1955, p. 281.
[5] Erosion and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962.
[6] J. H. Brunton, "Deformation of Solids by Impact of Liquids at High Speeds,"
Erosion and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, pp.
83-98.
[7] Deformation of Solids by the Impact of Liquids, The Royal Soc., London,
May, 1965. A summary of this meeting is reported by F. J. Heymann, De-
velopment Engineering, Steam Div., Westinghouse Electric Corp., Lester, Pa.,
June, 1965.
8] J. H. Brunton, "High Speed Liquid Impact," Deformation of Solids by the
Impact of Liquids, The Royal Soc., London, May, 1965.
i[9] D. Tabor, "A Simple Theory of Static and Dynamic Hardness," Proceedings
of the Royal Soc., London, 192A, 1947, pp. 247-274.
10} G. P. Thomas, "Multiple Impact Experiments and Initial Stages of Deforma-
tion," Deformation of Solids by the Impact of Liquids, The Royal Soc.,
London, May, 1965.
[11] J. B. Marriott and G. Rowden, "The Deformation of Steam Turbine Ma-
terials by Liquid Impact," Deformation of Solids by the Impact of Liquids,
The Royal Soc., London, May, 1965.
[12] S. M. DeCarso and R. E. Kothman, "Erosion by Liquid Impact," Erosion

and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, pp. 32-45.
[13] A. Thiruvengadam, "Intensity of Cavitation Damage Encountered in Field
Installations," Technical Report 233-7, Hydronautics, Inc., February, 1965.
(See also Symposium on Cavitation in Fluid Machinery, Am. Soc. Mechani-
cal Engrs., Winter Annual Meeting, Chicago, 111., November, 1965.)
[14] O. G. Engel, "Impact of Liquid Drops," Erosion and Cavitation, ASTM STP
307, Am. Soc. Testing Mats., 1962, pp. 3-13.
[75] O. G. Engel, "Pits in Metals Caused by Collision with Liquid Drops on Soft
Metal Spheres," Journal of Research, Nat. Bureau of Standards, Vol 62, No.
6, June, 1959, pp. 229-246.
[16] P. Eisenberg, "Cavitation Damage," Technical Report 233-1, Hydronautics,
Inc., December, 1963.
[17] R. E. Peterson, "Introduction," Erosion and Cavitation, ASTM STP 307,
Am. Soc. Testing Mats., 1962.
[18] A. Thiruvengadam, "High Frequency Fatigue of Metals and Their Cavitation
1964. (See also Paper 65-WA-UnT-4, Journal of Engineering for Industry,
Transactions, Am. Soc. Mechanical Engrs., Vol 88, Series B, No. 3, August,
1966.)
[19] A. Thiruvengadam and H. S. Preiser, "On Testing Materials for Cavitation
1963. (See also Journal of Ship Research, Vol 8, No. 3, December, 1964.)
[20] J. F. Ripkin, J. M. Killen, S. D. Christ, and R. M. Kuha, "A New Facility
for Evaluation of Materials Subject to Erosion and Cavitation Damage,"
Project Report 77, St. Anthony Falls Hydraulic Laboratory, University of
Minnesota, March, 1965.
[27] S. Waring, H. S. Preiser, and A. Thiruvengadam, "On the Role of Corro-
sion in Cavitation Damage," Technical Report 233-4, Hydronautics, Inc.,
February, 1964.
[22] A. Thiruvengadam, C. Couchman III, and H. S. Preiser, "Cavitation Damage
Resistance of Materials in Liquid Sodium," AIAA First Annual Meeting and
Display, 1964, Washington, D. C. (See also Journal of Space Craft and Rock-
ets, Vol 2, No. 2, 1965, pp. 267-269.)
[23] A. Thiruvengadam and H. S. Preiser, "On Cavitation Damage in Liquid-Metal
Pumps," Transactions, Am. Nuclear Soc., Vol 8, No. 2, 1965 Winter Meeting,
Washington, D. C., November, 1965, p. 405.
DISCUSSION
W. C. Leith1 (written discussion)—The erosion strength postulated in

this paper is based on five ductile metals (Type 316 stainless steel, Monel,
Tobin bronze, 2024 aluminum, and 1100 aluminum) for which the strain
energy from a static stress-strain tension test is related to the reciprocal
of the rate of volume loss in a magnetostriction test; but an opposite
relation is indicated for higher carbon steels and hardened metals as
shown in Fig. 5. Hence, the erosion resistance of ductile metaJs appears
related to the strain energy, but the erosion resistance of brittle metals
appears to be inversely related to the strain energy.
The strain-energy theory of erosion strength assumes that the yield
strength is the same in compression as in tension, which holds for ductile
materials, but the yield strength in compression for brittle materials is
usually at least double the yield strength in tension.
The statement that the strain energy of erosion strength applies to sand
erosion in water turbines appears inaccurate when Ref. 14 by Leith and
Mcllquham states that 202 stainless steel, aluminum bronzes, and 150A
titanium have poor sand erosion resistance and, incidentally, consider-
able strain energy.
Olive G. EngeP (written discussion)—With regard to Eq 3 in the paper,
at the time that this equation was derived3 it was thought that the coeffi-
cient a was a function of the impact velocity and that, if the impact
velocity were to be increased sufficiently, the coefficient a would approach
unity. Since that time, from a consideration of particle velocities and
crater depths produced in solid-solid and liquid-solid collisions, it has
been deduced4 that a is not a function of velocity but is instead a function
of the acoustic impedances of the drop liquid and of the solid. The maxi-
mum value that the coefficient a can have appears to be about 0.4. Con-
sequently, the pressure, p, given by Eq 3 can be at most piCiUf/5, which
will affect Eqs 4 and 5 and the agreement found between Eq 5 and the
data of Table 1.
Equation 3 gives the pressure exerted by a liquid-drop impact against
an unyielding solid (infinite modulus of elasticity). It would seem that,
1
H. G. Acres & Company, Ltd., Niagara Falls, Canada.
2
Evendale, Ohio.
3
O. G. Engel, "Waterdrop Collisions With Solid Surfaces," Journal of Research
of the Nat. Bureau of Standards, Vol 54, 1955, p. 281.
4
O. G. Engel, "Note on Particle Velocity in Collisions Between Liquid Drops
and Solids," Journal of Research of the Nat. Bureau of Standards, Vol 64A, 1960,
p. 497.
DISCUSSION ON CONCEPT OF EROSION STRENGTH 37
in writing a general threshold criterion, the pressure exerted by a liquid-

drop impact against solids that yield (finite modulus of elasticity) should
be used. This pressure is4
FIG. 5—Strain energy related to reciprocal of rate of volume loss.
If the threshold criterion is to be used only for water-drop impacts

against metals, the quantity apiCi/pmCm is negligible. However, for water-
drop impacts against polymers such as neoprene and for mercury-drop
impacts against metals it is not negligible. To make a general threshold
criterion that is applicable to rain erosion of neoprene coatings and drop-
impact erosion in mercury-vapor turbines as well as for water-drop-
impact erosion of metals, the quantity mCm should not be dropped.apiCi/p
In going from Eq 4 to Eq 5 and in going from Eq 7 to Eq 8, the dy-

namic yield strength has been replaced by the endurance limit in one-to-
one correspondence. For steels and for a few other body-centered metals
it is an empirical rule that the ultimate tensile strength is about equal to
twice the endurance limit, and in some cases the ultimate tensile strength
is about equal to the yield strength. However, it would appear that re-
placement of the dynamic yield strength by the endurance limit in one-
to-one correspondence as a generality requires justification.
Finally, in going from Eq 6 to Eq 7, use has been made of the plane-
wave sound speed or rod speed, mC = (Em/pm)11*, where Em is Young's
modulus of the solid and pm is the density of the solid. It has been pointed
out5 that the sound speed that must be used in connection with the crater
depth equation is the speed of sound in infinite medium given by Cm =
[£"m(l — v)/(p m l+ j>)(l — 2?)]1/2, where v is Poisson's ratio for the solid.
However, even by using Cm = (Em/pm)112 I am unable to obtain Eq 7
from Eq 6 except by neglecting the acoustic impedance of the liquid,
Cipi. In the case of water-drop impacts against metals, Cipi <3C Cmpm ,
but for water-drop impacts against polymers such as neoprene and for
mercury-drop impacts against metals, the acoustic impedance of the
liquid begins to approach that of the solid, or dpi ~ Cmpm . If the acoustic
impedance of the liquid has been neglected in obtaining Eq 7 from Eq 6,
then Eq 7 should be given with an approximation sign, the extent of the
approximation should be indicated, and the use of Eq 7 should be re-
stricted to water-drop impacts against metals.
J. B. Marriott6 (written discussion)—There is considerable evidence to
show that erosion occurs by single or multiple pressure pulses which
may be the result of a collapsing bubble or an impinging droplet. These
events occur over very localized areas, and the effect of the pressure pulse
will depend on the properties of the material in this area. In his approach to
the subject, the author considers the material being eroded to have con-
tinuous properties as would be found in single-phase alloys. However,
many of the more erosion-resistant materials are multiphase. In most
cases, the properties of each phase will be significantly different and will
be affected to differing degrees by changes in strain rate.
The consequence of each eroding pressure pulse will, therefore, de-
pend on its relation to particles of each phase in the alloy. In addition,
the importance of each phase in determining the rate of metal loss may
vary during the course of exposure.
For a theoretical treatment to be of greatest assistance it should be
capable of accommodating these more complex alloys.
6
O. G. Engel, "Pits in Metals Caused by Collision With Liquid Drops and Soft
Metal Spheres," Journal of Research of the Nat. Bureau of Standards, Vol 62,
1959, p. 229.
6
The English Electric Co. Ltd., Central Metallurgical Laboratories, Whetstone,
Nr. Leicester, England.
Gerhard Langbein7 (written discussion)—The concept of erosion

strength given in our paper, "Material Destruction Due to Liquid Im-
pact," and some previous papers8 is similar to the proposal of Thiruven-
gadam. We define as erosion strength, however, the quotient of the
energy offered to the material surface and the eroded volume. Thiruven-
gadam uses instead of the offered energy the absorbed energy, which is
less than the offered energy by a material-dependent efficiency factor. A
proposal could be to call the first definition (offered energy) "experi-
mental erosion strength" and the second (absorbed energy) "true erosion
strength."
A. Thiruvengadam (author)—The author is grateful to the discussers
for their useful remarks. It should be pointed out that some changes were
made in the present version of the paper on receipt of comments by the
referees on the preprint version. It is important to note that the water
hammer theory and Engel's theory agree closely according to the experi-
TABLE 3—Prediction of threshold velocities for hot-pressed alumina and

white sapphire.
pCmercury = 1.966 X 106 g/(cm2. SCC)
Compressive
Pm^m j ^threshold,
Material g/(on2 sec) Strength, YC , cm/sec YC/PI
dyne/cma
Hot-pressed alumina. . 4.345 X 10« 2.8 X 1010 4.276 X 10* 0.95

White sapphire 4.378 X 106 2.1 X 1010 3.514 X 104 0.88
mental data shown in Table 1. This agreement, despite the different value
of a suggested by Engel's discussion, shows that in these experiments the
drops were probably not perfectly spherical in shape.
Engel's comment that the ratio of acoustic impedances between liquid
and solid may not always be negligible is well taken. There is a surpris-
ingly close correlation between the water hammer theory and the experi-
mental data reported by Engel9 on the resistance of white sapphire and
hot-pressed alumina to collision with mercury drops. Since the ratio of
impedances are not negligible for this case, the impact pressure Pt is
given by
1
Dornier System GmbH, Friedrichshafen, Germany.
8
"Regenerosion von Kunststoffen," Z. Kunststoffe, Vol 56, Heft 1, Jahrgang,
1966.
8
O. G. Engel, "Resistance of White Sapphire and Hot-Pressed Alumina to Col-
lision With Liquid Drops," Journal of Research of the Nat. Bureau of Standards—
A. Physics and Chemistry, Vol 64A, No. 6, November-December, 1960.
It can be shown that the value of Pz is roughly equal to the compressive

strength of hot-pressed alumina and white sapphire when the damage
on these materials is initiated. (See Table 3.)
For repeated impacts, Eq 5 simply implies that one would not observe
any damage even after an infinite number of impacts if the water hammer
pressure produced by the impact was less than the endurance limit of
the material.
With regard to Marriott's question whether the concept of erosion
strength is applicable to heterogeneous materials, it should be pointed
out that erosion itself is a statistical process and that the mean depth of
erosion measured in an erosion test would reflect this. Hence, the con-
cept of erosion strength is equally applicable to complex materials.
Any attempted correlation of erosion with strain energy should satisfy
the following three conditions:
1. The strain energy values should be obtained for the specific speci-
men by measuring the area of the stress-strain curve from an actual ten-
sion test conducted on the same rod stock from which the erosion speci-
mens were machined. It could be misleading to estimate strain energy
from "nominal" values of yield strength, ultimate strength, and ultimate
elongation, as Leith has done.
2. The interaction of test duration should be eliminated. It is now well
known that the rate of erosion depends most significantly on the duration
of test, and only the time-independent rate of volume loss should be cor-
related with any material property. It is apparent that Leith does not
recognize this.
3. The materials used for such correlations should not exhibit strain
rate sensitivity. It is also well known that some of the hardened materials
exhibit high strain rate sensitivity.
The comment by Langbein with regard to his proposal of "offered
energy" and "absorbed energy" is well taken. In fact, we call them
"input intensity" and "output intensity".10 The efficiency tj is given by
/2//i = 1}-, where /i is the input intensity, and 72 = iSe/t is the output
intensity. While the output intensity h can be expressed in terms of rate
of depth of erosion and erosion strength, irrespective of the type of ero-
sion test such as cavitation erosion, rain erosion, turbine erosion, sand
erosion, and so on, the input intensity I\ cannot be expressed explicitly
as a general case. For example, the input intensity of cavitation erosion11
is defined as I\ = p?/piCi, where pc is the bubble collapse pressure, and
piCi is the acoustic impedance of the liquid.
10
A. Thiruvengadam, "A Comparative Evaluation of Cavitation Damage Test
Devices," Symposium on Cavitation Research Facilities and Techniques, Am. Soc.
Mechanical Eng., New York, 1964 (see author's closure).
"• A. Thiruvengadam, "On Modeling Cavitation Damage," Technical Report
233-10, Hydronautics, Inc., 1966.
For liquid drop collisions, /i may be defined as
where:
n = number of drops impinging on a unit area per unit time,
Pi = density of liquid,
d = diameter of the drops, and
Ut = impact velocity.
Similarly, for each specific erosion test the input intensity has to be de-
rived and the efficiency of erosion should be determined. While this is
the overall objective of erosion research, the determination of erosion
strength may be easily accomplished from the output intensity alone.
G. Hoff,1 G. Langbein* and H. Rieger*
Material Destruction Due to Liquid Impact
REFERENCE: G. Hoff, G. Langbein, and H. Rieger, "Material Destruc-

tion Due to Liquid Impact," Erosion by Cavitation or Impingement,
ASTM STP 408, Am. Soc. Testing Mats., 1967, p. 42.
ABSTRACT: A survey is given of the rain erosion behavior of all classes
of materials in the drop-impact velocity range between 200 and 400 m/
sec. The results are compared with the cavitation behavior of matter
and with the results of single water jet impact test. The dependence of
the erosion properties upon velocity, angle of incidence, and drop size
has been studied. If the materials are to be classified by increasing rain
erosion strength of the most resistive representative of a material class,
the order is the following: glasses, plastics, ceramics, and metals. The
process which produces cracks and the material breakup is explained.
The net result of this explanation provides guides for developing mate-
rials with improved erosion strength. By the energy balance of the erosion
process, relations between the erosion strength and the mechanical prop-
erties of materials are given, and characteristic parameters for describ-
ing the erosion properties are proposed.
KEY WORDS: rain erosion, erosion, cavitation, liquid impact, impact,
cold working, crack propagation
Already in 1959 material damage caused by rain erosion of the F 104

G made evident in Germany the importance of basic investigations of
the rain erosion behavior of solids at supersonic speeds. Under sub-
contract to the German Ministry of Defense, a rain erosion test rig was
built up by the Dornier-System GmbH for velocities up to 470 m/sec.
Due to this facility, investigations could be carried out on the influence
of the different test parameters, for example, velocity of the specimens,
angle of incidence of the drops, and rain density, as well as the behavior
of the complete material spectrum with a view to rain erosion destruc-
tion. These investigations were carried out to develop materials with ex-
treme rain erosion resistivity and to propose construction principles for
rain-erosion-resistant supersonic aircraft by means of the discovered laws
governing rain erosion. A water gun has been developed for investiga-
1
Research Dept., Dornier-System GmbH, Friedrichshafen.
2
3
42
HOFF ET AL ON MATERIAL DESTRUCTION DUE TO LIQUID IMPACT 43
tions at extreme impact velocities, by which the individual impact proc-

esses can be tested up to velocities of Mach 6.
Rain erosion is only a special kind of material destruction by liquid
impact. Similar destruction of material occurs in the case of cavitation;
therefore, we have also started investigating cavitation problems.
A laboratory is being built up at Dornier-System for general investi-
gations concerning the behavior of solids under mechanical shock
stresses, where the impact processes from micrometeorite impact, the
resistance against gun projectiles, and liquid impact will be treated with
uniform methods to find the analogs of the different processes.
Investigations of the damage of materials by impacting water drops
were carried out earlier by other authors [1-14}.4 In two review articles
FIG. 1—Test rig (schematic).
of Rheingans [15] and Nowotny [16], a survey is given of the work done
on cavitation.
Test Equipment
Our rain erosion test rig has been constructed according to the
principle of the rotating arm. Figure 1 shows a schematic representation
of the rain erosion test facility. A one-armed rotor of 1.2 m radius, oper-
ated by a gasoline engine, carries at the end of the arm the material speci-
men to be investigated. Eight spraying nozzles for the creation of water
drops are installed above the specimen path, thus allowing a variation
of rain density and drop size. The maximum circumferential speed to be
obtained with this rotor is 470 m/sec.
For the measurement of rain erosion at low speeds and for the in-
* The italic numbers in brackets refer to the list of references appended to this
paper.
FIG. 2—Water gun (schematic).
FIG. 3—Pendulum-shot equipment.

vestigation of the influence of aerodynamic effects, a small evacuating

rain erosion test facility has been constructed with an arm radius of 15
cm and a maximum circumferential speed of 330 m/sec.
Material destruction by cavitation is tested by means of an ultrasonic
generator. The specimens to be investigated are vibration excited in the
cavitation liquid by a magnetostrictive oscillator, with a frequency of
20,000 cps at amplitudes of about 40 /*.
In order to investigate the behavior of materials exposed to the im-
pact of single water jets with velocities up to 2000 m/sec, a water gun
has been constructed, similar to the facility at the Cavendish Laboratory
[17, 18]. Here a lead bullet ejects a small quantity of water out of a
FIG. A—Erosion curve.
funnel-shaped container with velocities to Mach 6. A velocity ratio of

about 5 between the lead bullet and the water jet takes place. Figure 2
shows the basic construction of the water gun.
The energy balance can be established with the pendulum shot equip-
ment and the aid of the energy and impulse equation for the impact of a
steel ball on a material specimen attached to the pendulum. Figure 3
shows the principle of this equipment.
Performance of the Tests
, A method for studying rain erosion is the measurement of the weight
loss A<J. The average erosion depth, em , is: em = AG/y-F, where 7 is
the specific weight and F the specimen area. Figure 4 shows a typical
erosion curve, beginning with a horizontal tangent and turning into the
linear destruction increase after a certain transition period. The variables
which characterize the rain erosion behavior of such a curve are: the
incubation time, tK , the erosion rate, dem/dt, and the erosion resistivity
/* for 0.1 mm average erosion depth. The intersection point of the
backward prolongation of the linear destruction increase with the time
axis represents the incubation time. The time for removal of a layer of
mean depth of 0.1 mm is called erosion resistivity, t*. The following
standard conditions have been established in order to classify the
materials according to their resistivity against rain erosion:
v = 410 m/sec,
drop diameter = 1.2 mm,
pw/L = 1 X 10~5 = water volume/air volume, and
6 = 0 (normal drop impact).
FIG. 5—Dependence of erosion depth upon impact velocity.
All investigations have been carried out on plane disks of 16.8 mm

diameter. Their thickness was dimensioned such that within the ex-
posure time the thickness would not be completely penetrated. The
same specimens of 16.8 mm diameter are used for cavitation tests, too.
Material comparison is performed at a frequency of 20,000 cps and an
average amplitude of 40 p. The destruction of material is measured with
respect to the weight loss. The average erosion depth is calculated the
same as for rain erosion.
The destructive effects on materials caused by limited water jets with
high velocities are tested by the water gun. Velocity measurements are
performed with wire barriers or a high-frequency camera. The crater
depth as a function of the impact velocity is determined.
HOFF ET AL ON MATERIAL DESTRUCTION DUE TO LIQUD IMPACT 47
FIG. 6(b)—Relationship between critical flight velocity and droplet incidence

angle.
Test Parameters
Impact Velocity
The predominant factor in material destruction by rain erosion is the
velocity of the specimen. In considering the energy flux through the
exposed specimen surface, one would predict an increase in rain erosion
destruction with the third power of velocity, since the kinetic energy
increases as v2 and the number of impacts increases again with v. How-
ever, our measurements have shown that the increase in destruction for
the majority of metals, ceramics, and high polymers takes place between
the 5th and 7th power and even reaches the 13th power for glasses.
Figure 5 shows the observed velocity dependence for different mate-
rials. Due to the different velocity dependence, it will become obvious
that a classification of materials according to their rain erosion resistivity
depends on the testing velocity.
Angle
Rain erosion depends severely upon the drop impact angle. Figure 6a
shows the definition of the angle of incidence, 9, and the relationship
between angle of attack and flight direction. Rain erosion in relation to
the drop impact angle and the specimen velocity has been measured for
different materials.
Figure 6b shows several curves of equal rain erosion destruction after
a running time of 12 min at a rain density of 1 X 10~5, measured in the
vO diagram. These curves can be represented with good approximation
as VQ/COS 0, whereby v0 of the curve in question means the velocity at
0 = 0. It can be concluded that only the normal component of the
velocity of impacting rain drops will be responsible for destruction by
rain erosion. Therefore, the angle dependence of rain erosion can be
attributed to the velocity dependence referring to the geometry of the
area impacted. These measurements confirm that the law given by Fyall
[73] for lower velocities is also valid for the highest velocities we could
obtain. According to this law, only the normal component of velocity is
decisive for material destruction.
Drop Size
With regard to the energy flux, it could be assumed that the eroded
volume at given impact velocity and water quantity will be independent
of drop size. However, within the local pressure region ahead of moving
bodies, frictional forces are acting on the free-falling water drops. Since
these frictional forces decrease with the 2nd power, while the inertia
forces increase with the 3rd power of drop diameter, the acceleration of
drops within the pressure region ahead of the specimen increases with
decreasing drop diameter, that is, the relative velocity between drop
and specimen decreases with decreasing drop diameter. Therefore, the
essential part of drop-size dependence at rain erosion can be attributed
to a velocity dependence. For our test equipment, these aerodynamic
effects are particularly efficient at drop diameters smaller than 0.1 mm.
The velocity alteration with varying drop diameter may be neglected
within the diameter range above 1 mm. Within this range, however, the
roughness of the eroded specimens increases with rising drop diameter,
while the erosion velocity of both specimens seems to be the same. Figure
7 shows two eroded specimens after exposure to different drop diameters,

all other test conditions being constant. In most cases, a more smooth
surface will be connected with more surface_energy per eroded volume.
Têrelolre7therel:o^6rrre~sisfa^c^may beThigher with little drops due to
this effect also.
Temperature
The temperature of a surface moving in the rain will be determined by
two effects acting in opposite directions. The adiabatic compression
within the pressure region ahead of the moving specimen will cause a
velocity-dependent temperature rise, whereas the impacting water drops
FIG. 7—Drop-size effect.
result in specimen cooling. At a specimen velocity of 410 m/sec, a speci-

men temperature of 85 C has been measured, which decreases to ap-
proximately 35 C for exposure to a rain of the volume concentration of
1 X 10~5. Within the measurement inaccuracy, this cooling is even in-
dependent of rain temperature hi the range 10 to 80 C. This effect can
be explained in the following way. The kinetic energy of a water drop
with a velocity of 410 m/sec could not evaporate more than 3 per cent
of its volume. In case of the impact process, most of the kinetic energy
of the water drop will be used to create a new drop surface. This surface
increase caused by several factors will lead to an instantaneous evapora-
tion of the whole drop and thus to the temperature effect measured.
Theoretically, a cooling proportional to ram density and the 3rd power
of velocity can be expected within the velocity range accessible to us,
because the creation of a new surface for one- drop is proportional to
FIG. 8—Comparison of rain erosion behavior of the different material classes.
TABLE 1—Characteristic values for rain erosion resistance of different materials.

Material <K , sec Cm , M/sec lo.i mm > sec
Glass (triplex twin grd. qual.) 0.5 1360 0.5

Plexiglas 1.2 400 1.1
Aluminum 7 14 15
Polyamid 77 6 83
Aluminum alloy 90 2.5 130
Polyurethane 9486 16 0.8 137
AlzOs Sintox alumina . . 270 0.4 530
Iron 320 0.5 500
Titanium Lt31 540 0.5 760
Steel X20Crl3 1500 0.35 1 800
Electrodeposited nickel layer 5580 0.03 6 600
Tempered steel 100Cr6 4200 0.01 13 000
the 2nd power of the velocity, and the impacting rain quantity is again
proportional to the velocity.
Material Comparison
Figure 8 shows, in general, the rain erosion behavior of the total ma-
terial spectrum by an em(f) diagram plotted on a log-log scale. The ma-
terial comparison has been performed with an impact velocity of 410
FIG. 9—Comparison of cavitation behavior of some materials.
m/sec and a rain density of pw/L = 10~5. Of all material classes, some
characteristic representatives had been selected for this purpose. The
group of inorganic glasses is represented by only one material, since all
types tested so far have nearly the same erosion strength. If rain erosion
resistivity of the individual material classes is classified according to
the best material of each class, the following increase of rain erosion
resistivity will be recognized: glasses, plastics, ceramics, and metals.
Rain erosion resistance increases by more than 105 from glasses up
to the most resistant hard metals. Table 1 gives, in general, rain erosion
characteristics of the most important materials under our standard con-

ditions.
Figure 9 gives a survey of cavitation resistance of some materials by a
linear representation within the em(f) diagram. The measurements have
been performed by the oscillator at 20,000 cps with an amplitude of 40
ft at 20 C in distilled water.5
The sequence of resistivity of materials against rain erosion and cavi-
tation is similar. Deviations of the resistivity sequence can, among others,
be recognized hi Al CuMg and macrolone and are partly due to the cor-
rosion influence, which will be more severe in the case of cavitation.
FIG. 10(a)—Effect of a single water jet on pure aluminum.
The results obtained with the water gun cannot directly be compared
with our rotating-arm ram erosion investigations, since the impact
velocity of the water jet is beyond the velocity range presently achievable
by the rotating arm; but the order of resistivity of the different material
classes is the same as for rain erosion.
Figure 100 shows the material destruction of aluminum by a water
jet with a total volume of approximately 100 mm3 and 2000 m/sec. Fig-
ure 10£ shows the crater formation dependence with impact velocity of
Plexiglas. The velocity increases linearly from 600 to 1300 m/sec.
8
The results on cavitation behavior of different materials have kindly been
placed at the authors' disposal by R. Hammesfahr, Dornier System.
FIG. 10(fe)—Water jet impact effect as a function of speed on Plexiglas.

The Destruction of Solids by Impacting Drops of Water and Improving

the Resistivity of Materials Against Drop Impact
Metals
To improve the resistivity of metallic materials against drop impact
by suitable means, it is first required to clarify the mechanism of destruc-
tion of metallic materials by the impacting water drops. Thereby, the
following methods of investigation have been applied [22]: determination
of weight loss; measurement of microhardness and coercive field
strength; and light- and electron-microscope investigation.
Measurements of the weight loss of the specimens and their depend-
FIG. 11—Correlation between microhardness, coercive field strength, and

weight loss as a function of exposure time.
ence upon exposure time to rain are suitable for a fast general view of
the behavior under rain erosion of the material to be investigated. From
such em(t) curves (Figs. 4 and 8), for example, the incubation time or
the maximum erosion rate can be taken as suitable parameters to de-
scribe the rain erosion behavior of the material in question.
For clarification of the destruction processes which happen at impact
of drops on metals, this method is not quite suitable, however, since in
this case it is required to investigate the changes of the material right
from the beginning of exposure to rain. Therefore, among other things,
it must be decided whether the destruction of metals by drop impact is
corrosive or mainly erosive in nature, that is, whether there will occur
mainly a plastic deformation and thereby strength hardening of the
metal.
Measurement of microhardness and coercive field strength may con-
tribute to the solution of this open question. In Fig. 11 the increase of

field strength6 and microhardness of pure nickel is plotted as a function
of exposure time in comparison with the weight loss.
The remarkable point is that right at the beginning of exposure time
a strong increase of coercive field strength and also of microhardness
occurs. Both properties reach their maximum value near the end of the
incubation time of the AG(f) curve. These measurements show that the
internal stresses will increase very strongly at the beginning of the ex-
posure time due to plastic deformation of the surface of the metal, ow-
FIG. 12—Slip lines, deformation twins, and breakout on the surface of a co-
balt single crystal.
ing to the impacting rain drops. Toward the end of the incubation time
on the AG(0 curve the internal stresses reach their maximum value de-
fined by the rupture strength of the metal and which cannot be ex-
ceeded. On reaching this maximum level of the internal stresses, cracks
will form, spreading out during further impact exposure and thereby
causing weight loss.
Light-microscope investigations, carried out during the incubation
6
In ferromagnetic materials having a magnetostriction constant unequal to
zero, the coercive field strength in some cases will be severely changed by internal
stresses. In such materials, measurement of the coercive field strength, therefore,
is a sensitive method to prove internal stresses.
time of the AG(f) curve, indicated that in some cases the surface of the
metal shows many slip lines. Figure 12 is an example of such light-
microscopic observation. It shows slip lines and deformation twins on the
surface of a cobalt single crystal which has been exposed to rain for 20
sec at a velocity of the specimen of 410 m/sec. Furthermore, from Fig.
12 one can recognize a distinct connection between breakouts of mate-
rial and deformation twins, respectively slip lines. In general, breakouts
of material occur hi areas of intense formation of slip lines or at defor-
mation twins, that is, at locations where high internal stresses are present
within the material.
To supplement the light-microscope observations, electron-microscope
investigations have been carried out. They offer the advantage of higher
resolution and the possibility of observing the behavior of defects in the
crystal structure within the interior of the metal by use of transmission
observations of electrolytically thin-polished metal foils. These investi-
gations show that during the incubation tune many lattice defects will be
formed, mainly in form of dislocations.7 Toward the end of the incuba-
tion time on the AG(0 curve the dislocation density will amount at 1012
to 1013 dislocations/cm2. Such high dislocation densities are observed
hi metals only immediately before rupture of the metals.
In order to increase the resistivity of metals against rain erosion, the
following statements can be made.
It has to be avoided that during exposure to ram impact strong local
stress peaks will be formed within the metal, or that considerable localized
deformations will occur, giving rise to material contractions. Since both
effects, that is, the creation of peaks of internal stress and strong slip,
are connected with moving dislocations, it is necessary to decrease their
mobility as far as possible by suitable dislocation obstacles. These dis-
location obstacles should be as small and as homogeneously distributed
as possible within the material. Thus, large concentration of dislocations
can be prevented at these obstacles and hinder the formation of cracks.
Dislocation obstacles which decrease the mobility of dislocations are
precipitations, grain boundaries, martensite needles, and other disloca-
tions. In the following, the effectiveness of such dislocation obstacles
will be demonstrated by some examples. The common feature of these
tests is the variation of number and size as well as spatial distribution
of dislocation obstacles and the investigation of the changes of rain
erosion resistivity caused thereby.8
The size and spatial distribution of precipitations can be changed best
by suitable heat treatments of age-hardenable alloys. Figure 13 shows
7
Dislocations are linear defects in the crystal structure, which are, by reason
of the strains within the crystal structure, surrounded by a field of mechanical
stresses.
8
Rain erosion resistivity is difficult to determine by only one single measure-
ment datum. Therefore, in the following, the maximum temporary weight loss
or the incubation time have been selected as dimensions for rain erosion resistivity.
FIG. 13—Rain erosion of beryllium copper.
FIG. 14—Relation between erosion rate and tensile strength.
the rain erosion behavior for copper as well as for various age-hardened
beryllium-copper alloys (2 per cent Be). The tensile strength, <rB, has
been chosen as a measure for the age hardening, caused by a certain
heat treatment. It is recognized that rain erosion resistivity increases
with increased tensile strength. This result is concluded since the dislo-
cafion~oBstacie7lnThe tornTof precipitations cause not only an increase
in rain erosion resistivity, but simultaneously an increase of tensile
strength and microhardness. From Fig. 14, which shows (d^G/df)m&JL
as a function of <TB in a double-logarithmic scale, the following propor-
tionality results:
This trend of increased rain erosion resistivity for rising tensile strength,
or microhardness, is also observed for other age-hardenable alloys, as,
for example, AlCuMg-, Be-Ni-, Ti-6Al-4V-alloys and maraging steels.
Our investigations have shown that not only precipitations formed by
heat treatments of age-hardenable alloys will cause an increase of rain
erosion resistivity. For example, an increase can also be effected by small
solid oxide, carbide, or nitride particles, if they are distributed homo-
geneously within the material in the form of small particles. Experimental
examples are the increase of rain erosion resistivity of dispersion-hard-
ened metals and carburized and nitrided steels.
Martensite needles are particularly suitable obstacles for mobile dis-
locations. Their influence on rain erosion resistivity has been investigated
in plain carbon steels mainly by Herbert [23]. He obtained the following
results: The martensite formation due to quenching the specimen in
water will cause a five to sixfold increase of incubation time of these
steels in comparison with the solution-treated state. With some few
exceptions, all carbon steels, solution treated as well as in the hardened
state, the rule has been confirmed that rain erosion resistance increases
with growing hardness, respectively tensile strength.
Investigations of the influence of gram boundaries on rain erosion re-
sistivity of metals have been carried out for sintered iron of different
grain size. It was observed that the rain erosion resistivity increases with
decreasing grain size. This tendency was expected since the number of
gram boundaries, which are effective as dislocation obstacles for mobile
dislocation, increase to the same extent as the mean grain size is dimin-
ishing. In the case of all other metals tested, the investigations have
shown that rain erosion resistivity of the metal in question will increase
with the fineness of the grain structure.
In order of their mechanical stress field, the dislocations interact with
each other. This dislocation characteristic, acting as a dislocation ob-
stacle to itself, also results in increased rain erosion resistivity. For ex-
ample, investigations show that the incubation time of cold-strengthened
copper specimens is approximately 25 per cent greater than the incuba-
tion time of annealed copper specimens. These test results show that
rain erosion resistivity can be increased by means of homogeneous de-
formation of a strain-hardenable metal.
The examples of experimental investigations mentioned here are, on

the one hand, a confirmation of the destruction model of metals in rain
erosion as discussed before; however, they supply directions for im-
proved rain erosion resistivity of metallic materials by application of
certain treatments. The examples given show that hi general the follow-
ing rule applies to a certain metal or within a series of alloys: All treat-
ments causing an increase of microhardness or tensile strength will
simultaneously lead to an increase of rain erosion resistivity.
Thus, microhardness, critical shearing stress, and tensile strength are
suitable reference parameters in order to determine roughly the qualities
of a certain metal as to rain erosion resistivity.
FIG. 15—Rain erosion of some glasses, ceramics, and a special sapphire.
Ceramic Materials and Glasses

Dielectric ceramic materials are used hi aeronautical engineering, be-
cause of their electrical transmission ability, to protect radar equipment;
glasses serve as material for windows as well as protective covers for
optical instruments. Thus, investigations on the destruction of these ma-
terials by drop impact are also of great technical interest.
Ceramic materials and glasses are brittle. Mechanical stresses result-
ing from plastic deformation remain concentrated in a narrow space so
that cracks can easily develop and propagate. This is the reason for the
brittleness of these materials, a fact clearly recognized in the destruction
in rain erosion too. The causes for this brittle-fracture behavior are
somewhat different for ceramics and glasses. Ceramic materials have
a crystalline structure; dislocations, however, are almost immovable.
Glasses are amorphous materials and therefore have no dislocations at
all. Therefore, interior stresses in glasses as well as in ceramic materials

remain considerably localized and give rise to the formation of cracks.
The AG(0 curves of glasses and ceramics generally also show an in-
cubation time. However, weight loss occurring is extremely inhomo-
geneous, and the measurement results fluctuate rather severely, since
large parts of the specimen fly off during the exposure to rain because
of severe crack expansion. The formation and propagation of cracks in
these brittle materials is dependent to a great extent on already existing
microcscopic or macroscopic faults of the specimen. Surface scratches
or inhomogeneities theoretically decrease the expected mechanical
strength of these materials to a larger extent than is the case for metals or
plastics.
Figure 15 shows, for example, that the resistivity of glass-plastic com-
binations9 will be higher than that of pure glass. This is intelligible, be-
cause crack expansion within the plastic layers of such glass-plastic
combinations will be severely restricted. One can recognize, from Fig.
15, that rain erosion resistivity of glasses is lower than that of ceramic
materials and that the resistivity of ceramic materials increases in the
order magnesum-oxide zirconium-oxide, aluminum-oxide. This had
been expected, for the mechanical strength of these materials also in-
creases in the same sequence.
Investigations with sintered AÔs specimens showed that the resistiv-
ity increases also with decreasing porosity, that is, with increased density
of the specimen and the attendent increase of mechanical strength.
Within the series of specimens investigated, a sapphire crystal with a
c-axis parallel to the surface has proved most resistant. This indicates
that the monocrystalline material in the case of ceramics is more re-
sistant to rain erosion than polycrystalline ones.
Figure 15 also shows that sapphire, used as safety windows for op-
tical equipments, has a far better resistivity to rain erosion and somewhat
better optical characteristics than the ultrared windows manufactured of
quartz glasses, used now on the F-104.
A comparison of the velocity dependence of rain erosion rate by
drop impact of metals compared with glasses and ceramics shows that
ceramic materials and generally also glasses are more resistant at lower
velocities than soft metals, for example, pure aluminum. This behavior
can be explained by the level of the critical tensile strength and the
brittle-fracture behavior of these materials. The critical tensile strength
(a criterion for the beginning of plastic deformation under mechanical
stress) is generally somewhat higher for glasses and ceramics than for
soft, ductile metals. Thus, the force of impacting water drops at lower
velocities is sufficient to plastically deform materials such as aluminum;
'These combinations are materials with stratified structures, consisting of glass

and synthetic laminations.
it is not sufficient, however, for the deformation of glasses and ceramics.

At greater velocities, when resulting stresses exceed the critical tensile
strength of glass, ceramics, and metals, the velocity of crack expansion
will represent the decisive destruction phase. Whereas metals in this
velocity range can still plastically deform before complete destruction
takes place, a fast destruction occurs in glasses and ceramics, due to
severe crack propagation as a result of the afore-mentioned brittle-frac-
ture behavior of these materials.
FIG. 16—Relation between erosion rate and notch impact strength.
Plastics
Considering the resistivity of plastics including Plexiglas, Plexidur,
macrolone, and polyethylene (see Fig. 8), it will be observed especially
that the brittle materials are less resistive than the tough substances.
Thus, the notch-impact strength should be a criterion for the resistivity
of materials. Figure 16 shows the relation between notch impact strength
and erosion rate, which is inverse to the resistivity. Indeed, a decrease
of erosion velocity can be observed with increasing notch-impact strength.
The notch-impact strength increases with increasing molecular weight
of polymers. This molecular weight effect is easy to recognize in the
low-pressure polyethylene series (see Fig. 16). Since the workability of
polyethylene decreases with increasing molecular weight, we have tried
to obtain the effect of higher molecular weight by radiation with electrons
after processing. An increase in resistivity up to a maximum at 20Mrad10
10
Mrad= 108erg/g.
can be obtained with low-pressure polyethylenes as well as with high-

pressure polyethylenes. For extended radiation, polyethylene will em-
brittle, thus causing decreased values of resistivity. The described rela-
tion between notch-impact strength and rain erosion resistivity will be
established theoretically in the following section.
Elastomeric material, especially polyurethane (see Fig. 8), has ex-
TABLE 2—Relation between Shore hardness and weight loss for some elastomers.
Weight Loss Shore A- Material Exposure
Hardness Time, sec
15 80 polyurethane 40
22 74 ethylene-propylene-copolymerisat with soot 5
20 70 ethylene-propylene-copolymerisat with soot 5
5 62 ethylene-propylene-copolymerisat with SiO2 5
2 60 ethylene-propylene-copolymerisat with SiO2 5
7 75 natural rubber 5
4 66 natural rubber 5
FIG. 17—Dependence of erosion strength upon softener concentration at poly-

vinylchloride.
cellent rain erosion resistivity. It is natural to correlate the resistivity of

this kind of material with its compliance, respectively the E-modulus.
Thus, Table 2 gives the Shore hardness in relation to the loss of weight.
In all cases described, the "softer" materials with lower Shore hardness,
respectively lower E-modulus, are more resistive. The reason is that in
softer rubbers the drop impact causes a lower stress than in hard rubbers.
Now, the relationship between erosion behavior and relaxation be-
havior will be described by the example of softened polyvinylchloride.
Erosion resistivity of polyvinylchloride softened with dibutylsebacate
first decreases on increase in softener concentration (see Fig. 17) to a

minimum at 10 per cent softener content, then increases to a maximum
of approximately 20 per cent softener content, and decreases again be-
cause of diminishing strength of the material. The initial decrease of
resistivity is in accordance with the well-known alteration of relaxation
behavior of this substance when softeners are added. As a result of the
quasi-cross-linking effect of the polar softener molecules, an embrittle-
ment of the material will occur. This embrittlement is associated with a
movement of the temperature maximum of the mechanical-loss factor
toward higher temperatures. If the softener concentration is increased
further, the softener causes a decrease of viscosity of the substance.
This results in a shifting of the tan 8 maximum to lower temperatures
and simultaneously hi an increase of resistivity against drop impact. It
FIG. 18—Explanation of the relation of temperature and frequency integral

of mechanical-loss factor.
shows that the decrease of characteristic temperatures of the relaxation

spectrum causes an increase of resistivity of the substance. We think
the explanation is the following: The more transport processes are
available for the removal of impact energy from the impact area, the
less absorption will take place in the surroundings of the impact area.
All these processes, of which the time constants are smaller than the
velocity of the relative deformation of the substance at impact, <o
106 cps, are to be considered. A measure for the amount of these proc-
esses should be the integral of the mechanical-loss factor tan 8 from w0 to
infinity
Measurements of the mechanical losses at high frequencies are incom-

plete in most cases. However, it is possible to deduce the energy absorp-
tion at high frequencies from torsion pendulum tests at low frequencies,
for example, 1 cps, and low temperatures. Therefore, the loss-factor
integral from 0 K up to a certain temperature T*
should be an equivalent measure for the amount of relaxation processes,

whereby the temperature T* is defined by the Arrhenius relation for
the time constant TO ^ lA>o of relaxation processes
This relation is illustrated in the frequency temperature diagram (Fig.

18) as the dashed line, which corresponds to the "phase-limit," explained
in Ref 19, between the flexible and rigid state of a material. The tem-
perature integral of the loss factor will increase with falling temperature
rmax of the loss factor maximum. Therefore, with decreasing tempera-
ture Tjnax, the erosion resistivity of a number of similar substances
should increase.
Material outbreak begins in all classes of materials for the normal
rain erosion test conditions only after one surface element has repeat-
edly been hit by drop impact. Thus, the destruction mechanisms of the
material classes mentioned have in common a critical quantity. If this
value is exceeded, the outbreak will begin. This quantity is the interior
stress arising with deformation of material, or the plastic energy con-
nected herewith. As soon as the interior stress reaches certain critical
values for the creation of a crack, cracks are formed, propagate, and
combine, and finally the material will break out.
Energy Balance of Erosion

In order to find relations between the erosion behavior and other me-
chanical properties of matter, it will be useful to observe the energy
balance of the rain erosion process in the stationary state [20, 21}, that
is, at a constant erosion rate. The energy flu mpacting the material
surface per time and area will be
A portion <£X of the energy flux <£ will be absorbed by the material,
whereas the remainder will be reflected, or transmitted through the
material. The energy absorption leads to erosion. It is assumed that for
erosion the energy per volume of the material, e, will be necessary.
According to this, the following equation must apply:
where: F = exposed area and AF = volume element eroded within the

time A?.
It can be concluded:
The quotient of e and X will be specified as erosion strength, /, in the

following. The value of/can be calculated with reference to the afore-
mentioned erosion velocity as follows:
Now, we can further separate the values e and X. The quantity e, the
energy per volume necessary for erosion, is a measure for the energy
receptivity of a material until complete destruction occurs. Therefore, it
should be proportional to the energy of fracture of a material which
can formally be divided up into a critical stress, <r 0 , and a critical strain,
So:
The absorption factor, X, can be separated into that portion of energy

which passes through the surface, and that portion, X', which is ab-
sorbed below the surface. The first quantity can be determined by
means of calculations for the Hookean range of material deformation,
.if the sound impedances, Z, of the material and, Zw , of water are
known:
Thus, the following value is obtained
The factor X' will be a function monotonically increasing with the

relation of drop-impact pressure, p, to the critical tension <TO . A simple
formulation for this function can be obtained by assuming propor-
tionality of X' and the energy below the stress-strain curve from zero
to the impact pressure/?:
The impact pressure;? will be proportional to.the velocity
if one calculates within the Hookean range of deformation, which is

valid for the first instant of impact.
The more relaxation mechanisms which are available for energy

removal, the smaller will become the factor /3. Thus, it will decrease
monotonically with the loss-factor integral stated before. Reciprocal
proportionality will be the simplest hypothesis for this relation. After
all the aforementioned have been substituted in the equation for rain
erosion strength, the relation with the mechanical properties of a material
is:
where:
o-Q = critical tension,
£o = critical strain,
Z = sound impedance of the material,
Zw = sound impedance of the water,
v = impact velocity,
co = measuring frequency,
T = temperature,
T* = critical temperature, and
tan 8 = mechanical-loss factor.
It will be possible, by means of this formulation, to explain the
relations determined empirically.
The equation
relating tensile strength and the maximum erosion rate has been found
for copper and a number of copper-beryllium alloys. This corresponds
to the proportionality of erosion strength with the square of the critical
tension <r0 in Eq 13.
The linear increase of erosion strength with the energy a0£o under the
stress-strain curve at constant tension, o-0 , will correspond to the ex-
perimentally determined decrease of the erosion rate with increasing
notch-impact strength. Therefore, the notch-impact strength will present
a better measurement value for erosion strength than the energy of
fracture during static test does, since higher strain rates will occur at
i notch impact. Thus, a better approximation to the conditions at high
\strain rates of 106 cps, at drop impact, can be obtained.
A third relation can be recognized from the general formulation.
When the sound impedance Z is comparable to that of water, as occurs
in elastomeric materials, / will become a function monotonically in-
creasing with Z, since Z appears twice within the denominator. Erosion

strength has to increase with decreasing Shore hardness, because Z is
proportional to the square root of the Zs-modulus, and since the Shore
hardness represents a measurement value for the E-modulus. This corre-
sponds to the results obtained for elastomeric materials as given in Table
2.
It can be recognized that the higher the Shore hardness, the higher will
be the weight loss. In most cases it will not be possible to turn to rather
low Shore hardnesses since this will lead to a decrease of tensile strength.
Therefore, a maximum of erosion resistivity seems to be obtained for an
intermediate value of Shore hardness. The optimum Shore hardness we
observed at 70 Shore, whereas Schmitt [24] has found the optimum
Shore hardness at 80 Shore. This can be explained by the increasing in-
fluence of the E-modulus with rising impact velocity, as may be seen
from Eq 13. Our measurements have been performed at 400 m/sec;
those of Schmitt at 200 m/sec.
The influence of relaxation behavior as discussed before is expressed
by the proportionality of erosion strength to the loss-factor integral.
Finally, it should be emphasized that the increase of erosion with
drop-impact velocity, which will almost evoke the impression of velocity
limits for erosion, can also be recognized from Eq 13. With increasing
velocity, the impact pressure will be raised. When the impact pressure
reaches the limit of elastic behavior of the material, the integral
of the energy below the stress-strain curve, increases extremely. This

means a severe decrease of erosion resistivity will occur at the velocity
corresponding to this pressure.
Proposals for Characteristic Erosion Behavior Describing Parameters
For a comparison of the results obtained on different test rigs, it will
be important to characterize the materials with the aid of data which are
as independent as possible of special test conditions.
The erosion resistivity /:
represents such a value for the characterization of erosion in a stationary

state, which better conforms to the aforementioned condition than the
erosion rate em . Whereas em is proportional to the energy flux impacting
the surface, the erosion strength will thus become independent.
For the construction of the erosion curve, a second parameter is
necessary, which defines how strong the incubation of erosion is marked.

Therefore, the so-called "incubation depth" is suited best; it is impact
energy per area during incubation divided by the erosion strength:
The tangent to the erosion curve can be constructed by means of/and

eK . The quantity eK has to be established on the negative em-axis. Then,
a straight line with a slope of em = <£//must be drawn through this
point.
A value which is important in practice, the erosion resistance t*, can
easily be derived from eK and /. The term t* means the time necessary
FIG. 19—Different roughness at equal erosion depth.
for erosion of a certain critical depth, e, for example, 100 n. The following
formula is valid for this quantity:
The incubation depth, eK , is related to the average depth, em , to which

energy will enter below the surface during incubation and thus also to
the roughness of specimens at the beginning of erosion. Figure 19 dem-
onstrates this in specimens of polypropylene and of high-pressure poly-
ethylene. Both specimens show almost the same loss of weight. The
polypropylene specimen has considerably greater roughness than the
polyethylene specimen. According to this, eK of polypropylene is up to
1100 fi, whereas it is only 200 p in the case of polyethylene. Yet, it must
be stated that the relation applying in this case will not be generally
valid, since the roughness of a material is also determined by its inhomo-
geneity, whereby fractures may occur.
References
[1] E. Honegger, "Uber Erosionsversuche," BBC. Mitt., 14, 1927, pp. 74-95.
[2] J. Ackeret, "Uber Hohlraumbildung (Kavitation in Wasserturbinen)," Escher
WyssMitt. 1, 1928, pp. 40-45.
[3] J. Ackeret and P. de Haller, "Untersuchung uber Korrosion durch Wasser-
stoss," Schweizer Bauz., 98, 1931, p. 309.
[4] M. Vater, "Das Verhalten metallischer Werkstoffe bei Beanspruchung durch
Fliissigkeitsschlag," Z. VDI, 81, 1937, pp. 1305-1311.
[5] M. Vater, "Das Verhalten verschiedener Stable bei Fliissigkeits-Schlagbean-
spruchung infolge Olkavitation," Verlag Technik, Berlin, 1951, p. 24.
[6] M. Vater, "Wasserschlag-Dauerversuche an reinem Eisen," Z. VDI, 82, 1938,
pp. 672-674.
[7] M. v. Schwarz, W. Mantel, and A. Steiner, "Tropfenschlaguntersuchungen zur
Feststellung des Kavitationswiderstandes (Hohlsog)," Z. Metallkunde, 33,
1941, pp. 236-244.
[8] M. v. Schwarz and W. Mantel, "Die Zerstorung metallischer Baustoffe durch
Wasserschlag," Korrosion und Metallschutz, 13, 1937, pp. 375-379.
[9] W. Mantel, "Untersuchungen mit einem Tropfenschlagapparat zur Erfor-
schung der Zerstorung metallischer Baustoffe durch Wasserschlag," For-
schungsarbeiten uber Metallkunde und Rontgenmetallographie, Hauser Ver-
lag, Miinchen, Germany, 21, 1937, p. 62.
[10] O. G. Engel, "Impact of Liquid Drops," Erosion and Cavitation, ASTM STP
307, Am. Soc. Testing Mats., 1962, p. 3.
[11] S. M. de Corso and R. E. Kothmann, "Erosion by Liquid Impact," Erosion
and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, p. 32.
[12] J. H. Brunton, "Deformation of Solids by Impact of Liquids at High Speed,"
Erosion and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, p. 83.
[13] A. A. Fyall, R. B. King, and R. N. C. Strain, "Rain Erosion Aspects of Air-
craft and Guided Missiles," Journal, Royal Aeronautical Soc., 1962, p. 447.
[14] H. Busch and G. Hoff, "Untersuchungen uber Regenerosion bei Uberschall-
geschwindigkeiten bis zur Machzahl 1, 4," Jahrbuch WGLR, 1962, p. 544.
[75] W. J. Rheingans, "Resistance of Various Materials to Cavitation Damage,"
Report of 1956 Cavitation Symposium, Am. Soc. Mechanical Engrs., New
York, 1957, p. 27.
[16] H. Nowotny, "Erosionspriifung metallischer Werkstoffe," Handbuch der
Werkstoffprufung, Herausgeber E. Siebel, Springer-Verlag, Berlin-Gottingen-
Heidelberg, 1955, pp. 601-621.
[17] F. P. Bowden and J. H. Brunton, 'The Deformation of Solids by Liquid Im-
pact at Supersonic Speeds," Proceedings, Royal Soc., London, Vol 263, 1961,
p. 433.
[18] F. P. Bowden, "The Brittle Fracture of Solids by Liquid Impact by Solid
Impact and by Shock," Proceedings, Royal Soc., London, Vol 282, 1964, p.
331.
[19] G. Langbein, "Die Anwendung der dielektrischen Methode bei der Unter-
suchung von Kunststoffen," Z. Kunststoffe, 51, 1961, p. 503.
[20] G. Hoff and G. Langbein, "Regenerosion von Kunststoffen," Z. Kunststoffe,
56, 1966, p. 2.
/\21] H. Busch, G. Hoff, and G. Langbein, "Rain Erosion Properties of Materials,"
Philosophical Transactions, A, Vol 260, 1966, p. 168.
[22] H. Rieger, "Uber die Zerstorung von Metallen beim Aufprall schneller Was-
sertropfen," Z. f. Metallkunde, 57, 1966, p. 693
[23] W. Herbert, "Einfluss der Warmebehandlung von Stahlen auf deren Regenero-
sionsbestandigkeit," im Konferenz-bericht: Regenerosions Konferenz 1965,
herausgegeben vom BMVtdg, Germany.
[24] H. Schmitt, "Regenerosionsversuche an metallischen und nichtmetallischen
Rotorblattbelagen," im Konferenzbericht: Regenerosions-Konferenz 1965,
herausgegeben vom BMVtdg, Germany.
F. J. Heymann1
On the Time Dependence of the Rate

of Erosion Due to Impingement or
Cavitation
REFERENCE: F. J. Heymann, "On the Time Dependence of the Rate

of Erosion Due to Impingement or Cavitation," Erosion by Cavitation or
Impingement, ASTM STP 408, Am. Soc. Testing Mats., 1967, p. 70.
ABSTRACT: Increasing attention is being given to the observed varia-
tions in the time rate of weight loss during impingement and cavitation
erosion testing. It is generally recognized that a fuller understanding of
these variations is needed for more meaningful use of the test results and
more insight into erosion mechanisms. A pattern often observed is that of
an initial period of little or no weight loss, successively followed by a
period of high rate of loss and one of diminished rate of loss. Some re-
sults, however, follow other patterns, such as a continuously declining
rate or a continuing sequence of fluctuations. There has been no full
agreement concerning the causes for, and the significance of, the various
phases.
Part I of this paper reviews the pertinent findings in the literature and
discusses each of the following influences in some detail:
(a) The effect of the geometrical changes in the eroded surface on the
macroscopic flow, cavitation, or impingement conditions which determine
impact severity.
(b) The effect of the metallurgical and geometrical changes in the sur-
face on the resistance of the surface to the impacts.
(c) The relative significance of the various material-removal mecha-
nisms, such as single-impact failure, fatigue failure, and corrosion.
Part II develops a tentative statistical model of the erosion process for
the case where fatigue is the predominant failure mechanism. The analy-
sis predicts rate-time patterns similar to many of those observed, and it
leads to the conclusion that the instantaneous erosion rates during non-
steady periods are strongly dependent on the scatter associated with finite
fatigue life and with test parameters such as bubble or drop diameters.
While the analysis predicts the eventual attainment of a steady-state rate
independent of that scatter, this state is probably often unattainable in
practice because the increasing roughness of the surface itself affects the
erosion rate.
KEY WORDS: erosion, impingement, cavitation, work hardening, rough-
ness, surfaces, mathematical models, statistical analysis, cracking, fatigue
(materials)
1
Senior engineer, Development Engineering Dept., Steam Div., Westinghouse
Electric Corp., Lester, Pa. Personal member ASTM.
70
HEYMANN ON TIME DEPENDENCE OF RATE OF EROSION 71
The recent literature dealing with the resistance of materials to im-

pingement and cavitation erosion has become increasingly concerned
with the fact that the rate of material loss is not uniform in time. While
this, as a fact, had been noted for many years, some of its consequences
have only lately been emphasized. Thus, as Thiruvengadam and
Preiser [I]2 have pointed out, the comparison of test results can be very
misleading if not based on corresponding phases of the rate-time curve,
and therefore the rather common practice in the earlier literature, of
testing all specimens for the same length of time, is subject to criticism.
FIG. 1—Characteristic rate-time curve according to Thiruvengadam [1].
FIG. 2—Characteristic rate-time curve according to Plesset and Devine [2],

Hobbs [3], and Pearson [4].
The authors of Ref 1 proposed that characteristic erosion-time curves

could be described in terms of four zones: (1) "incubation zone" with
no weight loss, (2) "accumulation zone" with loss rate increasing to a
peak, (3) "attenuation zone" with decreasing loss rate, and finally (4)
"steady-state zone" with constant loss rate (Fig. 1). They do not at-
tempt any detailed explanation for these zones, but suggest that the
first three zones are influenced by the initial condition of the surface
and that only the final zone is truly characteristic of the material itself
and that it should be used for comparison or correlation purposes.
This particular suggestion was disputed by Plesset and Devine [2],
2
paper.
who showed photographically that in a magnetostrictive oscillator the

attenuation zone is associated with a cavitation cloud of much reduced
intensity, and attributed this to hydrodynamic damping effects due to
the heavily roughened specimen surface. Moreover, the authors of Ref
2 held that the maximum erosion rate persists as a steady-state rate
for some time rather than forming a narrow peak, as described by
Thiruvengadam and Preiser [1], and that there is no real indication of
any final steady-state zone. A characteristic curve of this type is shown
by Fig. 2. -
Similar statements have been made by a number of recent investiga-
tors. Thus, both Hobbs [3], using a magnetostrictive oscillator cavita-
tion test, and Pearson [4], using a drop impingement erosion rig, have
called the region of maximum erosion rate the "steady-state" period,
and have based then: correlations of erosion with material properties
and test conditions (such as oscillation amplitude or impingement veloc-
ity) on this maximum loss rate. Both have associated the declining loss
rate of the final period with heavy surface damage, as did Plesset and
Devine [2], and feel that it is not a practicable measure of the erosion
resistance. This point of view is also supported by Hammitt [5], whereas,
on the other hand, a point of view similar to that of Thiruvengadam
and Preiser [1] has been adopted by Evans and Robinson [6].
The object in Part I of this paper is to review in some detail the ob-
servations concerning erosion rate-time patterns which can be found in
the literature and to discuss these in relation to the possible influences
of surface-characteristics and material-removal mechanisms. The object
in Part II is to show that a simple statistical model of the erosion proc-
ess—which regards erosion as a multiplicity of fatigue failures—can
predict characteristic rate-time curves of most of the observed types,
and to discuss some of the implications of this model in relation to the
measurement and correlation problem.
Part I—Review of Factors Influencing Erosion Rate
Observed Erosion Rate-Time Patterns

As early as 1928, Honegger [7] presented his impingement erosion
results in terms of "specific erosion" (erosion rate) curves, which gen-
erally exhibited a rising and then falling pattern, although not enough
readings were taken to establish the curves very precisely. His ex-
planation was entirely a geometric one: "As long as the surface of the
specimen is smooth, it offers an unfavorable surface for the impinging
water to attack; hence, the water flows off to either side. Erosion does
not take place therefore for some time. As soon as any roughness is
formed, however, the erosion proceeds rapidly as the jet impinges with
great force in the unevennesses. If, finally, the unevenness has attained
a considerable depth, a layer of water adheres to the now completely

roughened surface. This water absorbs part of the impact of succeeding
drops of water so that the force of the jet is not so effective as formerly.
The specific erosion consequently decreases after a certain depth has
been attained." This explanation, especially of the reduction of erosion
rate as the surface gets roughened, is essentially the same as that ad-
vanced in current papers on impingement erosion.
Curves of weight loss versus time given by Kerr [8], obtained hi a
magnetostriction cavitation device, do not show a distinct zero-loss in-
cubation time, but do exhibit a slope which at first increases and later
decreases. In tabulating the data, he acknowledged this by listing "ini-
tial erosion" based on the first 30 min of test and "relative resistance"
FIG. 3—Typical cumulative erosion-time curves from jet impingement tests,

adapted from Fig. 4 of Ref 10. (u = impact velocity, m/sec.)
based on the subsequent 60 min of test. By contrast, many of the

loss-time curves obtained with wheel-and-jet type of impingement de-
vices, during the 1930's and 1940's, exhibit a very marked incubation
stage with zero or exceedingly small slope, followed by a stage of very
much steeper slopes, but often of rather irregular and unpredictable
shape.3 Such curves were given, for example, by von Schwarz et al [9]
3
A possible explanation for this irregularity lies in the fact that these results
were obtained at relatively low impingement velocities (generally less than 250
ft/sec) against the side of a relatively large diameter jet (up to 10 mm diameter).
Thus, if we accept a fatigue mechanism of failure, the low impact stresses should
result in a pronounced incubation time, and the large jet size—that is, large stress
field applied on each impact—could be expected to result in large erosion frag-
ments when failure does occur. Thus, the loss rate should exhibit a larger random
fluctuation than if erosion took place by the formation of many smaller fragments,
as it might in accelerated cavitation tests where the size of the individual bubbles
is small but the impact stresses are much higher and occur at much greater fre-
quency.
and Brandenberger and de Haller [10] (Fig. 3). However, even in these
data one sees a definite tendency for the loss rate to decrease eventually.
Nevertheless, many authors over the years have smoothed their results
in such a manner as to show the erosion as progressing from an incuba-
tion stage, through an acceleration stage, to a presumed linear or steady-
state stage, and have ignored the eventual diminution of the erosion rate.
Cavitation erosion curves of what might be called the conventional
pattern were shown by Leith and Thompson [11] (Fig. 4). In the discus-
sion to that paper, Plesset held that even during the incubation stage
there is a small but increasing rate of material loss and that surface dam-
age will eventually result in a nonlinearity of the erosion rate.
Most authors have presented then* results in terms of cumulative
FIG. 4—Typical cumulative erosion-time curves from cavitation tests, adapted

from Fig. 7 of Ref 11. (Magnetostriction device, in distilled water.)
loss-time curves, rather than by rate-time curves which require the dif-
ferentiation of empirical data and will magnify all the scatter and uncer-
tainty of those data. Whether, from a curve such as the upper one in
Fig. 4, one would draw a rate-time curve such as Fig. 1 or one such as
Fig. 2 will depend very much on one's belief of what the rate-time curve
should look like. From the practical point of view of quantifying the
results in terms of one or two numbers, it is convenient and not too in-
accurate to draw a straight line through the general region of maximum
steepness, and to specify its slope as a rate parameter and its intercept
as an incubation parameter as done, for instance, by Mathieson and
Hobbs [12]. This empirical procedure is not much affected by whether
the "true" shape is a straight line leading to the rate curve of Fig. 2 or
an S-shaped line leading to that of Fig. 1. (From a theoretical point
of view the distinction is of interest, and this will be discussed in Part
II.) The major criticism which can be leveled at some of the earlier in-
vestigators is that they have not even followed the aforementioned pro-
cedure, but they have tabulated and correlated their results on the basis
of cumulative loss or of loss rates measured at one arbitrarily chosen
point in time which may fall within completely different stages of the
erosion-time pattern for different materials or conditions. The author
has discussed some of the foregoing points more fully in Ref 13.
Less Frequently Observed Rate-Time Patterns
All the previously mentioned results exhibited what may be called
the "conventional" pattern or some minor variation thereof. However,
FIG. 5—Cumulative cavitation erosion-time curves which begin at maximum

rate, adapted from Fig. 24 of Ref 14. (Rotating-disk device at 150 ft/sec)
there are erosion results which do not follow this pattern at all. Thus,
Lichtman et al [14] presented loss-time curves of which many exhibit
no apparent incubation or acceleration stages, but rather begin with a
maximum rate which declines thereafter (Fig. 5). These results were
obtained in a rotating-disk cavitation device.
Exactly the same type of result has been obtained in the spray im-
pingement erosion test facility at the author's laboratory. Erosion rates
invariably seem to begin at a maximum value and then decrease—rapidly
at first, and then more gradually leading into or approaching a lower
steady-state value. Figure 6 shows some "characteristic erosion-rate
curves," obtained by curve fitting through points obtained from several
specimens for each material. One might suspect that incubation and
acceleration stages lie in the region to the left of the curves, as shown,
and were simply missed because initial weight-loss readings were gen-
FIG. 6—Typical erosion rate-time curves obtained in Westinghouse Steam

Division spray impingement facility during 1956-1959.
FIG. 7—Early loss measurements for a titanium (6AI, 4V) alloy tested in the
Westinghouse Steam Div. Facility.
erally not taken until about 2 hr of exposure. In order to check this, we

made measurements on one specimen—a titanium alloy of fairly good
erosion resistance—after 5 min of exposure and several more times dur-
ing the first hour of testing. (In all of the titanium specimens which we
have tested, the erosion rate has continued to decrease for at least 30 hr.)
The result is shown in Fig. 7 and gives every appearance that the erosion
rate does in fact begin at a maximum value, or that if there is an incuba-
tion stage it occurred within the first minute. The latter alternative is
supported by some of the analytical results to be discussed in Part II.
However, it may be worth noting that Thiruvengadam [75] has shown
the rotating disk to be the most intensive cavitation damage device and
that our test facility produces impingement of probably rather small
droplets at a high velocity, probably exceeding 2000 ft/sec. Thus, some
single-impact damage may be occurring in both cases, contributing to
the lack or deemphasis of an incubation period.
At the other extreme of a rather low-intensity cavitation damage de-
vice, deviations from the "conventional" rate-time pattern are also found.
Thus, Hammitt [16,17] has described results obtained in a cavitating
venturi, in which a brief initial stage of high loss rate is followed by a
sort of "displaced" incubation period of little additional loss, again
followed by a stage of significantly higher rate of erosion which may
take on a variety of patterns. In some cases, for example, there is a
series of increasing fluctuations. It should be mentioned, however, that
the initial "high" loss rate represents a total material loss which is too
small to be weighed and must be determined by pit-counting methods.
It is possible that this stage—said to represent the removal of isolated
very weak spots on the surface—occurs so rapidly in accelerated tests
that it simply cannot be measured, and that the whole extent of these
results corresponds to what would be the incubation and acceleration
stages of an accelerated test.
In summary, erosion rate-time patterns such as described in Refs 1
and 2 have long been noted in both impingement and cavitation erosion
test results. Any final explanation of this pattern, with a view to estab-
lishing which characteristic of it should be used for reporting the data
and making correlations, should also be able to account for those re-
sults which deviate from the conventional pattern. Parenthetically, the
comparison between different erosion results and the interpretation of
the variations found in the erosion-time patterns would be made easier
if these results were expressed in a rationalized form, such as "mean
depth of penetration" (MDP), wherever this is feasible. This is becom-
ing more common but is unfortunately not yet standard practice.
Factors Influencing the Erosion Rate

The erosion-time pattern can be influenced by at least three kinds of
effects:
(a) The effect of the geometric condition of the surface on the cavi-
tation or impingement severity and thus possibly on (c).
(6) The effect of the physical and geometric condition of the original
surface, and of the changes in that condition, on the resistance of the
surface to impingement.
(c) The relative significance or interaction of the various material-
removal mechanisms, such as single-impact failure, fatigue failure, and
sometimes corrosion, as determined by the impact severity and fluid
properties, etc.
The desirability of expressing the material loss by a rationalized
parameter was mentioned earlier. Of no lesser desirability is that the
test duration be expressed in some rationalized manner, by which re-
sults may be more readily compared between different tests and extrapo-
lated to operating conditions. In impingement erosion tests, the number
of impacts has often been used, though for many correlation purposes
the volume of water impinged per unit area, as used by Pearson [4], is
preferable and results in a nondimensional erosion rate when combined
with the MDP as a measure of loss. In cavitation erosion, the problem
is somewhat more difficult, though Thiruvengadam [18] has made an
attempt at formulating such a parameter.
For the purpose of investigating the nature of the erosion rate-time
relationship, however, the most fundamental independent variable is
surely some measure of the damage to the surface itself—the cumula-
tive erosion or MDP, or the surface roughness, or perhaps some other
measure. Clearly, any time variation of erosion rate (presuming the con-
stancy of the gross environment) must somehow or other be related to
changes in the material surface, and the discovery of which surface
property provides the best correlation would in itself help the under-
standing of the causality involved. (Part II of this paper implies that the
median lifetime of erosion fragments provides a significant time scale.)
Let us briefly discuss each of the three previously mentioned effects,
with primary reference to ductile, work-hardening materials.
Effect of Surface Geometry on Impingement Severity

The manner in which the roughness of an eroded surface can reduce
the intensity of cavitation attack has been referred to earlier and is dis-
cussed in detail by Plesset and Devine [2] for magnetostriction devices
and by Hammitt et al [17] for cavitating venturi and rotating-disk de-
vices. In impingement erosion devices, a protective liquid layer held in
the depressions of a roughened surface has been supposed to account
for a diminution of erosion rate with time, by Honegger [7] and many
subsequent investigators. One might argue, however, that if this were
indeed so, then while the valleys of the rough surface were protected
the peaks would not be; thus, one would expect this process to converge
rapidly to an equilibrium condition in which both the roughness and the

erosion rate remain constant. This does not generally occur.
Some authors (for example, Hammitt et al [17]) have also supposed
that in the earlier stages of roughening—before the previously men-
tioned effects become operative—the increased surface area due to the
roughening exposes more surface to erosion attack and may therefore
partially account for an initially increasing erosion rate. This seems un-
likely, since if the number of droplets impinging, or number of bubbles
collapsing, against the projected surface area remains constant, an in-
crease of actual surface area would only result in an effective reduction
of the concentration of impacts on the exposed surface.
In impingement erosion one may postulate other reasons why slight
FIG. 8—Schematic illustration of possible effects of small and large surface

roughness.
roughness may tend to increase the erosion rate and heavy roughness
may tend to decrease it. Bowden and Brunton [79,20] have demonstrated
that the high-speed impact of liquid on a ductile metal produces a ring
of roughening, due to plastic deformation, around the impact point.
They have stated that actual material removal in a single such event
is by shear failure of these or previously existing minute surface steps,
due to the high-speed radial outflow of the drop after impact. Even at
much lower impact velocities, one may suppose that small surface
irregularities of whatever source can provide stress raisers for the im-
pact itself and may help to initiate fatigue cracks due to the radial flow-
induced shear forces on the "peaks," which in turn could result in tensile
stresses combined with stress concentrations in the "valleys." (See Fig.
80.)
On the other hand, when the surface damage is gross enough so that
the irregularities exceed the drop size, then the impact in many cases
80 EROSION BY CAVITAT1ON OR IMPINGEMENT
may be effectively a glancing one with a much reduced velocity com-

ponent normal to the local surface, and it has been shown by a number
of investigators recently [4,21] that the normal component of impact
velocity is of paramount significance in determining erosion damage.
However, a droplet hitting directly into the bottom of a depression or
pit (if this is not protected by retained liquid within it)4 may produce a
multiplied or prolonged impact force, due to the lack of a free boundary
from which pressure-release waves can immediately begin to propagate
inward to relieve the water-hammer pressure in the impacting drop,
as is believed to occur (see Ref 20) in impact on a flat surface. Thus,
the foregoing may help explain why erosion tends to take the form of
deep pits rather than a general wearing away of the surface. (See Fig.
86.) Pohl [22] has presented excellent photomicrographs of crack for-
mations at the bottom of erosion pits.
Another geometric effect leading to a reduced impingement severity
and erosion rate is believed to occur in steam turbines, where it has long
been observed that the erosion of low-pressure blades may proceed to a
rather severe stage in the first few months of operation and thereafter
may slow down greatly or even cease completely. Here erosion is due
to large liquid drops, which are swept slowly off the trailing edge of the
previous stationary blade, and do not have time to be fragmented or to
be accelerated to full steam velocity before they are impacted by the
moving blades at a high relative speed and a negative angle of attack.
As the blade is eroded, its "mean" leading edge recedes, and the path
traveled by the drops may be sufficiently lengthened so that the addi-
tional fragmentation and acceleration of drops allows them to better
follow the steam path through the blades. This effect, of course, is dif-
ficult or impossible to reproduce in conventional test facilities.
It is clear that some of the factors discussed in the foregoing influence
the effective impingement velocity and thereby the relative significance
of single-impact and multiple-impact or fatigue-type damage. There is
also an interaction with the drop-size distribution, since a certain rough-
ness may, according to the previous arguments, increase the damage po-
tential of large drops and decrease that of small drops.
Effect of Surface Condition on Erosion Resistance

The surface conditions which affect the resistance of the material itself
may include conditions of prestress, the degree and extent of work
hardening, the prevalence of surface or immediate subsurface flaws
which can act as stress raisers and crack-initiation points, and the
4
Nor am I fully convinced that the presence of a small "pool" of liquid in an
existing pit would necessarily mitigate the stress or force felt by the surface when
a drop impinged therein. Some analytical work in this direction would be en-
lightening.
manner in which the surface geometry facilitates local stress patterns

conducive to rapid formation of detached fragments.
Work hardening of the surface material could be imagined to have
two contradictory influences: on the one hand, a higher hardness is
generally associated with higher erosion resistance; on the other hand,
in ductile materials, erosion due to multiple impacts is generally held
to involve repeated plastic deformation causing work hardening, which
leads to eventual brittle failure or to formation of cracks which propagate
like fatigue fractures. (See, for instance, Refs 20 and 23.) By the first
criterion, initial work hardening of the surface should retard the onset
of erosion; by the second, it should promote it. Hobbs [3] has observed
that in specimens work hardened by machining or grinding the incuba-
tion period is lengthened, and it is his practice to remove the work-
hardened layer by hand polishing prior to testing. We have seen evidence
of similar behavior. The apparent contradiction can probably be resolved
by recognizing that in the foregoing case a cohesive work-hardened layer
is formed over the whole surface, which reduces the amount of plastic
deformation due to the impacts, and thereby retards the rate at which
the amount of work hardening can be increased to the point where
work hardenability is exhausted and fractures begin. The work harden-
ing induced by erosion itself will, on the other hand, be localized at
certain points, which for various reasons are susceptible to the greatest
plastic deformation, and will not significantly inhibit the continuation
of this process to the failure condition. Plesset and Devine [2] have
shown by X-ray analysis that the plastic deformation in a surface sub-
jected to cavitation damage reaches a stable depth almost immediately
after the beginning of exposure and that this depth remains essentially
constant thereafter as erosion progresses. They therefore conclude that
subsequent changes in erosion rate are not attributable to changes in the
metallurgical condition of the surface.
Most investigations of the effect of various surface treatments have
shown an influence on the early stages only. This includes the effect of
shotpeening [12] and etching [9,24]. Etching increased the erosion rate
initially when compared to polished specimens. It was observed that, in
the latter, one of the early effects of erosion attack is a "bringing out" of
the grain boundaries and the creation of something like an etched ap-
pearance, due to minute failures in the weaker intercrystalline material.
This observation has also been made by others, for example, Wheeler
[25]. Keller [26] has shown that by periodic polishing of the surface
the erosion rate can be maintained indefinitely at a value corresponding
to that of the incubation period.
As far as the effect of the surface geometry on erosion resistance is
concerned, it seems evident that the irregVilar shape of an eroded sur-
face not only brings more inclusions, discontinuities, and other possible
stress raisers and crack-initiation points to the surface, but also provides
a geometry which makes it easier for the impacting drops to apply local
bending or tensile stresses which, in combination with the aforementioned
stress raisers, will induce failure. This is well exemplified by Fig. 9 which
shows a section through an eroded surface.
Effect of Material-Removal Mechanisms on Rate-Time Pattern

For the sake of argument, the spectrum of erosion mechanisms in a
ductile material may be divided into several regimes as a function of im-
pact intensity, or, in the case of droplet impingement, as a function of
impact velocity if drop size is held constant. These regimes obviously
merge one into the other; there are no sudden transitions between them.
FIG. 9—Photomicrograph of a section through an eroded stellite specimen

' (X250).
For very low velocities below some "first threshold" value, no meas-
urable damage or material loss will occur during any practical exposure
time, or material loss is confined to isolated weak spots. Such threshold
velocities, empirically deduced from test or operating experience or
arbitrarily derived from the endurance limit of the material by some
safety factor, have been used as design guides in some phases of steam
turbine and condenser design. It is not fully established whether there
actually is a velocity below which erosion will never occur. Honegger
[7] doubted it; and Vater [27], who suggested that the dependence of
erosion on velocity could be regarded and plotted analogously to the
dependence of fatigue life on applied stress, regarded the erosion proc-
ess as one somewhat similar to corrosion fatigue (in which there is no
endurance limit). He therefore stated that the "threshold velocity" had
to be defined as that velocity below which no measurable weight loss
occurred after some specified number of impacts. In any case, one might
say that in this first regime the erosion, if any, corresponds to that in the
incubation stage of the conventional rate-time pattern; that is, it will be
low, possibly gradually increasing with some random fluctuations, and
will be highly influenced by the initial surface conditions, as previously
discussed, and by the possibility of simultaneous corrosion as shown by
Wheeler [25].
As the velocity exceeds the first threshold, something akin to fatigue
failure becomes the predominant failure mechanism. Metallurgical ob-
servations substantiating this, and descriptions of the probable sequence
of events leading to failure and the formation of loose fragments, have
been provided by many investigators including Mousson [28], Vater
[27], Pohl [22], von Schwarz et al [9], Plesset [23], Wheeler [25],
Brunton [20], Thomas [29], and Marriott and Rowden [30]. Some in-
vestigators have found more plastic deformation in the surface than
might be expected. Thomas [29] noted small plastic depressions in the
surface during the early stages of exposure at velocities whose presumed
impact pressures were less than the yield point of the material. Branden-
berger and de Haller [10], on the basis of extensive radiographic studies,
concluded that fracture in erosion is neither like static fracture nor like
fatigue fracture, but is accompanied by a degree of damage to the crystal
structure which is intermediate between that associated with those fail-
ure modes. It must be remembered, though, that the stress-geometry
condition—at least when the surface is still relatively smooth—is not
of such a nature as to make "static" rupture easily possible. Thus the
general regime of predominant fatigue or repeated-impact rupture will
extend well into the velocity range where each drop could be expected
to produce noticeable plastic deformation.
In this regime one may expect to find rate-time curves exhibiting the
"conventional pattern;" that is, an incubation stage related to the fact
that a certain number of impacts are required before fatigue failures
occur, an acceleration stage, possibly a steady-state stage, an attenuation
stage, and possibly a final steady-state stage though probably no generali-
zations should be made about the behavior when gross surface damage
has set in. The possibility of relating these phases in the erosion rate-time
curve more specifically to the fatigue properties of the material will be
explored in Part II of this paper.
A second threshold velocity may be associated with that velocity at
which the material loss due to single-impact damage process becomes
significant. This is probably related to the "visible damage threshold"
described by DeCorso and Kothmann [31,32], above which a single
impact leaves a distinct crater in a smooth material surface. This re-
gime eventually must merge into the regime of hypervelocity impact.
The exact determination of the second threshold velocity from the point
of view of material removal is difficult, because in single-impact experi-
ments—such as those performed by DeCorso, and also by Brunton [20],

Engel [33], and others—the actual material removed from the surface
could not be reliably established, although crater depths or crater pro-
files were measured. From two curves reproduced in Ref 34, one can
deduce that, for hypervelocity impact of %6-in.-diameter aluminum
spheres on an aluminum surface, the ratio of target volume loss to crater
volume is approximately 0.15 at a velocity of 7 km/sec (23,000 ft/sec),
reducing to about 0.09 at 4 km/sec (13,000 ft/sec). One may cautiously
infer from this that at the velocities of interest to us, say 1000 to 4000
ft/sec, the corresponding ratio will be very much smaller yet.5 Of course,
this must be balanced by the fact that such loss occurs with each im-
pinging drop, whereas many repeated impacts over some finite area are
required to generate one erosion fragment by the fatigue failure mecha-
nism. For any quantitative estimate of the relative significance of the
two mechanisms, more data are needed on each.
Qualitatively, one may say that as single-impact erosion becomes
significant the incubation period can no longer be a zero-weight-loss
period, but rather will begin by exhibiting an erosion rate corresponding
to the single-impact erosion, this rate increasing in time as additional
fatigue-type erosion sets in. Fatigue in this instance probably corresponds
more to low-cycle fatigue due to strain cycling than to high-cycle fatigue
due to stress cycling. The geometry of the eroded surface will now be
affected by the heavy plastic deformation due to each drop as well as by
the breaking away of larger erosion fragments due to fatigue fractures.
More work is needed on the relationship between surface geometry and
impact damage under these conditions before the rate-time pattern in
the regime of single-impact damage can be reliably described.
Even when initial conditions are still in the "fatigue regime," once the
surface is slightly eroded it may become susceptible to single-impact
failures because of geometrical effects discussed earlier.
Part II—A Statistical Erosion-Rate Model
Qualitative Description of Proposed Model

As we saw hi the previous sections, the "conventional" erosion-rate-
versus-time pattern is that associated with a predominant fatigue mecha-
nism for material removal. It is hi this regime that most of the test data
and the practical experience lie. As is well known, fatigue is intrinsically
a statistical process exhibiting a considerable scatter, and this fact will be
made use of hi developing an analytical model for the erosion rate-time
6
This inference should be valid qualitatively although the actual material-
removal mechanism in the hypervelocity regime is a liquid-like flow of the target
material accompanied with some "splashing out," whereas that in our regime of
interest is related to the shear effect of radial outflow, as described earlier.
pattern applicable to this regime. The qualitative results have interesting

implications with reference to the previously reviewed findings and to
previously attempted correlations between erosion and fatigue data.
The approach to be described, though numerical in nature, can at this
time predict no more than qualitative trends and should be considered as
exploratory.
The mathematical and logical formulation of the model, in a prelimi-
nary form, is given in detail in the Appendix. The basic reasoning of
the model is as follows: We assume that each small element of sur-
face is subjected to an impact fatigue environment and that after a cer-
tain time (that is, a certain number of impacts) it will be detached from
the surface as an erosion fragment, due to subsurface fatigue failure.
Further, we assume that when many such surface elements are con-
sidered, the individual times required for their removal would be de-
scribed by some statistical distribution function, much as the number
of cycles to failure of a large number of fatigue specimens (stressed to
the same level) can be described by a distribution function. When ero-
sion fragments are removed and expose "fresh" surface to impingement
attack, the time to remove elements of this new surface will likewise be
described by a distribution function and so on. The time-to-failure distri-
bution function for these newly exposed surfaces will probably not be
the same as that for the original surface, since they will have been
subjected to some subsurface stress condition, even before being ex-
posed to direct impingement, and since the surface geometry will be
different.
In the case of conventional fatigue specimens, the distribution occurs
primarily as a result of the statistical nature of the fatigue process itself.
In the case of erosion fragments, it must ultimately reflect the variations
in the concentration and the severity of impacts (that is, droplet velocities
and sizes), variations in the local surface geometry and properties, and
variations in the size of fragments formed. At present, however, one
arbitrary distribution curve is assumed to represent all of these sources
of scatter.
Qualitatively, it can be seen that if these distributions had very little
scatter or dispersion, that is, if the lifetimes of all surface elements were
about equal, then the erosion rate would be zero until that lifetime was
reached, at which instant a very high rate would be exhibited while all
of the original surface flaked off, to be followed by another interval of
zero rate until the second layer flaked off and so on.
If, however, these distributions have a significant dispersion, one can
intuitively predict that this will result in a rate-time curve which up to a
first peak looks somewhat like the distribution curve, but in which sub-
sequent peaks and valleys are attenuated and a steady-state rate is ap-
proached. An "incubation period" will exist if the dispersion is not ex-
FIG. 10—Typical computed erosion rate-time curves from preliminary statisti-

cal model, using Normal distribution functions.
cessive. One might think of the variation in the surface-element lifetimes

as "dispersing" the periodicity associated with one layer being removed
after another.
Results of Preliminary Model
The preliminary mathematical formulation and computer program
considered one distribution function applicable to the original surface
and one other applicable to each of the subsequently exposed "surfaces."

Both were specified as Normal distributions truncated and normalized
over a finite time span. Thus the significant input parameters were the
nominal mean lifetime, MF , and standard deviation, o-F , for the original
surface, and the corresponding values, MG and <TG , for the "undersur-
faces." Figure 10 shows some rate-time curves obtained by this program,
with the distribution parameters as indicated. Note that the attaining
FIG. 11—Experimental erosion rate-time curves, computed from cumulative

erosion curves given in Ref 35.
of a steady-state rate is hastened both by increasing the dispersion of

the functions and by specifying a shorter mean lifetime for the under-
surfaces as compared to the original surface.
Fluctuations such as shown in Fig. 10 have occasionally been ob-
served, as illustrated by Fig. 11 which shows rate-time curves com-
puted from experimental cumulative erosion curves presented by Kent
[35]. Moreover, fluctuations which would appear quite prominent in
rate-time curves are not nearly as evident if the same data are plotted
as cumulative erosion versus time—which, after all, is how the data are
actually obtained. Therefore, it seems quite conceivable that in many

cases such fluctuations would barely have been noted and would have
been "smoothed" out of the raw data, or might have been lost entirely
through the data points being too far apart in tune.
The fluctuations, however, are by no means an inevitable consequence
of this model if nonsymmetrical distribution functions are used, as will
be seen in the results obtained from the elaborated formulation of the
model, described in the following section.
Description of Elaborated Model

The elaborated model permits the specifying of a different distribu-
tion function for each "level" below the original surface and of two dif-
ferent functions for the original surface: one for the "unaffected surface,"
in which erosion takes place by the initiation of new pits, and one for
the "affected surface," which is that surrounding existing pits and in
which erosion is presumed to take place by the growth of these pits. The
program computes the rate of erosion, the cumulative erosion, and the
exposed area at each level, from which, in turn, it computes an average
surface profile and surface roughness at selected time points. The formu-
lation and programming approach is described in detail in Appendix A of
Ref 46.
For the elaborated analysis, we have adopted the log-Normal distribu-
tion; that is, one which would appear as a Normal distribution if the
frequency of failures (based on logarithmic time increments) were plotted
on a logarithmic time scale. This has been shown to provide a reasonable
description of fatigue life data by, for instance, Stulen [36], Roeloffs
and Garofalo [37], and Epremian and Mehl [38], and is convenient to
handle mathematically. It is generally recognized, however, that a more
rational and accurate description would be provided by three-parameter
distributions such as the Weibull distribution [39] or extreme value dis-
tribution [40]. The use of one of these distributions should be con-
sidered in further developments of this model.
A rigorous analysis intended to give quantitative results would also
have to take into account the variations in drop sizes and velocities which
may be encountered under real impingement conditions. This would
bring the theories of cumulative damage into the picture. The physical
and statistical aspects of fatigue life under repeated stress cycles of
varying amplitude have been discussed, for instance, by Freudenthal
[41], who presents a number of different possible approaches to the
problem. Several of them lead to the conclusion that "high stress am-
plitudes shorten the fatigue life out of all proportion to their number
of application or their cycle ratio." This would seem to argue against
the adoption of a simplified approach such as a superposition of the
erosion rates due to various droplet size or velocity ranges computed

independently.
For our present, explorative purposes, we remain with the log-Normal
distribution or loss-rate function. On a logarithmic time scale, the log-
Normal is described by its mean, m, and its standard deviation, <r, both
FIG. 12—Computed rate-time curves based on log-Normal distributions,

showing effect of varying dispersion a while keeping median at constant value
M - 1.0.
m and a of course being logarithmic magnitudes. For our purposes, we

must transform the distribution onto a linear or "real-time" scale. It
then becomes a skewed distribution and its mean, median, and mode
values no longer coincide, as they do in a symmetrical distribution.
The real-time value whose logarithm is m, which we shall denote by
M = 10"*, establishes the median value of the log-Normal distribution,
that is, that value of t at which half of the specimens (or surface ele-
ments) will have failed. This is the value generally used to establish a
point of an engineering S-N curve. The mode, or peak in the distribution

curve, will occur at a value less than M. The mean value, E, or arithme-
tic average of all lifetimes, will occur at a tune value greater than M,
namely at time E = M X 101 -i5ffZ. For purpose of discussion, we will
characterize all distributions by their values of o-, and either M or E.6
FIG. 13—Computed curves based on log-Normal distributions, showing effect

of varying dispersion a while keeping mean at constant value E — 1.0.
From Refs 38 and 41 one may infer that, for fatigue data, o- ranges from
0.13 to 0.40, with 0.25 a representative value. For erosion fragments
one might well expect even higher dispersions. These statistical aspects
are discussed in more detail in Appendix B of Ref 46.
6
The log-Normal distribution function and its properties are discussed on
pages 115 and 117-118 of Ref 42, but the formula given there for the probability
density function (that is, our rate function) seems to be incorrect. A correct ex-
pression can be found on p. 220 of Ref 43.
Results of Elaborated Model

The number of variations which could be investigated with this program
is unlimited, and all we can demonstrate here are some of the important
effects. The most significant of these is the effect of the dispersion
parameter a-. Figure 12 shows computed erosion-time curves for various
values of a from 0.15 to 0.80, with the median M held constant. Figure
13 shows a similar set of curves with the mean E held constant. In each
case, the same distribution is assumed for all surfaces and levels. Since
hi such cases the eventual steady-state erosion rate must be proportional
to the reciprocal of the mean lifetime, all curves in Fig. 13 approach the
same steady-state rate.
Two striking results appear from these curves: (a) The maximum
FIG. 14—Example of the effect of using a higher median value for the "un-
affected" surface (Mu — 3.0) than for all other surfaces (M = 1.0). Compare with
Fig. 12d, but note difference in vertical scale.
erosion rates vary considerably, and (b) almost all of the experimentally
found rate-time patterns (discussed hi Part I of this paper) can be at
least qualitatively generated by proper choice of the dispersion parame-
ter a-. When a is small, the curves exhibit damped fluctuations similar to
those of Fig. 10. When <r is increased, the fluctuations die out and the
steady-state rate is attained quite quickly. When a is further increased, a
single peak appears in the curve. At very high values of a, this peak may
occur so early that the time resolution is just not fine enough to show
the acceleration stage of the rate-time curve, and the curve therefore
appears to begin at its maximum value. The same is probably true for
experimental data like those of Figs. 5 and 7. It does not seem unreasona-
ble to suppose that erosion due to very small droplets, where each impact
stresses only a minute portion of the surface area, would be characterized
by a high dispersion in the fragment lifetimes.
In many of the curves of Figs. 12 and 13 the ratio of the erosion-rate
peak to the expected steady-state value is not as great as sometimes found

in practice. But it should be recognized that at time values greater than
the median, the surface has suffered heavy erosion damage, and one may
therefore expect that geometric effects such as described in Part I may
have set in by this time and have caused an additional diminution of the
erosion rate and possibly suppression of further fluctuations. Certainly
one would expect the results predicted by this analysis to be at least
modified by the geometric effects. Thus, Figs. 12d and I3b may cor-
respond to experimental results of the type of Fig. 1 and Figs. I2b and I3a
to results of the type of Fig. 2. It is possible, however, that some appropri-
ate combination of distribution functions for the different surfaces could
FIG. 15—Examples of computer "surface profile" curves, showing the un-

eroded area as a function of level below the original surface, at various values of
time T.
result in an elongated hump such as in Fig. 2, which then again would not
constitute a steady-state value.
Figure 14 shows an example of slowing down the loss rate from the
"unaffected" surface as compared to that of all other surfaces—which
are presumed to be more susceptible to erosion because of the irregular
geometry, as discussed in Part I. This case is identical to that of Fig. I2d
except that for the unaffected surface the median lifetime has been in-
creased to 3.0. Note that the shape of the rate curve has been made more
similar to that typified by Fig. 1; the cumulative loss rate is also shown
and is quite similar to typical curves such as Fig. 4.
Figure 15 shows "surface profile" curves at various values of the time
T, computed for some of the previous cases. The ordinates indicate the
surface "level," with 0 representing the original surface. The abscissas
represent the area not yet eroded at each level. (Thus the difference in
abscissa between adjacent levels represents the area "exposed" at the
lower of the two levels.) Note that in Fig. 15a, a case of low dispersion
value (a- = 0.25), the erosion is shallower and more evenly distributed
than in the other two cases which represent high dispersion values (a- =
0.8). This suggests that the geometric effects due to severe roughness—
which tend to reduce the erosion rate—are delayed in the former case,
which may explain why the maximum erosion rate in such a case may
persist for some time and give rise to rate curves typified by Fig. 2.
Figure 16 shows the computed surface roughness, versus computed mean
depth of penetration, for the same three cases, confirming the lower rough-
ness associated with a lower dispersion value.
Discussion and Conclusions

Let us now examine the implications of this model with respect to
correlations of incubation times and erosion rates. Since the incubation
FIG. 16—Computer rms surface roughness versus mean depth of penetration

(cumulative erosion) for the examples of Fig. 15. The letters (a), (6), and (c)
correspond to the similarly designated cases in Fig. 15.
time seems obviously related to the fatigue nature of erosion, several

investigators have attempted correlations reflecting this. Thus Leith and
Thompson [11] correlated the incubation times of several materials with
the corrosion fatigue limit for 107 cycles of these materials. Mathieson
and Hobbs [12] made a similar correlation with the conventional en-
durance limit for several aluminum alloys. In both cases, the results were
reasonably consistent, but the approach is hardly logical since the incuba-
tion time in erosion surely should be related to a finite lifetime to failure
rather than to a stress value at which no failure occurs. Thus the success
of these correlations surely depended on a second, implicit correlation
between the finite fatigue lives at the test stress, and the endurance limits,
valid for the group of materials compared. Ripken et al [44] have used a
more logical approach and have correlated the number of impacts cor-
responding to the incubation time at a given impact velocity, with the
number of cycles to failure in bending fatigue at an equivalent stress level.
The stress level was assumed to be given by the well-known water-hammer

pressure, p = pCV. The incubation period was defined by the intercept
on the time axis of the straight-line tangent to the steepest slope of the
cumulative weight loss curve.
If the previously developed model is valid, this procedure is still not
quite correct. The statistical model implies that the apparent incubation
period depends not only on the mean lifetime of the erosion fragments
but also on the scatter in these lifetimes. The erosion rate becomes non-
zero when the first "element" fails and continues to increase until approxi-
mately the mode or most probable value of the lifetime is reached on the
top surface. But it is the median value—which may occur later yet if the
distribution is skewed—which corresponds to the nominal lifetime at the
appropriate stress as obtained from a conventional S-N fatigue curve.
Whether either the median lifetime or the associated scatter in erosion
fragments corresponds to that of full-scale bending or pull-type fatigue
specimens is at present a moot question. However, the discrepancies in
the correlation of Ref 44 are in the direction which the foregoing argu-
ment would predict.
If one stipulates a steady-state erosion process, then the erosion rate
would certainly be inversely proportional to the mean lifetime of erosion
fragments (provided their size distribution remained constant). This is
the basis from which one can draw the analogy between the loss-rate
reciprocal versus impact velocity in erosion, and cycles to failure versus
stress level in fatigue, as proposed by Ref 27. This appears to provide a
rational basis for attempting to predict an erosion-speed relationship on
the basis of known fatigue data for the material, although to my knowl-
edge this attempt has not been made. But here, again, the statistical model
suggests that the "obvious" approach is not quite correct. It implies that
the maximum erosion rate—by which many investigators have, for practi-
cal reasons, reported and correlated their results—does not necessarily
represent a steady-state erosion process at all, but rather the "deluge" of
erosion fragments from the top surface layer which takes place in the
vicinity of the "most probable" fragment lifetime from the beginning of
exposure. Thus the maximum instantaneous erosion rate is, again, not
merely a function of the average fatigue life of the surface elements but
also of the scatter in lifetimes. Consequently, anything which influences
that scatter will influence the maximum erosion rate, even though it may
not affect the eventual hypothetical steady-state rate.
What can this model contribute toward the resolution of the dispute
referred to in the beginning of this paper? It implies that Thiruvengadam
and Preiser [1] are correct in claiming that the erosion rates during the
stages encompassing the first peak in the rate-time curve are not charac-
teristic merely of the material under test, since, as we have seen, the shape
of this curve depends on the shape of distribution functions which, in
turn, depend in part on characteristics of the test method such as the dis-
tribution of bubble or droplet sizes, etc. It implies, also, that while the
erosion rate would, in the absence of other influences, tend toward a
steady-state value as postulated in Ref 1, this generally occurs only after
most of the original surface had been eroded, at which time the surface
damage will be so severe as to make the erosion conditions susceptible to
geometry effects such as described in Ref 2. In short, the instantaneous
erosion rate may never be characteristic of only the material, and for
meaningful correlations it will become necessary to standardize the test
method very carefully, or to use properly chosen cumulative erosion
measurements (for example, time required to attain some specified mean
depth of penetration, of significance to the intended application).
The view of erosion as a fatigue process is not new, and yet it carries
with it a number of other implications which perhaps have not been
sufficiently emphasized in the past and to which I should like to draw at-
tention, even though they are only peripheral to the present topic. These
include:
1. There is little likelihood of finding one specific independently
measurable material property which will predict erosion resistance, since
none has been found to predict fatigue strength uniquely although far
more research has been done on fatigue than on erosion.
2. In fatigue, the relation between stress and endurance is determined
by test for each material, and is not expressible in simple analytical form.7
Similarly, the relation between impact velocity and erosion very likely
does not follow any specific law but must be established uniquely for each
material.
3. Although erosion is the result of many failures, and some of the
statistical scatter found in fatigue data may well average out in an erosion
test, yet to obtain valid results (or results with calculable confidence limits)
many more data points must be taken and many more replications must
be run than has been customary to date. Related to this is the need to
establish accurately the erosion-versus-exposure curve, and to carry out
all tests to corresponding degrees of cumulative erosion if one wants to
draw any quantitative comparisons from them.
Finally, for further development of the present approach, one would
have to learn more about the actual sizes or size distributions of erosion
fragments and how this affects then: individual and statistical fatigue
lives under given impact stresses, and to learn more about the distribution
of impact intensities and impact areas and how these can be accounted
for by cumulative damage theories.
7
Empirical formulas to represent the fatigue S-N relationship have been pro-
posed and are reviewed by Weibull [45], pages 174-183. These, however, are
curve-fitting attempts rather than expressions of a physical law.
A cknowledgments
The author wishes to acknowledge the contribution of R. I. Shrager,
who was responsible for the mathematical formulation and computer
programming of the preliminary model, and of L. B. Godio, who assisted
in the formulation and programming of the elaborated model. W. D.
Pouchot of the Westinghouse Astronuclear Laboratory, F. G. Hammitt
of the University of Michigan, and J. C. Freche and S. G. Young of
NASA-Lewis Research Center all contributed helpful discussions and
criticisms. A part of this work was supported by NASA under contract
NAS 7-390, with technical supervision through Jet Propulsion Labora-
tory.
APPENDIX
Mathematical Formulation of Model

Let any surface exposed to erosion be thought of consisting of elementary
areas (or volumes, if their thickness is considered) whose lifetimes under the
erosion attack can be described by a normalized distribution function /(?).
Thus, by definition
and the distribution function for a specific surface area A, exposed to erosion
from time t = 0, is therefore
Since a surface element is lost from the surface when its lifetime is reached,
Eq 2 can equally well be regarded as a loss-rate function for the area A.
Equation 2 may be further generalized by stating that the loss rate from
an area A-^, first exposed to erosion at time t = T^, is thereafter given by
Let us now consider the original or "top" surface of a body exposed to

erosion. We may take its area to be unity, and every portion of its area is
simultaneously exposed to erosion at time t = 0. Thus*/(0 adequately de-
scribes the loss rate from the top surface. As surface area is eroded, or lost,
from the top surface, an equal area is created, or exposed, at the "second"
level, located a distance h below the surface where h is assumed the thickness
of erosion fragments. For convenience, the thickness h will also be assigned
a numerical value of unity on some appropriate scale. In turn, the second-
level surface will be eroded to expose a third-level surface, and so on. But, in
computing the actual loss rates from all of the "undersurfaces," one must
recognize that the lifetimes of surface elements must be measured from the
time they were first exposed, and the total loss rate from all surface elements
which were first exposed during a time increment dT at time T depends on
the total area which was first exposed during that time interval.
Let Y(t) be the total rate of erosion, from all levels, at time t. This is what
we desire to compute. But Y(f) is, ipso facto, also equal to the rate at which
new surface area is exposed, at all levels below the top surface, at time
t. (Strictly speaking, it is proportional to it, but with h = 1.0 it is numerically
equal.)
Thus, the total surface area first exposed during increment dT at time T
is Y(T) dT, and the loss rate from this area at time t is, by Eq 3,
The total loss rate at time t, from all undersurfaces, is composed of contribu-
tions from all undersurface areas first exposed during all time increments
from T = 0 to T = t, or
The total loss rate or erosion rate, Y(t), is the sum of that from the top
surface and that contributed by all undersurf aces, or
The fact that the contributions from the undersurfaces and from the top
surface from two distinct terms in Eq 5 makes it conveniently possible to
assign a different distribution function for the top surfaces as compared to
all undersurfaces. This is desirable if one wants to reflect the fact that the
top surface has in many ways a different nature and history than the under-
surfaces exposed as a result of erosion. We finally, then, state
where:
f(f) = distribution function for top surface and
g(t) = distribution function for undersurfaces.
For the initial explorations Eq 6 was computer programmed directly, using
Normal distributions for functions f(t) and g(t), normalized over specified
time spans rather than between the limits of plus and minus infinity as sug-
gested by Eq 1.
The formulation of the elaborated model, which keeps track of the area
eroded at each "level," is described in detail in Appendix A of Ref 46.
The log-Normal frequency distribution function as programmed for the
elaborated model is:
This function has the following properties:

TO is a "delay time" which may be introduced to ensure that no failures
occur before t = T0. The mean, or expected value, is
The median value is

The mode, or most probable value, is
The input may be prescribed in terms of T0, m, and a directly; the latter two
may also be prescribed in terms of the equivalent logarithms to base 10, or
in terms of the equivalent real-time quantities and
References
1963.
/[2] M. S. Plesset and R. E. Devine, "Effect of Exposure Time on Cavitation
Damage," Journal of Basic Engineering, Transactions, Am. Soc. Mechanical
Engrs., Vol 88D, No. 4, 1966, pp. 691-705.
[3] J. M. Hobbs, private communication.
[4] D. Pearson, private communication. See also, D. W. C. Baker, K. H. Jolliffe,
and D. Pearson, 'The Resistance of Materials to Impact Erosion Damage,"
A Discussion on Deformation of Solids by Impact of Liquids, Philosophical
Transactions, Royal Society of London, A, Vol 260, part No. 1110, 1966,
pp. 193-203.
[5] F. G. Hammitt, private communication.
[6] A. Evans and J. Robinson, "Erosion Experiments Related to Steam Turbine
Blading," Parsons Journal, Vol 10, No. 61, Christmas, 1965, pp. 464-473.
[7] E. Honegger, 'Tests on Erosion Caused by Jets," The Brown Boveri Review,
Vol 14, No. 4, April, 1927, pp. 95-104.
[8] S. L. Kerr, "Determination of the Relative Resistance to Cavitation Erosion
by the Vibratory Method," Transactions, Am. Soc. Mechanical Engrs., Vol
59, 1937, pp. 373-397.
[9] M. von Schwarz, W. Mantel, and H. Steiner, 'Tropfenschlaguntersuchungen
zur Feststellung des Kavitationswiderstandes (Hohlsog)," Zeitschrift fur
Metallkunde, Vol 33, 1941, pp. 236-244.
[10] E. Brandenberger and P. de Haller, "Untersuchungen iiber Tropfenschlagero-
sion," Schweizer Archiv, Vol 10, 1944, pp. 331-341 and 379-386.
[11] W. C. Leith and A. L. Thompson, "Some Corrosion Effects in Accelerated
Cavitation Damage," Journal of Basic Engineering, Transactions, Am. Soc.
Mechanical Engrs., Vol 82D, 1960, pp. 795-807.
[12] R. Mathieson and J. M. Hobbs, "Cavitation Erosion: Comparative Tests,"
Engineering, Vol 188, Jan. 22, 1960, pp. 136-137.
[13] "Second Quarterly Progress Report; Oct. 1, 1965 through Jan. 15, 1966;
Basic Investigation of Turbine Erosion Phenomena; Contract NAS 7-390,"
WANL-PR(DD)-007, Westinghouse Astronuclear Laboratory, 1966.
[14] J. Z. Lichtman, D. H. Kallas, C. K. Chatten, and E. P. Cochran, Jr., "Study
of Corrosion and Cavitation-Erosion Damage," Transactions, Am. Soc. Me-
chanical Engrs., Vol 80, 1958, pp. 1325-1341.
[/5] A. Thiruvengadam, "A Comparative Evaluation of Cavitation Damage Test
Devices," Symposium on Cavitation Research Facilities and Techniques, Am.
Soc. Mechanical Engrs., New York, N. Y., 1964, pp. 157-164.
[16] F. G. Hammitt, "Observations on Cavitation Damage in a Flowing System,"
Journal of Basic Engineering, Transactions, Am. Soc. Mechanical Engrs.,
Vol 85D, 1963, pp. 347-359.
[17] F. G. Hammitt, M. J. Robinson, C. A. Siebert, and F. A. Aydinmakine,
"Cavitation Damage Correlations for Various Fluid-Material Combinations,"
ORA Report No. 03424-14-T, Department of Nuclear Engineering, The
University of Michigan, October, 1964.
'[18] A. Thiruvengadam, "A Unified Theory of Cavitation Damage," Journal of

Basic Engineering, Transactions, Am. Soc. Mechanical Engrs., Vol 85D,
1963, pp. 365-376.
[19] F. P. Bowden and J. H. Brunton, 'The Deformation of Solids by Liquid
Impact at Supersonic Speeds," Proceedings, Royal Society of London, A, Vol
263, 1961, pp. 433^50.
[20] J. H. Brunton, "Deformation of Solids by Impact of Liquids at High Speeds,"
Erosion and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, pp.
83-98.
[21] H. Busch, G. Hoff, and G. Langbein, "Rain Erosion Properties of Materials,"
A Discussion on Deformation of Solids by Impact of Liquids, Philosophical
Transactions, Royal Society of London, A, Vol 260, Part No. 1110, 1966,
pp. 168-178.
[22] E. Pohl, "Der Einfluss der Dampfnasse auf die Schaufeln im Niederdruckteil
einer Turbine (Schluss)," Der Maschinenschaden, No. 3, 1937, pp. 37—43.
\u3] M. S. Plesset and A. T. Ellis, "On the Mechanism of Cavitation Damage,"
Transactions, Am. Soc. Mechanical Engrs., Vol 77, 1955, pp. 1055-1064.
[24] M. von Schwarz, "Tropfenschlaguntersuchungen zur Feststellung des Kavita-
tionswiderstandes von Edelmetallen," Zeitschrift fur Metallkunde, Vol 35,
1943, pp. 73-78.
K?5] W. H. Wheeler, "Mechanism of Cavitation Erosion," Cavitation in Hydro-
dynamics, National Physical Laboratory Symposium, Teddington, England,
1955, Philosophical Library, New York, N. Y., 1957.
[26] A. Keller, "Kavitation und Tropfenschlag," Schweizer Archiv, Vol 23, No.
11, November, 1957, pp. 346-361.
[27] M. Vater, "Das Verhalten Metallischer Werkstoffe bei Beanspruchung durch
Fltissigkeitsschlag," Zeitschrift, Verein Deutscher Ing., Vol 81, 1937, pp.
1305-1311.
[28] J. M. Mousson, "Pitting Resistance of Metals Under Cavitation Conditions,"
Transactions, Am. Soc. Mechanical Engrs., Vol 59, 1937, pp. 399-408.
[29] G. P. Thomas, "The Initial Stages of Deformation in Metals Subjected to
Repeated Liquid Impact," A Discussion on Deformation of Solids by Impact
of Liquids, Philosophical Transactions, Royal Society of London, A, Vol
260, Part No. 1110, 1966, pp. 140-143.
[30] J. B. Marriott and G. Rowden, "The Erosion of a Cobalt-Chromium Alloy
by Liquid Impact," A Discussion on Deformation of Solids by Impact of
Liquids, Philosophical Transactions, Royal Society of London, A, Vol 260,
Part No. 1110, 1966, pp. 144-149.
[31] S. M. DeCorso and R. E. Kothmann, "Erosion by Liquid Impact," Erosion
and Cavitation, ASTM STP 307, Am. Soc. Testing Mats., 1962, pp. 32-45.
[32] S. M. DeCorso, "Erosion Tests of Steam Turbine Blade Materials," Proceed-
ings, Am. Soc. Testing Mats., Vol 64, 1964, pp. 782-796.
[33] O. G. Engel, "Pits in Metals Caused by Collision with Liquid Drops and
Rigid Steel Spheres," Journal of Research, Nat. Bureau of Standards, Vol
64A, 1960, pp. 61-72.
[34] R. J. Eichelberger, "Hypervelocity Impact," Behavior of Materials Under
Dynamic Loading, Am. Soc. Mechanical Engrs., 1965, pp. 155-187.
[55] R. P. Kent, "Some Aspects of Metallurgical Research and Development
Applied to Large Steam Turbines," Parsons Journal, Vol 10, Christmas, 1964,
pp. 285-295.
[36] F. B. Stulen, "On the Statistical Nature of Fatigue," Statistical Aspects of
Fatigue, ASTM STP 121, Am. Soc. Testing Mats., 1952, pp. 23-44.
[37] R. Roeloffs and F. Garofalo, "A Review of Methods Employed in the Statis-
tical Analysis of Fatigue Data," Proceedings, Am. Soc. Testing Mats., Vol
56, 1956, pp. 1081-1090.
[38] E. Epremian and R. F. Mehl, 'The Statistical Behavior of Fatigue Properties
and the Influence of Metallurgical Factors," Fatigue with Emphasis on Sta-
tistical Approach, ASTM STP 137, Am. Soc. Testing Mats., 1953, pp. 25-57.
1 00 EROSION BY CAVITAtlON OR IMPINGEMENT
[59] W. Weibull, "A Statistical Distribution Function of Wide Applicability,"

Journal of Applied Mechanics, Transactions, Am. Soc. Mechanical Engrs.,
Vol 73, pp. 293-297.
[40] A. M. Freudenthal and E. J. Gumbel, "Distribution Function for the Pre-
diction of Fatigue Life and Fatigue Strength," International Conference on
Fatigue of Metals, Institution Mechanical Engrs. and Am. Soc. Mechanical
Engrs., Session 3, Paper 5, 1956.
[41] A. M. Freudenthal, "Physical and Statistical Aspects of Cumulative Damage,"
IUTAM Colloquium on Fatigue (Stockholm, 1955), Springer-Verlag, Berlin,
1956, pp. 53-62.
[42] N. L. Johnson and F. C. Leone, "Statistics and Experimental Design, Vol. 1,"
John Wiley & Sons, Inc., New York, 1964.
[43] H. Cramer, "Mathematical Methods of Statistics," Princeton University
Press, Princeton, N. J., 1957.
[44] J. F. Ripken, J. M. Killen, S. D. Crist, and R. M. Kuha, "A New Facility for
Evaluation of Materials Subject to Erosion and Cavitation Damage," Project
Report 77, St. Anthony Falls Hydraulic Laboratory, University of Minnesota,
March, 1965.
[45] W. Weibull, "Fatigue Testing and Analysis of Results," Pergamon Press,
Oxford, 1961 (published for and on behalf of AGARD).
[46] F. J. Heymann, "On the Time Dependence of the Rate of Erosion Due to
Liquid Impact or Cavitation," Report E-1448, Steam Divisions Engineering,
Westinghouse Electric Corp., June, 1966.
DISCUSSION
Olive G. Engel1 (written discussion)—The picture presented is one of

layers of cells whose lifetimes are described by statistical distribution
functions. This is a good approach because the general features of the
erosion-rate-versus-time curve can be obtained from a model of this kind.
In my opinion, however, a model based on the real physical picture is
preferable because it is more informative.
Let us consider the same picture of a test specimen consisting of layers
of cells. Let us first consider that the test specimen is made of a low-
strength brittle material that does not work harden. When drop impinge-
ment begins, the cells of the surface layer contain no cracks. We can
represent them as c0-cells, where c represents a cell and the subscript-0
notation indicates that the cell contains no cracks. If the impact energy
delivered by a drop is sufficient to initiate a crack in a surface cell, we can
represent the event in the following way:
1
Evendale, Ohio.
DISCUSSION ON TIME DEPENDENCE OF RATE OF EROSION 101
where: e is the energy per unit volume needed to form a crack in the ma-
terial being considered and c\ is a cell that contains one crack.
As drop impact is continued, the density of Cj-cells in the surface layer
will increase and the chance that a Ci-cell will be struck by another drop
and develop a second crack will become increasingly more probable. This
event can be described as
The c2-cells that form contain two cracks; the two cracks may or may not
intersect. If the cracks in a c2-cell do intersect, it is a c2*-cell or a critical
site. This distinction is made because if just one more intersecting crack
should form in a c2*-cell in such a way as to complete crack formation
around a triangular piece of the solid, this piece will break away as an
eroded fragment. The event can be described in the following way:
an eroded fragment
The Co-cell which is formed in the foregoing event is composed of the
underlayer material that is exposed when the eroded fragment breaks
away. Consequently, the ejection of an eroded fragment simultaneously
regenerates starting material. The newly exposed cell of underlayer ma-
terial is not exactly the same as a cell of the original surface material,
however; the angle that the newly exposed surface presents to the imping-
ing drops differs from that presented by the original surface, and, in
addition, the newly exposed material may contain residual crack ends.
The simplified picture of the erosion process presented so far is re-
stricted to consideration of the cracking damage produced by the impact
energy; it has not taken into account the damage that results from the
radial flow of the liquid contained in the drops. It is applicable to an
erosive environment in which flow is minimized, such as cavitation
erosion produced with use of a magnetostriction oscillator, and, with a
little reflection, it can be seen that this simple model is able to interpret
the characteristic features of curves of erosion rate plotted against time
for data obtained with the use of a magnetostriction oscillator.
There can be no loss of material until a minimum of three cracks inter-
sect in such a way as to circumscribe a triangular pyramid. Consequently,
there must be an initial period during which no erosion takes place (incu-
bation period). During this period, the density of Ci-cells, c2-cells, and
c2*-cells in the surface layer will build up.
When the density of c2*-cells becomes high enough so that the proba-
bility is substantial that one of them will be struck by a drop, there will be
a large number of cells that contain one or more cracks. In fact, the
density of c2*-cells will be greater at this time than at any later tune be-
cause up to this time there has been no process operating that removes
c2*-cells. Consequently, as soon as one of them is hit by a drop and a
fragment of solid is lost, this event will quite suddenly become a frequent
occurrence, and the rate of erosion plotted against time will exhibit a
sharp rise (accumulation period).
Concomitant with the process of destroying c2*-cells by the loss of
eroded fragments is the process of regeneration of c0-cells as the under-
layer material is exposed. The initial rapid rate of erosion contains a check
against itself which slows it down, since not only are the essential c2*-cells
being destroyed but also c0-cells are being produced, and these must be
converted to Ci-cells and then to c2*-cells before they .can again become
sites of material loss. Consequently, after the first rapid rise, the rate of
erosion will begin to fall off (attenuation period).
The state of a system that is subject to competing processes will pass
through a maximum or a minimum. The rate of erosion plotted against
time will pass through a maximum and then fall until the rate of produc-
ing c2*-cells and the rate of destroying them become equal. At this time,
the rate of erosion should be essentially constant (steady-state period).
If we apply this model to a ductile metal, which work hardens to the
point of embritflement and then cracks, the essential features are the
same but there are some modifications. In the case of a ductile metal, the
first drops that impinge produce craters (by plastic flow) rather than
cracks. As the cratering process continues, the surface becomes covered
with craters and new craters are superposed over old craters until the
worked metal becomes brittle enough to crack. When this point is reached,
the previously described model will operate as for a brittle material until
eroded fragments are broken loose. When the underlayer material is
exposed, however, it will have to be work hardened to the point of em-
brittlement before cracks will form in it to produce new Ci-cells, c2-cells,
and c2*-cells. This new feature will result in an oscillation in the curve of
erosion rate plotted against time; it will be particularly evident in the
steady-state period.
What has just been described is the simplest form of the model; it is
restricted to the case where fluid flow of a drop liquid or of a cavitating
liquid is essentially absent. But usually fluid flow occurs, and there are
then two damage-producing attributes of an impinging drop. These are
(a) the impact pressure that it exerts and (£) the radial flow of the liquid
of the drop.
If the relative impact velocity is sufficiently high, the impact pressure
can produce cracks in the surface of a brittle solid. These cracks are circu-
lar in isotropic solids and polygonal in anisotropic solids which have pre-
ferred cleavage planes. Forces exerted by the radial flow of the drop
liquid bearing against the raised edges of these cracks are able to break
pieces of solid away. This constitutes an erosion mechanism that operates
without the necessity of intersecting cracks and, consequently, occurs
before intersecting cracks are formed.
For this reason, the very first drop that strikes the solid with energy
sufficient to crack it may produce erosion loss due to the radial flow of
its liquid against the raised edge of the crack it produced. In fact, even at
impact velocities that are too low to provide sufficient energy to crack the
solid, an impinging drop can erode protruding surface irregularities from
the solid by the shear stress that its radially flowing liquid exerts when it
bears against them. In the light of these considerations, there can be no
zero erosion period (incubation period) where rapid fluid flow is present.
When a low-strength brittle material is eroded, the surface is uniformly
roughened. For a material of this kind, the progress of long-term erosion
will involve the gradual movement of an eroded surface layer through the
thickness of the test specimen. The general surface roughness that is pro-
duced as erosion progresses will reduce the rate of erosion, but it appears
that the rate of erosion should eventually become constant since once
the surface roughness has reached a certain degree of coarseness there
should be little change in this degree of coarseness.
The case of a high-strength material is different; it will start to fail at
weak spots. When eroded fragments are ejected at these spots, residual
ends of cracks remain. The residual crack ends will go on propagating and
erosion will continue over restricted areas around the separate weak spots
until pits are formed at these spots. The pits will deepen until they
eventually pierce the test plate. For materials of this kind, the rate of
erosion can be expected to decrease progressively with time because there
is no evidence to indicate that it should ever reach a steady-state value. It
is possible, however, that a nearly steady value may be reached when the
test plate is peppered with deep pits or even with holes and the sloping
walls of the remaining material between the pits or holes have a roughly
similar angle of inclination.
1 am currently working on the further development of this model of
the erosion process. At this time I am only able to share with you the
thoughts about erosion rate and the outline of the model of the erosion
process that are given hi the foregoing. I hope to be able to complete the
model and to present it to you at a future time.
F. G. Hammitt2 (written discussion)—The author is to be congratulated
for this clear and comprehensive summary of the present situation relating
to damage rate versus test duration effects in cavitation and impingement
erosion, as well as for his very original and significant statistical model of
the"erosion process when fatigue is the predominant failure mechanism.
It is \very interesting to note that all the various rate-time curves which
have been observed experimentally can be explained without reference to
the effect of accumulated damage on the flow pattern and hence upon the
cavitation regime. In my opinion, however, this latter effect is of sub-
2
Nuclear Engineering Dept., The University of Michigan, Ann Arbor, Mich.
FIG. 17—Experimental erosion rate-time curves from two identical tests on CEGB impingement erosion rig.
stantial importance in most cases, and hence should somehow also be in-
cluded hi a predicting model.
Very qualitatively, it seems to me that all the observed rate-time curves,
including those showing several maxima and minima, can also be ex-
plained simply on the basis of flow-pattern changes through the effect of
accumulated damage. In given situations, these changes are capable of
both triggering local cavitation which will then spread the damage region
and increase damage rates in a generally unpredictable manner and of
protecting the surface by attenuating the cavitation shocks imposed on the
surface through increasing distance from the cavitating region as the
damage in a given region becomes substantial, or both. In various situa-
tions, there are also numerous other mechanisms which might be men-
tioned which are capable of either increasing or decreasing damage. Hence,
it would seem to me that the feedback between accumulated damage and
the flow pattern is capable of increasing, decreasing, or maintaining con-
stant damage rates, depending upon which of the mechanisms predomi-
nated at the moment. Since the dominant mechanism may change as dam-
age proceeds, it is conceivable that maxima and minima may be generated
in the rate curve by these flow effects as well as by the fatigue statistics
discussed in the paper.
D. Pearson* (written discussion)—Within Central Electricity and Gen-
erating Board (CEGB), erosion results have been reported as cumulative
mass loss plotted against time. In the published form, such data cannot
be differentiated accurately, and hi Fig. 17 I have plotted the mean
rate of mass loss between weighings against time for a pair of specimens
exposed to the same conditions. The points are not a close fit to any smooth
curve, and this must result in the interpretation being influenced by per-
sonal opinion, though the results would seem more consistent with Fig. 2,
rather than Fig. 1 of the paper.
A major problem in erosion research is to determine the relative im-
portance of the following on the detail shape of the observed erosion mass
loss-time curve:
1. inconsistency in test machine performance,
2. the effect of the material being detached in finite-size pieces,
3. inconsistencies hi erosion resistance between specimens made from
the same batch of material, and
4. genuine variations hi erosion rate as erosion progresses.
The author is only interested in the last, but it cannot be distinguished
from the other three spurious causes using the test results for a single
specimen. Results are required for many specimens of one material tested
using the same nominal conditions (however, to permit Effects 1 and 4 to
be separated, the specimens should not all be eroded together).
3
Central Electricity and Generating Board, Research and Development Dept.,
Marchwood Engineering Laboratories, Marchwood, Southhampton, Hants, Eng-
land.
Has the author obtained any such data, and has he been able to analyze
these data to determine the shape of the true erosion curve?
J. M. Hobbs4 (written discussion)—Mr. Heymann's excellent paper
has gone a long way toward explaining the effects of time on erosion rate
but seems to have painted a rather gloomy picture of the present state of
the art.
According to Mr. Heymann, it is still apparently a matter of opinion
whether any steady-state value of erosion rate has a definite significance.
Surely, even though the maximum erosion rate is nothing but "the 'deluge'
of erosion fragments from the top surface layer," the same is true under
field conditions. Thus in any test, provided that the maximum erosion
rate is maintained for sufficient time to indicate that the scatter in the
lifetimes of individual particles is evenly distributed, it could be used
for comparative rating of materials. This would seem to be preferable to
the time required to attain some specified mean depth of penetration,
observing that the latter method is sensitive to surface conditions and
would therefore necessitate extreme care in specimen preparation.
On the subject of correlations of incubation periods with fatigue limits,
I would agree that this approach is hardly logical, but add that in some
cases it is unavoidable. In liquid impact tests it is possible to vary the im-
pact velocity and the number of impacts corresponding to the incubation
time at each velocity. Thus for different materials it is possible to derive
some threshold or endurance limits of velocity which can be correlated with
their respective fatigue endurance limits.
It is not possible in any standard vibratory cavitation erosion test to
vary the stresses caused by cavity collapse, and hence tests must be con-
ducted with a nominally fixed stress system. To correlate the incubation
periods of different materials, fatigue lives of the same set of materials
would have to be determined all at the same stress. This would be im-
practicable for more than a very limited range of materials, as a stress
equal to the endurance limit of a strong material may well be greater than
the yield stress of a weaker one.
Hence, in this type of test for comparison of the behavior of different
materials there is little choice but to use fatigue endurance limit on the
one hand and some function of either the incubation period or the erosion
rate on the other.
R. I. Armstrong5 (written discussion)—Mr. Heymann is to be compli-
mented on his lucid review which provides some clarification of ideas on
the reasons for variations in erosion rate with time. His digital analog of
the erosion process, which, while it might appear to be oversimplified,
does result in similar curves to those obtained from tests.
4
Properties of Fluids Div., National Engineering Laboratory, East Kilbride,
Glasgow, Scotland.
5
Research and Development Div., C. A. Parsons and Co., Ltd., Newcastle upon
Tyne, England.
Accepting that Mr. Heymann's first models are tentative only, doubt
remains whether his computed rate-time curves (for example, Fig. 10 of
the paper) are consistent with experimental observations made at C. A.
Parsons and quoted by Mr. Heymann in support of his thesis in Fig. 11.
It must be granted that there is a superficial resemblance in the shape of
the curves, but this similarity breaks down when an attempt is made to
apply the basic premise of the theoretical curves to the experimental ones.
Mr. Heymann's premise is that one simple distribution curve should serve
for all sources of scatter on the "lifetimes of erosion fragments." In conse-
quence, the time taken to reach the first peak (time = 1 in Fig. 10 of the
paper) is given by the mode of the distribution curve. Viewed in this way,
the first peak in the erosion rate time curve is, as Mr. Heymann aptly
describes, "a deluge of erosion fragments from the top surface layer."
This breaking up of the surface is readily observable by microscopic ex-
amination. Therefore, to check the applicability of the model, the time
FIG. 18—Appearance of 18W-6Cr-0.7C tool steel after 2-hr erosion test on

the polished surface (X360).
taken for significant break up of the surface must be compared with the
time taken to reach maximum erosion rate, these being presumed equiva-
lent. From Fig. 11 of the paper, the tune to maximum rate for tool steel
is about 35 hr, which is typical of the current tests at C. A. Parsons. How-
ever, photographs of tool steel specimens eroded in current tests for
only 2 hr (Fig. 18) show the top surface to have reached a stage of rapid
disintegration, the amount of material lost by this time being only about
2 per cent of that lost in 35-hr test.
The observed increasing rate, which follows and which has been shown
previously to be by no means coincident with disintegration of the top
suface layer, must either be the result of a more rapid production of

particles from the subsurface layers or of an increased size of fragment.
No evidence for the latter has so far been discovered at C. A. Parsons
where stereoscopic microexamination of eroded surfaces with comparison
photographs at successive stages have been used to detect such fragmenta-
tion. A limited number of actual measurements of pit depth changes and
of eroded particles collected by the specimen guard rings have also failed
to reveal the presence of materially larger fragments. While this evidence
is of a negative nature and the loss of larger particles may still have oc-
curred undetected, it appears more likely that increasing numbers of
smaller particles are actually produced. The explanation should then
FIG. 19—Vnetched section through an undercut erosion pit in 18W-6Cr-0.7C

tool steel after 100-hr test (X160).
be in terms of factors which can produce equivalent increases in the rate of

fatigue damage. These are surely tied up with surface geometry, for ex-
ample, Mr. Heymann in Part I of his paper suggests that the damage rate
would be increased by the development of irregularities smaller than the
impinging droplets and reduced where the scale of roughness is larger
than the droplet size (Fig. 8 of the paper). The occurrence of considerable
undercutting in Parsons' impingement tests, Fig. 19, suggests that sec-
ondary water flow effects are responsible, such as Bowden and Brunton
have postulated in Refs 19 and 20 of the paper. The tensile loading es-
sential to produce fatigue would be more readily obtained in this manner.
F. J. Heymann (author)—The discussion is gratifying and has served
to put the statistical model into a clearer perspective and to throw more
light on some of its limitations.
It is encouraging to see that Dr. Engel has been thinking along some-
what similar lines, and I am grateful that she has given us her early
thoughts in some detail. She has examined and modeled the actual physi-
cal occurrences more specifically than I have dared to do. What I have
proposed is a phenomenological model in the sense that it can predict
consequences whose real physical causes are much more intricate than
the model explicitly admits. This is useful up to a point, in that it may
show what significance statistical properties of the material and of the
erosive environment do have for the erosion-time relationship, without
requiring a precise or quantitative knowledge of the physical progress of
damage leading to material loss. If one seeks to develop a more realistic
model based explicitly on the physical processes believed to be going on,
one treads on much more dangerous ground. But this, eventually, has to
be done and surely there are few people better acquainted with this
ground than is Dr. Engel, or more qualified than she to lead the way
over it.
I should like to take up Mr. Pearson's comments next. There can be
no doubt that the factors which he lists play a role, but I do not agree
that I am interested only in the last or that all the others are "spurious
causes." It is, in fact, part of my thesis that such factors as random (not,
of course, systematic) fluctuations in test machine performance, the size
distribution of erosion fragments, and strength variations within the ma-
terial, all could help mold the shape of the erosion-time curve and could
affect the maximum erosion rate, independently of the eventual steady-
state rate which may be determined mainly by the average conditions.
I do not believe, therefore, that it is meaningful to talk of a "true" erosion
curve from which these effects are subtracted, although it would certainly
be instructive to be able to separate and evaluate the various influences.
To answer Mr. Pearson's question, we have not made any experiments
with this objective, and I must admit that at present we have no test fa-
cility which is controllable enough to permit it to be done. Mr. Pearson's
graph (Fig. 17) is very interesting in that both curves show the same
fluctuations. As I understand it, they represent two specimens exposed
simultaneously in the same test rig; thus, the fluctuations could be due to
slightly different conditions during the various runs, though one might
then have expected greater fluctuations during the rising portion of the
curve (accumulation or acceleration stage). It may therefore be appropri-
ate to recall Dr. Engel's comment that oscillations "will be particularly
evident in the steady-state period."
I certainly cannot dispute Professor Hammitt's point that flow-pattern
changes due to accumulated damage will surely affect the erosion rate
and could explain many rate-time patterns. In Part I of the paper I have,
in fact, referred to a report by Hammitt et al which discusses this in detail.
What intrigued me, however, was the similarity between the commonly
observed rate-time behavior in impingement erosion and in cavitation
1 10 EROSION BY CAPTATION OR IMPINGEMENT
erosion, which occur in very different flow regimes. It would seem to be

a great coincidence if this behavior were entirely due to feedback of the
surface geometry on the flow patterns. I feel sure that both statistical
effects and flow effects must play some role in all of the erosion regimes;
however, which of these is predominant or what their relative importance
is in any particular erosion regime, this is a question which awaits answers.
Mr. Armstrong's discussion is a contribution toward this. His observa-
tions are valuable and undeniably constitute evidence for attributing the
shape of Fig. 11 more to geometry effects. Another significant difference
between Figs. 10 and 11 is that hi the former the peaks occur at about
equal time intervals from the beginning of exposure, whereas in the latter
the time at the second peak is about four times the tune at the first peak.
Mr. Armstrong's last two sentences corroborate the last paragraph in the
section on "Effect of Surface Condition" in Part I of the paper. It would be
good H the effects of surface geometry could somehow be incorporated
into the mathematical model, but I have not so far been able to devise a
convincing way of doing that.
Lastly, as Dr. Hobbs points out, the practical necessity of using erosion
test data for comparative or predictive purposes remains and cannot be
bypassed. I am not entirely happy with any of the criteria which have been
proposed, including the one tentatively suggested in this paper and com-
mented on by Dr. Hobbs. However, ft the original surface preparation
greatly affects the time to reach a given depth of erosion in a test, then
the same would be true hi actual service, provided the specified depth
of erosion is the same in each case. Thus, it would seem to be more real-
istic to include this effect than to try to evade it. The problem is that the
total testing time to obtain the specified depth of erosion may, of course, be
very long for a good material. If, on the other hand, the specified depth
of erosion for the test is chosen to be much less than that tolerable in
service, then I do not believe that accurate predictions can be made from
the test results toward service performance. If that is being "gloomy" then
I must confess to it, but corresponding conclusions have had to be
accepted in fatigue and creep testing.
With respect to the determination of velocity thresholds or endurance
timits and their correlation with fatigue endurance limits, this I agree
is sound in principle, but there is probably also a size effect which makes
the velocity threshold a function of drop or jet diameter. The results of
Refs 4,10, and 32, among others, suggest this quite strongly.
In summary, I am grateful for the interest evidenced and the thoughts
presented by the discussers. I believe that the statistical effects must be
present, and even if it can be shown that they are overshadowed by other
effects something will have been learned.
D. J. Beckwith1 and J. B. Marriott1
Water Jet Impact Damage in a Cobalt-

Chromium-Tungsten Alloy
REFERENCE: D. J. Beckwith and J. B. Marriott, "Water Jet Impact

Damage in a Cobalt-Chromium-Tungsten Alloy," Erosion by Cavitation
or Impingement, ASTM STP 408, Am. Soc. Testing Mats., 1967, p. 111.
ABSTRACT: The mechanism of water jet erosion in a wrought cobalt
base alloy used for shielding steam turbine low pressure blades is shown
to be a three-stage process. Detailed metallographic examination of
lightly eroded polished surfaces has confirmed the presence of carbide/
matrix interface cracking and slip line cracking during the initial stage.
Similar damage is observed during conventional fatigue deformation.
The relative importance of these two modes of cracking during the
subsequent stage is indicated, and the over-all kinetics of the process are
discussed in terms of the alloy structure and the nature of liquid jet
impact.
KEY WORDS: erosion, liquid impact, cavitation, hydraulic jets, fatigue
(materials), metallography, cobalt, dispersed phase alloys, tungsten car-
bide, cracking, impingement
Erosion by impacting water droplets can occur on the leading edge of

the rotating blades in the low pressure stages of a steam turbine. The
damage is most severe on the last row of blades, and the problem has be-
come more acute as the peripheral speed has increased. It is current prac-
tice to protect the leading edge with an erosion shield. The shielding
alloys used in many large machines, however, show a significant rate of at-
tack, and a more erosion resistant material is required.
The events leading to the formation and impact of water droplets
against the moving blades have been described by many authors, notably
Gardner [I],2 and the origin of the large drops responsible for damage
has been attributed to areas of condensed steam on the stator blades. Sig-
nificant contributions to the physics of the impacting drop have been
made by Bowden and Brunton [2], Bowden and Field [5], and Engel
1
Metallurgist and head of materials group, respectively, Central Metallurgical
Laboratories, English Electric Co., Whetstone, Leicester, England.
2
paper.
Ill
1 12 EROSION BY CAVITATION OR IMPINGEMENT
[4]. The impacting drop produces a high pressure pulse, P, normal to

the metal surface represented in its simplest form by the water hammer
equation
P = Pcv
where:
p = density of liquid,
c = velocity of the compressive waves in the liquid, and
v = velocity of the drop.
The pressure is maintained for a period determined by the time required
for release waves to pass through the compressed liquid. Once this has
occurred the water will be free to flow outward from the point of impact
at a velocity at least equal to the initial velocity of the drop [5],
Comparatively little work has been reported, however, on the behavior
of materials under actual or simulated conditions, and it is not possible,
in all cases, to assess the resistance to erosion from a knowledge of con-
ventional properties. Therefore, it has been necessary to study the mecha-
nism of erosion damage in an alloy of interest. The shields used in service
are frequently made from a cast or wrought cobalt-chromium-tungsten
alloy. The wrought version, Haynes 6B, has similar over-all erosion re-
sistance to the cast material, and since it also has better uniformity of
structure and properties, it was selected for this study.
The initial results of this work were reported by Marriott and Rowden
[6], who described observations of surface deformation and cracking
during the early stages of erosion by water jet impact. They also showed
that the type of erosion damage found in service could be reproduced in
the test rig. The results of further observations made on this test rig will
now be given, and the metallographic changes occurring correlated with
those observed during more conventional modes of deformation.
The sequence of events occurring during growth of erosion damage in
this alloy is described, and the importance of alloy structure in controlling
the kinetics of the process is indicated.
Experimental Details
The erosion test equipment used is a rotating disk type and has been
described in the previous paper [6]. Specimens are attached to the pe-
riphery of the disk and on each revolution cut through a water jet % 4 in.
diameter. Testing was carried out at a pressure of 0.5 psi absolute at
relative impact velocities of 1030 and 1400 ft/sec; conditions which are
found in service. A track of erosion damage is produced on the test
face of the specimen, and this deepens as the test proceeds. According
to the water hammer equation, the stress at the point of impact is about
90 ksi. A stress of 145 ksi can be computed using an alternative equa-
tion.3 Test specimens, polished before testing, were used to study the
8
D. Pearson, Central Electricity Generating Board, private communication.
BECKWITH AND MARRIOTT ON WATER JET IMPACT DAMAGE 11 3
early stages of erosion. The surface of the eroded test specimen was ex-
amined by specialized optical and electron metallographic techniques,
including Normarski interference contrast and scanning electron micros-
copy. Mechanically polished and electropolished specimens were used
and, although no differences were found in the development of damage,
the study of matrix deformation was facilitated by the latter technique.
TABLE 1—Typical analysis of Haynes alloy 6B.

c, Si, Mn, s, p, Ni, Cr, W, Fe, Mo, Co,
% % % % % % % % % % %
Chemical analysis of
bulk material. . . . 1.14 0.62 0.84 0.010 0.009 0.69 28.6 4.95 2.34 0.02 balance
Electron probe micro-
analysis of:
Matrix 75 5 S 7 ? 3 67.6
Carbide 74 0 7 8 0 7 13.4
FIG. 1—Erosion curve for Haynes 6B at 1030 ft/sec illustrating the type of
damage observed during each stage.
The structure of the alloy in the wrought condition consists of a dis-

persion of approximately 10 volume per cent of a coarse carbide in a
highly alloyed cobalt base matrix. The compositions of matrix and
carbide have been determined by electron probe microanalysis and are
given in Table 1.
FIG. 2—Shadowed plastic replica showing carbide/matrix interface cracking

around Particle A and slip line cracking at B in Haynes 6B during Stage 1 (X9400).
Reduced one third for reproduction.
FIG. 3—Carbide particle loss and slip line cracking in Haynes 6B during Stage
2 (X2700). Arrows indicate where carbide particles have been removed.
BECKWITH AND MARRIOTT ON WATER JET IMPACT DAMAGE 11 5
FIG. 4—Section showing carbide particle loss in Haynes 6B during Stase 2

(X1175).
FIG. 5—Extended cavity in Haynes 6B during Stage 2 (X1800).
Discussion of Observations
Erosion Damage
Erosion in this alloy is a three-stage process which is related to the
weight loss v endurance curve (Fig. 1). This figure also indicates the
profile of the test specimen associated with each stage. The first indica-
tions of damage are carbide/matrix interface cracking and cracking of
persistent slip lines in the matrix (Fig. 2); both forms of cracking are caused
1 16 EROSION BY CAV1TATION OR IMPINGEMENT
by deformation of the matrix by the force of direct impact, P. Evidence of

typical carbide/matrix interface cracking is shown around particle, A,
and a typical example of matrix cracking at B.
Examination at the subsequent stage of damage, but before significant
weight loss, shows that the first metal loss is carbide particle removal
(Fig. 3). Depending on the extent to which the interface is cracked the
carbides will be totally or partially removed (Fig. 4). The action of water
flowing across the surface at high velocity is thought to be significant at
this stage. Surface examination after further attack reveals the presence
FIG. 6—Damage on a persistent slip line in Haynes 6B during Stage 2 (X8000).
of larger cavities which would seem to be produced by the removal of

blocks of material by surface water impacting against the sides of the
original carbide cavities. No directional effect was observed; cavities and
cavity shape were orientated randomly with respect to the test direction.
Some of the edges of these larger cavities are parallel to slip directions
indicating that this process is assisted by the presence of slip line crack-
ing in the matrix.
An example seen at a f airly early stage of enlargement is shown in Fig.
5. Further evidence of the role of fast flowing surface water during the
early stages of the erosion process was found during examination of
the slip markings in lightly eroded material (Fig. 6). Damage has oc-
BECKWITH AND MARRIOTT ON WATER JET IMPACT DAMAGE 117
curred at the edge of a persistent slip line. This could only have been
caused by a force acting along the specimen surface. A similar effect
was observed by Bowden and Brunton [2] in their single impact work on
70/30 brass. We do not believe that this type of damage contributes to
the over-all erosion to any significant amount but is valuable in demon-
strating that such forces are present. With continuing attack, the type of
damage shown in Fig. 5 becomes more extensive but without an ap-
preciable increase in the depth of damage. This indicates that at this
FIG. 7—Cracking extending from the base of the eroded track during Stage 3
in Haynes 6B (X400).
stage of the process, direct impacts are effective only on the remaining
parts of the original surface.
When erosion has occurred to a stage at which little of the original
metal surface remains, the impacting jet becomes increasingly enclosed
by the eroded track, and the liquid is not free to release compressive
stresses by surface flow. Forces are then exerted at stress concentrations
within the eroded area sufficient to propagate cracks into the main body
of the material (Stage 3). The distance travelled by this cracking will de-
pend on the duration and magnitude of the pressure pulse and a number
of material factors influencing crack propagation, such as work harden-
ing capacity and fracture toughness. These cracks quickly link up under
repeated application of stress causing substantial metal loss. Typical ex-
amples of cracks running into the main body of the material from the
eroded cavity are shown in Fig. 7. As the cavity deepens, cracking be-
comes more concentrated at the base where the stresses are highest. Some
slowing down of the erosion rate hi Stage 3 has been found after con-
siderable erosion has taken place. This effect has been observed to a vary-
ing degree on a number of test rigs and is believed to occur under service
conditions. The reason for this behavior is not yet fully understood.
Impact velocity has an effect on the relative importance of the three
stages of the erosion process. The difference between erosion curves for
Haynes 6B at 1030 ft/sec and at 1400 ft/sec is shown in Fig. 8. The ero-
FIG. 8—Erosion curves for Haynes 6B at 1030 and 1400 ft/sec.
sion process is faster at the higher impact velocity, and the second stage of
the process is less effective in reducing the over-all rate of metal loss. The
1400 ft/sec curve is a portion of the full test curve shown in Fig. 14.
Damage After Mechanical Stressing
In order to achieve some understanding of the deformation processes
taking place during erosion by liquid impact, metallographic changes
occurring during more conventional modes of deformation were studied
hi this alloy.
Surface metallographic examination of fatigue cracks formed during
high-stress reverse bending and push-pull testing revealed features (for
example, cracking along slip directions) commonly associated with
FIG. 9—Fatigue crack in Haynes 6B propagating in slip directions and circum-

venting a carbide particle (X1800).
FIG. 10—Carbide particle cracking during plain bending in Haynes 6B (X1800).

FIG. 11—Cracking in carbide particles behind the fracture face of a fractured

tension test piece in Haynes 6B (XI800).
FIG. 12—Carbide particle cracking observed during the early stages of erosion
in Stellite 20 (X1800).
fatigue damage in pure metals [7]. Figure 9 shows an example of crack-

ing in slip directions; this type of cracking is also observed during the
early stages of erosion.
In the conventional fatigue test, carbide/matrix interface cracking
and carbide particle cracking are both associated with the propagation
of the crack. An example of carbide/matrix interface cracking is also
shown in Fig. 9, where the main fatigue crack has propagated around a
carbide particle at A, but cracking across carbide particles was more
generally observed. Carbide particle cracking is also observed after a
FIG. 13—Erosion curves for Haynes 6B and a 12 per cent chromium steel at
1030 ft/sec.
single bend to 30 deg (Fig. 10) and in uniaxial tension (Fig. 11). At the
early stages of erosion carbide particle cracking has not been seen in this
alloy, but it has been observed in cobalt-chromium-tungsten alloys con-
taining a high percentage of carbide phase (Fig. 12) and also in sintered
tungsten carbide-cobalt alloys.
This difference hi behavior of Haynes 6B in mechanical testing and
erosion may be associated with the nature of the stressing during liquid
impact. The work of Gurland and Plateau [8] has shown that, in a uni-
formly stressed system, a second phase particle in a ductile matrix
produces a stress concentration effect at the particle resulting in local
plastic yielding. Further stressing results in fracture of the component
least able to absorb the additional stress. In conventional modes of stress-
ing we have observed that fracture of the carbide particle occurs. During
liquid impact, a localised area of the surface receives a short duration
stress pulse. Within this area both components of the structure are equally
stressed. The second phase carbide particle, in this case because of its
elastic properties, may transmit a proportion of the stress pulse to under-
lying matrix material where it may be absorbed by plastic strain. That
part of the surface which consists of matrix material will absorb part of
the initial impact by plastic strain, as we have observed. In this situation
the tendency for carbide particle cracking is reduced, and matrix or inter-
face cracking could occur.
In alloys with a high percentage of carbide particles, the matrix is
more highly constrained and the over-all plasticity reduced. The necessary
FIG. 14—Erosion curves for Haynes 6B and a sintered tungsten carbide/10 per
cent cobalt alloy at 1400 ft/sec.
transference of strain during liquid impact cannot occur to a sufficient ex-

tent, and cracks form in the carbides located at the surface.
Improved Resistance to Erosion

From a study of the erosion curves for a number of materials we ob-
serve that the presence of Stage 2 (Fig. 1) is important in reducing the
extent of erosion after a specific time. This is shown in Fig. 13 by the ero-
sion curves for Haynes 6B (420HV) and for a 12 per cent chromium steel
(530HV) in which Stage 2 is not observed. Stage 2 is associated with the
presence of carbide particles in Haynes 6B and is also found in other
materials containing similar dispersed phase particles.
However, the presence of a large volume content of dispersed phase
particles, although contributing significantly to the reduction of erosion
at early stages, may result hi a low resistance to crack propagation in the
bulk material. This will produce a high rate of erosion in Stage 3. Be-
havior of this type is exhibited by some sintered tungsten carbide-cobalt
alloys and is illustrated in Fig. 14 for an alloy containing 10 per cent
cobalt.
In tests on this alloy Stage 2 extended to 500 X 103 impacts, after
which the weight loss increased rapidly until 550 X 103 impacts was
reached. After examination of the test specimen at this point it was con-
cluded that further testing might lead to fracture of the test specimen.
FIG. 15—Typical Stage 3 cracking in a sintered tungsten carbide/10 per

cent cobalt alloy (X1800).
The extensive nature of cracking was confirmed by metallographic ex-

amination. It is clear from the photomicrograph in Fig. 15 that the propa-
gation of this cracking is assisted by carbide particle cracking.
The following aspects of dispersed phase alloys are worth further
examination with a view to improving the erosion resistance.
1. The dispersed phase/matrix interfacial energy.
2. The matrix strength and resistance to crack propagation.
3. Variation in the dispersed phase particle size and shape and in the
volume fraction present.
It should be emphasized that in addition to erosion resistance the me-
chanical and physical properties of any proposed shielding alloy must be
compatible with service conditions at the tip of a long L.P. blade.
Conclusions
The mechanism of erosion damage in Haynes 6B is a three-stage proc-
ess.
1. Carbide/matrix interface cracking under the action of impact pres-
sure accompanied by fatigue cracking of the matrix.
2. Carbide particle removal and shallow surface damage caused prin-
cipally by the action of water flowing across the surface at high velocity.
3. Mass metal removal caused by intersecting cracks propagating un-
der the action of a mass of water impacting in the eroded cavity.
A cknowledgments
The assistance of the University of Leeds in producing Fig. 6 is ac-
knowledged with thanks. We wish to thank the Director of Research,
English Electric Co., for permission to publish this paper.
References
[1] G. C. Gardner, "Events Leading to Erosion in the Steam Turbine," Proceedings,
Institute of Mechanical Engrs., Vol 178, 1963-1964, pp. 593-623.
[2] F. P. Bowden and J. H. Brunton, "Deformation of Solids at Liquid Impact at
Supersonic Speeds," Proceedings, Royal Society of London, Series A, Vol 282,
1961, pp. 433-450.
[3] F. P. Bowden and J. E. Field, 'The Brittle Fracture of Solids by Liquid Im-
pact and by Shock," Proceedings, Royal Society of London, Series A, Vol 282,
1964, pp. 331-352.
[4] O. G. Engel, "Water Drop Collisions with Solid Surfaces," Journal of Research,
Nat. Bureau of Standards, Vol 54, 1955, pp. 281-298.
[5] G. Taylor, "Oblique Impact of a Jet on a Plane Surface," Philosophical Trans-
actions, Royal Society of London, Series A, Vol 260, 1966, No. 1110, pp. 96-
100.
[6] J. B. Marriott and G. Rowden, 'The Erosion of a Cobalt-Chromium Alloy by
Liquid Impact," Philosophical Transactions, Royal Society of London, Series A,
Vol 260, 1966, No. 1110, pp. 144-149.
[7] N. Thompson, N. J. Wadsworth, and N. Louat, "Origin of Fatigue Fracture
in Copper," Philosophical Magazine, Series 8, Vol 1, 1956, pp. 113-126.
[5] J. Gurland and J. Plateau, "Mechanisms of Ductile Rupture of Metals Contain-
ing Inclusions," Am. Society Metals Transactions Quarterly, Vol 56, 1963,
September, pp. 442-455.
Allen Smith,1 R. P. Kent,1 R. L. Armstrong1
Erosion of Steam Turbine Blade Shield

Materials
REFERENCE: Allen Smith, R. P. Kent, and R. L. Armstrong, "Erosion

of Steam Turbine Blade Shield Materials," Erosion by Cavitation or Im-
pingement, ASTM STP 408, Am. Soc. Testing Mats., 1967, p. 125.
ABSTRACT: The physical processes preceding blade erosion by wet
steam in low-pressure turbines have been traced from the collection of
water droplets larger than 30 p by the stationary blading to the stripping
of this water into drops at the blade trailing edges and the impact of these
drops with the following moving blade row.
Experiments indicate that the maximum stable drop diameter before
the last row of a 500-Mw turbine is 800 M, being a function of downstream
Mach number as well as the Weber number.
Stresses produced by impact of drops have been calculated by existing
theories to be sufficiently high to initiate damage in shield materials.
Turbine experience and our experiments show that erosion rate is time
dependent, with three successive zones: a primary zone with a low erosion
rate, a secondary zone where the rate rises to a maximum, and finally, a
tertiary zone, where the rate diminishes to a steady-state value. Experi-
ments on artificially roughened specimens suggest that this reduction in
erosion rate may be partially caused by droplet breakup on the peaks of
the eroded surface and partially by the retention of water in the cavities.
The terminal rate dictates the life of a turbine erosion shield, and
materials must be tested well into the tertiary zone before a valid
assessment can be made.
KEY WORDS: steam, turbines, impact, erosion, hardness, liquid impact,
impingement
Nomenclature
C Acoustic velocity
Cd Drag coefficient
d Droplet size
dc Critical droplet size
g Gravitational constant
/ Impact number
1
Head, LP turbine and compressor group, chief metallurgist, and research
metallurgist, respectively, Research and Development Div., C. A. Parsons and Co.,
Ltd., Newcastle Upon Tyne, England.
125
L Characteristic length (taken as blade chord)

M Weight loss, g/cm2
Mn Mach number, Mn = V/C
p Pressure
Rm Ratio of weight loss of test material to standard
Rv Ratio of volume loss of test material to standard
T Time, hr
U Reference velocity (taken as vector mean steam velocity through
blading)
V Fluid velocity
W Theoretical weight of impacting water per unit area, g/cm2
We Weber number, We = (pVdc}/a
v Kinematic viscosity
p Density
ps Density of steam
pw Density of water
a Surface tension
An important stage in the development of large, high-speed turbine

generators in Great Britain was marked by the recent commissioning of
a single-line, 3000-rpm, 500-Mw unit, which was the first of some 50
units of a similar rating under construction by various manufacturers.
An important design requirement of these turbines was to obtain a small
leaving loss from the last blade row, which demanded the maximum
exhaust area consistent with blading reliability. The exhaust area was
accommodated by using three double-flow LP turbines, making six ex-
hausts in all, and by increasing the blade tip diameter to 136 in. from
the previous 120 in. The higher centrifugal blade loading and transonic
aerodynamic performance imposed by the increased tip diameter were
accommodated by careful design. The higher tip speeds, however, in-
creased the erosion potential of the wet steam as the impact speeds
of the blade and drops were greater. Tests therefore became necessary
to determine whether the increased erosion due to high impact speeds
would more than offset the reduction in erosion resulting from the lower
moisture content of LP steam brought about by the general adoption of
reheat in steam cycles. For this reason the Central Electricity Generating
Board (CEGB) and turbine manufacturers began collaborative investi-
gations into erosion. Both the physical process of erosion and the erosion
resistance of different materials are being studied at C. A. Parsons and
Co., Ltd.
In tracing the physical causes of erosion of LP turbine moving blades
assessment of the effectiveness of stationary diaphragm blades in collect-
ing water droplets of different sizes was necessary. A wet air tunnel was
therefore built. This tunnel, with other rigs, also enabled the maximum
FIG. 1—Contra-rotating erosion test rig.
size of drop torn from the blade trailing edges to be estimated; it is

these drops which are considered to be responsible for erosion.
Design of a test rig for comparing the erosion resistances of blade
shield materials at impact speeds of up to 2000 ft/sec started at C. A.
Parsons in 1959. This erosion rig was the third to be constructed by the
Company: the first (in 1910) was capable of 1000 ft/sec and the second
(in 1930) of 1300 ft/sec. The new machine, in which the sprayers and
specimens operated in centra-rotation, has now been in service for four
years. It has been used to compare the erosion resistance of several
blade shield materials and to study the variations hi the rate of weight
loss during erosion. Recently an attempt has been made to explain ob-
served differences in initial and terminal erosion rates by testing speci-
mens with then* surfaces artificially roughened to simulate the peaks and
craters on partly eroded surfaces. Their weight losses have been com-
pared with those for specimens with unroughened surfaces.
Apparatus
Wet Air Tunnel

Water collection by turbine blades and the drops produced from the
shearing action of the flow on the resulting water sheet were studied on
a wet air tunnel. This consisted of an arcuate cascade of diaphragm
blades supplied through a trapezoidal shaped steadying length from a
6000 ft/min3, 60-in. WG fan. Water was injected about 3 ft upstream
from the blading through an air blast nozzle which distributed the drop-
lets fairly uniformly across the duct. Those droplets reaching the tunnel
walls were ducted away before reaching the cascade.
Erosion Testing Machine

The machine has two separately driven contra-rotating shafts, each
carrying a mild steel overhung disk; the specimens are mounted at the
perimeter of one disk and the sprayers on the other, so that impact speeds
of up to 2000 ft/sec can be obtained (Fig. 1). Both jets and specimens
are enclosed hi a vacuum chamber, the shaft glands being steam-packed
so that vacua of up to 29-in. Hg can be controlled by bleeding air into
the primary of a two-stage ejector.
Each shaft is driven from a two-pole, water-cooled induction motor,
rated at 90 hp at 12,000 rpm. Independent variable frequency supplies
are provided to test materials at different impact speeds, since the rota-
tional speed of the jets as well as the water supplied must be kept con-
stant to ensure consistency in the size of the water droplets.
Water for the sprayer is fed along the hollow center of the shaft and
passes through radial holes into two nozzles mounted on the smaller disk,
SMITH ET AL ON STEAM TURBINE BLADE SHIELD MATERIALS 1 29
the flow being equally divided between them. The sprayers are mounted
diametrically opposite to one another, the radius of the nozzles being
9l/s in.; these nozzles are 1A FJFH Watson fan type, pressure-atomized
units, and both their flow rate and droplet size are checked regularly on
a static rig.
The specimens are 1A ifl. diameter and J/4 in. thick, terminating in an
integral shank, and are mounted in tool steel guard rings to prevent frag-
mentation and unrepresentative weight loss at the edges (Fig. 2). They
FIG. 2—Specimen and sprayer.
are held in holders by circlips and are mounted in diametrically opposite

positions, 23 in. apart.
The standard conditions for comparing materials in the rig are a
vacuum of 28.5 in. Hg, a flow of 1.5 gal/min, and an impact speed of
1730 ft/sec; this corresponds to the speed at the mean height of the
outer erosion shields of a 136-in. tip diameter final blade row in a 500-
Mw turbine. The temperature of the test chamber atmosphere was usually
one or two degrees above the saturated steam value of 92 F, and the
mean inlet water temperature to the nozzles was 60 F.
In comparing the erosion rates hi the test rig with those in turbines
the rates in the rig were known to be ten times greater at impact speeds
of 1570 ft/sec.
FIG. 3—Sprayer water droplets in contra-rotating erosion rig during test run
under standard conditions (XllVz). Reduced one third for reproduction.
Photographic Techniques
Photographic apparatus for establishing the size and movement of the
droplets was arranged parallel to the axis of the rig with a camera on one
side of the vacuum chamber and a light source on the other, so that the
droplets were silhouetted against the light passing through electrically
heated sight glasses in the plane sides of the vacuum chamber.
SMITH ET AL ON STEAM TURBINE BLADE SHIELD MATERIALS 131
The light source was a 10-kv flash unit, triggered from an auxiliary
5-kv circuit operated from either the sprayer shaft or from both shafts.
Under atmospheric conditions high-speed flash photographs initially
taken of the Watson sprayers established the mean droplet size as 120-/X,
Sauter mean diameter: [SMD = ^ds/^d2]. The spray diverged to about
75 deg in one plane, producing a thin fan about 1A in. thick, 2 in. from
the nozzle, Evans and Robinson [I].2
Photographs taken under standard test conditions within the rig at
10,000 rpm and a vacuum of 28.5 showed the droplet sizes to be 90 to
100-/X SMD (Fig. 3) or rather less than under static atmospheric condi-
tions. The angle of divergence of the jet also fell to about 21 Yz deg, as
deduced from the erosion markings on the specimen holders.
FIG. 4—Relative impact velocity of drops on to last row moving blades.
Physical Aspects of Erosion
Process
Turbine blade erosion is initiated by a relatively small proportion of
the water droplets in the steam being collected on the concave pressure
face of the cylinder blades and forming a film which is drawn towards the
trailing edge by the drag of the steam. Here the film grows and in a region
of separated flow may even pass around the trailing edge on to the convex
face before being torn away by the shearing action of the main steam
flow. Relatively large drops (50 to 800 /* diameter compared with a mean
size of less than 0.5 /*) are consequently produced and have to be ac-
celerated from rest by the steam. These larger drops arrive at the inlet
plane of the moving blade row at only a fraction of the absolute velocity
of the steam and are subsequently struck by the convex surfaces of the
2
paper.
following moving blade row. In the case of drops of 200 /*, and above,
the terminal velocity before impact may be under 300 ft/sec, but the
relative impact velocity on to the blading can approach the blade speed
(1780 ft/sec for 136-in. blade tip diameter at 3000 rpm, Fig. 4). The
position of damage to actual blading (Fig. 5) confirms that such large
drops are responsible for erosion.
These large drops probably form only a small part of the total water
present at the exhaust of a turbine; most are probably 0.5 /* or less in
FIG. 5—Eroded turbine blades after 12 years' service at a tip speed of 1115
ft/sec and a mean wetness of 11 per cent.
diameter, since the blue end of the spectrum is absorbed if white light
is shone through the steam in a turbine exhaust.
Droplet Collection
Operating experience has shown that erosion damage to the stationary
blades in a turbine is relatively insignificant compared with that to the
moving blades. This suggests that the water film deposited on the moving
blades is centrifuged to the periphery, as described by Gardner [2] and
drained along the cylinder wall, only a very small part reaching the blade
trailing edges where it is broken up into drops. On the stationary blades,
however, the collected water only drains under the influence of gravity,
which may be only one ten thousandth of the centrifugal field in the mov-
ing blades and, consequently, the greater part of the water reaches the
trailing edge to be torn into relatively coarse drops.
The collection of the drops by the blading in a turbine is difficult to
reproduce experimentally, since the acceleration and terminal velocity
of individual drop sizes vary, with the result that the angle at which the
drops approach the stationary blading is a function of their size. An
attempt, however, was made to estimate the collection effectiveness on a
FIG. 6—Breakup of water sheets into droplets.
cascade of LP turbine diaphragm blades in an open jet wet air tunnel,

although this could only be approximate because the injected water drops
were travelling in the same direction as the air flow.
If the injected water droplets exceeded 60 p. in diameter, the collection
efEectiveness of the concave blade surfaces approached 100 per cent over
blading Mach and Reynolds' number ranges of 0.15 to 0.5 and 2.5 x
106 to 8.3 X 105 (based on chord and efflux velocity), respectively.
To estimate the minimum size of droplet collected in the turbine, use
was made of the impact number, I, where:
a nondimensional expression derived from viscous drag assumptions in

FIG. 7—Effect of vacuum on weight loss.

the trajectory equations of a droplet in a gas [3]. Martlew [4] subse-

quently adopted this expression to correlate the collection of small par-
ticles in his work on gas turbine blade fouling. A similar approach was
justified in our own case because the potential flow regime around the
blading was essentially the same in both cascade and turbine; the Reyn-
olds' number of the smallest droplet deposited on the cascade was also
well within the viscous flow region, and the turbine droplet Reynolds'
numbers were, in general, expected to be even lower because of the low
steam density. Thus, from a cascade impact number of 3.5, the minimum
droplet size one would expect to collect on the last stage diaphragm
blades of a 500-Mw set was estimated to be 30 p..
Breakup of Water Film

The breakup of water films by air streams at different velocities has
been studied photographically by a colleague, D. H. McAllister, under
atmospheric discharge conditions. Care was taken in the experimental
work to establish the maximum stable droplet size by photographing the
spectrum at several distances downstream from three sources: a hypo-
dermic needle, a square-edged plate, and a cascade of LP turbine dia-
phragm blades (Fig. 6). It was impossible in these cases to correlate the
critical diameter in terms of a specified Weber number [W
a] based on the bulk flow velocity, since the velocity exponent appeared
to be considerably less than two.
The best correlation was found by plotting Weber number against
Mach number (MJ which gave the relationship W
the blading. This suggests a diameter of 800 /* as the maximum stable
droplet size leaving the trailing edge of the last stage diaphragm of a
500-Mw turbine, but further work is necessary at lower gas densities
before an accurate prediction can be made. The CEGB [5] have used
an introscope inserted behind the last moving blade row to determine the
maximum drop size in a 120-Mw turbine, and the largest drops seen
leaving the last stage diaphragm blades in this machine appear to be
about 400 /*.
Impact By Water Drops

The impact between the water drops and the moving blades is the last
stage in the physical process of erosion, and the most important, in that
it dictates the stress produced in the erosion shield materials which pro-
tect the moving blades' leading edges. The true impact velocity is less
than the theoretical value because the steam density and specimen shape
both tend to protect the surface from the blows of the drops. This has
been demonstrated in the contra-rotating erosion rig where the maximum
weight loss can be reduced by a factor of 4 (Fig. 7) if the pressure inside
the chamber is raised from 2 to 5 in. mercury absolute while the shaft
FIG. 8—Typical test results from tool steel.
speeds and water flows are kept constant, the mean droplet/diameter
being between 90 and 110 /^ under all conditions.
The theories used to assess the stress caused by impact, in most cases,
have been based on a water jet or a spherical droplet colliding normally
with an inelastic flat plate, Bowden and Brunton [6], The stress level
in the material is assumed to equal the pressure generated in the water,
which is given by the water hammer equation P = p C V. This yields a
maximum stress of 1.2 X 105 psi for an impact velocity of 1730 ft/sec.
Engel [7] and others have modified the water hammer equation to take
into account the elasticity of the target, and Kirkwood and Montroll [8]
have also made allowance for the thermodynamic properties of water at
high pressure levels, above 3000 psi. Impact stresses calculated by this
latter method are 60 per cent higher than those derived using the un-
modified water hammer equation for an impact velocity of 1730 ft/sec,
FIG. 9—Artificially roughened specimen surfaces (X25).
being 1.9 X 105 psi which is sufficient to initiate fatigue damage in cur-
rently used materials. None of these theories, however, explains the re-
duction in erosion rate, dM/dT, observed both in turbines and in the
test rig after prolonged running (Fig. 8). Initially, during the incubation
or primary zone of erosion in which damage is initiated at slip planes,
there is little or no weight loss [9]. During the following secondary zone,
however, the cracks propagate and link together, releasing small pieces
of surface material, causing the erosion rate to increase to a maximum.
Then follows a tertiary zone, where the surface appears to be uniformly
pitted; this is characterized by a reduction in the erosion rate toward a
steady terminal value. Moreover, it is this tertiary region which is im-
portant to designer and operator alike, rather than the initial and second-
ary zones, since it is in this region that the turbine erosion shields operate
for most of their lives.
In the tertiary region the maximum stress level over the metal surface
is thought to be reduced because some droplets hit peaks on the eroded
surface and are split, while others fall in craters containing water which
distributes the loading.
FIG. 10—Weight losses of stainless steel gauzed specimens.
The forces required to split a droplet are probably related to the in-
ternal pressure resulting from the surface tension at the boundary. A
100-/* droplet would have an internal pressure of Ys psi and is, therefore,
likely to split should it hit a point. Thereafter, the damage to the flanks
of the peak by the two water flows would probably be less than that
caused by normal impact, because experiments by Baker et al [10] sug-
gest that only the normal component of the impact velocity is responsible
for erosion.
The centrifugal acceleration at the specimen in the contra-rotating
(left) Weft horizontal. (right) Warp horizontal.
FIG. 11—Damage to stainless steel gauzed specimens after ¥2. hr test (X40).
erosion test rig is of the order of 40,000 g (near the last row blade tip in
an LP turbine it is 17,000 g). Since this is more than sufficient to centri-
fuge any deposited water from the surface, a regime in which the surface
depressions are water filled is difficult to imagine. Experimental evidence,
however, suggests that some water is retained, the amount probably
depending on the shape of the crater and on the rate of supply of water
droplets. To verify that water drainage from an eroded surface is an
important feature hi the tertiary erosion zone, experiments were con-
FIG. 12—Weight losses of grooved maraged steel specimens.
ducted on the contra-rotating erosion rig using specimens artificially

roughened in four different ways.
Experiments with Artificially Roughened Surfaces
In each of the four experiments on specimens with artificially
roughened surfaces, four specimens, two roughened and two unrough-
ened, were tested simultaneously. The standard speed and water flows of
1730 ft/sec and 1.5 gal/min were maintained, but a vacuum of 28 in.
Hg was used instead of 28.5 in. Hg to suit the working range of a vacuum
controller. In the first experiment stainless steel gauze with a warp and
weft of 28 X 140 wires per inch was brazed to the surfaces of two speci-
mens of similar composition (Fig. 9). The specimens were fitted so that in
one the weft was radial to the machine axis and in the other, the warp.
Both gauze specimens initially exhibited a lower rate of weight loss
than the plane ones, but after 3 hr the rates became virtually equal (Fig.
10), and only isolated fragments of gauze remained. Water drainage
through the gauze under the action of the centrifugal force was easier
when the warp was aligned radially to the machine axis, and this explains
the increase in erosion rate over that of the other gauzed specimen.
Photographs of the gauzes (Fig. 11) during the early stages of erosion
revealed fairly extensive damage to the flanks of the wires (Fig. 9), some
FIG. 13—Weight losses of etched tool steel specimens.
damage to the peaks, and virtually no damage to the exposed wires in the
lower section of the mesh. The base wires, moreover, remained un-
damaged until the surface wires had disintegrated, which suggests that
entrapped water was protecting them.
A 14.5Cr-5.75Ni-2Cu-1.6Mo-0.85Nb maraging steel was used for
the second experiment, the specimens having grooves machined in their
surfaces in the unhardened condition. This material was chosen so that
a hardness of 450 DPH could be attained by both the serrations and the
base material after aging. The peaks were of triangular section, the pitch
of the grooves being 250 /* (Fig. 9). One specimen had grooves horizontal,
that is, parallel to the axis of rotation, and the other vertical grooves, so
that drainage was minimized in the first and encouraged in the second.
After 10 hr the weight loss of the specimen with horizontal grooves
FIG. 14—Tool steel chevron specimens (Xll).

TABLE 1—Weight losses of typical materials relative to standard tool steel.

Final
Impacting Ratio of Refer-
Average Water (ap- Weight ence
Description Material Composition, weight % Hardness. proximate), Loss to
DPH kg/cm2 Standard Flg
f.-X"18
1«
Specimen, '
Rm
Tungsten high-
speed tool
steels Fe-18W-6Cr-0.7O 610 180 X 10' 1.2
630« all values0 LOG- 4
830 170 ONS
Fe-10.1W-4.3Cr-10Co- 920 70 0.28 11
3.8V-3.5MO-1.25C
Fe-9.4W-4Cr-4.9Co-2.8V- 890 90 0.43 10
1.25C
Chromium
steels.... .Fe-18.5Cr-9.4Ni-l.4Mn- 165 5 120
0.35Si-0.04C 290 5 78
Fe-18Cr-2Ni-14.8Mn- 250 5 6
0.57N-0.06C 380 5 2.6
Fe-14Cr-4W-4V-1.4C 800 120 0.46 9
Fe-12Cr-0.25V-0.8Mo-l .5C 510 100 1.41 1
620 100 0.94 5
760 100 0.46 8
Ni-Co maraging
steel Fe-18Ni-8Co-3Mo-0.2Ti- 530 20 3.5
0.1A1
Cobalt base. .. .Co-31Cr-14W-lC, cast 515 230 0.8 6
Co-30Cr-9W-1.2C, wrought 490 120 0.34 12
Co-26Cr-5W-lC, cast 380 100 1.1 2
cast 405 180 1.0 3
wrought 450 100 0.58 7
Co-31Cr-4.5W-l.lC, 380 60 1.2
wrought
Co-29Cr-4.5W-1.2Si-lC, 500 90 0.82
cast
Cemented car-
bide WC-6Co 1570 80 0.85
TiC-30Ni/Mo 1500 70 1.75
a
Standard tool steel comparator.
was significantly lower than that with vertical grooves (Fig. 12), both
being superior to the plane specimens. Moreover, the improvement was
maintained until the grooves were eroded away, after which the terminal
erosion rates of all four specimens equalized.
For the third experiment, specimens of a 12Cr-l.5C-0.8Mo-0.25V
tool steel were used, and serrations were photoetched onto the surface
after hardening to 685 DPH. The serrations were pitched at 300 //, instead
of at the 250 /* of the machined specimens, and they appeared more
irregular and shallower (Fig. 9). Again one specimen was tested with
the grooves vertical and the other with the grooves horizontal. These
FIG. 15—Effect of hardness on erosion of 12Cr-l.5C-0.8Mo-0.25V tool steel.

specimens exhibited the same trends in weight loss as the machined speci-
mens (Fig. 13), the horizontally grooved specimens losing least, the
vertically grooved specimen, more, and the two plane specimens, most.
Chevron patterns were photoetched onto the surface of 18W-6Cr-0.7C
tool steel for the fourth test after hardening to 830 DPH (Fig. 14). In the
FIG. 16—Comparison of materials in the secondary stage of erosion.
test one of these specimens had the chevrons pointing outward and the
other inward.
The rate of weight loss of the plane specimens was again higher than
that of the grooved ones, but the duration of the test was limited and
prevented the erosion being taken into the tertiary zone. Photographic
enlargements of the grooved specimens (Fig. 14) indicated that the sur-
face was eroded more severely along the center line of the specimen
when the chevrons were pointing radially inward and central drainage
was encouraged.
Evaluation of Erosion Resistance of Materials

Several important properties must be considered in selecting materials
for erosion shields, such as capability of being cast, forged, or rolled to
shape, erosion and corrosion resistance, ductility, and temper resistance
to permit attachment to turbine blades by brazing or other means. Only
erosion resistance lies within the scope of this paper.
A standard 18W-6Cr-0.7C tool steel shield material taken from one
cast and with a common heat treatment was used as a comparator in
each run to ensure consistent evaluation of the erosion resistance of any
new material.
A wide range of steels, cobalt alloys, and cemented carbides (Table 1),
also nickel and titanium alloys, have been tested at an impact velocity of
1730 ft/sec. Specimens with hard coatings produced by surface harden-
ing, plating, and spraying were also tested, but the coatings exfoliated
relatively quickly.
Gardner [11] showed in 1932 that hardness was an important criterion
of erosion resistance in steels. This was confirmed when the erosion resist-
ance of a 12Cr-l.5C-0.8Mo-0.25V steel at several hardnesses was plotted
(Fig. 15). When the erosion rate was plotted against the weight of im-
pacting water per unit area, the terminal rate showed a decrease with in-
creasing hardness. This concept of a hardness-erosion resistance rela-
tionship does not apply, however, when alloys strengthened by different
mechanisms are compared, since comparable terminal erosion rates are
approached by steels, cobalt alloys, and cemented carbides of greatly
differing hardnesses. In Fig. 16 the relative weight losses under standard
test conditions are plotted against hardness for several steels, cobalt,
nickel, and titanium alloys; and cemented carbides after impacting with
approximately 10 kg/cm2 of water. (This value was calculated from a
water flow to the jets of 1.5 gal/min, assuming that the trajectory of the
droplets was unaffected by the atmosphere hi the test chamber because
of its low density.) All these alloys are still in the secondary stage of
erosion (that is, they have increasing rates) where the greatest variation
in weight loss occurs. This variation is probably related to metallurgical
differences such as grain size, dispersion, and particle size of carbides and
other phases, and nonmetallic inclusions and, in some degree, to varia-
tions in the test conditions. For steels a relationship is shown between
hardness and erosion resistance, although there are substantial variations
between test results. The hardness erosion relationship is difficult to pre-
dict for the cobalt alloys because the test results lie on an almost vertical
line, and, in the case of the carbides, the relationship is masked by
scatter in the results.
The significance of variations in secondary and maximum erosion rates

cannot be assessed, however, without relating them to terminal erosion
rates. Twenty-six standard high-speed tool steel specimens (18W-6Cr-
0.7C) from one batch of material formed the largest group, heat treated
and tested under identical conditions. All the test results from these
FIG. 17—Computed mean weight losses and erosion rates for all standard 630
DPH, 18W-6Cr-0.7C tool steel comparator specimens.
specimens, expressed as weight losses for theoretical values of impacting

water at 1730 ft/sec, were statistically analyzed giving the following
relationship
The erosion rate was initially zero, increasing to a maximum after

FIG. 18—Weight loss trends for representative materials.
about 20 hr, and then decreasing toward a steady terminal rate (Fig. 17).
Extrapolation of the results, using Eq 5, showed that impact with about
2000-kg/cm2 water would be required to remove 0.1-in. metal (the
approximate thickness of turbine erosion shields). Tool steels, cobalt
base alloys, and cemented tungsten carbides, impacted with more than
50-kg/cm2 water, tend toward terminal erosion rates between 0.3 X
10~6 and 2.0 X 10~6 g/g. For these materials the terminal erosion rate
would account for most of the weight loss resulting from 0.1-in. penetra-
tion; the early peak in the rate curve is thus of comparatively little im-
portance. This is demonstrated by a 31Cr-14W-lC cobalt alloy which
was initially (after impact with 12-kg/cm2 water) approximately 40 times
FIG. 19—Comparison of volumetric losses for tungsten carbide tool steels and
a cobalt base alloy.
as erosion resistant as the standard tool steel but after impact with 100-
kg/cm2 water was less than one and a half times as resistant.
Erosion ratios, Rm , the ratios of the weight losses of the test material
to those of the standard high-speed tool steel tested under identical
conditions, are given in Table 1 for typical materials. When more than
40 kg/cm2 of water had impacted with the specimens, the erosion rates of
materials hardened by different mechanisms converged (Fig. 18). Most
results for cemented carbides expressed on a weight loss basis were
higher than standard tool steel, but when plotted on a volume basis, Rv ,
the curve for the most resistant material (tungsten carbide/cobalt
cemented carbide) more closely approached those for the most promising
steels and cobalt base alloys (Fig. 19).
In the early stages of the tertiary zone (that is, a penetration of 0.01 in.)
the most resistant materials, hardened W-Co tool steels, high carbon
12 to 14 Cr-Mo-V steels, wrought cobalt base alloys, and the tungsten

carbide/cobalt cemented carbide were about two to four times better
than the comparator specimens of 18W-6Cr-0.7C tool steel hardened to
630 DPH. These materials, however, were not tested long enough to
establish their terminal erosion rates with sufficient accuracy to claim
significant advantages in turbine shield life.
Conclusions
1. The collection effectiveness of the last stage diaphragm blading
of a 500-Mw LP turbine is likely to approach 100 per cent if the droplets
exceed 30 p, diameter.
2. The final water droplet sizes cannot be uniquely predicted from a
specific Weber number based on the bulk velocity of the steam. The
relationship We = 200 (MJ1-75 seems more applicable to turbine blading
and yields a maximum stable drop size of 800 /* before impact with the
last moving row of blades.
3. By present estimates impact forces are sufficient to initiate erosion
damage in the materials currently used for shielding blades. In all mate-
rials, however, the terminal erosion rate lies considerably below the
maximum. The decrease is thought to be caused by reduction in surface
stresses resulting partly from the breakup of droplets striking the peaks
on the eroded surface and partly from cushioning by water in the craters.
This theory was supported by the experimental erosion of artificially
roughened specimens.
4. Microstructures and mechanical properties of materials have been
difficult to correlate with erosion resistance because differences in the
physics of impact at successive stages of erosion were not fully appreci-
ated at the time.
5. Further testing will be necessary to establish the terminal erosion
rates of the more promising materials, such as hardened W-Co tool
steel, high carbon 14Cr-W-V steel, wrought cobalt based alloys, and
tungsten carbide/cobalt cemented carbides, before significant advantages
can be claimed in terms of turbine blade shield life.
A cknowledgments
The authors wish to acknowledge the assistance of all who have helped
in the preparation of this paper, in particular that of D. H. McAllister
and J. Robinson who carried out the experimental programs. Their
thanks are extended to the following manufacturers for their co-opera-
tion in the erosion program and for permission to publish compositions
and results of the materials: Associated Electrical Industries Ltd., Deloro
Stellite Ltd., Hall & Pickles Ltd., Jessop-Saville Ltd., Metro-Cutanit
Ltd., Samuel Osborn and Co. Ltd., The Brown-Firth Research Labora-
tones, and Wall Colmonoy Ltd. Finally the authors thank the directors
of C. A. Parsons and Co. Ltd. for permission to publish this paper.
References
[1] A. Evans and J. Robinson, "Erosion Experiments Related to Steam Turbine
Blading," Parsons' Journal, Vol 10, No. 61, Christmas, 1965, pp. 465-473.
[2] G. C. Gardner, "Events Leading to Erosion in the Steam Turbine," Proceed-
ings, Institute Mechanical Engrs., Vol 178, Part 1, No. 23, 1963-1964, pp.
593-623.
[3] G. I. Taylor, "Notes on Possible Equipment and Technique for Experiments
on Icing on Aircraft," R & M 2024 ARC Technical Report, Aeronautical
Research Council, pp. 1-2.
[4] D. L. Martlew, "The Distribution of Impacted Particles of Various Sizes on
the Blades of a Turbine Cascade," Aerodynamic Capture of Particles, E. G.
Richardson, editor, Pergamon Press, London, 1960, pp. 104-111.
[5] D. G. Christie and G. W. Hayward, "Observation of Events Leading to the
Formation of Water Droplets Which Cause Turbine Blade Erosion," Philo-
sophical Transactions, Royal Soc., Series A, No. 1110, Vol 260, pp. 183-192.
[6] F. P. Bowden and J. H. Brunton, 'The Deformation of Solids by Liquid Im-
pact at Supersonic Speeds," Proceedings, Royal Soc., Series A, No. 1315,
Vol 263, 1961, pp. 433-450.
[7] O. G. Engel, "Waterdrop Collisions with Solid Surfaces," Journal of Re-
search, Nat. Bureau of Standards, Vol 54, No. 5, 1955, pp. 281-298.
[8] J. G. Kirkwood and E. W. Montroll, "Progress Report on the Pressure Wave
Produced by an Underwater Explosion," II, OSRD No. 676, Office of Scientific
Research and Development, 1942.
[9] J. B. Marriott and G. Rowden, 'The Deformation of Steam Turbine Materials
by Liquid Impact," Philosophical Transactions, Royal Soc., Series A, No.
1110, Vol 260, pp. 144-150.
[10] D. W. C. Baker, K. H. Jolliffe, and D. Pearson "Resistance of Materials to
Impact Erosion Damage," Philosophical Transactions, Royal Soc., Series A,
No. 1110, Vol 260, pp. 193-203.
[11] F. W. Gardner, "The Erosion of Steam Turbine Blades," The Engineer,
February, 1932, pp. 146, 174, and 202.
DISCUSSION
Olive G. Engel1 (written discussion)—The reported finding that the

rate of erosion on a grooved surface is lower than that on a planar sur-
face is significant. It is related to the observation of Honegger2 and others
that the rate of erosion is reduced as the surface becomes roughened.
The explanation of this observation advanced by the authors is that the
maximum stress level over the metal surface is reduced because some
drops hit peaks and are split, while others fall in craters containing water
which distributes the loading.
I would like to suggest that the amount of surface that exists as sharp
peaks on which drops could split is quite small in comparison with the
total surface and that the observed reduction in erosion rate is more
likely due to the reduced angle of attack afforded by the slanted walls
of the surface corrugations. The effect of reduced angle of attack and
that of the cushioning afforded by liquid layers have been pointed out by
others, but the explanation of these beneficial effects appears to me to
need more clarification.
With regard to the angle of attack, the normal impact velocity is the
product of the observed impact velocity and the sine of the angle of
attack if this angle is measured from the impact surface. If one con-
siders only the damage that results from the impact pressure, and if one
ascribes this damage to the normal component of the impact velocity,
the effect of the angle of attack can be accounted for by this relationship
because the sine of the impact angle becomes smaller as the impact angle
is reduced. Actually, the flow of the drop liquid also contributes to the
damage produced by an impinging drop and must be considered.
As long as the surface is planar, the radial flow of the drop liquid is
able to shear pieces of the solid out of the surface as it bears against
surface protrusions and against the elevated edges of cracks that are
produced in the solid as a result of the impact blows. After the surface
has become covered with pits, the drops impinge into the pits. When a
drop strikes into a pit, the drop liquid shoots backward in the direction
from which the drop approached. Not only does this cut off the damaging
effect of the radial flow of the liquid of the drop that impinged but it
may also mitigate the pressure exerted by an additional drop that is just
about to collide with the same pit by striking against this drop and thereby
Chemical physicist, Space Power and Propulsion Section, General Electric
Co., Evendale, Ohio.
2
E. Honegger, "Erosion Experiments," Brown Boveri Review, Vol 14, 1927,
p. 95.
DISCUSSION ON STEAM TURBINE BLADE SHIELD MATERIALS 153
lowering the relative velocity at which it will hit the surface. The effect
of spray rebounding from an eroded surface in decreasing the relative
impact velocity of additional drops that are about to strike the surface (or
to be struck by it) was pointed out by Deal and Wahl.3
With regard to the cushioning effect of liquid layers, if the layer is
deep enough so that the drop does not strike bottom and if the impact
velocity is low enough so that the sound energy given to the liquid layer
is negligibly small,4 then essentially all of the impact energy of the drop
will be used in accelerating the liquid of the layer and the liquid of
the drop; this kinetic energy will later transform into potential energy
due to gravity and to potential energy of generated surface.5 At high
impact velocities the sound energy given to the target liquid may become
appreciable and, if this is the case, the above generalization will break
down; a sound pulse or pressure wave initiated in the liquid will be
transmitted to the solid. If the layer is so thin that the impinging drop
strikes bottom, then the drop will give part of its energy directly to the
solid, that is, an impact pressure will be exerted against the solid. The
part of the drop energy that is given to the solid will depend both on the
magnitude of the impact velocity and on the thickness of the liquid
layer; it will increase as the impact velocity increases and as the thick-
ness of the liquid layer decreases.
It has been postulated that if a drop should strike into a pit that is
already full of water, the impact pressure transmitted to the solid would
be substantially increased because of the somewhat conical shape of the
pit. Whatever pressure pulse is initiated in the trapped liquid as a result
of the impact will indeed be magnified if this liquid is contained in a
conical cavity. However, as pointed out above, at low impact velocities
the sound energy radiated into a liquid as a result of impact of a liquid
drop is small.4 Unfortunately, the expression on which this prediction
is based does not hold at high impact velocities, and the relative im-
portance of this effect at high impact velocities remains unknown.
Allen Smith, R. P. Kent, and R. L. Armstrong (authors)—Dr. Engel
has made a most interesting contribution on the impaction process of
droplets on eroded surfaces. Whether a droplet impinges on an erosion
peak or not, however, is dependent on the size of the droplet in relation
to the pitch of the peak. In the case of our erosion tester the Sauter mean
diameter of the droplets is 100 m/* and the mean pitch of the peaks on
an eroded tool steel surface 300 m//,. The chances of a droplet striking a
3
J. L. Beal and N. E. Wahl, "A Study of the Rain Erosion of Plastics and Other
Materials," Air Force Technical Report No. 6190, Wright Air Development Center,
Wright-Patterson Air Force Base, Ohio, April, 1951.
4
G. J. Franz, "Splashes as Sources of Sound in Liquids," Journal, Acoustical
Society of Am., Vol 31, 1959, p. 1080.
5
Olive G. Engel, "Crater Depth in Fluid Impacts," Journal of Applied Physics,
Vol 37, 1966, p. 1798.
peak are, therefore, more probable than Dr. Engel would suggest. That
oblique impacts on the sides of the peaks reduce the severity of erosion,
however, is not disputed, and it may be of interest to note that Gardner
(Ref 11 of the paper) shaped his blade erosion shields in the thirties to
take advantage of this principle. It should be mentioned, however, that
microsections of the damaged area show that the pits have steeply
sloping sides, and, in some cases, the peaks have considerable overhang
caused by undercutting.
W. K. Fentress® (written discussion)—We would like to congratulate
the authors on their excellent spray-cascade test results. We hope that
these tests will be continued and extended to the submicron range of
drop size by the use of turbine exhaust steam or possibly steam-air mix-
tures. From calculations in WANL-PR(DD)7 it appears that particles
formed in expansion of steam are initially about 0.4 p diameter and in
the downstream stages grow to about 0.9 /«, diameter. These calculations
are based on the condensation, nucleation, and growth of the moisture
particles in a typical low-pressure turbine. Drops of the same general
size are also implied by Gardner.8 It therefore appears that particles less
than l-fj. size account for the greater part of the collected moisture in
well-drained turbines, with adequate moisture extraction.
The following comments are with respect to droplet collection and are
intended to complement the authors' coverage of the topic.
The authors consider the collection of drops on the surface of stator
blades from the standpoint of inertial impaction. The conventional trajec-
tory equations, based on the laws of motion, are used. It appears that
the quantity q in the equations is a correction for slip flow to allow for
the fact that Stokes' law drag for continuum flow does not apply when
the flow about the miniature particles is in the slip-flow regime. In fact,
the slip-flow drag is only about one half the continuum drag at 0.3
Knudsen number,9 corresponding to about I-/* drop size for conditions
in typical turbines and only approaching continuum drag at about 20-//,
drop size.
By calculations in WANL-PR(DD),7 0.7 and 1.6 per cent of the
drops are collected on the surface of the stator blades for 0.4 and 0.9-/*
estimated drop size. As to the larger drops, all drops greater than roughly
18 fj. are collected. Generally, these calculations are by the equations re-
ferred to in Mr. Smith's report with correction for slip-flow and simplify-
ing assumptions.
a
Development engineer, Westinghouse Electric Corp., Lester, Pa.
7
"Basic Investigation of Turbine Erosion Phenomena," WANL-PR(DD), West-
inghouse Astronuclear Laboratory (Contract NAS 7-390), 1966.
8
G. C. Gardner, "Events Leading to Erosion in the Steam Turbine," Proceed-
ings, Institute Mechanical Engrs., Vol 178, Part 1, No. 23, 1963 and 1964.
9
H. W. Emmons, Fundamentals of Gas Dynamics, Princeton University Press,
Princeton, N. L, 1958, Section H.
DISCUSSION ON STEAM TURBINE BLADE SHIELD MATERIALS 1 55
Martlew's experimental work is cited by the authors in justification of

this approach. It is well to note however that Martlew's tests are not
highly relevant to the collection of submicron drops. These tests were
performed with molten wax particles in the range of 2 to 8 /t diameter,
while, as noted above, our calculations indicate that most of the turbine
drops are in the range of 0.4 to 0.9 p. diameter. While this difference
may not seem as important, note that in the range of 0.4 to 2-p. drop size
there is an approximate 2 to 1 change in the correction for slip flow,
corresponding to a near 2 to 1 difference in collection efficiency.
Messrs. Smith, Kent, and Armstrong—Stokes' drag relationship was
assumed in determining a minimum droplet size of 32 m//, for 100 per
cent collection by the blades, no correction being made for a "slip-flow
regime" where the diameter of the droplets approach the mean free path
of the vapor. Since the slip flow-drag is lower than the continuum drag,
one would expect the droplet size for 100 per cent collection to be below
32 m/t and Mr. Fentress' value of 18 m/u, is possibly more representative
than our own.
It is also interesting to note that in the absence of other forces, the
slip drag continues to fall with increase in Knudsen number (mean free
path/droplet diameter) which suggests a lower threshold diameter where
all droplets collide with the concave surfaces of the blades. This, there-
fore, may explain how blades become initially wetted by droplets al-
though it may also be argued that such small droplets would rebound
back into the bulk steam flow.
F. J. Heymann10 (written discussion)—The effect of surface roughness
on impingement erosion severity is something which has often been
speculated upon but seldom investigated experimentally in any controlled
manner. The authors are to be congratulated for carrying out very
carefully thought out and instructive experiments. The results do tend to
confirm the supposition that under favorable circumstances a water
film can remain in the surface "valleys" and thus help to protect the sur-
face. In my own contribution11 to this symposium I had speculated that
this supposition was inconsistent with the often observed fact that erosion
tends to progress by deepening already existing pits—which are precisely
the places which one would think to be protected if water was held there.
It would therefore be very interesting to find out, if possible, whether
erosion in the authors' artificially roughened specimens took place mainly
in the valleys or the peaks of the surface.
The lack of a distinct "hump" in the erosion rate curves for the
roughened specimens can be regarded as consistent with the statistical
model which I offered, for an uneven surface will result in a variety of
10
Senior engineer, Westinghouse Electric Corp., Lester, Pa.
u
See p. 70.
impact angles, and, since impact severity is strongly dependent on

impact angle (see Ref 10 of the paper), this will effectively result in a
greater scatter in "surface element lifetimes." The model predicts that
increasing this scatter (or the dispersion of the lifetime distribution func-
tion) can result in the damping out of fluctuations such as shown in Fig.
8 of the paper, and a more rapid approach to a uniform erosion rate.
(See Figs. 10a and c, and Figs. 12a and b of my paper. Admittedly, a
still further increase in scatter can lead to a heightened single peak,
Fig. 12d of my paper, if the distribution function is indeed log-normal as
hypothesized.)
However, the difference in the erosion rates which the authors found
for different orientations of the roughness patterns certainly cannot be
explained by the above model, if the mean impact direction of the drops
is indeed normal to the average surface. The truth of the matter is
probably a complicated interaction of many effects: thus, for instance,
the broader the valleys are with respect to drop size, the more likely is an
impact attenuation effect due to any retained water, but perhaps the less
likely is retention of the water.
These thoughts may suggest some further useful experiments but in
no way detract from the remarkable contribution which the authors have
already made.
Messrs. Smith, Kent, and Armstrong—Mr. Heymann postulates that
erosion progresses by the deepening of existing pits. This may be true in
the secondary zone of erosion but we have found no evidence of this in
the tertiary zone.
Photographic enlargements have shown that erosion on the gauzed
specimens was initiated on the sloping flanges of the wires with only
minor damage to the peaks and no damage to the exposed lower wires in
the gage. We had intended to pin-point the areas of erosion attack on the
machined, maraged steel specimens, but unfortunately the erosion rate
was unexpectedly high and completely roughened the surface before the
first inspection; the tool steel specimens were, of course, already rough-
ened in the craters by the etching and an assessment of the water damage
could not therefore be made. It has been established, however, on plain
tool steel specimens that the crater depths do increase in the secondary
zone, as suggested, but in the tertiary zone the mean pit depths remain
sensibly constant.
Microsections taken through representative specimens have also shown
a tendency for the peaks to be undercut at right angles to the direction
of impact, and this may result in some peaks becoming detached. The
primary source of weight loss, however, is thought to be caused by the
scouring action of the water on the sides of the peaks rather than the
detachment of the peaks.
The lack of a distinct hump in the rate of weight loss versus time
*Kr. 20—Comparison of the texture of erosion damage between turbine anderosion test ring (X3).
curves for the roughened specimens does seem to be consistent with the
statistical treatment proposed by Mr. Heymann, although, from a physi-
cal point of view, it is probably oversimplified. Mr. Heymann, however,
has made a significant contribution to the mathematical treatment of the
erosion problem, and it is our intention to investigate his model more
closely.
D. Pearson12 (written discussion)—The size range of drops observed in
wet-steam turbines (or predicted by the authors' colleague D. H. Mc-
Allister), is much greater than that obtained in the Parsons erosion
machine. Bearing in mind the effect of drop size on the texture of an
eroded surface (Ref JO of the paper), it is unlikely that the tertiary zone
would be reached during the erosion of a turbine blade, though it might
be the major part of the erosion of a specimen in the Parsons erosion
machine because of the small drop size produced in it.
Have the authors any evidence that their machine produces the same
scale of erosion damage, as that observed in typical steam turbines?
Messrs. Smith, Kent, and Armstrong—As yet we have been unable to
measure the droplet sizes responsible for erosion in our turbines. If,
however, it is postulated that the texture of the eroded surface is related
to the droplet size for any one material then eroded turbine shields and
test rig specimens indicate a marked similarity (Fig. 20) the mean pitch
of the peaks in each case being approximately 300 m/x. An estimate of
the minimum size of droplet causing erosion can also be deduced from
the depth of penetration of damage along the convex blade surfaces on
the assumption that the water film is broken up instantaneously into fixed
sizes at the trailing edges of the preceding nozzles. This would indicate a
minimum size of 70 to 100 m/x or rather less than the mean size of 100
mju. of our erosion tester.
If, by scale of erosion damage, Mr. Pearson means the rate of weight
loss in the tester compared to that in the turbine, then the tester is
many times more severe. This, in all probability, is caused by the larger
number of collisions in the tester compared with the turbine.
12
Central Electricity Generating Board, Marchwood Engineering Laboratory,
Marchwood, Southhampton, Hants, England.
J. M. Hobbs1
Experience With a 20-kc Cavitation

Erosion Test
REFERENCE: J. M. Hobbs, "Experience With a 20-kc Cavitation

Erosion Test," Erosion by Cavitation or Impingement, ASTM STP 408,
Am. Soc. Testing Mats., 1967, p. 159.
ABSTRACT: A modified 20-kc ultrasonic drill unit has been found to be

highly suitable for determining the relative rates of erosion of metals
exposed to Cavitation in water to an accuracy of about ±12 per cent.
With %-in.-diameter specimens at an amplitude of 0.002 in. (peak-to-
peak) in water at 20 C, all materials tested eroded at a fairly steady rate
after 5 mm3 had been removed, this rate being maintained until 10 mm3
were eroded. Twenty-four steels and four nonferrous metals were tested
under standard conditions, and an attempt has been made to correlate
their erosion resistance with the various forms of strain energy derived
from tension tests on the same materials. The best correlation was ob-
tained with the "ultimate resilience" (that is, the limiting strain energy
beyond the elastic limit), expressed as V2-(tensile strength)2/elastic modu-
lus.
KEY WORDS: cavitation, erosion, wear, vibration, metals, steels, physical
properties, mechanical properties, strain energy
For about thirty years the vibratory cavitation erosion test normally
used in laboratories has been based on the nickel tube transducer first
exploited by Gaines [I]2 in 1932. This was further developed by Hun-
saker [2] and was used extensively for the comparative testing of mate-
rials by Kerr [3], Rheingans [4], Beeching [5], and Nowotny [6], to
mention only a few early investigators. No standard test equipment or pro-
cedure was specified until 1956 [7], by which time many different devices
had been put into operation, as indicated in Table 1.
The nickel tube transducer, although cheap and easy to build, was not
without its shortcomings. In addition to the unpleasant noise emitted, the
efficiency was low owing to eddy-current losses and the open magnetic
loop. Moreover, to obtain satisfactory magnetostrictive properties it was
1
Senior scientific officer, Fluids Group, National Engineering Laboratory, East
Kilbride, Glasgow, Scotland.
2
paper.
159
TABLE 1—Examples of vibratory test equipment (a) during 1932-1956 and (b) from 1956 onwards.
Name or Organization Country Frequency, kc Amplitude, in. Specimen Type
(peak-to-peak)
(a) Gaines, 1932 U.S.A. 8.9 1.2° X 10-"

Hunsaker, 1935 U.S.A. 6.5 3.4 %-in. dia, flat
Kerr, 1937 U.S.A. 6.5 3.4 %-in. dia, flat
Beeching, 1942 Britain 9.0 %-in. dia, concave
Nowotny, 1942 Germany 9.0 3.6 ~%-in. dia, flat
Kornfeld, 1944 Russia 7.5 1.6 18-mm dia, flat
Escher Wyss Switzerland 9.0 ~2 10-mm dia, flat
Rheingans, 1950 U.S.A. 6.5 %-in. dia, flat
Glickman Russia 8.0 18-mm dia, flat
Pisarevskii, 1952 Russia 8.0 18-mm dia, flat
Wheeler, 1956 Britain 8.0 1.4 0.4-in. dia, grooved
(b) Dominion Engineering Canada 6.5 3.5 %-in. dia, flat
Pisarevskii and Eroshov Russia 20.0 1.2 18-mm dia, flat
Plesset U.S.A. 14.2 2.0 %-in. dia, rimmed
Budapest Technical University Hungary 6.5 3.3 %-in. dia, flat
National Engineering Laboratory Britain 20.0 2.0 £f$-in. dia, flat
Stone Manganese Marine Britain 20.0 2.0 %-in. dia, flat
New York Naval Shipyard U.S.A. 22.0 1.0 %-in. dia, rimmed
Hydronautics U.S.A. 14.0 to 16.5 2.5" %-in. dia, flat
Shell Research Britain 20.0 2.0
Monsanto U.S.A. 20.0 2.0 >£-in. dia, flat
0
Maximum value.
HOBBS ON 20-KC CAVITATION EROSION TEST 161
necessary to sacrifice strength, and hence fatigue failures of the tube were
common. The amplitude of vibration, which was the main factor con-
trolling the rate of erosion, was also difficult to measure and control.
In recent years, ultrasonic devices have been developed for machining
and drilling operations, and robust transducers suitable for cavitation
erosion testing are now commercially available. Some of these have al-
ready been tried in various parts of the world. In Britain, the National
Engineering Laboratory (NEL) and a number of other organizations are
proposing to adopt a certain type of 20-kc transducer [8]. In Russia,
FIG. 1—Apparatus used at NEL.
some tests also at 20 kc have been reported [9]. The United States has
already made use of similar vibrators operating at frequencies ranging
from 14.2 to 22 kc [10-12]. The latter part of Table 1 gives a number
of examples of known work using modern transducers, and the variety
of test conditions should be noted.
The time is now ripe for consideration of a new standard test device
for research purposes to take full advantage of these more efficient units.
Although it may subsequently be found that vibratory testing is not the
best method to use for cavitation erosion determination, it has proved
to be very convenient and—within its limitations—very valuable in the
past. Therefore, even though a superior type of test may eventually be
evolved, there is likely to be still a place for the vibratory test. In view of
the difficulty of relating results obtained under quite different conditions
1 62 ER6SION BY CAPTATION OR IMPINGEMENT
(especially of frequency and amplitude), it is most desirable to adopt

a common set of test conditions for work already in progress.
At NEL, the 20-kc vibrator has been subjected to an extensive series
of proving tests to determine its suitability for comparative erosion test-
ing. As a result of these tests, which were most encouraging, it has been
possible to specify tentative test conditions for the comparison of metals
hi water.
FIG. 2—Total erosion and rate of erosion versus time.
Using these conditions, the behavior of a number of common engineer-

ing materials has been investigated, and an attempt has been made to
correlate the results with the mechanical properties thought most likely
to affect the erosion resistance.
Magnetostriction Tests at NEL
Apparatus
The equipment consists of a water-cooled 20-kc transducer built up
of a stack of nickel laminations and driven by a 500-w ultrasonic gen-
FIG. 3—Erosion-time curves for different materials.
FIG. A—Rate of erosion-time curves for different materials.
HOBBS ON 20-KC CAVITATION EROSION TEST 1 65
erator, Fig. 1. To obtain the required amplitude of vibration, a double

cylindrical stepped velocity transformer is attached, and this hi turn
carries the specimen. Thus, the nickel laminations are not highly stressed,
and the amplitude at the specimen is limited only by the strength of the
material used for the velocity transformer.
Automatic control of the frequency to give resonant operation is ob-
tained by means of a feedback signal from a piezoelectric crystal strain
gage cemented to the side of the transducer. By measuring the voltage
developed by this strain gage, an indication of the amplitude is given.
This is calibrated directly by means of a micrometer microscope.
Flat-faced specimens % in. diameter and 1A in. thick have been
widely used in the past, and this size was chosen for the present investiga-
tions. The specimens are screwed into the lower end of the velocity
transformer, the combined length being designed to give a natural funda-
mental frequency of longitudinal oscillations of 20 kc. The test liquid
is contained in a glass beaker standing in a water bath to maintain con-
stant temperature during a test.
Erosion is assessed after successive exposures to cavitation by weight
loss, using an analytical balance capable of measuring to ±0.1 mg. The
rate of erosion is found to depend upon a number of variables, the effects
of which are discussed hi the following section.
Variables Affecting Test Results
Effect of Test Duration—In the tentative test conditions adopted by
NEL and other British workers [8], an amplitude of 0.002 in. (peak-to-
peak) was proposed, and all tests reported here were made using this
value unless otherwise stated. Results of successive exposures to cavita-
tion were obtained, and a typical graph of cumulative weight loss plotted
against total time of exposure is shown hi Fig. 2 (full line). A second
curve, obtained by graphical differentiation of the first, shows the rate
of erosion as a function of time. Four separate phases of the process can
be distinguished, namely:
1. an initial period, usually termed the "incubation period," during
which there is no measurable loss of material, although plastic deforma-
tion can occur and cracks may be formed in the specimen;
2. a transition period as erosion begins locally at first, but then in-
creases in rate as it gradually spreads across the whole effective test area;
3. a period, which may be termed the "constant rate" period, during
which the erosion rate is sustained at a near constant value; and
4. a period during which the erosion rate decreases gradually as pits
deepen and the cavitation collapse is attenuated by trapped air or water.
Results with six different materials are shown plotted hi the same
manner in Figs. 3 and 4. In general, the results were qualitatively simi-
lar, although there were marked differences hi both the rates of erosion
FIG. 5—Rate of volume erosion versus volume eroded for different materials,
in the constant-rate period and in the durations of this period. Moreover,

whereas the less resistant materials passed straight from the constant-rate
period into the final period of decreasing rate of erosion, the more
resistant Monel metal and stainless steel first showed a slight increase in
erosion rate before the final falloff. This slight increase was found to be
FIG. 6—Rate of erosion versus amplitude.
accompanied by a small increase hi the eroded area of the specimen.

The latter effect may, therefore, be the cause of the former.
Two general observations were true for all six materials tested,
namely, that the final appearance of the eroded surface was similar for
quite different materials and that the beginning of the decrease in erosion
rate was definitely associated with formation of a honeycombed texture.
Therefore, a new basis of comparison was tried, the volumetric rate of
erosion being plotted on a base of total volume removed (see Fig. 5).
This revealed that for all these materials the volumetric rate of erosion
under standard conditions was sensibly constant over the range 5 to 15
mm3. The rates of erosion of different materials should therefore be com-

pared within this range. It should be noted, however, that the erosion
rate-time behavior of materials under different frequency-amplitude
conditions will not necessarily be the same.
Effect of Surface Finish—It has been suggested [11] that the initial
surface condition of the specimen will affect the erosion rate. This was
found to be true only in the early stages of attack. Specimens whose
surfaces were either work hardened or highly polished suffered less
damage in a given time than others of the same material that had been
electropolished or were somewhat rougher, respectively. It therefore
might take different times to reach the constant-rate period, but the
rate of erosion in that period is independent of the original surface con-
dition. It was found expedient to minimize testing time by removing any
work-hardened layer and producing a consistently smooth but not highly
polished surface by rubbing down the as-turned test face on fine (Grade
600) emery paper.
Effect of Vibration Amplitude—It had always been found in vibratory
cavitation tests that the rate of erosion was extremely sensitive to ampli-
tude. This was investigated for four materials, and the results are shown
in Fig. 6, taking the rates of erosion in the constant-rate period.
It was observed that a certain finite movement was necessary to initi-
ate cavitation (and hence cause erosion). With soft metals, the damaged
area increased with amplitude until it covered about 80 per cent of the
area of the specimen. This variation in area was less pronounced in the
case of the more resistant metals. Over the range of amplitude covered
(0.001 to 0.003 in.), the erosion rate increased with amplitude to the
power 1.5.
For most purposes, the amplitude of 0.002 in. tentatively adopted
proved to be a satisfactory compromise. This did not overstress the
spigot of the specimen for most materials of interest, gave a fairly good
rate of erosion over about 75 per cent of the specimen area, and re-
quired a power input to the transducer of about 200 w.
Effect of Vibration Frequency—Although the frequency is virtually
fixed by the mechanical resonance of the system, it is possible for small
departures from the nominal 20 kc to occur, and the effect of these was
investigated for an amplitude of 0.002 in. It was found that the rate of
erosion tended to decrease slightly with increasing frequency. There are
a few contributory causes of this.
As the frequency increases at constant amplitude, the maximum
velocity and acceleration will have to increase. These determine the
pressures causing bubble growth and collapse, which will also increase.
Thus, if these were the only factors concerned, the erosion rate would
tend to increase. However, the time available for bubble growth is in-
versely proportional to the frequency, and the rate of growth is limited
HOBBS ON 20-KC CAPTATION EROSION TEST 1 69
by finite rates of heat transfer. Therefore, the maximum size that a

bubble can attain under a given pressure amplitude will be an inverse
function of the frequency. The energy released on collapse, and conse-
quently the damage produced, is a function of bubble volume. Of all
these effects, the last mentioned seems to be predominant and may ex-
plain the observed behavior.
Effect of Temperature—The effect of the mean temperature of the
cavitating liquid on the rate of erosion was investigated over a range of
temperatures from 10 to 60 C. The results were in agreement with those
obtained from previous tests by Leith [13] and others [14], who used
different frequency and amplitude conditions. They showed that the rate
of erosion in water was a maximum at a temperature of about 50 C,
and indicated the need for fairly close control of the temperature of the
test liquid.
The results of the tests just described are valid only if the tempera-
ture near the face of the specimen is the same as the measured tempera-
ture of the test liquid, and this seems to have been assumed without ques-
tion in the past. In order to verify this assumption, specimens were fitted
with thermocouples both just beneath the specimen surface and also
protruding slightly into the cavitation cloud. Other thermocouples indi-
cated the ambient temperature and the temperature at a point in the test
liquid (as the test liquid was vigorously agitated by the transducer, its
temperature was assumed to be uniform). All readings were recorded at
intervals of 3 min throughout a normal test run. At no time did the
difference between the temperature of the test liquid and the temperature
inside the cavitation cloud exceed 2 C.
Test Conditions
The following test conditions have been adopted as standard at NEL:
Frequency 20 ± 0.1 kc
Amplitude 0.002 ± 0.00005 in. (peak-to-peak)
Diameter of specimen 0.625 ± 0.001 in.
Thickness of specimen 0.250 ± 0.005 in.
Submergence of specimen 0.125 — 0.185 in.
Depth of test liquid 3 in.
Temperature of test liquid 25 ± 1 C
Standard test liquid distilled water
Although 25 C is now standard, the results in Table 2 were obtained

at 20 C.
"Rate of erosion" is defined as the mean loss of weight in milligrams
per hour hi the period between the removal of 5 and 10 mm3 from the
specimen. These volumes correspond to weight losses of about 40 and
80 mg, respectively, for most alloys likely to be encountered.
The specimens are coupled to the transducer by a %6~ m « UNF
screwed spigot % in. long. Silicone grease is applied to the machined

mating surface to minimize acoustic losses.
A 400-ml capacity beaker contains the test liquid, of which 300 ml
are used. Although distilled water is specified, in tests of mild steel no
detectable difference in the erosion rate is found with tap water.
The dissolved gas content of the test liquid is reduced to approximately
70 per cent saturation immediately prior to a test, either by degassing
under a vacuum or by means of a 30-min run using a dummy specimen.
This gas content is found to be the equilibrium value obtained by initially
saturated water after about 30 min of cavitation under the standard test
conditions.
Accuracy
The limits imposed on the test conditions were fixed such that errors
in setting should not cause changes in erosion rate greater than ±5 per
cent. It has been found, however, that the standard deviation of the
mean rate of erosion lies between ±10 and ±15 per cent. This larger
error must be attributed partly to variations in the material tested.
Relation Between Erosion Resistance and Other Properties
Mechanical Factors Affecting Cavitation Erosion
Over the years, several theories have been put forward to explain the
mechanism of cavitation erosion, and numerous attempts have been
made to investigate the possible correlation of erosion rate with certain
mechanical properties. Investigations of this type are still being made as
more materials are developed.
There is considerable evidence to show that hardness and tensile
strength generally give a good indication of erosion resistance, but many
exceptions to the general rule indicate that elasticity, ductility, and
fatigue also play a part. It seems unlikely that any single property can
control the resistance to erosion, and hence a combination of properties
must be considered. The actual properties used may depend upon the
type of material, that is, brittle or ductile, and mode of failure, for ex-
ample, tensile, shear, fatigue, and so forth. The mode of failure will be
affected by the strength of the material and the severity or "damage
potential" of the cavitation to which it is exposed.
Field measurements of cavitation damage have been made by Knapp
[15], using relatively weak ductile materials and pit-counting techniques
to estimate its intensity. More recently, Thiruvengadam [16] has made an
analysis of many examples of erosion and has deduced the various
intensities of cavitation3 responsible. The intensity was based upon
3
Here defined as the power absorbed per unit area of the damaged material
surface.
HOBBS ON 20-KC CAV1TATION EROSION TEST 171
estimations of the energy necessary to cause the material deformation

or volume removed. This has shown a wide range of cavitation intensities,
some of which are apparently much greater than those obtained in
laboratory tests.
Under normal comparative test conditions, the severity of cavitation
is sensibly constant or may be varied over a limited range only. On the
other hand, the strengths of the materials being evaluated may cover a
wide range. This basic difference from reality must be recognized when
analyzing the results of such tests.
The magnitude of stresses produced by cavity collapse will depend
upon a number of factors, including the cavity diameter, the excess
pressure causing collapse, the amounts of air and vapor in the cavity,
and the elastic properties of both the specimen and the liquid.
A material can absorb the energy released by a collapsing cavity in
three ways:
1. by elastic deformation,
2. by plastic deformation, and
3. by fracture.
The relative contributions of these will depend upon the limit of pro-
portionality between stress and strain, the tensile strength, the ductility,
and the elastic modulus of the material under consideration. Provided
that assumptions are made concerning the relative importance of Fac-
tors 1 to 3, various expressions estimating the ability of a material to
store or absorb energy can be derived. Three such expressions are proof
resilience, "ultimate resilience," and work done to cause fracture.
Proof Resilience
A material having a high proof stress and low modulus of elasticity
would be able to deform and absorb perhaps all the available energy
without exceeding the elastic limit. A measure of the ability of a material
to do this is its proof resilience, which is the greatest strain energy that
can be stored per unit volume of a material without permanent strain.
Proof resilience is given by the expression
^ (proof stress )2/elastic modulus (1)
As the energy is stored elastically it is released once the stress is re-
moved; but in certain cases after repeated stressing, fatigue failure may
occur. Hence, a criterion that there should be no damage at all is that
the stress caused by cavitation must be less than the fatigue endurance
limit of the material in the environment in which it is to be used.
Ultimate Resilience
If the elastic limit is exceeded, additional energy will be absorbed in
deforming the material plastically. If this overstraining is repeated
cyclically, certain materials which are capable of work hardening, for

example austenitic stainless steels, will develop successively new and
slightly higher yield points. During plastic deformation, the ratio of
stress to elastic or nearly elastic strain is almost equal to the original
elastic modulus. The strain energy required for each loading cycle once
the material has been overstrained will therefore be greater than that
needed for elastic strain.
For low rates of strain, the strain energy beyond the elastic limit may
tend to a limiting value (here termed the ultimate resilience) given by
^(tensile strength)2/elastic modulus (2)
Only part of this energy is released on unloading, and therefore it cannot
be described as true resilience. The balance of the strain energy is ab-
sorbed by the progressive plastic deformation of the material, which will
cause work hardening. This may be followed by fatigue failure once the
capacity for work hardening has been exhausted.
Work Done to Cause Fracture
The total energy required or the work done to cause failure of a
material is normally much greater than the energy stored elastically. In
the very simple case of tensile fracture, the area of the stress-strain
diagram is a measure of this work per unit volume. For ductile materials,
it may be derived approximately from the expression
[Yield load + %j(maximum load-yield load)] total extension. ..(3)
This approximation neglects elastic strain up to the yield point and takes
the remainder of the stress-strain curve as parabolic. For unit volume
[Proof stress + %(tensile strength — proof stress)] X elongation. .(4)
Effect of Strain Rate

Values of proof stress and tensile strength are determined from tension
tests conducted at low rates of strain, and the effect of strain rate should
be considered. Increased strain rate tends to give higher values of yield
stress (proof stress) and tensile strength, but lower values of elongation
at failure. Thus, there may be a tendency to underestimate the resilience
and the work done to cause fracture.
In applying the results of static tests to what is most definitely a
dynamic phenomenon, it would be very desirable to be able to apply
corrections for differences in behavior caused by strain rate. Unfortu-
nately, not all materials are equally sensitive to strain rate, but within
a class of materials the differences are not found to be large. In general,
low-melting-point alloys are the most sensitive to strain rate; hence,
brasses and bronzes will be very sensitive to strain rate, steels will be
HOBBS ON 20-KC CAPTATION EROSION TEST 173
moderately sensitive, and high-melting-point alloys such as Stellite will

be the least sensitive.
As the comparative performance of a limited range of materials is
being considered, effects of the different strain-rate sensitivity of a broad
range of materials will not necessarily be relevant. We can, therefore,
expect some meaningful correlation of erosion resistance with some
static physical attributes of materials of a given type, but not necessarily
with the same attributes of other dissimilar materials.
TABLE 2—Erosion data and mechanical properties of materials tested.
Condition
Mater- Young's
ials Material Specification or Alloy Type Modulus,
Refer- Alloy Designation psi
ence o
pt.
1... M.S.S. F.V.F.G. 13% Cr 0.2% C W T 30 X 106

2... M.S.S. B.S. 970 En 56C 13% Cr 0.2% C w T 30.6
3... M.S.S. F.V. 566 11.5/2.3/1.4 Cr/Ni/Mo W T 31.1
4... M.S.S. B.S. 970 En 57 17/2 Cr/Ni 0.25% C w T 30.6
5... M.S.S. B.S. 970 En 57 17/2 Cr/Ni 0.25% C C T 30.6
6... M.S.S. B.S. 970 En 57 17/2 Cr/Ni 0.25% C C H 30.6
7... A.S.S. B.S. 970 En 58B 18/9 Cr/Ni 0.08% C w A 32.4
8... A.S.S. B.S. 970 En 58J 18/9/3 Cr/Ni/Mo w A 29.1
0.07% C
9... M.S.S. B.S. 1630A" 13% Cr 0.15% C (max) C T 30.4
10... M.S.S. B.S. 1630A" 13% Cr 0.15% C (max) C H 30.4
11... M.S.S. B.S. 1630B" 13%Cr0.2%C(max) C T 30.4
12... M.S.S. B.S. 1630B" 13%Cr0.2%C(max) C H 30.4
13... A.S.S. B.S. 1631B Nb° 18/9/1 Cr/Ni/Nb C A 30.0
0.07% C
14... A.S.S. B.S. 1632D0 18/9/3 Cr/Ni/Mo C A 30.0
0.07% C
15... F.S.S. B.S. 1648B" 30% Cr 0.4% C C A 29.1
16... F.S.S. B.S. 1648O 30%Crl.7%C C A 29.1
17... P.H.S.S. F.V. 520B 15/5.5/1.5/1.5 Cr/Ni/ w T 31.1
Mo/Cu
18... P.H.S.S. Paralloy MPH 17/4/2.5 Cr/Ni/Cu C T 30.0
19... P.H.S.S. Paralloy MPH 17/4/2.5 Cr/Ni/Cu C H 30.0
20... P.H.S.S. Paralloy MPH-2 15/4/2/2.5 Cr/Ni/Mo/ C T 30.0
Cu
21... P.H.S.S. Paralloy MPH-2 15/4/2/2.5 Cr/Ni/Mo/ C T 30.0
Cu
22... P.H.S.S. Paralloy MPH-2 15/4/2/2.5 Cr/Ni/Mo/ C H 30.0
Cu
23. .. H.T.L.A.S. B S 970 En 24 1.5% Ni 0.5% Mn w H 30.4
0.35% C
24... H.T.L.A.S. B.S. 970 En 26 2.4% Ni 0.5% Mn w H 30.4
0.38% C
25... Gunmetal B.S. 1400 G-l 10%Sn2%Znl.5%Pb C A 12.6
26... Novoston Brit. Pat. No. 12% Mn 8% Al 3% Fe C 17.0
bronze 727021
27... Aluminum B.S. 1400 AB-1 10%A12.5%Fel%Ni C 19.3
bronze
28... Aluminum B.S. 1400 AB-2 10% Al 5% Fe 5% Ni C 15.9
bronze
TABLE 2—Concluded.
Erosion
Work Done Water Rate in
Tensile 0.1% to Cause at 20 C
Strength, Proof Fracture,
psi Stress, psi in.-lb/cu in. 3
mg/hr mmhr
/
105 X 103 63 X 103 33.5 30.1 X 10s 16.1 2.08
*
s
®
o\
c"
oo
1—»
N>
N>
Ul
(o
2... 114 73 26 27 253 90 215 26.2 9.1 1.17 51.3
3... 145 114 21 696 325 210 345 28.3 8.7 1.12 53.6
4... 159 125 17 36 321 256 414 25.0 7.6 0.98 61.2
5... 129 85 19 34 293 119 278 21.9 9.5 1.22 49.2
6... 182 155 1 2 457 392 545 1.8 2.9 0.38 158.0
7... 95 39 66 80 183 25 139 34.1 11.7 1.48 40.6
8... 90 38 59 113 228 22 123 42.8 11.3 1.43 42.0
9... 98 76 27 28 228 94 159 26.2 15.0 1.94 30.9
10... 168 119 4 0 398 233 466 6.1 5.6 0.72 83.3
11... 99 74 26 19 228 92 163 23.7 11.7 1.51 39.8
12... 176 153 1 0 439 383 511 1.7 3.3 0.43 139.5
13... 83 32 38 46 197 16 105 25.1 7.1 0.90 66.7
14... 81 34 48 54 176 18 103 31.5 9.5 1.20 50.0
is... 72 59 1 0 228 54 81 0.7 30.0 3.80 15.8
16... 110 75 1 2 353 87 186 1.0 4.2 0.53 113.3
17... 141 121 25 756 320 251 341 33.6 8.6 1.10 54.6
18... 141 112 19 19 321 208 327 25.0 9.6 1.23 48.8
19... 180 133 9 7 422 287 529 14.7 5.1 0.65 92.4
20... 153 142 21 35 343 327 383 31.2 8.6 1.10 54.6
21... 167 119 17 29 353 231 455 25.6 6.4 0.82 73.2
22... 196 154 12 11 476 390 632 21.8 4.8 0.61 98.5
23... 269* 86 86 583 1190 2.05 0.26 241.0
24... 269* 106 10* 575 1190 2.5 0.32 187.5
25... 42 18 22 94 13 69 7.5 34 4.0 15.0
26... 10P 436 30ft 306 183 54 307 24.4 11 1.47 40.8
27... 68 23 12.5 16 155 14 121 6.6 5.6 0.74 81.0
28... 95 43 6.5 10 211 56 284 5.0 4.2 0.55 109.0
NOTE—M.S.S. = martensitic stainless steel; A.S.S. = austenitic stainless steel;
F.S.S. = ferritic stainless steel; P.H.S.S. = precipitation-hardening stainless
steel; H.T.L.A.S. = high-tensile low-alloy steel; w = wrought; c = cast; T =
tempered; H = hardened; A = annealed.
0
Now merged with other relevant British Standards in B.S. 3100: 1957.
6
Typical values.
Test Results
The rates of erosion in the constant-rate period of the materials tested
are given in Table 2, together with details of each alloy type and its
mechanical properties. Values of proof stress, tensile strength, modulus
of elasticity, and elongation were obtained from tension tests. Bars from
the same melt were used for both the erosion and the tension specimens.
In analyzing the results of the erosion tests, the relations between the
erosion resistance and each of the following factors were examined in
turn:
1. work done to cause fracture (Eq 4),
2. proof resilience (Eq 1), and
3. ultimate resilience (Eq 2).
HOBBS ON 20-KC CAPTATION EROSION TEST 175
These quantities were all derived from the tension test data, using the
equations mentioned. Figures 7, 8, and 9 show each plotted against
"erosion resistance," which is defined as the reciprocal of the volumetric
erosion rate.
FIG. 7—Erosion resistance versus work done to cause fracture.
The linear correlation coefficient r for each of the three sets of data
was calculated from the equation
Equations for the regression lines were also derived for each figure,
using the method of least squares. Values of the correlation coefficients
and equations for the regression lines are as follows:
Figure Correlation Coefficient Regression Line
7 -0.54 y = 96.9 - 1.587*

8 0.60 y = 39.9 + 0.1646*
9° 0.65 y = 24.8 + 0.1403jc
Omitting points 23 and 24, as they were not in other figures.
Thiruvengadam and Waring [17] found a positive correlation between

the reciprocal rate of volume loss and the work done to cause fracture
(which he called "strain energy"). For the range of materials examined
in the present work, however, which were mostly stainless steels, there
seems to be a negative correlation with work done to cause fracture (see
Fig. 7). In some respects, the results indicate that the materials that re-
quire the least work to cause fracture (Materials 6, 12, and 16) generally
FIG. 8—Erosion resistance versus proof resilience.
have very high erosion resistance. These materials have practically no

ductility (1 per cent elongation) and low impact strength, however, and
therefore would not be considered in practice. Even neglecting these
materials, it is impossible to deduce a significant trend.
The correlation of erosion resistance with proof resilience (Fig. 8)
appears more promising and indicates a general tendency for the two
to be related. Proof resilience is a rather conservative property and is
strictly applicable only to materials loaded for the first time and within
the elastic limit. A more realistic approach, which permits some plastic
deformation under repeated loading, is considered to be the use of the
FIG. 9—Erosion resistance versus ultimate resilience.
ultimate resilience (Fig. 9). This gives the best general correlation ex-
cepting the three materials (6, 12, and 16) with 1 per cent elongation and
the two aluminum bronzes (Materials 27 and 28), all of which have an
inexplicably high erosion resistance. There are also some materials of low
ultimate resilience which possess an erosion resistance above average.
They comprise a group of austenitic steels (Materials 7, 8, 13, and 14)
which have exceptionally good corrosion-resisting properties and a great
capacity for work hardening. In particular, they are less susceptible to
corrosion cracking than the martensitic steels.
The excellent resistance of the two aluminium bronzes (Materials 27
and 28) may also be associated with their good corrosion resistance and
the possibility that then" mechanical properties are less affected by a
corrosive environment than those of the steels. There is a need, however,
to extend the work to include further nonferrous metals and to examine
the correlation of then* erosion resistance with the three factors consid-
ered here. This work has in fact been started at NEL.
In this discussion, consideration has been given to certain measured
mechanical properties. To show in more detail the mechanism of the
failure of the different alloys it would be necessary to examine the
microstructure of the specimens and to identify the metallographic proc-
esses leading to failure.
Conclusions
1. A magnetostrictive transducer of the type used for ultrasonic drills,
operating at 20 kc with an amplitude of 0.002 in., provides an effective
means of conducting vibratory cavitation erosion tests. The reproduci-
bility of this test is at present about ±12 per cent.
2. Under these conditions, the rate of erosion after 5 mm3 has been
eroded until 10 mm3 has been eroded remains fairly steady with all
materials examined so far. The mean rate of erosion over this range and
its reciprocal, the erosion resistance, provide indications of the ability
of a material to resist cavitation erosion.
3. Twenty-four steels and four nonferrous alloys have been tested
in this way, and attempts have been made to correlate their erosion re-
sistance with the various forms of strain energy that can be derived from
tension tests on the same materials. Fair correlation was obtained with the
"ultimate resilience," expressed as l/2 (tensile strength)2/elastic modulus.
A cknowledgments
This paper is published by permission of the Director of the National
Engineering Laboratory, of the Ministry of Technology. It is Crown
copyright and is reproduced by permission of the Controller of Her
Brittanic Majesty's Stationery Office.
References
[1] N. Gaines, "A Magnetostriction Oscillator Producing Intense Audible Sounds
and Some Effects Obtained," Physics, Vol 3, 1932, pp. 209-229.
[2] J. C. Hunsaker, "Cavitation Research," Mechanical Engineering, Vol 57, 1935,
pp. 211-216.
[3] S. L. Kerr, "Determination of the Relative Resistance to Cavitation Erosion by
the Vibratory Method," Transactions, Am. Soc. Mechanical Engrs., Vol 59,
1937, pp. 373-397.
[4] W. J. Rheingans, "Accelerated Cavitation Research," Transactions, Am. Soc.
Mechanical Engrs., Vol 72, 1950, pp. 705-724.
[5] R. Seeching, "Resistance to Cavitation Erosion," Transactions, Instn. Engrs.,
Shipbld., Scotland, Vol 85, 1942, pp. 210-276.
[6] H. Nowotny, Werkstoffzerstorung durch Kavitation (Material Damage by
Cavitation), VDI-Verlag, Berlin, Germany, 1942.
[7] L. E. Robinson, B. A. Holmes, and W. C. Leith, "Progress Report on Standardi-
zation of the Vibratory Cavitation Test," Transactions, Am. Soc. Mechanical
Engrs., Vol 80, 1958, pp. 103-107.
[8] J. M. Hobbs, "Vibratory Cavitation Erosion Testing: Proceedings of a Meet-
ing on 13th June, 1963," NEL Report 149, National Engineering Laboratory,
1964.
[9] M. M. Pisarevskii and E. F. Erashov, "Determination of the Resistance of
Materials to Cavitation by Means of Magnetostriction Vibrators" (in Rus-
sian), Energomashinostroenie, Vol 9, 1957, pp. 38-39. (Department of Scien-
tific and Industrial Research Translation No. 492.)
[10] M. S. Plesset, "On Cathodic Protection in Cavitation Damage," Journal of
Basic Engineering, Transactions, Vol 82, Series D, 1960, pp. 808-820.
Damage Resistance," Journal of Ship Research, Soc. Naval Architects and
Marine Engrs., Vol 8, 1964, pp. 39-56.
[12] A. E. Hohman and W. L. Kennedy, "Corrosion and Materials Selection Prob-
lems on Hydrofoil Craft," Materials Protection, September, 1963, pp. 56-68.
[13] W. C. Leith and A. Lloyd Thompson, "Some Corrosion Effects in Accelerated
Cavitation Damage," Journal of Basic Engineering, Transactions, Am. Soc.
Mechanical Engrs., Vol 82, Series D, 1960, pp. 795-807.
[14] R. E. Devine and M. S. Plesset, "Temperature Effect in Cavitation Damage,"
Report 85-27, Division of Engineering, California Institute of Technology,
1964.
[15] R. T. Knapp, "Accelerated Field Tests of Cavitation Intensity," Transactions,
Am. Soc. Mechanical Engrs., Vol 80, 1958, pp. 91-102.
[16] A. Thiruvengadam, "Intensity of Cavitation Damage Encountered in Field
Installation," Technical Report 233-7, Hydronautics Inc.
[17] A. Thiruvengadam and S. Waring, "Mechanical Properties of Metals and
Their Cavitation Damage Resistance," Technical Report 233-5, Hydronautics
Inc.
DISCUSSION
W. C. Leith1 (written discussion)—-This paper is an excellent sum-

mary of the present status of cavitation erosion testing, and it proposes
an international standard test procedure which includes the usual
ASME test button together with an industrial-type, water-cooled, nickel-
stack transducer operating at a frequency of 20,000 cps and an amplitude
of 0.002 in.
It is interesting to note that a specific rate of loss occurs in the period
between the removal of 5 to 10 mm3 from the specimen.
The availability of the 20,000-cps transducer from industrial manu-
facturers such as Mullard in England and Branson in the United States
makes it economical for small research projects in cavitation erosion
testing.
''R, Garcia2 and F. G. Hammitt2 (written discussion)—The author is
to be congratulated for an excellent and comprehensive investigation of
the cavitation resistance of a very broad range of ferrous and nonferrous
metals in water, including correlations of the experimentally determined
cavitation data with carefully measured mechanical properties. We believe
that he has contributed significantly to the improved understanding of
the damage phenomenon in a vibratory device and its relationship to me-
chanical properties of the test materials.
The test conditions we use in our own laboratory for water investiga-
tions are practically identical to those employed by the author with re-
spect to frequency, test specimen diameter, amplitude (measured in our
case with an accelerometer), and water temperature (tap water was used
in our experiments). Our submergence and depth of test fluid are some-
what greater than those used by the author. His comments regarding
standardization of test conditions for this type of cavitation investigation
are well taken, and we agree fully in this respect.
The curves of cumulative weight loss versus test duration and rate of
weight loss versus test duration obtained by the author are similar to
those we have obtained in our investigations in water, which involved
stainless steels, carbon steel, aluminum, brass, copper, nickel, and various
refractory alloys. While some of this is included in our own paper in this
symposium,3 it was previously reported in greater detail.4 We found in
1
H. G. Acres & Co. Ltd., Consulting Engineers, Niagara Falls, Canada.
2
Nuclear Engineering Department, The University of Michigan, Ann Arbor,
Mich.
3
See p. 239.
4
R. Garcia, R. E. Nystrom, and F. G. Hammitt, "Ultrasonic-Induced Cavitation
Studies in Mercury and Water," ORA Technical Report 05031-3-T, Department of
Nuclear Engineering, The University of Michigan, December, 1965.
DISCUSSION ON 20-KC CAVITATION EROSION TEST 181
our experiments that a constant maximum plateau rate of weight loss

existed in the range from 1 to 3 mils mean depth of penetration (MDP).5
This corresponds very closely to the author's results wherein the constant
plateau rate of weight loss was noted in the range from about 5 to 15 mm3
volume loss. Thereafter, the rate began to decrease due to the roughened
surface, characterized by deep pitting, as was also the case in our tests.
For high-density liquid metals such as lead bismuth and mercury (as also
reported in our paper in this symposium3), we found that the damage rate
was constant from about 1 mil MDP out to approximately 50 or 60 mils
MDP. In this case, the damage pattern was uniform throughout the test,
and material was evenly removed from the surface of the specimen. We
strongly agree with the author that this constant plateau rate of volume
loss for each material can best be used to rate the cavitation resistance of
the various materials. Generally, the author found the constant rate of
weight loss for many types of steel to be about 10 mg/hr, whereas in our
investigations we reported a value of approximately 3 mg/hr for Types
304 and 316 stainless steel. The values are not directly comparable due
to differences in the test materials, but the range of values seems to indi-
cate the same order of magnitude for the degree of cavitation intensity
in both facilities.
Our least-mean-square-fit computer correlations of our water cavitation
data4 with applicable mechanical properties show that tensile strength
and hardness are the most successful single correlating parameters. In
correlating our complete set of data obtained in water, mercury, lead
bismuth, and lithium with applicable mechanical and fluid properties, we
have investigated, prompted by the author's papers, elastic energy proper-
ties, defined by the author, of proof resilience and ultimate resilience.
Previously we had investigated the author's "work done to cause frac-
ture," that is, "engineering strain energy to failure" or energy under the
engineering stress-strain curve. In each case our complete body of data
was correlated in terms of nine mechanical properties plus the fluid
density. The nine mechanical properties were tensile strength, yield
strength, energy parameter (one of the three mentioned in the foregoing),
hardness, elongation, reduction in area, true strain energy considering
plastic elongation,6 true strain energy considering plastic reduction in
area,7 and elastic modulus. The correlation efforts which included the
author's proof resilience and work done to cause fracture as possible
correlating parameters were unsuccessful in that these properties did not
5
The mean depth of penetration (MDP) is computed assuming the weight loss is
smeared uniformly over the cavitated specimen surface.
6
A. Thiruvengadam and H. S. Preiser, "On Testing Materials for Cavitation
Damage Resistance," Technical Report 233-3, Hydronautics, Inc., December, 1963.
7
C. A. Harrison, M. J. Robinson, C. A. Siebert, F. G. Hammitt, and J. Lawrence,
"Complete Mechanical Properties Specifications for Materials as Used in Venturi
Cavitation Damage Tests," ORA Internal Report 03424-29-1, Department of
Nuclear Engineering, The University of Michigan, August, 1965.
appear in the predicting equations generated by the program. The cor-

relation of our data with the author's ultimate resilience was more suc-
cessful. A typical predicting equation applicable to our complete data
set in this case is as follows:
Average MDP rate = 0.510 + 6.045(UK)-1'* - 0.47l(p)-2
Coefficient of determination = 0.658
Average absolute per cent deviation = 54.5 per cent
where:
UR = ultimate resilience and
p = fluid density.
We have also correlated the author's data with his mechanical proper-
ties using our least-mean-squares regression analysis. The initial effort
consisted of correlating the experimentally determined cavitation data
with each of the mechanical properties listed in Table 2 of the author's
paper on an individual basis. This would indicate those properties best
able to predict cavitation resistance singly. It was found that the ultimate
resilience was most successful in this regard and resulted in a predicting
equation of the form:
Average MDP rate = 6.41 +9.21 X lOe(UR)~s

Average absolute per cent deviation = 25.8 per cent.
The author also found the ultimate resilience to be the most successful
correlating parameter in his studies. The other eight properties considered
in our correlation were somewhat less successful in this respect in the
following order: tensile strength (second best), hardness, yield strength,
proof resilience, elastic modulus, work done to cause fracture, impact
strength, and elongation (least successful). It is interesting to note,
particularly, the failure of the impact strength to predict cavitation
damage, in that this property might be expected to be useful in reflecting
the effect of the high transient loading rates which characterize cavita-
tion attack.
We have also correlated all of the author's data with the full set of
nine mechanical properties allowed to enter the predicting equation.
Under these circumstances, the statistically best predicting equation ob-
tained is as follows:
DISCUSSION ON 20-KC CAVITATION EROSION TEST 183
Average MDP rate = -94.49 + 1.39 X 105(UR)~2 - 6.70 X 10-5(#)2

+ 2.64 X lO'CrS)-1 + l.69(TS)lis + 0.27 (7S)1/2
- 3.79 X IW(PR)-2
where:
UR = ultimate resilience,
H = hardness,
TS = tensile strength,
IS = impact strength, and
PR = proof resilience.
It is seen that the cavitation damage is an inverse function of the ulti-
mate resilience, hardness, and tensile strength, and proportional to the
impact strength and proof resilience. Although this equation represents
all of the author's data very well, the large number of terms present
limits its utility. Other predicting equations (less representative statisti-
cally) indicate that the cavitation resistance is an inverse function of the
work done to cause fracture. The author also reported this somewhat
surprising finding in his correlations.
The fact as reported by the author and also noted by Young and
Johnston8 and Leith9 that many high-strength brittle materials show very
high cavitation resistance, if compared on the basis of energy under the
stress-strain curve as a measure of cavitation resistance, is interesting.
This tends to confirm our conclusion that both strength and energy
properties must be considered in order to arrive at a meaningful correla-
tion of cavitation damage with mechanical properties. A hypothetical
situation may be imagined where the material is so strong that no bubbles
will cause stresses even above the endurance limit. Then, regardless of
strain energies, no damage would occur.
Also, when several fluids are involved, fluid-material coupling param-
eters must be used. This is emphasized by the fact, as reported in our
paper in this symposium,3 that the damage ratio between mercury and
water in an identical vibratory test ranged between about 3 and 20.
Similar results are reported in another paper8 of the symposium in com-
paring sodium and mercury tests.
In further confirmation of the very high cavitation resistance of strong
8
See p. 186.
*W. C. Leith, "Discussion of Intensity of Cavitation Damage Encountered in
Field Installations," by A. Thiruvengadam, Cavitation in Fluid Machinery, Am.
Soc. Mechanical Engrs., 1965, pp. 32-46.
brittle materials as compared on the basis of energy under the stress-

strain curve, we have recently completed a series of tests in mercury
on tool steels and other materials. It was found at 500 F that the tool
steels were generally about ten times more resistant than the tantalum-
base alloy T-l 11, although the energy under the stress-strain curve of the
latter was about five times that of the tool steels. Hence, the discrepancy
in this case is by a factor of about 50.
Further, in Fig. 8 of our paper in this symposium,3 it will be noted
that Type 316 stainless steel is about Vis less resistant to cavitation erosion
than is T-l 11, but the stainless has a higher energy under the stress-
strain curve by about 20 per cent (Table 10 of the paper).
Allen Smith10 (written discussion)—The relatively large widths of the
plateaus of the erosion rate-versus-time curves for Monel and steel speci-
mens are of interest. Can these be explained by uneven wear on these
specimens so that the maximum erosion rate is not obtained simultane-
ously across the full specimen face?
/. M. Hobbs (author)—The author is grateful for Mr. Leith's remarks
and his support for the proposed test conditions. As he states the com-
mercial availability of suitable transducers is a big help, and as time
goes by other manufacturers are entering this market giving the research
worker some choice.
The enthusiastic support of Dr. Garcia and Professor Hammitt for the
views expressed in the paper and their analysis of test data is very
heartening. However, although they obtain similarly shaped curves of
erosion rate versus time with the plateau of maximum erosion extending
over the same mean depth of penetration range, there seems to be a sig-
nificant difference between the damage rates in the two test devices.
This may be accounted for by the 30 per cent greater area of the NEL
specimens and the probability that because of the greater quantity of
cavitation with this larger specimen the equilibrium level of dissolved gas
content will be lower, leading to a more energetic bubble collapse and
higher rates of attack.
The computer correlations of erosion rate with the mechanical proper-
ties are most interesting, and they confirm the conclusion that the ulti-
mate resilience is a significant factor in determining erosion resistance.
It is a little disquieting that the author's and the discussers' data give
different predicting equations, but this may be a consequence of the
statistically insufficient sample provided by the author's results. As
further test data become available, these correlations can be verified.
The lack of correlation with the impact strength at high rates of strain
on the one hand and with work done to failure at low rates of strain on
the other hand seems to indicate that the erosion rate is determined pri-
10
C. A. Parsons & Co. Ltd., Heaton Works, Newcastle upon Tyne, England.
DISCUSSION ON 20-KC CAVITATION EROSION TEST 1 85
manly by strength considerations, energy being of secondary importance.

It also seems to suggest that properties obtained at high rates of strain
may not correlate with erosion rate any better than properties obtained
in a normal tension test.
The wear on the specimens might be uneven as Mr. Smith suggests,
but this could also be true of metals other than Monel and steel. Our
observations indicate that the weaker materials are the most prone to
uneven pitting in the early stages of attack, whereas the surface texture
of strong materials is more uniform. If the wear is uneven and the maxi-
mum rate is not obtained simultaneously across the entire specimen
face, it is a truly remarkable coincidence that the peaks average out to
give an apparently constant rate for all the materials that have been tested.
Nevertheless, there is scope for further and more detailed investigation.
S. G. Young1 and J. R. Johnston1
Accelerated Cavitation Damage of

Steels and Superalloys in Sodium and
Mercury
REFERENCE: S. G. Young and J. R. Johnston, "Accelerated Cavitation

Damage of Steels and Superalloys in Sodium and Mercury," Erosion by
Cavitation or Impingement, ASTM STP 408, Am. Soc. Testing Mats.,
1967, p. 186.
ABSTRACT: An investigation was conducted to study cavitation damage
in liquid-metal environments of materials under consideration for com-
ponents of liquid-metal power conversion systems. The materials in-
vestigated included AISI Types 316 and 318 stainless steels, Sicromo 9M,
Inconel 600, A-286, Hastelloy X, L-605, Rene 41, and Stellite 6B. A
magnetostrictive apparatus was used to perform accelerated cavitation
damage tests with liquid sodium at 800 F and mercury at 300 F. Cavita-
tion damage determined by volume loss and surface roughness measure-
ments was used to rank the various materials and to compare the effects
of the different fluids on the degree of damage sustained. Metallographic
studies were made to determine the nature of the early stages of cavita-
tion damage. The materials tested in both sodium and mercury ranked
in the same order of resistance to cavitation damage, but the degree of
damage to all materials was consistently greater when tested in mercury.
The most resistant material was Stellite 6B; the least resistant material
was annealed Sicromo 9M. Surface roughness measurements provided
the same ranking of materials as that provided by conventional volume
loss measurements. Visual observation of sodium pump impeller blades
of three materials operated under cavitating conditions for 250 hr indi-
cated the same ranking of the materials with respect to resistance to
cavitation damage that was determined from the accelerated cavitation
tests.
KEY WORDS: cavitation, liquid metals, ultrasonics, magnetostriction,
sodium, mercury, high-temperature steels, nickel alloys, cobalt alloys,
space power systems, corrosion
In advanced space power conversion systems that use liquid metals

for the heat-transfer medium, cavitation damage to both rotating and
stationary components of the system may be expected. Damage can occur
1
Research engineer, Materials and Structures Div., Lewis Research Center,
National Aeronautics and Space Administration, Cleveland, Ohio.
186
YOUNG AND JOHNSTON ON ACCELERATED CAVITATION DAMAGE 1 87
in pumps and in stationary sections of the system where local pressure

fluctuations occur in the fluid. Components such as pump impellers or
venturi sections, tested for relatively short times in liquid-metal loops,
have shown damage in the form of pitting and surface erosion [1—3].2
Damage has also been observed in the turbine component of liquid-metal
systems when the vapor quality is less than 100 per cent [4]. It has been
reported that materials which resist cavitation attack are also resistant
to liquid impingement attack [5]. Because proposed space power systems
must function for 10,000 hr or longer, it is important to establish the
resistance to cavitation damage of materials that may be used in various
components of such systems.
Extensive research has been conducted to study the mechanism of
cavitation and cavitation damage of materials in water [6-75]. In moving
fluids where local pressures fall below the vapor pressure of the fluid,
cavities form. When these cavities are swept into regions of higher pres-
sure, they collapse with high velocity. If the collapse is on a metal sur-
face, localized high pressures will be transmitted to the metal, resulting in
severe damage. Although much of this damage is of a mechanical nature,
corrosion can be a contributing factor [10,15]. Conventionally used
engineering properties such as hardness, tensile and yield strengths,
fatigue limit, and corrosion resistance have all been considered as means
of ranking materials with respect to resistance to cavitation damage.
None of these properties individually provides a satisfactory criterion for
rating materials; however, recent work indicates that strain energy may
correlate with the intensity of cavitation damage for a number of mate-
rial [16].
In order to study a great number of materials in relatively short times,
various accelerated test methods for producing cavitation damage have
been devised. These include the rotating-disk method [13], venturi sys-
tems [17], and magnetostrictive systems [75]. All these methods have
been adapted for use in liquid-metal environments [14,18,19]. Of these
methods, the magnetostrictive apparatus was selected for this investiga-
tion because it afforded a relatively low-cost and versatile means of
producing highly accelerated cavitation damage. Materials currently
under consideration for use in components of liquid-metal space power
conversion systems were studied in both sodium and mercury. The pur-
poses of the investigation were to rank materials according to their resist-
ance to cavitation damage, to determine the nature of the early stages of
cavitation damage by metallographic studies, and to correlate accelerated
cavitation test data with cavitation damage sustained in actual pump
operation in liquid-metal loops.
2
paper.
1 88 EROSION BY CAVITAT.ON OR IMPINGEMENT
Materials, Apparatus, and Procedure
Materials
The materials tested for resistance to cavitation damage were the iron-
base alloys Sicromo 9M, A-286, and AISI Types 316 and 318 stainless
steels; nickel-base alloys, Inconel 600, Hastelloy X, and Rene 41; and
cobalt-base alloys, L-605 and Stellite 6B. The nominal chemical composi-
tion of each alloy is listed in Table 1. The heat treatments employed
as well as the densities of these alloys are listed in Table 2.
Reactor grade sodium (99.95 per cent purity) and triple-distilled
mercury were used as the test fluids. Chemical analyses indicated an
initial oxygen level of 10 to 41 ppm for the sodium. The mercury had
less than 0.2 ppm total impurity content.
Accelerated Cavitation Damage Test Facility

The apparatus used is shown schematically in Fig. 1. A more complete
description of the facility and test procedure is given in Ref 20. This
figure illustrates the dry-box arrangement, magnetostrictive transducer
assembly, and separately sealed liquid-metal test chamber. The dry box
and test chamber were designed to be evacuated to a pressure of approxi-
mately 10~3 torr (1 //,), backfilled with high-purity argon prior to testing.
The transducer was a commercial unit modified for use within the
vacuum dry box. The specimen was attached to the end of a resonant
system consisting of the transducer, exponential horn, and an extension
rod specimen holder. The horn served as a displacement amplifier and
provided a convenient attachment for a nodal flange vapor seal. The
amplitude and frequency of vibration were detected by a magnetic pick-
up. A signal was sent to an oscilloscope and to an automatic feedback
system that maintained a constant amplitude of vibration irrespective of
variations in resonant frequency induced by temperature changes. The
output of the magnetic pickup was calibrated against measurements of
amplitude made optically with a 200 power microscope.
When the transducer assembly was lowered into position, the sleeve
attached to the nodal flange on the amplifying horn sealed the liquid-
metal test chamber from the dry box, and the test chamber pressure was
regulated through a separate argon line.
Test Conditions
During all test runs an argon pressure of 1 atmos was maintained above
the liquid metal. Sodium tests were run at 800 ± 10 F, and mercury
tests at 300 ± 30 F. At these temperatures the vapor pressures of both
fluids were less than 0.1 psia. The frequency of vibration of the test
specimens was nominally 25,000 cps, and the peak-to-peak displacement
amplitude was 0.00175 (±0.00005) in.
TABLE 1—Nominal chemical compositions of test materials.
Composition, weight per cent
Material
Iron Nickel Cobalt Chromium Molyb- Tungsten Colum- Tita- Alumi- Carbon Manga- Phos-
Silicon phorous Sulfur Other
denum bium nium num nese
Stellite 6B6 3° j balance 30 1.5° 4.5 1.1 2° 2"

Rene 41 (AMS
5712)c balance 10 to 12 18 to 20 9 to 10.5 3 to 1.4 to 0.12" 0.10° 0.50° 0.015° 0.003 to 0.01
3.3 1.8 (boron)
L-605 (AMS
5759 B)c 3° 9 to 11 balance 19 to 21 14 to 16 0.05 to 1 to 2 1.0° 0.04" 0.03"
Hastelloy X 0.15
(AMS5754D)'. 17 to 20 balance 0.5 to 20.5 to 8 to 10 0.2 to 0.05 to 1.0° 1.0° 0.04° 0.03°
2.5 23 1.0 0.15
A-286 (AMS
5736 B)° balance 24 to 27 13.5 to 1 to 1.5 1.9 to 0.35° 0.08° 1 to 2 0.4 to 0.04 0.03 0.003 to 0.01
16 2.3 1 (boron)
0.1 to 0.5
(vanadium)
Inconel 600
(AMS 5665 G)". 6 to 10 balance 1° 14 to 17 1 0.5° 0.35° 0.15° 1.0° 0.5° 0.015" 0.5"
(copper)
AISI 318 stain-
less steel"* balance 13 to 15 17 to 19 2 to 2.75 0.8 0.08" 2.5° 1.0"
(min)
AISI 316 stain-
less steel (AMS
5648 C)c balance 12 to 14 17 to 19 2 to 3 0.08 1.25 to 1.0° 0.04" 0.03" 0.50«
e
2.0° (copper)
Sicromo 9M . . . . balance 8 to 10 0.9 to 0.2° 0.35 to 1.0° 0.04" 0.04°
1.2 0.65
a
6
Maximum.
Anonymous, "Wear Resistant Alloys," Bulletin F30-1-33-A, Haynes Stellite Co., 1962.
« Aerospace Material Specifications, "Ren6 41-AMS 5712; L605-AMS 5759 B; Hastelloy X-AMS 5754 D; A286-AMS 5736 B; Inconel-AMS
5665
d
G; and Type 316 Stainless Steel-AMS 5648 C."
e
Anonymous, "Stainless Steel Handbook," Allegheny Lundlum Steel Corp., 1951.
Anonymous, "High Alloy Castings," Bulletin 261, Duraloy Co., 1961.
TABLE 2—Heat treatments and densities of test materials.

Material Heat Treatment Density,
g/cm»
Stellite 6B solution heat treated at 2250 F; air 8.38

cooled
Rene 41 solution heat treated at 1975 F; rapid 8.25
quenched
L-605 solution heat treated at 2250 F; water 9.13
quenched
Hastelloy X solution heat treated at 2150 F; rapid 8.23
air cooled
A-286 solution heat treated at 1800 F; water 7.94
quenched; aged at 1325 F for 16 hr
Inconel 600 annealed 8.43
AISI 318 stainless steel annealed 7.99
AISI 316 stainless steel annealed 7.98
Sicromo 9M annealed; heat treated at 1750 F for 1 7.61
hr, then at 1350 F for 1 hr; air cooled
FIG. 1—Schematic diagram of magnetostrictive cavitation apparatus.
Test Procedure
The type of specimen used is shown in Fig. 2. The test surface of
each specimen was metallographically polished and then marked with a
series of microhardness impressions to delineate specific areas for metal-
lographic examination.
YOUNG AND JOHNSTON ON ACCELERATED CAVITATION DAMAGE 191
Specimens were cleaned, photographed, weighed, and subjected to

cavitation damage by vibration for varying intervals. After each period
of operation, the specimens were cleaned, weighed, and photographed.
Weight loss measurements were divided by density to obtain volume loss.
Surface roughness traces of the uniformly damaged portions of the speci-
mens were obtained with a linear profiler having a diamond stylus with
0.0005 in. radius and a cone angle of 51.5 deg. Usually a single trace
approximately 1A in. in length was taken. When several different traces
were taken on the same specimen, the arithmetic average surface rough-
ness values were in agreement within approximately 12 per cent.
Cavitation Damage Results

Volume Loss
Cavitation damage is expressed as volume loss for nine materials hi
sodium and five materials in mercury in Figs. 3a and b, respectively.
FIG. 2—Cavitation test specimen.
Volume loss data for all materials are also summarized in Table 3. The
materials tested in sodium were ranked in order of increasing damage
as follows: Stellite 6B, Rene 41, L-605, Hastelloy X, A-286, Inconel 600,
AISI Type 318 stainless steel, AISI Type 316 stainless steel, and
annealed Sicromo 9M. A wide range of damage was observed for the
various materials. For example, after 4 hr the most resistant material,
Stellite 6B, exhibited approximately 15 per cent of the damage sustained
by L-605, another of the more resistant alloys, but only approximately
1 per cent of the damage sustained by annealed Sicromo 9M, the most
heavily damaged material.
The materials tested in mercury were ranked in order of increasing
damage as follows: Stellite 6B, hardened Sicromo 9M, L-605, Hastelloy
X, and annealed Sicromo 9M. Again, a wide range in the degree of
damage was observed. For example, after 4 hr, the most resistant mate-
rial, Stellite 6B, showed approximately 16 per cent of the damage sus-
tained by L-605; and after 1 hr, approximately 2 per cent that of
annealed Sicromo 9M.
Annealed Sicromo 9M with an original hardness of Rockwell B-80 was
heat treated to a hardness of Rockwell C-40. The hardened alloy at 1
hr showed only about 6 per cent of the damage sustained by this alloy
in the annealed condition. In this particular instance, increasing the hard-
ness substantially increased resistance to cavitation damage.
The materials tested in both sodium and mercury ranked in the same
order of resistance to cavitation damage; however, the severity of cavita-
tion damage experienced by all materials in mercury at 300 F was two
to seven times greater on the basis of total volume loss than that ex-
perienced by the same materials in sodium at 800 F.
FIG. 3—Cavitation damage of materials in liquid metals.
Volume Loss Rate

Curves of volume loss rate for all the materials investigated are shown
in Figs. 4a and b. These curves represent the change of the volume loss
with respect to time as determined from the volume loss curves faired
through the actual data presented in Figs. 3a and b. Smooth curves were
drawn through the volume loss data for AISI Type 316 stainless steel,
Inconel 600, A-286, Hastelloy X, and Stellite 6B, in order to reduce the
effect of scatter. The first derivatives of these curves were used to obtain
the volume loss rate curves shown in Figs. 4a and b. The same procedure
was employed for the remaining materials in Figs. 3a and b, but because
the volume loss data points for the latter materials were more widely
spaced in time and because the exact shape of these curves is uncertain,
portions of the rate curves (Figs. 4a and b) for these materials are dashed.
A recent investigation [21] shows that materials tested for long times
in water first reached a relatively steady-state damage condition but
showed a decreasing damage rate after very long test times. In order to
limit test times to a reasonable length and at the same time to achieve a
meaningful ranking of materials with respect to their resistance to cavita-
TABLE 3—Cavitation test results.
Volume Loss, mm30 Steady- Surface Roughness,
State Min.°
Material Time, hr Volume Time, hr
]Loss 3Rate:1
i 2 3 4 mm /hr 1 2 3 4
(a) Sodium at 800 F

Stellite 6B 0.04 0.13 0.39 0.56 50
Rene 41 0.20 1.12 2.42 1.3' 15 40 65 85
L-605 0.22 1.25 2.58 1.4' 35 55 75 100
Hastelloy X 0 .66 3.10 5.78 8.14 2.4
A-286 1.40 4.70 7.67 10.4 2.8
Inconel 600 2 .70 6.40 10.3 13.4 3.1
AISI 318 stainless
steel 2,.90 7.16 11.8 15.9 4.1 80 100 115 120
AISI 316 stainless
steel 3,.64 8.63 13.8 18.0 4.2
Annealed Sicromo
9M 6 .40 15.5 24.7 33.9 9.2 235 290 320 335
(b) Mercury at 300 F
Stellite 6B 0 .34 0.75 1.28 1.81 0.6' 65 100 125 150
Hardened Sicromo
9M 1.07 3.45 6.80 10.5 3.4' 155 260 450 700
L-605 1.12 4.00 7.60 11.3 3.6' 265 365 525 750
Hastelloy X 5 07 8.2' 510
Annealed Sicromo
9M 18 ? 17.4 765
0
Values taken from curves faired through actual data.
6
Determined after 10 hr.
' May not have reached steady state.
tion damage, the steady-state damage rate in the present investigation

was taken as that region of the rate curves where the damage rate was
relatively constant. In most cases, this occurred after a peak damage rate
was observed.
In order to establish the steady-state condition of the most resistant
material, Stellite 6B, the material was run for a total test time of 10 hr
(Fig. 5). The damage rate curve for Stellite 6B, determined from the
curve of Fig. 5, is shown in Fig. 6.
When the materials are compared on the basis of their steady-state
damage rate, they rank in the same order with respect to resistance to
damage as when compared on the basis of total volume loss. Steady-

state volume loss rates are listed in Table 3.
Surface Roughness Measurements
Because cavitation damage is usually measured quantitatively in terms
of weight or volume loss, damage to system components, such as tubing
or impellers, is sometimes difficult to measure accurately because of
FIG. 4—Rate of cavitation damage of materials in liquid metals.
limited accessibility. If a correlation should exist between volume loss

and surface roughness measurements, the latter might possibly be used
to measure quantitatively the cavitation damage to these components.
Cavitation damage was measured in terms of surface roughness for all
the materials tested in mercury and five of the materials tested in sodium.
These measurements are also summarized in Table 3. Table 4 shows a
typical example of surface roughness as compared with volume loss meas-
urements for L-605 after testing in mercury at 300 F.
A comparison of the arithmetic average surface roughness with volume
loss as a function of test time after testing in sodium and in mercury is
shown in Figs, la and b, respectively. These materials are ranked in the

same order on the basis of both surface roughness and volume loss. Sur-
face roughness measurements are extremely sensitive, and a clear distinc-
tion among the materials as to relative cavitation damage resistance can
be made by this method during the early stages of damage, even though
very little volume loss has occurred.
A cross plot between the volume loss and surface roughness measure-
ments for given times (Fig. 8) indicates that, in the case of specimens
FIG. 5—Cavitation damage of Stellite 6B in sodium at 800 F.
FIG. 6—Rate of cavitation damage of Stellite 6B in sodium at 800 F.
damaged in sodium, the surface roughness tends to approach constant

values with increasing volume loss. These constant values of surface
roughness correspond to the steady-state volume loss rates illustrated hi
Fig. 4a.
For specimens damaged in mercury, however, the surface roughness
data, in general, tend to increase with increasing volume loss and do not
appear to approach constant values within the times tested. These results
may be caused by the fact that some of the materials tested in mercury
have not reached a true steady-state damage rate; thus, further test time
may be necessary to delineate more clearly the rate curves shown in
TABLE 4—Typical surface roughness traces compared with volume loss

measurements for L-605 in mercury at 300 F.
Time, Volume Arithmetic Surface

min. loss, average roughness
mm 3 surface trace
roughness,
juin.
0 <5
1 <0.01 30
4 0.05 50
16 0.15 110
60 1.12 260
Fig. 4b. On the other hand, these results may reflect the nature of cavita-
tion attack by mercury, which is quite different from that by sodium.
Evidence to this effect will be shown in the next section. Briefly, damage
to materials in sodium is evidenced by a general attrition of the surface,
which is shown by the relatively fine-textured surface that erodes
uniformly. Damage due to cavitation in mercury is evidenced by the
formation and continual deepening of wide craters.
FIG. 7—Comparison of surface roughness and volume loss for alloys exposed
to cavitation in liquid metals.
Metallography
Macrographs were taken of all materials after various exposure times.
The damaged surfaces of one of the least and one of the most cavitation-
resistant materials are illustrated after different exposure times to sodium
in Figs. 9 and 10, respectively. AISI Type 316 stainless steel exhibits
early cavitation damage as indicated by the macrograph for this material
taken after 5 min of testing. After 4 hr of testing, the cavitation damage to
this specimen in quite severe. On the other hand, the Stellite 6B specimen
still retained most of its original polish after 5 min of testing, and even
after 4 hr showed a relatively small amount of surface damage.
FIG. 8—Relation between volume loss and surface roughness.
Macrographs of specimens of all the alloys tested in sodium after 4 hr

are shown in Fig. 11. The alloys can be arbitrarily separated into three
groups, each displaying a different degree of cavitation damage: (1)
severe damage—annealed Sicromo 9M; (2) intermediate damage—AISI
Type 316 stainless steel and 318 stainless steel, Inconel 600, A-286,
and Hastelloy X; (3) slight damage—L-605, Rene 41, and Stellite 6B.
Macrographs of the materials after cavitation in mercury are shown
in Fig. 12. Again, these materials can be separated by visual observation
into three groups: (1) severe damage—annealed Sicromo 9M and
Hastelloy X; (2) intermediate damage—hardened Sicromo 9M and L-
605; and (3) slight damage—Stellite 6B. Comparison of Figs. 11 and 12
illustrates a marked difference between damage patterns caused by
sodium and those caused by mercury. After testing in sodium, the speci-
men surface was finely textured, and the rim of the specimen was rela-
tively undamaged. The specimen surface damaged by mercury was very
coarse and deeply cratered with heavy damage near the rim. These differ-
ences in surface damage probably resulted from differences in the com-
plex fluid flows and liquid impact forces due to the widely dissimilar
FIG. 9—Damaged surfaces of AISI 316 stainless steel specimens after exposure
to cavitation in sodium at 800 F.
properties of sodium and mercury, primarily density and surface ten-

sion.
Photomicrographs of the same area on specimens tested in sodium for
different times are shown hi Figs. 13-15. At relatively high magnifica-
tions, all specimens that were tested in sodium showed a selective damage
pattern. Three specific examples are shown in the figures. AISI Type 316
stainless steel showed severe matrix attack after only 5 min while some
grain or twin boundaries stood out in relief (Fig. 13). On the other
hand, in L-605, both grain and twin boundaries were attacked more
heavily than the matrix. This damage is evident after 45 min of test-
ing (Fig. 14). Stellite 6B showed very slight matrix attack after 5
min; only a few carbide particles were dislodged in the early phases of
test (Fig. 15). As test time was increased, more carbide particles were
FIG. 10—Damaged surfaces of Stellite 6B specimens after exposure to cavita-

tion in sodium at 800 F.
dislodged, leaving deep pits. These pits, which widened with time, evi-
dently served as sites for increased cavitation attack. These photomicro-
graphs indicate that although some of the carbide particles were dis-
lodged, most of them remained intact, and their presence evidently is a
major factor in making Stellite 6B highly resistant to cavitation damage.
In mercury, no particular portion of the microstructure of any of
the materials except Stellite 6B appeared to be attacked preferentially.
Figure 16 shows that the carbide particles in Stellite 6B are particularly
FIG. 11—Damaged surfaces of specimens after exposure to cavitation in

sodium at 800 F for 4 hr.
resistant to cavitation attack by mercury, whereas the softer matrix shows

definite attack.
Relation Between Accelerated Cavitation Damage and Material Proper-
ties
The ability to predict which materials have superior resistance to
cavitation damage from mechanical property data would obviously be
useful. Thus, a method of correlating cavitation damage with readily
available material properties, even though empirical in nature, might
serve as a guide to designers and as a substitute for accelerated cavitation
tests.
One of the most recent attempts to achieve a method of ranking
materials with respect to cavitation damage resistance in liquid metals
[16,18] indicates that the severity of cavitation damage may be inversely
FIG. 12—Damaged surfaces of specimens after exposure to cavitation in

mercury at 300 F.
related to the strain energy of materials. Strain energy is approximately

equivalent to the area beneath the stress-strain curve. When the stress-
strain curves are not available, strain energy can be approximated by
the following equation:
where:
YS = yield strength,
TS = tensile strength, and
e = elongation.
The necessary properties for calculating the strain energy of materials
at 800 F are given in Table 5; Fig. 17 shows the relation between strain
energy and the reciprocal of the steady-state volume loss rate of mate-
FIG. 13—Photomicrographs of AISI 316 stainless steel damaged surfaces

exposed to cavitation in sodium at 800 F (X250). Reduced one third for reproduc-
tion.
rials subjected to cavitation damage in sodium. The volume loss rate

values were obtained from Fig. 4a at 4 hr except that of Stellite 6B, which
was taken from Fig. 6 at 10 hr. The exact shape of the volume loss rate
curve for L-605 is not certain, as was mentioned previously. Although
this material may not have reached a steady-state condition after 4 hr,
the value for this test time was used as an approximation of the steady-
state rate.
Figure 17 shows that most of the data fall close to a straight line;
FIG. 14—Photomicrographs of L-605 damaged surfaces exposed to cavitation in

sodium at 800 F (X250). Reduced one third for reproduction.
FIG. 15—Photomicrographs of Stellite 6B damaged surfaces exposed to cavita-

tion in sodium at 800 F (X250). Reduced one third for reproduction.
however, the datum point for the material that performed most favor-
ably, Stellite 6B, is very far removed from the data of the other materials.
Thus, the use of the strain-energy parameter would have resulted in omit-
ting from consideration one of the most cavitation damage resistant ma-
terials.
Evidently, strain energy alone is not entirely representative of the
properties that control the resistance of a material to cavitation damage.
The hardness, elastic modulus, and the fatigue limit are other readily
FIG. 16—Photomicrographs of damaged surface of Stellite 6B exposed to

cavitation in mercury at 300 F (X250). Reduced one third for reproduction.
measurable material properties that might be expected to have some

effect on cavitation damage resistance. In order to completely evaluate
their role, extensive additional data for many materials are needed. Fi-
nally, the validity of any such parameter would be affected by corrosion
variables that differ for different environments and that might in some
cases be overriding. In any event, much additional research is needed to
achieve a better understanding of the relations between cavitation damage
and readily measurable material properties.
TABLE 5—Material properties at 800 F and calculated strain energy

parameter for sodium tests.
Ultimate Yield •PI^ ng Calculated
Material Tensile Strength (0.2 Ho°n £~ Strain Energy,
Strength, psi % offset), psi tion, % tg/mmja'
Stellite 6B* . . . . 138 000 71 000 29 0 21 3

L-605C 119 000 35 000 75 0 40 6
Hastelloy Xc 100 000 47 000 50 0 25 8
A-286' 137 000 92 000 21 0 16.9
Inconel 600C 88 000 29 000 49 0 20 1
AISI 316 stainless steelc 70 000 28 000 40 0 13 8
0
Strain energy = [(yield strength + tensile strength)/2] X elongation.
6
"Wear Resistant Alloys," Bulletin F30-1-33-A, Haynes Stellite Co., 1962.
c
V. Weiss and J. G. Sessler, editors, Aerospace Structural Metals Handbook,
Syracuse University Press, 1963.
FIG. 17—Relation of cavitation damage in sodium with strain energy param-

eter.
Relation Between Accelerated Cavitation Damage and Pump Im-

peller Tests
A qualitative comparison was made between the damage experienced
by three materials, Rene 41 and AISI Types 316 and 318 stainless steels,
tested in the accelerated cavitation damage facility and that experienced
by the same materials when used as pump impeller vanes in sodium loop
tests conducted at the Lewis Research Center.3 Visual observations in-
dicated that the materials ranked in the same order with respect to re-
3
W. S. Cunnan, D. C. Reemsnyder, and C. Weigle, unpublished data.
FIG. 18—Cavitation damage to pump impeller blades operated in liquid sodium

for 250 hr up to 1500 F.
sistance to cavitation damage after accelerated tests as after actual pump

loop operation.
Macrographs of specimens subjected to accelerated cavitation damage
are shown in Fig. 11. Macrographs of the damaged surfaces of impeller
blades that were operated for 250 hr under cavitating conditions at tem-
peratures up to 1500 F are shown in Fig. 18. The Rene 41 impeller blade
showed virtually no cavitation damage, whereas the AISI Types 318
and 316 stainless steel blades had marked regions of damage. The degree
of damage for the two stainless steel blades was not appreciably differ-
ent. When the materials are considered on the basis of volume loss in
the accelerated tests (Fig. 3d), Rene 41 shows considerably less damage
than either of these steels. Both steels, however, ranked very close with
respect to volume loss. It is significant that a qualitative agreement be-
tween the results of accelerated cavitation tests and full-scale impeller
operation was obtained. Surface traces were taken of the damaged areas
of the impeller blades in an attempt to determine a quantitative meas-
ure of the cavitation damage. However, the extent of general corrosion
of the blade surfaces masked the degree of localized cavitation damage.
Although further verification is required for other materials, and under
other operating conditions, these results suggest that the magnetostrictive
type of accelerated cavitation test can provide a useful means of selecting
materials suitable for longtime operation under cavitating conditions.
General Remarks
Corrosion Effects
Previous investigations [7,10,15] have indicated that corrosion can
be a contributing factor to cavitation damage. In pump impeller tests,
blades tested for 250 hr under cavitating conditions in sodium showed
corrosion over the entire blade surface as well as clearly delineated cavi-
tation damaged areas (Fig. 18). In the accelerated cavitation tests, the
contribution of corrosion to the total damage may be less than that nor-
mally found in longtime service under cavitating conditions. A pulsing
technique for determining the effect of corrosion in accelerated cavita-
tion tests in aqueous solutions has been introduced in an earlier investiga-
tion [10\. The introduction of hold periods substantially lengthens the
accelerated type of test; however, the method may be useful in deter-
mining the additional contribution of corrosion to the total damage
effect experienced by materials in accelerated liquid-metal cavitation
tests.
Effect of Oxygen Contamination in Sodium

Although reactor-grade sodium with very low initial oxygen content
was used in this investigation, chemical analysis indicated that the oxygen
level was near the limit of solubility after prolonged test times. In order
to reduce the possibility of subjecting the individual materials to sub-
stantially different fluid environments, each material was, in turn, sub-
jected to cavitation damage for a short increment of time, rather than
conducting a complete test with one material before beginning a test with
the next material. Also, fresh charges of sodium were introduced at
several intervals. The volume loss curves for individual materials did
not show significant discontinuities when tests were resumed with the
fresh charges of sodium.
For the test times and temperatures employed in this investigation,
oxygen contamination is not considered to have particularly adverse
effects upon material properties of the steels and superalloys; however,
certain refractory metals are adversely affected by exposure to oxygen
in sodium at high temperatures. Columbium and tantalum, for example,
are subject to severe embrittlement as a result of oxygen pickup. Further-
more, these metals are subject to accelerated corrosion attack when oxy-
gen is present. Adequate measures must therefore be taken to reduce oxy-
gen absorption when subjecting refractory metals to cavitation tests in
sodium.
Selection of Materials for Use Under Cavitating Conditions

The materials considered in this investigation show wide differences
in their resistance to cavitation damage; however, resistance to cavita-
tion damage is not the sole factor to be considered when selecting a
material for use as a pump impeller in liquid-metal systems. As in most
engineering applications, a compromise among various material proper-
ties will probably be required. Stellite 6B is far more resistant to cavita-
tion damage than all the other materials considered; however, it is notch
sensitive and has relatively low ductility. In contrast, L-605, which has
less resistance to cavitation damage, has greater ductility. This could be
an overriding factor to a designer in choosing a material for pump im-
peller applications in space power conversion systems. The preceding ex-
ample, of course, illustrates only one other aspect besides resistance to
cavitation damage that must be investigated in making a material selec-
tion. Others, such as rupture strength, creep resistance, and stability upon
longtime exposure to elevated temperature, must all be taken into con-
sideration. The net result could well be that the order of resistance to
cavitation damage would not necessarily be the order of choice of ma-
terials for pump impellers in space power systems.
Summary of Results
The resistance to cavitation damage of candidate materials for com-
ponents of liquid-metal space power conversion systems was investigated
in two liquid-metal environments. A magnetostrictive-type apparatus was
used to achieve an accelerated rate of cavitation damage. Tests were
run in sodium at 800 F and in mercury at 300 F. The following results
were obtained:
1. In all cases, the materials that were tested in both sodium and
mercury ranked in the same order based on resistance to cavitation
damage.
2. The severity of the cavitation damage experienced by all materials
YOUNG AND JOHNSTON ON ACCELERATED CAPTATION DAMAGE 21 1
in mercury at 300 F was consistently greater than that experienced by

the same materials in sodium at 800 F.
3. Stellite 6B was far superior to all other materials investigated. For
example, in sodium, when compared on a volume loss basis after 4 hr,
the damage sustained by this alloy was approximately 1 per cent that
of annealed Sicromo 9M, the most heavily damaged material. In mercury,
the volume loss sustained by Stellite 6B after 1 hr was approximately 2
per cent that sustained by annealed Sicromo 9M.
4. Surface roughness measurements provided a ranking of materials
with respect to cavitation damage resistance similar to that obtained by
conventional volume loss measurements.
5. Metallographic examination at high magnifications during the
early stages of cavitation damage indicated that cavitation in sodium
resulted in nonuniform damage to all materials, as evidenced by the
delineation of twin and grain boundaries. Cavitation in mercury, on the
other hand, resulted in a uniformly damaged surface with no apparent
preferential attack except for Stellite 6B. In this alloy, the carbides were
more resistant than the matrix. Examination at low magnifications after
appreciable damage had occurred indicated that cavitation in sodium re-
sulted in a fine-textured (matte) surface, whereas exposure to mercury
resulted in very coarse, deep craters.
6. Visual observations of pump impeller blades of AISI Types 316
and 318 stainless steels, and Rene 41 tested under cavitating conditions
for 250 hr at temperatures up to 1500 F in sodium indicated the same
ranking of these materials with regard to cavitation damage resistance as
that determined in the accelerated tests of this investigation.
References
[1] R. S. Kulp and J. V. Altieri, "Cavitation Damage of Mechanical Pump Im-
pellers Operating in Liquid Metal Space Power Loops," Pratt and Whitney
Aircraft, NASA CR-165, 1965.
[2] P. G. Smith, J. H. DeVan, and A. G. Grindell, "Cavitation Damage to
Centrifugal Pump Impellers During Operation with Liquid Metals and Molten
Salt at 1050°-1400° F," Journal of Basic Engineering, Am. Soc. Mechanical
Engrs., 85, No. 3, September, 1963, pp. 329-335; discussion, pp. 335-337.
September, 1963, pp. 347-359.
[4] Anonymous, "Sunflower Power Conversion System." Report No. ER-5163,
Thompson Ramo Wooldridge, Inc., NASA CR-56206, 1962.
[5] Anonymous, "SNAP-8 Topical Materials Report for 1963," Report No. 2822
(Special), Vol II, Aerojet-General Corp., March, 1964.
[6] R. T. Knapp and A. Hollander, "Laboratory Investigations of the Mechanism
of Cavitation," Transactions, Am. Soc. Mechanical Engrs., Vol 70, No. 5,
July, 1948, pp. 419-431; discussion, pp. 431-435.
[7] H. S. Prieser and B. G. Tytell 'The Electrochemical Approach to Cavitation
Damage and its Prevention," Corrosion, Vol 17, No. 11, November, 1961,
pp. 107-115; discussion, pp. 115-121.
[5] R. T. Knapp, "Recent Investigations of the Mechanics of Cavitation and
Cavitation Damage," Transactions, Am. Soc. Mechanical Engrs., Vol 77,

No. 7, October, 1955, pp. 1045-1054.
[9] M. S. Plesset and A. T. Ellis, "On the Mechanism of Cavitation Damage,"
Transactions, Am. Soc. Mechanical Engrs., Vol 77, No. 7, October, 1955,
pp. 1055-1064.
[10] M. S. Plesset, "Pulsing Techniaue for Studying Cavitation Erosion of Metals,"
Corrosion, Vol 18, No. 5, May' 1962, pp. 181-188.
[11] C. F. Naude and A. T. Ellis, "On the Mechanism of Cavitation Damage by
Non-Hemispherical Cavities Collapsing in Contact with a Solid Boundary,"
Journal of Basic Engineering, Am. Soc. Mechanical Engrs., Vol 83, No. 4,
December, 1961, pp. 648-656.
[12] W. H. Wheeler, "Indentation of Metals by Cavitation." Journal of Basic
Engineering, Am. Soc. Mechanical Engrs., Vol 82, No. 1, March, 1960, pp.
184-191; discussion, pp. 191-194.
[13] J. Z. Lichtman and E. R. Weingram, "Cavitation Design Handbook," Report
NR 062-314, U.S. Naval Applied Science Lab., Brooklyn, N. Y., September
30, 1964.
[14] F. G. Hammitt, L. L. Barinka, M. J. Robinson, R. D. Pehlke, and C. A.
Siebert, "Initial Phases of Damage to Test Specimens in a Cavitating Venturi
as Affected by Fluid and Material Properties and Degree of Cavitation,"
June, 1965, pp. 453-464.
[75] W. C. Leith and A. L. Thompson, "Some Corrosion Effects in Accelerated
Cavitation Damage," Journal of Basic Engineering, Am. Soc. Mechanical
Engrs., Vol 82, No. 4, December, 1960, pp. 795-807.
[16] A. Thiruvengadam, "A Unified Theory of Cavitation Damage," Journal of
Basic Engineering, Am. Soc. Mechanical Engrs., Vol 85, No. 3, September,
1963, pp. 365-376.
[17] F. G. Hammitt, "Cavitation Damage and Performance Research Facilities,"
Symposium on Cavitation Research Facilities and Techniques, J. William
Holl and Glenn M. Wood, editors, Am. Soc. Mechanical Engrs., 1964, pp.
175~1«4.
[18] H. S. Preiser, A. Thiruvengadam, and C. Couchman, "Cavitation Damage
Research Facilities for High Temperature Liquid Alkali Metal Studies,"
Symposium on Cavitation Research Facilities and Techniques, J. William
Holl and Glenn M. Wood, editors, Am. Soc. Mechanical Engrs., 1964, pp.
146-156.
[19] R. W. Kelly, G. M. Wood, H. V. Marman, and J. J. Milich, "Rotating Disk
Approach for Cavitation Damage Studies in High Temperature Liquid
Metal," Paper No. 63-AHGT-26, Am. Soc. Mechanical Engrs., March, 1963.
[20] S. G. Young and J. R. Johnston, "Accelerated Cavitation Damage of Steels
and Superalloys in Liquid Metals, NASA TN D-3426, Nat. Aeronautics and
Space Administration, 1966.
[21] M. S. Plesset and R. E. Devine, "Effect of Exposure Time on Cavitation
Damage," Report 85-31, California Institute of Technology, 1965. (Available
from DDC as AD-471191.)
DISCUSSION ON ACCELERATED CAVITATION DAMAGE 213
DISCUSSION
F. G. Hammitt1 and R. Garcia1 (written discussion)—The concept ad-

vanced by the authors of correlating cavitation damage with surface
roughness is interesting and could well be highly useful before the dam-
age became too extensive. Judging at least from some of our own tests
in a vibratory facility, as reported in our paper2 of this symposium,
the surface roughness may well reach a steady-state condition during the
latter portion of a test, and this would, of course, contribute to the rela-
tively constant damage rate sometimes observed during this portion of a
test.
The present series of tests in mercury and sodium are quite similar
to our own in mercury, lead bismuth, and water, reported in our paper.2
In addition, we have completed a series using the same materials in
lithium.3 Unfortunately, direct quantitative comparisons with the pres-
ent paper are not possible since we have only one material hi common
(Type 316 stainless steel) so that the rankings obtained in the two facil-
ities, which differ in detail in frequency and amplitude, cannot be com-
pared. This situation, of course, indicates the desirability in future of
tests in different laboratories at least including several "standard" ma-
terials, perhaps to be defined by ASTM for this purpose.
A qualitative comparison of the appearance of the mercury damage
obtained in the present investigation with our own shows that it is indeed
very similar. Also, there is some resemblance between the cavitated sur-
faces from the sodium tests presented in the present paper and those we
have obtained with water and lithium, both fluids in the same density
range.
W. C. Leith4 (written discussion)—The test data presented by the
authors have been obtained with a unique transducer operating at a fre-
quency of 25,000 cps and an amplitude of 0.00175 in., which makes
it difficult for comparison to other reported data, but it is interesting to
note that the volume loss after 2 hr given in Table 3 is about 18 times
greater in mercury than in sodium, and, strangely enough, this volume-
loss ratio agrees with the theoretical predicted ratio calculated elsewhere.5
1
Nuclear Engineering Dept, The University of Michigan, Ann Arbor, Mich.
2
See p. 239.
3
R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation Studies in
Lithium at Elevated Temperatures," ORA Report 05031-5-T, Department of
Nuclear Engineering, The University of Michigan, Ann Arbor, Mich., May, 1966.
* H. G. Acres & Company Ltd., Consulting Engineers, Niagara Falls, Canada.
5
W. C. Leith, "Prediction of Cavitation Damage in the Alkali Liquid Metals,"
Proceedings, Am. Soc. Testing Mats., Vol 65, 1965, pp. 789-800.
FIG. 19—Predicted cavitation damage in alkali liquid metals.
The predicted cavitation damage for alkali liquid metals is given in

Fig. 19 and for sodium and mercury in Fig. 20.
Test Material Ratio of Volume Loss

(mercury/sodium)
Stellite 6B 0.75/0.04 = 18.75

L-605 4.0/0.22 = 18.2
Theoretical (100) 925/50 = 1 8 . 5
H. S. Preiser6 (written discussion)—The authors are to be compli-

mented for making available additional cavitation damage data to ma-
terials exposed to liquid metals. Such data are still sparse due to the
complexity of the experimental technique required.
6
Principal research scientist, Hydronautics, Inc., Laurel, Md.
FIG. 20—Predicted cavitation damage in sodium and mercury.
There are several aspects of the paper which need emphasis to place
the conclusions drawn in the proper perspective—first, the use of the
term "accelerated" cavitation damage. While it is true that the magneto-
striction test produced damage of higher intensity than that experienced
with the sodium pump impeller tests, there are other examples in the
literature which show the damage rate of field devices to be of the same
order or greater than that of conventional magnetostriction devices.7
7
A. Thiruvengadam, "Intensity of Cavitation Damage Encountered in Field
Installations," Technical Report 233-7, Hydronautics, Inc., Laurel, Md., February,
1965. (See also Symposium on Cavitation Problems in Fluid Machinery, Am. Soc.
Mechanical Engrs., Winter Annual Meeting, Chicago, El., November, 1965.)
Therefore, since the word "accelerated" is a relative term, the writer

would prefer not to generalize the magnetostriction test as an accelerated
test.
The authors indicate that the order of merit of cavitation damage
resistance of materials was in the same order whether the cumulative loss
or steady-state volume-loss rates were used as a basis of comparison.
While this is true for the data they have presented, care should be ex-
ercised in ascribing this circumstance to the general case. Other data8 for
cavitation damage of materials hi liquid metals show that peak rates of
damage are not in the same order of merit as the steady-state rates of
damage. Therefore, for relatively short periods of exposure, total cumu-
lative damage may not order the cavitation damage resistance of materials
properly. For safety in design, the steady-state damage rates should be
used to rate the resistance of the material to cavitation attack at a
given intensity.
Although surface roughness measurements used by the authors cor-
related with volume-loss measurements, it cannot be emphasized too
strongly that such data are qualitative at best, since the surface roughness
is not a finite indication of the amount of material removed during the
cavitation damage process.
In regard to the authors' attempt at correlating the strain energy of the
material with the reciprocal of its steady-state volume loss rate, it is
gratifying to see that most of their data correlate well with strain energy.
It is suggested that the apparent anomaly observed for Stellite can be
explained by the following:
1. Stellite has not fully reached its steady-state damage rate, and
therefore its rate of damage could be higher.
2. The weight-loss measurements of the order of 0.1 mg in a typi-
cal 30-min test run on Stellite are unreliable at the intensity of damage
obtained with a magnetostriction device. We have sometimes meas-
ured a weight gain after cavitation exposure due presumably to moisture
accumulation on the specimen surface during weighing.
3. The estimated strain-energy value may be much lower than the
actual area of the stress-strain diagram. The approximate value of strain
energy is derived from an engineering stress-strain diagram which is
idealized to the shape of a trapezium; however, in actuality, the stress-
strain diagram could vary considerably from this shape, depending upon
the material and temperature of the test. It is clear that more work is
required to verify the strain-energy correlation; however, it seems short-
sighted to discard its importance at this juncture.
The authors' point about cavitation damage resistance of a material
not necessarily being the controlling property for its selection in a given
8
P. Eisenberg, H. S. Preiser, and A. Thiruvengadam, "On the Mechanism of
Cavitation Damage and Methods of Protection," Paper No. 6, Soc. Naval Architects
and Marine Engrs., Winter Annual Meeting, November, 1965.
engineering application is well taken, but where cavitation damage re-

sistance is found to be overriding, the correct measurement of this param-
eter is essential.
The authors deserve to be complimented for their very thorough pres-
entation; however, the conclusions drawn relate to this specific experi-
ment and should not be generalized until more data become available.
/. B. Marriott9 (written discussion)—It is encouraging to compare the
curves obtained by Young and Johnston for cavitation in liquid metals
with those obtained for water droplet impact and observe the similarity
in shape.
In Fig. 14 of the paper, the development of cavitation damage in L-605
is shown.
From the appearance of the specimen after 45 min attack, metal loss
would seem to be associated with grain boundaries and persistent slip
lines. In the study of water jet erosion damage of Haynes 6B reported
by Mr. Beckwith and myself, we find that slip-line cracking plays a sig-
nificant role in the erosion process (Figs. 3 and 5 of our paper illustrate
this.) We have not, however, studied the behavior of an alloy free from
the dispersion of coarse carbides. Alloy L-605 would seem to be fairly
similar to the matrix of Haynes 6B, and it is interesting to note that it
behaves as would be expected.
In Fig. 3 of their paper, Young and Johnson report that Rene 41 and
L-605 have a similar cavitation resistance. These two alloys have rather
similar compositions with the transposition of nickel and cobalt. In the
solution-treated condition, they should also have fairly similar hardness.
Can the authors say whether the progress of cavitation damage in Rene
41 is similar to that in L-605?
R. W, Wilson10 (written discussion)—There are three comments I wish
to make with regard to the paper by Young and Johnston and the
paper by Garcia et al.11
1. Results obtained on short-time cavitation tests may be misleading
when applied to practical systems because short-time tests emphasize
the mechanical component of cavitation damage and minimize the con-
tribution made by corrosion. In service, cavitation damage usually arises
from the combined effects of these two factors, and one augments the
other. Even liquids which are not usually regarded as corrosive can be-
come very reactive toward metals, due to the very high local pressures
and temperatures which are attained during cavity collapse.
2. I am most interested to see the various damage patterns produced
on circular disk-type specimens exposed to cavitation in different liquid
metals. I have observed similar effects on disk specimens exposed in
9
Leicester, England.
"Principal scientist, Shell Research Ltd., Thornton Research Centre, Chester,
England.
11
See p. 239.
various liquids.12-13 Thus, benzene gives a "spoked wheel" pattern with

an undamaged center zone, whereas water gives a spoked-wheel pattern
with severe damage at the center of the disk. Heptane gives a single ring
of damage. The remarkable similarity of these patterns to vibrational pat-
terns observed on circular plates (Chladni Figures14'15) suggests that on
cavitation test disks these damage patterns are a consequence of the
modes of vibration of the disks.
3. The amount of mechanical cavitation damage sustained hi different
liquids by a particular metal depends primarily on the product of the
density and velocity of sound in the liquids, when other test variables such
as amplitude and frequency of vibration of specimens, and vapor pres-
sure of test liquids are kept constant. This product is usually termed the
pV factor and is known in physics as the acoustic impedance. The relation
between the amount of damage and the pV factor can be derived from
Rayleigh's expression16 for the pressure developed in a liquid during the
collapse of a cavity and has been verified experimentally.
S. G. Young and J. R. Johnston (authors)—The authors appreciate
the reviewer's comments and feel that the discussions presented will
provide the reader with additional insight into the complexities of the
liquid metal cavitation problem. We are pleased to note the similarities
between our data and those of Messrs. Hammitt and Garcia, as well as
the theoretical predictions by W. C. Leith and the droplet impact ob-
servations of J. B. Marriott. In answer to Mr. Marriott's question con-
cerning Rene 41, the progress of cavitation damage in Rene 41 was simi-
lar to that in L-605, and grain boundary attack was observed. This was,
however, not^s pronounced, as in the case of L-605.
There is indeed a definite need to combine and compare the results ob-
tained by the several laboratories that are testing materials for resistance
to cavitation damage. The use of "standard" materials, as indicated by
Dr. Hammitt, to achieve meaningful comparisons would be helpful.
In answer to Mr. Preiser's comments, the term "accelerated" has
usually been associated with the magnetostrictive test (ASME report on
Standardization).17 The authors do not state that this is the only method
12
R. W. Wilson and R. Graham, "Cavitation of Metal Surfaces in Contact
With Liquids," Proceedings of the Conference on Lubrication and Wear, Institute
Mechanical Engrs., London, England, October, 1957.
18
R. W. Wilson, 'The Influence of the Physical Properties of Liquids on the
Severity of Cavitation Damage," Compressed Air and Hydraulics, Vol 27, 1962,
p. 382.
" Lord Rayleigh, Sound, Vol 1, Chapters 9 and 10,1877.
15
M. D. Waller, Chladni Figures, G. Bell and Sons Ltd., London, England, 1961.
18
Lord Rayleigh, "On the Pressure Developed in a Liquid During the Collapse
of a Spherical Cavity," Philosophical Magazine, Vol 34, 1917, p. 94.
17
L. E. Robinson, B. A. Holmes, and W. C. Leith, "Progress Report on Stand-
ardization of the Vibration-Cavitation Test," Transactions, Am. Soc. Mechanical
Engrs., January, 1958, pp. 103-107.
which can be used to rapidly damage materials by cavitation. In answer

to the question raised as to whether it is best to rate materials on a cumu-
lative damage or a steady-state damage rate basis, the authors believe that
both methods are needed to give a complete picture of the nature of the
cavitation damage.
We agree that surface roughness measurements in these tests are
qualitative. However, as indicated in the paper, such measurements can
be useful in those instances where conventional weight loss measure-
ments cannot readily be made.
In reference to Mr. Preiser's comments dealing with the results ob-
tained for Stellite 6B, this material has clearly shown a steady-state
damage rate zone from volume loss measurements, as indicated in Figs.
5 and 6. Further evidence of this steady-state zone was noted from the
surface roughness versus volume loss measurements (Fig. 8a). The weight
loss measurements for Stellite 6B were on the order of 4.0 to 4.8 mg/1
hr during the last 6 hr of testing, not 0.1 mg/30 min.
The authors agree that the strain energy concept should not be dis-
carded because the Stellite 6B data point does not fit. The use of the
true area under the stress-strain diagram would, of course, provide a
more accurate determination of the strain energy than the estimated value.
However, it is unlikely that the value of strain energy so determined
would be five times the estimated value, as is required to conform to the
strain energy concept in this instance.
The corrosion aspect of cavitation damage as mentioned by R. W.
Wilson is very important. Materials that are especially subjected to
corrosion can probably be separated from corrosion-resistant materials
by the use of a pulse-type of test, as suggested by Plesset (Ref 10 of the
paper). The authors appreciate the additional comments offered by Mr.
Wilson and feel that these contribute to the limited information avail-
able regarding damage patterns and effects of fluid parameters on cavi-
tation damage.
K. K. Shalnev,1 J. J. Varga* and G. Sebestyen?
Scale-Effect Investigation of Cavitation

Erosion Using the Energy Parameter
REFERENCE: K. K. Shalnev, J. J. Varga, and G. Sebestyen, "Scale-

Effect Investigation of Cavitation Erosion Using the Energy Parameter,"
Erosion by Cavitation or Impingement, ASTM STP 408, Am. Soc.
Testing Mats., 1967, p. 220.
ABSTRACT: The authors define the energy parameter as the ratio of
the volume loss of specimen material during erosion test to a certain
fraction of the work done by the Cavitation drag force. On the grounds of
the analysis of the initial formula of the energy parameter as well as
on the basis of experimental investigations, a method is suggested to
allow for the influence of geometrical similarity (scale effects) upon cavita-
tion damage when flow velocity or the characteristic dimension of the
model and certain' physical properties of the fluid, such as viscosity,
surface tension, or density, are varied.
KEY WORDS: erosion, Cavitation, scale (ratio), viscosity, surface ten-
sion, density
Up to recent times, the influence of the scale effect on the intensity of

cavitation damage has been investigated with particular respect to veloc-
ity scale. This research activity includes papers, among others, by Schro-
ter [1,2],4 Knapp [3,4], Kerr and Rosenberg [5], Rata [6], Govinda Rao
and Thiruvengadam [7], and Hammitt [8]. Our recent research work
[9-12] has rendered, by means of the energy parameter, an adequate
basis for the determination of the effect of velocity on erosion intensity.
Investigations of the influence of model geometrical scales have not
been conducted so far, unless the practical observations by Ackeret and
de Haller [73] could be considered as such.
The influence of the velocity, of the dimensions of geometrically
similar models, and of certain physical properties of fluids may be deter-
1
Institute of Mechanical Problems, Academy of Sciences of U.S.S.R., Moscow.
2i 3
Department of Hydraulic Machinery, Budapest Technical University, Hun-
gary.
1
paper.
220
SHALNEV ET AL ON SCALE-EFFECT INVESTIGATION 221
mined by analyzing the energy parameter expression. Although velocity

effects had been dealt with in our previous papers [9-72], with respect to
our having completed these studies by new data, it seems desirable to
review this side of the problem.
The experimental investigations have been conducted by the Institute
of Mechanical Problems, Soviet Academy of Sciences of U.S.S.R., and,
with the cooperation of the Working Group for Fluid Mechanics, Hun-
garian Academy of Sciences, in the cavitation tunnel of the Department
of Hydraulic Machinery, Budapest Technical University. The investiga-
tions were conducted with circular-cylinder models of diameter d and
length h located in plane-parallel flow in a working section of rectangular
section. Cylinder and working section dimensions are as follows:
Rectangular Working Section Circular Cylinder

Dimensions, mm d, mm h, mm
6 X 25 6,7.5 6
12 X 50 12, 15 12
24 X 100 24 24
48 X 200 48 48
The Energy Parameter of Cavitation Damage
Definition
The energy parameter E of the cavitation damage is the ratio of the
volume loss per unit time of the material exposed to cavitation erosion,
AF, to the work unit performed by a certain fraction of the cavitation
drag force of the body causing cavitation, AD.
where:
CD = drag coefficient of the body,
CDC = same but in case of cavity flow,
A Cz/ = a certain part of drag determined by direct extrapolation as
function of the Weber number ACD (We) (Weber number
We = pvM2d/<r, where p is fluid density and a means surface
tension extended up to its zero value) (see Fig. 1),
qx = 0 . 5 v^/g, where VM is flow velocity with respect to the con-
traction caused by the model in the working section, and
7 = specific weight of the fluid.
Methodological Problems
In the course of our previous investigations, the conditions to be
provided for in order to ensure the similarity of erosion intensity have
been pointed out. These conditions refer, first of all, to the following:
1. The stage cavitation X = lz/d, where lz represents cavity length.
The results of our investigations have indicated that the cavitation condi-
tion exerts a considerable influence on the intensity of cavitation erosion.
However, the place where the maximum erosion intensity occurs is also
influenced by the Reynolds number. In the critical Reynolds number
range (Re = 1 X 105 to 2.4 X 105) the maximum occurs at X = 3 (Fig.
FIG. 1—The approximative correlation of the drag coefficient (Co) of the circular
cylinder in function of the Weber number (We). CD , without cavitation; CDC > in cavity
flow.
2a), whereas in the range above the critical Reynolds number it can be
found at X = 1.5 (Fig. 2b).
2. The relative model dimensions, h/d or a/d. According to our test
results, the intensity of cavitation erosion, that is, the volume of material
eroded during unit time, is maximum when the width, a, of the test
section, that is, the length, h, of the cylinder, is exactly the same as the
diameter of the cylinder; that is, a/d = h/d = a/h = 1 (Fig. 3).
3. The water temperature within the flow device (see Fig. 4) [14],
However, only a few attempts have been made to discover the nature
of effects exerted by the test period and the surface-finish quality of the
test specimen on the volume of erosion.
The test period problem should be studied with the AG(r) or AF(r)
curve taken as a basis, where AG is the weight loss of the specimen sub-
jected to erosion test and is the volume loss due to

erosion during unit time, in which 7,, is the specific gravity of the speci-
men material and T is the given duration of the test. Almost all metal
types tested from the aspects of cavitation damage exhibit in the early
stages of testing two more or less clearly definable cavitation damage
zones as illustrated by diagram AG(r) (see Fig. 5). These are: (a) incuba-
(a) In case of low Reynolds numbers (d = 6 mm)

(b) In case of high Reynolds numbers (d = 48 mm, v = 12 m/sec)
FIG. 2—Influence of the cavitation conditions (X) on erosion intensity (AF, mnf/hr).
tion zone and (V) complete destruction (accumulation) zone. The border
point of these two zones defines the critical point. As revealed by the
experiments conducted with lead test specimens in the working section
of 48 by 200 mm dimensions, the incubation period is characterized by
the decreasing acceleration of erosion, whereas the total destruction-
accumulation zone is characterized by an increased erosion acceleration.
The critical point is particularly emphasized in such AG!T~2(r) diagrams.
In Fig. 6 the critical point is marked by the minimum value of the AG"
curve. This, in the given case, is r cr it = 7.8 h. Thus the assumption can
be arrived at according to which the incubation period of cavitation

erosion is brought about by actual cavitation phenomena effects while
the total destruction zone is due to the joint cumulative action of cavita-
tion and flow. For the examination of scale effect of cavitation erosion,
the incubation period should be used. In cavitation damage investigations
of different materials, however, the incubation period is usually neglected,
and experiments of identical period of time are conducted for the entire
material specimen set.
FIG. 2—Continued.
Neglecting the incubation period makes the determination of individual

test periods uncertain, which is inadmissible for scale effect studies. This
problem will be discussed in detail later.
Prior to discussing the influence of surface roughness, the configura-
tion of the damaged test specimen surface will be presented (Fig. 7).
Erosion may be divided in two clearly perceptible parts: (a) along the
wall surface of the cylinder and (fc) at the end of the cavitation zone.
Both parts consist of crater-shaped damages.
Over the bottom of the pits there are colored areas which can be
observed (Fig. 8). Not much is known about the origin of these. It could
be by heat effect. When the pits have arrived at a certain depth, further
FIG. 3—Influence of relative cylinder dimensions (h/d) on erosion intensity (AF,

mm3/hr); d = 6 mm, v = 17 m/sec, X = const.
FIG. 4—Influence of water temperature (t, C) on erosion intensity (AF, mnp/hr)

according to the results of experiments conducted in a working section of 6 by 25 mm,
X = 3.
increase of this depth is replaced by an increase in diameter. In case of

excessively large-size pits, there are secondary ones developing within
these pits. At the artificial grooves, displacing of the cavity walls can be
observed (Fig. 9).
The influence of surface roughness with lead specimens was also tested
in a manner identical to that employed for the cavitation damage experi-
ments described previously, in a working section of 48 by 200 mm.
Roughness was developed artificially by rolling it along a weight-loaded
FIG. 5—The two cavitation damage phases: (/) incubation zone; (2) total destruction
(accumulation) zone.
threaded cylinder. Roughness is thus represented by parallel grooves at

right angles to the direction of flow. The values of roughness of the three
samples expressed in conventional root mean square terms are hq =
12.37 and 45 n. The appearance of the surface roughness, including ero-
sion marks, is illustrated by Fig. 9. Typically, groove walls are displaced
during the initial phase of erosion. Diagram &GCTit (hq) reveals (Fig. 10)
that, up to hq = 30ju, roughness does not affect erosion to a significant
extent.
Verification of the Value E = Const
The reciprocal value of the energy parameter means the actual re-
FIG. 6—Determination of the critical point with the aid of erosion acceleration (AG" =
AG/r2) as a function of the time (T); (d = 48mm, v — 12 m/sec).
FIG. 7—Full view of the lead specimen following an extended cavitation effect. Size
of the test piece is 96 by 240 mm.
sistance of the material to cavitation damage which can be expressed by

the work required for the destruction of material unit volume. The
determination of the relationship between E and the hydrodynamic
similarity is of the utmost importance. The results of calculations per-
228 EROSION BY CAVITATION OR IMPINGEMENT'
FIG. 8—Erosion damage. Material of the test piece is lead (X8).
FIG. 9—Erosion exhibited on the rough surface lead specimen (X16).

FIG. 10—The effect of suâce roughness (hq) on the critical weight loss of erosion
(AG cr ,/ng).
FIG. 11—The energy parameter (E\) as a function of the Weber number (We).
formed in order to determine E gave, on the basis of test results obtained

by circular-cylinder model experiments in geometrical similar working
sections and under similar cavitation conditions, the following formula
(Fig. 11):
where subscripts 1 and 2 represent geometrically similar models and

fluids with different density, respectively, wherein the experiments have
been conducted. The ^-values for cylinders of d = 6 and 12 mm diam-
eter, respectively, have been obtained from certain A(j = const values.
The Is-values concerning cylinders of d = 24 and 48 mm diameter are
calculated with the A<j cr i t figures. Calculations performed for each
cylinder with the data rendered by the total destruction period would
reveal excessive scattering. Therefore, the results referred to in the fore-
going show that E = const if the foregoing methodological requirements
are only adhered to in the course of the experiments.
Energy Parameter as a Test Basis for Erosion
Scale Effect Investigations

Equation 1 representing the energy parameter may be used, if slightly
modified, for cavitation damage scale effect analysis purposes. Modifica-
tion is justified by the effort to determine a correlation between the energy
parameter of erosion and the very large number of quantities characteriz-
ing the physical properties of the fluid employed for cavitation erosion
test purposes as well as the hydromechanics of flow. The previous formula
for the energy parameter has only been presented for completeness' sake.
Modification of the Formula of E

Taking into consideration the correlation AC0 (We) (see Fig. 1), and
substituting it into Eq 1 of the energy parameter, we come to the correla-
tion
where: k = const.
Assuming that d = h, we get the following result:
Assuming that on changing from one model to the other, and as the
value of the energy parameter is constant for the material subjected to
erosion test, we can write the following equation:
In the course of the cavitation damage investigations using one of the

models (or prototypes), the erosion intensity of the test specimen may be
expressed, using nondimensional quantities, with the intensity of the
other specimen:
In Eq 6, Y represents the scale number of velocities (Y = Vi/v2); A is

the scale number for the characteristic dimensions of the model (in
experiments conducted with cylinders, A = di/dz}; P is the fluid density
scale number (P = PI/PS); S is the scale number of fluid surface tension
(S = 0-1/0-2).
In the following considerations, the results of experiments made in
geometrically similar test sections, with water, on lead specimens, the
purpose was to verify the validity of the velocity and the geometrical
scale numbers. As the experiments were made within definite, narrow
temperature limits of the liquid, the changes in liquid density and surface
tension are disregarded. Similarly, the changes in liquid viscosity are also
neglected.
In an earlier investigation [75] for determining the frequencies of wakes
shedding from a circular cylinder, the correlation between the Strouhal
number, S, and the cavitation number, K, was obtained as
Since our experiments were made with constant cavitation number of

definite value (which means the similarity of the cavitation conditions
as well), this, according to Eq 7, meant also identical Strouhal numbers.
Velocity Scale Effect

In order to explain the velocity scale effect on the basis of the energy
parameter, values A3 = P2 = S~] = 1 should be substituted into Eq 6.
In this case
Equation 8 is adequately verified by experimental data obtained in 6 by

25 mm and 12 by 50 mm [16] as well as in 48 by 200 mm working sections
(Fig. 12). In the incubation period, exponent a will increase, up to the
critical point, to a maximum value of a = 5. Then, in the total destruction
period, it will remain constant [77].
The Scale Effect of Model Dimensions

For the determination of the scale effect of model dimensions, values
75 = p2 _ 2-1 = l should be substituted into Eq 6. In this case
This correlation was controlled by means of results obtained by experi-

FIG. 12—Dependence of exponent a on weight loss (AG).
FIG. 13—Variation of exponent /3 in function of the relative test period (T/TCT).
ments conducted with 24 and 48-mm-diameter cylinders, respectively,

tested at a flow velocity of v = 12 m/sec. Exponent /3 was calculated in
both experiments using the ratio AG^/AG^ , that is, on the basis of the
erosion weight loss per T/TCT time of the test specimens. The calculated
results on /3 are presented by the diagram /3(T/rcr) (see Fig. 13).
Contrary to the correlation a(A(j), Fig. 12, where the a-value is grad-
ually increasing to 5 by about its reaching the vicinity of the critical point,
the value of j8 is constant along the incubation period with 0 = 3. After-
wards, the /3-value increases to approximately 4. This comparison reveals
that, testing the joint effect of velocities and model dimensions on erosion
intensity, the critical weight-loss values shown by the erosion specimen
should be made use of.
The Scale Effect of Fluid Quality

Investigations of the scale effect of fluid quality might encounter many
difficulties, particularly if each fluid characteristic is to be tested sepa-
rately. It is extremely difficult to select noncorrosive fluids differing only
by one property and identical from all other aspects. Such experiments
have been conducted by Nowotny [18] and Rheingans [79]. However,
Rheingans employed a magnetostriction device for his experiments con-
ducted with wave cavitation, and there is nothing to verify the identity
where cavitation and other conditions are concerned. The experiments
by Hammitt [20] seemed more appropriate to compare theory and prac-
tice. His experimental results obtained by using stainless steel test speci-
mens in water and mercury at a flow velocity of v = 20 m/sec for cavita-
tion damage test purposes will be utilized in this paper. Assuming that
the two experiments have been conducted under identical cavitation
conditions from the experimental conditions employed by Hammitt, it
requires the substitution of the values Y5 = A3 = 1 into Eq 6.
In cavitation damage experiments, regardless of their method, if
various fluid types are employed, besides the parameters mentioned in the
foregoing, an important influence is exerted by the compressibility of the
liquid expressed by the compressibility coefficient, \[/, and also by the
liquid viscosity, which may be taken into consideration with the kinematic
viscosity of the liquid, v. In the case of experiments conducted with
different fluid types, Eq 6 may be modified in erosion intensity calcula-
tions as follows:
where:
As a result of calculating Eq 10, the relation of intensities, if t = 20 C,

is AKjf/AJV = 49, where AVM is the volume loss per unit time in mer-
cury, and AVW is the same in water, whereas assuming a fluid temperature
of t = 30 C the former amounts to kVM/kVw = 72. The experimental
results arrived at by Hammitt are illustrated by Fig. 14
where: iM and iw indicate the mean depth figures of specimen erosion in

mercury and water, respectively. Within the incubation period, the
&VM/&VW values calculated according to Eq 10 comprise the values de
rived from measurement data.
Conclusion
The energy parameter of cavitation damage defines the absolute re-
sistance of the material damage caused by cavitation. This was studied in
case of cavitation erosion produced behind circular cylinder models,
that is, of that brought about by vortex-type cavitation. By taking the
FIG. 14—Cavitation damage to stainless steel in water and mercury, respectively,

expressed by in and iw erosion depth figures, according to experiments by Hammitt, and
erosion depth ratio (iu/iw) in function of the relative test period (T/TCT).
energy parameter as a basis, scale effect in case of cavitation not only

behind cylinder models but other bodies as well may be analyzed. Deter-
mination of the energy parameter was performed primarily by means of
experiments conducted in water. Obviously, such a determination could
be effected for cavitation in other fluid types too. Consequently, further
experimental investigations on this problem, by using other fluids, appear
desirable. A precondition of the comparison of theoretical and practical
results is the application of a method providing such a method which
was discussed in the foregoing.
References
[/] H. Schroter, "Korrosion durch Kavitation in einem Diffusor," Zeit. VDI, Vol
76, No. 21,1932, p. 511.
[2] H. Schroter, "Werkstoffzerstorung bei Kavitation," Zeit. VDl, Vol 78, No. 11,
1934, p. 349.
[3] R. T. Knapp, "Recent Investigations of the Mechanics of Cavitation and
Cavitation Damage," Transactions, Am. Soc. Mechanical Engrs., Vol 77, No.
7, 1955, pp. 1045-1054.
[4] R. T. Knapp, "Accelerated Field Test of Cavitation Intensity," Paper No.
56-A-57, Am. Soc. Mechanical Engrs., 1956.
[5] S. L. Kerr and K. Rosenberg, "An Index of Cavitation Erosion by Means
of Radio Isotopes," Transactions, Am. Soc. Mechanical Engrs., Vol 80, No.
6, 1958, pp. 1308-1314.
[6] J. M. Rata, "Erosion de Cavitation. Mesure de 1'erosion par jauges resistantes,"
Recherche sur les Turbines Hydrauliques Symposium de Nice, September,
1960, C4, pp. 1-10.
[7] N. S. Govinda Rao and A. Thiruvengadam, "Prediction of Cavitation Dam-
age," Proceedings, Journal Hyd. Div., Am. Soc. Civil Engrs., Vol 87, Septem-
ber, 1961, pp. 37-62.
Vol 85, Series D, No. 3, September, 1963, pp. 347-349.
[9] K. K. Shalnev, "Experimental Study of the Intensity of Erosion Due to Cavita-
tion," Cavitation in Hydrodynamics, Proceedings NPL Symposium, Stationery
Office, London, 1956, 22/1-37.
[10] K. K. Shalnev, "Masshtabnii effect cavitazionnoi erozii," Zh. P. M. T. F., No.
4, 1962, pp. 121-128.
[11] J. J. Varga, B. A. Tchernavskii, and K. K. Shalnev, "O metode issledovania
masshtabnovo effecta cavitazionnoi erozii," Zh. P. M. T. F., No. 3, 1963, pp.
122-129.
[12] J. Varga, Gy. Sebestyen, K. K. Schalnew, and B. A. Tschernavskij, "Unter-
suchung des Massstabeffektes der Kavitationserosion," Acta Technica, Ac. Sci.
Hung., Vol 51, 1965, pp. 361-379.
[13] J. Ackeret and P. de Haller, "Uber die Zerstorung von Werkstoffen durch
Tropfenschlag und Kavitation," Schweiz Bauzeitung, Vol 108, No. 8, 1936,
pp. 105-116.
[14] K. K. Shalnev, "Uslovia intensivnosti cavitazionnoi erozii," /zv. ANSSSR
OTN, No. 1, 1956, pp. 2-20.
[15] J. Varga and Gy. Sebestyen, "Determination of the Frequencies of Wakes
Shedding from Circular Cylinders," A eta Technica, Ac. Sci. Hung., Vol 53,
1966, pp. 91-108.
[16] K. K. Shalnev, "Gidromechanitcheskoe aspecti cavitazionnoi erozii," /zv.
ANSSSR OTN, No. 1, 1958.
[77] J. Varga and Gy. Sebestyen, "Observations on Cavitation Velocity-Damage
Exponent in a Flowing System," Periodica Polytechnica Engineering, Vol 8,
No. 3, 1964, pp. 343-352.
[18] H. Nowotny, Werkstoffzerstorung durch Kavitation, VDI Verlag, Berlin,
1942.
[19] W. J. Rheingans, "Accelerated Cavitation Research," Transactions, Am. Soc.
Mechanical Engrs., Vol 72, No. 5, June, 1950, pp. 705-724.
[20] F. G. Hammitt, L. L. Barinka, M. J. Robinson, R. D. Pehlke, and C. A.
Siebert, "Initial Phases of Damage to Test Specimens in a Cavitating Venturi,"
Transactions, Am. Soc. Mechanical Engrs., Vol 87, No. 2, June, 1965, pp.
453-464.
DISCUSSION
F. G. Hammitt1 (written discussion)—The concept of a correlation

between cavitation damage rates and the power expended by the stream
in overcoming the drag imposed by the cavitation around an object in
the stream is interesting, but would seem to me to be applicable only in
very special cases such as that studied by the authors. It seems obvious
that in many flowing situations there could be no such relation since the
cavitation cloud may not collapse upon a structural member. A super-
cavitating propeller is an example of a flow situation where the authors'
concept could not apparently apply.
I am happy to note that our damage results with mercury and water
in a cavitating venturi are consistent with the author's theoretical expecta-
tions. However, since only this single set of data is available for com-
parison, the close agreement they report may be partly fortuitous.
I agree with the authors that the damage velocity exponent in flowing
cavitation systems does, indeed, in many cases depend upon degree of
damage. This effect was found to be especially important in recent rotat-
ing-disk tests2 with which I was associated.
In conclusion, I think the authors are to be sincerely thanked for pre-
senting this paper on their research which is not as well known in this
country as one would wish.
Olive G. Engels (written discussion)—It is informative to observe, as
indicated on graphs drawn by the authors, that erosion does occur to
some extent during the incubation period if the cavitation-producing de-
vice involves fluid flow. Only in the case where fluid flow is minimized,
as in cavitation produced with a magnetostriction oscillator, is the incuba-
tion period of zero erosion.
/. B. Marriott* (written discussion)—The previous discusser (Miss
Engel) asked whether an incubation period was found in magnetostriction
testing. In testing Haynes 6B by this method we have seen an incubation
period, the progress of damage being similar in form to that found in
1
Nuclear Engineering Dept., The University of Michigan, Ann Arbor, Mich.
2
G. M. Wood, L. K. Knudsen, and F. G. Hammitt, "Cavitation Damage Studies
With Rotating Disk," Paper 66-FE-ll, to be published in Journal of Basic En-
gineering, Transactions, Am. Soc. Mechanical Engrs.
3
Chemical physicist, Space Power and Propulsion Section, General Electric
Co., Evendale, Ohio.
4
Nr. Leicester, England.
DISCUSSION ON SCALE-EFFECT INVESTIGATION 237
water jet impact testing. This point has been referred to in the discussion
on erosion held at The Royal Society, London, in 1965.5
Messrs. Shalnev, Varga, and Sebestyen (authors)—We wish to express
our gratitude for the comments of the discussers. We agree with Pro-
fessor Hammitt that our experimental results cannot be made use of in
the case of supercavitation. However, we would like to emphasize that,
in course of our investigations, the cavitation types considered—par-
ticularly those in pumps and hydraulic turbines—are typical and cause
undesirable erosion damage. In the course of other investigations con-
ducted by the authors, the results of which have not been published as
yet, it was ascertained that there is an unequivocal correlation between
the cavitation conditions in pumps and the length of the cavitation zone
produced in the blade channels. On the basis of experiments conducted
with prototype machines, it was demonstrated earlier that the condition
of such separating, vortex-type cavitation depends only on the cavitation
index, and this was verified by model experiments as well. Naturally, if
the shedding cavitation does not contact a solid body, no cavitation
damage would be encountered.
The question whether the good agreement between the authors' theo-
retical conclusions and Professor Hammitt's experimental results is a
mere coincidence will be determined only by further investigations and
the subsequent refinement of the theory.
Miss Engel's assumption that there is an erosion weight loss in the
incubation zone if the cavitation-producing device involves fluid flow
and none in other cases (such as with a magnetostriction oscillator where
there is only a minimum flow present) seems very interesting, but the
authors have a different opinion.
Although many define the incubation zone as that where no erosion
weight loss can be observed, the authors believe that plastic deformation
and, in addition to the appearance of cracks, a slight weight loss will
take place in this zone. However, when testing high-strength materials or
small-size specimens with magnetostriction equipment, this observation
is extremely difficult or impossible to make. This is why the opinion was
developed that there is no weight loss in the incubation period. In rela-
tively large-size flow devices, where specimens of similarly large dimen-
sions are tested, and with low-strength materials (such as lead, aluminum,
or cadmium), on the other hand, these phenomena can be readily ob-
served. According to the experiences gathered by the authors, the ero-
sion weight loss per unit time is of a constant value in the incubation
period and has a linear character plotted as a function of time. The illus-
5
Philosophical Transactions, Vol A250, No. 1110, 1966, pp. 73-315.
trations in the papers by Heymann6 and Hobbs,7 partially support the au-
thors' opinion.
As far as the contribution of Mr. Marriott is concerned, the foregoing
statements must be referred to again with, however, the addition that
these apply to water-jet impact testing as well.
Finally, the authors wish to express their gratitude to Professor Ripken
for presenting then" paper at the symposium and to Mr. Heymann, chair-
man of the symposium, for his efforts exerted in order to make the pres-
entation of the paper possible.
9
See p. 70.
7
See p. 159.
R. Garcia,1 F, G. Hammitt? andR. E. Nystrom*
Correlation of Cavitation Damage with

Other Material and Fluid Properties
REFERENCE: R. Garcia, F. G. Hammitt, and R. E. Nystrom, "Correla-

tion of Cavitation Damage with Other Material and Fluid Properties," Ero-
sion by Cavitation or Impingement, ASTM STP 408, Am. Soc. Testing
Mats., 1967, p. 239.
ABSTRACT: Ultrasonic-induced cavitation studies were conducted in
lead-bismuth alloy at 500 and 1500 F, in mercury at 70 and 500 F, and in
water at 70 F for a wide variety of materials, including refractory alloys,
steels, brasses, copper, nickel, and plastics. Correlations of the cavitation
damage with applicable mechanical and fluid properties were carried out.
Some conclusions are:
1. Damage rates in mercury were 3 to 20 times greater than in water,
depending upon the material. Hence, clearly no correlating equation
which considers only the mechanical properties of the material can apply
for both fluids.
2. There is no single material mechanical property which can be used
to correlate the damage, even if coupling parameters to account for fluid
property changes are included in the correlation.
3. In general, the best correlations include energy-type mechanical
properties, strength-type properties, and fluid coupling parameters.
4. No relatively simple single correlating equation applies well to all
the data. This may indicate the insufficiency of the statistically determined
mechanical and fluid properties for the correlation of cavitation damage
which is known to be a highly transient process.
KEY WORDS: cavitation, ultrasonics, vibration, refractory materials,
stainless steels, erosion, water, mercury, lead-bismuth, liquid metals, high
temperature, mechanical properties, fluid properties, corrosion, pitting
Cavitation can be described as a hydrodynamic phenomenon which

relates to the formation and collapse of vapor bubbles in a liquid under
essentially adiabatic conditions. However, according to present theory,
the cavitation damage process is very closely related to damage from
1
Formerly at The University of Michigan, Ann Arbor, Mich., presently, senior
engineer, Aerojet-General Corp., Von Karman Center, SNAP-8 Div., Azusa, Calif.
2
Professor of nuclear engineering, Department of Nuclear Engineering, The
University of Michigan, Ann Arbor, Mich.
3
Doctoral applicant, Department of Nuclear Engineering, The University of
Michigan, Ann Arbor, Mich.
239
droplet or particle impingement or conventional erosion [I]-4'5 Thus,

the cavitation damage data are also, to some extent, applicable to the
resistance of the same materials to these other forms of attack, so that
the fields of droplet erosion in wet vapor streams (as in turbines or other
two-phase flow passages), rain erosion of high-speed aircraft, micro-
meteorite bombardment of space vehicles, and so on, are involved.
The successful pumping and handling of high-temperature liquid
metals, wherein cavitation itself is a problem, is of considerable im-
portance to the space program, particularly SNAP liquid-metal Rankine
FIG. 1—Block diagram of the high-temperature ultrasonic vibratory facility.
cycle power-conversion equipment. As has been recently demonstrated,

damaging cavitation attack can occur in bearings [2], close-clearance
passages [3], etc., as well as pumps [4,5].
Accelerated Cavitation Studies

In a prototype system, the damage due to cavitation appears usually
only after fairly lengthy operation under design conditions. Hence, a sys-
4
paper.
5
Reference [1] includes many papers on the relations between these various
forms of attack, including one by one of the present authors.
GARCIA ET AL ON CORRELATION OF CAVITATION DAMAGE 241
tematic study of this type, covering a variety of materials and plant condi-
tions, will involve considerable cost and time. An alternate approach,
sacrificing direct applicability to some extent in the interests of economy,
is to accelerate the cavitation losses by employing any one of several labo-
ratory techniques which have been developed for this purpose. For the
present investigation we have followed this course using a vibratory cavi-
TABLE 1—Specimen material-fluid-temperature combinations investigated.
Fluid
Material Water, Mercury, Mercury, ™;B^
500 and
70 F 70 F 500 F ™0OF
1100-0 Al (U-M)° X5
2024-T351 Al (U-M) X
6061-T651 Al (U-M) X
304 Stainless steel (U-M) X X X X
316 Stainless steel (U-M) X X X X
Hot-rolled carbon steel (U-M) X X X
T-lll (Ta-8W-2Hf) (P & W) X X X X
T-222 (Ta-9.5W-2.5Hf-0.05C) (P & W) X X
T-222(A) (P & W) X X
Mo-MTi (P & W) X X X X
Cb-lZr(P&W) X X X X
Cb-lZr(A) (P &W) X X X X
Plexiglas (U-M) X X
Cu (60% cold worked) (U-M) X
Cu (900 F anneal, 1 hr) (U-M) X
Cu (1500 F anneal, 1 hr) (U-M) X
Cu-Zn (60% cold worked) (U-M) X
Cu-Zn (850 F anneal, 1 hr) (U-M) X
Cu-Zn (1400 F anneal, 1 hr) (U-M) X
Cu-Ni (60% cold worked) (U-M) X
Cu-Ni (1300 F anneal, 1 hr) (U-M) X
Cu-Ni (1800 F anneal, 1 hr) (U-M) X
Ni (75% cold worked) (U-M) X
Ni (1100 F anneal, 1 hr) (U-M) X
Ni (1600 F anneal, 1 hr) (U-M) X
a
The notations (U-M) and (P & W) following the specimen materials indicate
the source of the material, namely, The University of Michigan and Pratt &
Whitney Aircraft (CANEL), respectively; whereas the notation (A) denotes an
annealed condition of the material.
6
X indicates test conducted for this specimen material-fluid-temperature com-
bination.
tation device. In our own laboratory we also employ a flowing venturi

system [6]. While the venturi is reasonably similar to actual flowing sys-
tems, damage occurs only rather slowly. Although it is our eventual
purpose to compare results from the vibratory and venturi systems, this
aspect of our overall program is not covered in the present paper.
In addition to the cavitation testing program, it is essential to determine
the applicable mechanical properties of the materials tested at the test
temperatures so that a correlation between resistance to this form of two-
phase attack and some combination of the mechanical properties can be

obtained, if, indeed, such a general correlation with the statically de-
termined mechanical properties exists. Applicable mechanical properties
include ultimate tensile strength, yield strength, hardness, strain energy
to failure, elongation, reduction in area, impact resistance, and so on. If
several fluids are involved, the correlation would be expected to include
FIG. 2—Standard cavitation test specimen.
terms which are functions of applicable fluid properties, such as density,

surface tension, net positive suction head, bulk modulus, kinematic vis-
cosity, and acoustic impedance ratio.
High-Temperature Ultrasonic Cavitation Vibratory Facility

The University of Michigan high-temperature ultrasonic cavitation
vibratory facility has been described elsewhere [6,7]. However, the major
features of the facility will be reviewed here. Figure 1 is a schematic show-
ing the arrangement of the audio oscillator, power amplifier, transducer-
horn assembly, test specimen, oscilloscope, frequency counter, high-

temperature furnace and cavitation vessel, and accelerometer. For studies
at elevated temperatures, the transducer-horn assembly is attached to the
special cavitation vessel which is filled with the appropriate fluid. The
accelerometer, oscilloscope, and frequency counter are used to monitor
the amplitude and frequency of vibration of the horn tip during the tests
FIG. 3—Special Plexiglas cavitation specimen and mounting stud.
[#]. Argon cover gas for the fluid is used under suitable pressure to main-
tain uniform suppression pressure for all tests.
The cavitation facility has been operated at fluid temperatures in excess
of 1500 F at a nominal frequency of 20 kc and double amplitude of 2
mils. It is capable of operation with a variety of fluids.
Present Investigation
To date, cavitation erosion data have been obtained in lead-bismuth
alloy (70 per cent lead-30 per cent bismuth) at 500 and 1500 F [9-22],
in mercury at 70 and 500 F [13,14], and in water at 70 F [13] for a variety

of materials, utilizing this facility. Although somewhat related data have
also been obtained in our venturi facility, this paper is concerned ex-
clusively with the studies conducted in the ultrasonic vibratory facility.
Table 1 summarizes the specimen material-fluid-temperature combina-
tions which were studied in this investigation, and the meaning of the
various symbols used in the description of these materials.
Experimental Procedure
Test Specimens
Standard test specimens (Fig. 2) were machined from bar stock for all
materials tested except Plexiglas, copper, copper-zinc, copper-nickel,
and nickel. The required height dimensions A and B (Fig. 2) are varied for
each material to provide a standard specimen weight (9.4 ± 0.1 g) which
is necessary for resonance. For stainless steel, A and B are 0.250 and
0.625 in., respectively.
Due to its very low density and brittle nature, it was impractical to
utilize standard test specimens of Plexiglas. The low density required
an unfeasibly large height, while the brittleness of the material made it
impossible to attach a specimen to the ultrasonic horn with adequate firm-
ness without damage to the thread. A tight attachment is necessary so that
the ultrasonic energy is efficiently transmitted across the interface. Hence,
a design consisting of a Plexiglas test specimen with internal threads and
a separate stainless steel mounting stud, which provides adequate mass,
was adopted and proved satisfactory (Fig. 3).
It was desired to test the identical heat treats of copper, copper-zinc,
copper-nickel and nickel in the vibratory facility, which had been previ-
ously tested in the venturi loop facility, so that a very direct comparison of
results would be possible. Since these materials had been procured only
in %6~m- sheet stock for the venturi specimens, it was necessary to design
a special specimen for the vibratory test, consisting of an adapter of a
suitable material (similar in shape to the standard specimen of Fig. 2) and
a disk of the desired material. It was necessary, then, to provide a suitably
firm attachment between disk and adapter. The adapters were fabricated
from brass bar stock, and the disk of the desired material was attached
using soft solder, taking care not to heat the test material to a temperature
which would significantly change its properties. The bond provided by
various epoxy resins and cements had previously been found inadequate.
While the acoustic impedance of the soft solder is similar to that of both
the brass adapter and disk materials, this was not the case for epoxy resins
and cements. The design adopted results in the desired specimen weight.
Test Conditions
Initially, each of the specimens was weighed on a precision electronic
balance to an accuracy of 0.01 mg, and then attached to the tip of the
stainless steel exponential horn, whereupon the unit was assembled.
The tests in lead-bismuth alloy at 500 and 1500 F and in mercury at
70 and 500 F were conducted in the 316 stainless steel cavitation vessel
previously mentioned. The investigations in water at 70 F were conducted
in a Plexiglas cavitation vessel whose dimensions were identical to those
of the 316 stainless steel container. The Plexiglas vessel permits visual ob-
servation of the bubble cloud and the condition of the specimen surface
during a test.
At elevated temperatures the test fluid was maintained at the required
temperature by a suitable controller, which allowed temperature variations
of less than 5 F.
The specimens were oscillated at 20 ± 0.002 kc with the exception of
the 1500 F tests where the resonant frequency was 18 ± 0.002 kc. The
submergence of the horn tip was held constant at 1 l/z ± l/s in. in all fluids,
while the double amplitude at the specimen was maintained at 2 ± 0.1
mils for all the tests, as determined by a precision accelerometer [8].
The argon cover gas pressure was adjusted to maintain constant static
pressure above vapor pressure (suppression pressure) for all fluids at
the specimen face. The lead-bismuth tests at 500 and 1500 F and the
mercury tests at 70 F were conducted at a slight overpressure (0.5 psig) to
prevent inward leakage of oxygen, and the corresponding suppression
pressure was used for the remainder of the tests. The mercury tests at 500
F then required an argon pressure of 2.4 psig, and the water tests at 70 F
an argon pressure of 1.1 psig because of the different densities and vapor
pressures of these fluids. While a constant suppression head, rather than
pressure, may have been desirable, the pressure capabilities of the equip-
ment were not adequate to allow this course to be pursued.
Total test duration varied for the different materials and was always
sufficient to obtain a good determination of damage rate which, neglecting
the very early portion of the test, was found to be essentially linear within
the total accumulated damage obtained. Test duration was thus a function
of the fluid and fluid temperature and, as shown in the figures, differed
widely for different materials. The tests were terminated when the com-
plete face of the specimen was damaged, and an approximately uniform
rate of damage established. Frequent inspections and weighings were made
during the tests to monitor the condition of the specimen surface and
establish the rate of weight loss. In the mercury tests prior to weighing,
it was necessary to remove any excess mercury adhering to the surface by
heating in a vacuum furnace, thus eliminating oxidation of the specimen.
In the lead-bismuth tests, all transfers of the specimens to and from the
high-temperature cavitation vessel were made at a fluid temperature of

500 F. At each examination, excess lead-bismuth adhering to the test
specimen was removed by quickly heating with a propane torch to a
temperature just above the melting point of the lead-bismuth alloy
(~350 F). The excess fluid was then separated from the test specimen
FIG. 4—Effect of cavitation test duration on MDP at 500 F in lead-bismuth

alloy.
by a very brief blast of compressed air. The process was rapid enough
so that only negligible oxidation occurred.
Heating time from 500 to 1500 F for the lead-bismuth tests is approxi-
mately 1 Vi> hr. Cooling time from 1500 to 500 F is approximately 5 hr.
Since the piezoelectric crystals must be maintained at a temperature
below 150 F (Curie point), the top plate of the cavitation vessel is water
cooled. A fan provided additional cooling.
Experimental Results
General
The cavitation damage data is shown as accumulative mean depth of
penetration (MDP) versus test duration. We believe the mean depth of
penetration, computed assuming that the weight loss is smeared uniformly
over the cavitated specimen surface, is more physically meaningful than
weight loss, since it is generally the total penetration of a particular com-
ponent by cavitation erosion that would render it unfit for service. Of
course, neither weight loss nor MDP is sensitive to damage distribution and
form, that is, damage may vary from isolated deep pits to relatively uni-
form wear, depending on material-fluid combination. Obviously, as a
"figure of merit," MDP at least takes into account the large variation in
density that may occur within a set of test materials.
TABLE 2—Summary of cavitation results in lead-bismuth at 500 F.
Material Average Weight- Average MDP
Loss Rate, mg/hr Rate, mils/hr
T-lll (P & W ) . . . 49.1 0.72

T-222(A) (P & W) 51.6 0.76
Mo-^Ti (P & W) 30.7 0.78
316 SS (U-M) 26.6 0.88
304 SS (U-M) 28.2 0.93
Cb-lZr (P & W) 55.1 1.63
Cb-lZr(A) (P & W) 119.5 3.54
Lead-Bismuth Alloy at 500 F [12]

Figure 4 shows accumulative MDP versus test duration for the seven
materials tested in lead-bismuth alloy at 500 F. The slopes of these curves,
representing the average damage rate in mils per hour are listed in Table 2
along with rates of weight loss. These data from Ref 12 are included here
for completeness.
On the basis of MDP, the alloy T-lll 6 exhibited the greatest resistance
to cavitation in this experiment, followed in order by T-222(A),6 Mo-J/2
Ti,6 304 stainless steel,7 and 316 stainless steel.7 The Cb-lZr6 and the
Cb-lZr(A)6 were considerably less resistant than the others so that the
Cb-lZr(A) test was concluded after only 8 hr of testing versus 12 hr for
the others. It is clear from Fig. 4 that the rate of erosion for each indi-
vidual material was approximately constant. These tests cover a range
up to about 12 mils MDP, so that within this range approximately con-
stant rates can be expected. Note that at 500 F the stainless steels fared
almost as well as the refractory alloys T-lll, T-222(A), and Mo-V^Ti.
6
Pratt & Whitney Aircraft.
7
The University of Michigan.
FIG. 5—Photographs of specimens subjected to cavitation damage in lead-

bismuth alloy at 500 and 1500 F.
GARCIA ET AL ON CORRELATION OF CAPTATION DAMAGE 249
FIG. 5—Continued
Figure 5 shows photographs of the test specimens before exposure and

at the conclusion of the cavitation experiment.
Detailed examination of the 303 stainless steel exponential horn, the
316 stainless steel container vessel, and the sides of the various test speci-
mens, all of which are not subject to cavitation, but are submerged in the
test fluid, indicate that corrosion effects in the absence of cavitation in
these investigations were negligible.
Lead-Bismuth Alloy at 1500 F [14,25]
The materials tested at 1500 F were identical to those tested at 500 F.
The data obtained at 1500 F are displayed in Fig. 6 in terms of accumu-
lative MDP versus test duration, while Table 3 shows the rates.
FIG. 6—Effect of cavitation test duration on MDP at 1500 F in lead-bismuth

alloy.
TABLE 3—Summary of cavitation results in lead-bismuth at 1500 F.

T-lll (P & W) 57.1 0.84

T-222(A) (P & W) 59.9 0.88
Mo-^Ti (P & W) 42.6 1.08
Cb-lZr (P & W) 70.0 2.07
316 SS (U-M) . 83.3 2.80
Cb-lZr(A) (P & W) 128.4 3.80
304 SS (U-M) 342.0 11.30
On the basis of MDP rates, the refractory alloy T-lll exhibited the
greatest resistance to cavitation damage in this experiment, as it did at
500 F. The T-222(A) and the Mo-V^Ti follow closely. The Cb-lZr and
the 316 stainless steel rated well behind the tantalum and molybdenum
alloys, while both the Cb-lZr(A) and the 304 stainless steel were con-
siderably less resistant than the others. Both were grossly damaged,
especially the 304 stainless steel, where the test was concluded after only
5 hr, whereas 10 hr were used for the others. Again the rate of erosion for
each individual material was approximately constant during most of the
test.
FIG. 7—Effect of temperature on cavitation resistance in lead-bismuth alloy.
As expected, the refractory alloys T-lll, T-222(A), Mo-VfcTi, and

Cb-lZr are far superior to the stainless steels with respect to resistance to
cavitation erosion at 1500 F.
Photographs of the test specimens before and after exposure at both
500 and 1500 F are presented in Fig. 5 for comparison.
Detailed examination of the exponential horn, container vessel, and
the sides of the various specimens again did not indicate any evidence of
corrosion.
TABLE 4—Summary of cavitation results in mercury at 500 F.

TV/raton-ai
Matenal
Average Weight- Average MDP
T-1H(P&W) 29.48 0.43

T-222(A) (P & W) 31.52 0.46
Carbon steel (U-M) 18.60 0.61
316SS(U-M) 19.01 0.63
304 SS (U-M) 20.83 0.69
Mo-MTi (P & W) 43.16 1.09
Cb-lZr(P&W) 81.85 2.43
Cb-lZr(A) (P & W) 125.78 3.73
FIG. 8—Effect of cavitation test duration on MDP at 500 F in mercury.

FIG. 9—Photographs of specimens subjected to cavitation damage in mercury

at 500 F.
Comparison of Lead-Bismuth Results at 500 and 1500 F

The tantalum alloys, T-lll and T-222(A), were the most resistant to
cavitation at both 500 F and at 1500 F. The Mo-V^Ti ranks third at both
temperatures. Note the poor relative performance of the stainless steels at
1500 F as compared with 500 F. This is, of course, to be expected from a
consideration of the effect of temperature upon their respective mechani-

cal properties. The differences in the amount of attack on the various
specimens and the effect of temperature can readily be seen in the photo-
graphs of Fig. 5.
For each material tested, the amount of damage sustained by the speci-
men at 1500 F was greater than that sustained at 500 F for constant testing
tune, as expected. The difference was, of course, much greater for the
stainless steels than for the refractories since their mechanical properties
are much more temperature dependent.
The effect of temperature on the damage rates is shown specifically in
Fig. 7. It is almost negligible on the T-l 11, T-222(A), and Mo-V&Ti, while
its effect on the stainless steels is quite dramatic, as evidenced by the slopes
of the appropriate curves.
TABLE 5—Summary of cavitation results in mercury at 70 F.

304 SS (U-M) 9.82 0.32

316 SS (U-M) 9.88 0.33
T-lll (P & W) 23.71 0.35
1-222 (P & W) 28.92 0.43
Mo-MTi (P & W) 22.58 0.57
Cb-lZr (P & W) 31.04 0.92
Cb-lZr(A) (P & W) 54.22 1.61
Plexiglas (U-M) 19.00 3.99
Mercury at 500 F
Table 4 summarizes the cavitation results obtained on the eight ma-
terials tested in mercury at 500 F. Figure 8 shows accumulative MDP
versus test duration.
On the basis of either weight-loss rate or MDP rate, T-l 11 is again the
most cavitation resistant of the materials tested, while the T-222(A) is
again next (about 7 per cent less resistant). The hot-rolled carbon steel,
316 stainless steel, and 304 stainless steel rank third, fourth, and fifth,
respectively. Three refractory materials, Mo-VgTi, Cb-lZr, and Cb-
lZr(A), were the least resistant, with the Cb-lZr(A) considerably the
worst. These three materials suffered damage ranging from three to eight
times that of the tantalum-base alloys. Again the rate of erosion for each
individual material is approximately constant during the test.
Photographs of the specimens at the conclusion of the test are shown
in Fig. 9. The materials are arranged in order of decreasing resistance to
cavitation damage. Note the severe pitting of the Mo-YzTi, Cb-lZr, and
Cb-lZr(A) surfaces. In all cases, the damage is relatively uniform over
the specimen face rather than in the form of individual, isolated, deep
pitting. It is felt that the approximately constant rate of erosion noted for
all the materials tested in mercury at 500 F is due to this uniform damage
pattern, and the fact that the area presented to the collapsing-bubble
cloud is approximately constant for the duration of the test. A similar
comment applies to the lead-bismuth results. A photograph of a 304 stain-
FIG. 10—Effect of cavitation test duration on MDP at 70 F in mercury.
less steel specimen before exposure is included in Fig. 9 and serves to

indicate a representative initial surface condition for all specimens.
Detailed examination of the exponential horn, container vessel, and the
sides of the various specimens again did not indicate any evidence of cor-
rosion.
FIG. 11—Photographs of specimens subjected to cavitation damage in mercury

at 70 F.
Mercury at 70 F
Table 5 summarizes the cavitation results obtained in mercury at 70 F,
while Fig. 10 is a plot of accumulative MDP versus test duration for the
nine materials tested.
The 304 stainless steel and 316 stainless steel were the most resistant
FIG. 12—Effect of temperature on cavitation resistance in mercury.
to cavitation at 70 F based on average MDP rate, differing by only 3 per

cent. The alloys T-lll and T-222 were 6 per cent and 30 per cent less
resistant than the stainless steels, respectively, while the Mo-VsTi was 80
per cent less resistant. The Cb-lZr, hot-rolled carbon steel, and Cb-lZr(A)
all were considerably less resistant, with MDP rates approximately three
to five times greater than for the stainless steels. The Plexiglas suffered the
greatest average MDP rate (about ten times that of the stainless steels).
Once again the rate of erosion for each individual material is approxi-
mately constant.
Figure 11 shows photographs of the test specimens at the conclusion
of the test. The materials are arranged hi order of decreasing resistance
to cavitation damage. Again the damage is relatively uniform over the
specimen face. A photograph of a 304 stainless steel specimen before
exposure is included to indicate initial surface condition.
No evidence of corrosion was noted on the test specimens.
Comparison of Mercury Results at 70 and 500 F

The most cavitation-resistant materials at 70 F were the stainless steels
with the tantalum-base alloys ranking third and fourth. At 500 F, the
TABLE 6—Summary of cavitation results in water at 70 F—Subsets 1 and 2.

T-222 (P & W) 1 .05 0.02

T-lll (P & W) 4.33 0.06
Mo-MTi (P & W) 3.49 0.09
316 SS (U-M) 2.81 0.09
304 SS (U-M) 3.04 0.10
Cb-lZr (P & W) 5.10 0.15
Cb-lZr(A) (P & W) 6.10 0.18
2024-T351 Al (U-M) 6.13 0.57
6061-T651 Al (U-M) 7.73 0.72
Plexiglas (U-M) 6.60 1.39
1100-0 Al (U-M) 28.90 2.70
superior mechanical properties of the tantalum-base alloys at this mod-

erately elevated temperature are already evident as the T-lll and
T-222(A) rank first and second, respectively. The hot-rolled carbon steel,
316 stainless steel, and 304 stainless steel rank third, fourth, and fifth,
respectively, at 500 F. The Mo-VzTi, Cb-lZr, and Cb-lZr(A) all main-
tained the same relative position at both test temperatures. The hot-rolled
carbon steel which had fared well at 500 F with a third ranking was dam-
aged almost 70 per cent more at 70 F. This apparently anomalous be-
havior is due to the fact that several of the mechanical properties of the
hot-rolled carbon steel such as tensile strength and yield strength are
actually significantly greater at 500 F than at 70 F. The reverse is true of
the stainless steels.
With the exception of the hot-rolled carbon steel, all of the materials
tested sustained greater damage at 500 F than at 70 F, as would be ex-
pected. The stainless steel damage rate was about doubled over this
moderate temperature range, and the T-lll increased about 23 per cent.
FIG. 13—Effect of cavitation test duration on MDP at 70 F in water—Subset 1.
The damage rates for the Mo-VSTi, Cb-lZr, and Cb-lZr(A) were increased
by a factor of two to three. The effect on the T-l 11 and T-222 is almost
negligible. Figure 12 shows these effects for the mercury tests.
Water at 70 F
General—The 24 materials tested in water at 70 F have been divided
into three subsets. The first consists of those materials that have also
been tested in mercury and lead-bismuth, namely 304 stainless steel, 316
stainless steel, T-l 11, T-222, Mo-VfcTi, hot-rolled carbon steel, Cb-lZr,
and Cb-lZr(A). The second subset consists of the three aluminum alloys
and Plexiglas, while the third includes twelve alloys and heat-treat combi-
nations of copper, copper-zinc, copper-nickel, and nickel. The second

and third subsets contain materials that have been tested only in water
(with the exception of Plexiglas which was also tested hi mercury at 70 F).
Subsets 1 and 2—Table 6 summarizes the damage data in water at
70 F for these materials. Figure 13 is a plot of accumulative MDP versus
FIG. 14—Effect of cavitation test duration on MDP at 70 F in water—Subset 2.
test duration for the eight materials contained in Subset 1. Figure 14 is the
corresponding plot for Subset 2.
On the basis of average MDP rate, 1-222 is the most cavitation resistant
of the materials contained in Subsets 1 and 2. Alloy T-111, ranking second,
suffered about three times more damage than T-222. Materials Mo-i^Ti,
FIG.15-Photographs of specimens subjected to cavitation damage in water

at 70 F-Subset 1
316 stainless steel, 304 stainless steel, Cb-lZr, Cb-lZr(A), and hot-rolled
carbon steel follow in that order. The aluminum alloys and Plexiglas were
the least resistant among these materials. Considering only the three
aluminum alloys, the 2024-T351 alloy was the most resistant, while the
very soft 1100-0 alloy sustained the greatest damage.
Figure 13 indicates that the rate of erosion for the T-222, T-lll,
Mo-YzTi, 316 stainless steel, and 304 stainless steel is approximately
constant over the test duration, while the rate of erosion for the Cb-lZr,
Cb-lZr(A), and the hot-rolled carbon steel is approximately constant dur-
ing the early stages of the test and then begins to decrease as the accumu-
lative weight loss and the accumulative MDP increase to larger values.
Examination of the specimens indicated that those materials with a con-
stant damage rate exhibit a fairly uniform and fine-structure damage pat-
tern, whereas those showing a nonlinear response are characterized by
surface damage consisting primarily of heavy, isolated, deep pitting. This
FIG. 16—Photographs of specimens subjected to cavitation damage in water

at 70 F—Subset 2.
latter pattern would result hi a greatly reduced bubble population, thus

producing greatly reduced damage rates [15], Hence, for this type of
pitting, one would expect the erosion rate to decrease as the total weight
loss or MDP increased, as clearly pointed out in a recent paper by Plesset
and Devine [15].
Figure 15 shows photographs of the test specimens in Subset 1 after
the test, arranged in order of decreasing resistance. Note the deep, isolated
pitting of the Cb-lZr, Cb-lZr(A), and hot-rolled carbon steel surfaces. A
photograph of a 304 stainless steel specimen before exposure is included
in Fig. 15 and serves to indicate a representative initial surface condition
for all the specimens tested.
TABLE 7—Summary of cavitation results in water at 70 F—Subset 3.
Cu, cold worked 32.83 0 95

Cu, 900 F anneal 35.37 1.02
Cu, 1500 F anneal 33 32 0 95
Cu-Ni, cold worked 24.18 0.70
Cu-Ni, 1300 F anneal 21 97 0 63
Cu-Ni, 1800 F anneal 16.25 0 47
Cu-Zn, cold worked 12.74 0 38
Cu-Zn, 850 F anneal 23.88 0.72
Cu-Zn, 1400 F anneal 22.78 0.68
Ni, cold worked 15.27 0.44
Ni, 1100 F anneal 20.25 0.58
Ni, 1600 F anneal 16.69 0.48
FIG. 17—Effect of cavitation test duration on MDP at 70 F in water—Subset

3 (copper and nickel).
263
Figure 14 shows that the rate of erosion for the three aluminum alloys
is approximately constant during the test, in spite of the deep, isolated
pitting of the type which in the previous materials corresponded to a
nonlinear damage rate. The explanation for this anomaly is not known at
present.
FIG. 18—Effect of cavitation test duration on MDP at 70 F in water—Subset

3 (copper zinc and copper nickel).
Photographs of the test specimens in Subset 2 at the conclusion of the

cavitation experiment are presented in Fig. 16. A photograph of a 2024-
T351 aluminum specimen before exposure is included for comparison.
Subset 3—Table 7 summarizes the cavitation results obtained in water
at 70 F for the materials hi Subset 3, namely, the twelve copper, copper-
zinc, copper-nickel, and nickel heat-treat and alloy combinations. The

various heat treats of a given material are grouped together. Figure 17
shows accumulative MDP versus test duration for the six copper and
nickel materials, while Fig. 18 is the corresponding plot for the six copper-
zinc and copper-nickel materials.
Among these, on the basis of average MDP rate the copper-zinc (60
FIG. 19—Photographs of specimens subjected to cavitation damage in water

at 70 F—Subset 3.
per cent cold worked) was the most resistant. Nickel (75 per cent cold
worked) ranked second, while the copper-nickel (1800 F anneal, 1 hr)
and nickel (1600 F anneal, 1 hr) were third and fourth. The three copper
heat treats were the least resistant to cavitation damage in Subset 3 with
the copper (900 F anneal, 1 hr) ranking last.
Considering only the three copper specimens, the cold-worked material
was most cavitation resistant while the high-temperature heat treat ranked
second and the low-temperature heat treat was third. Identical rankings
apply to the three copper-zinc specimens and the three nickel specimens.
For copper nickel, the high-temperature heat-treated specimen was the
most cavitation resistant followed by the low-temperature heat-treated

material and the cold-worked specimen in that order.
Figures 17 and 18 indicate that the erosion rate for the Subset 3 ma-
terials is generally not constant. This is believed due to the pattern of the
surface damage which for these materials is characterized by heavy, iso-
lated, deep pitting (Fig. 19). As previously discussed [75], such a condi-
tion would result in changes in flow geometry giving a greatly reduced
bubble population. This preferential damage of the surface may be caused
by nonumformity of applicable mechanical properties, resulting in a sub-
stantial interaction between mechanical and chemical effects through the
setting up of galvanic cells. This may be consistent with the fact that for
stainless steel a uniform pitting distribution is obtained in lead-bismuth
and mercury, while in water, deep, isolated pitting results.
The lack of similarity in damage pattern may also be partially due to the
fact that the net positive suction head (NPSH) has not been modeled be-
tween tests so that similarity of flow regime would not be expected.
In all of the tests conducted in water, only the hot-rolled carbon steel
specimen showed definite visual indications of corrosion. Hence, the re-
sults for this material are significantly influenced by both corrosion and
mechanical attack.
Comparison of Mercury and Water Results at 70 F
A comparison of the results obtained in mereury and water at the same
temperature (70 F) is useful to observe effects of fluid properties on cavi-
tation damage. This particular case affords an observation of the effect of
density variation in a vibratory test conducted under fixed static suppres-
sion pressure. The comparison is made on the basis of average MDP rate
(see Tables 5 and 6).
1. Several differences exist in the comparative ratings of materials be-
tween the two fluids. In mercury, the stainless steels and tantalum-base
alloys were the most resistant to cavitation damage and only differed in
this respect by about 25 per cent. In water, the tantalum-base alloys were
the most resistant, while the stainless steels suffered damage two to five
times greater than the T-lll and T-222. The Mo-^Ti ranked third in
water and fifth in mercury. However, the first five rankings are occupied
by the same materials in both fluids. The Cb-lZr, carbon steel, Cb-lZr(A),
and Plexiglas suffered the most damage and had comparable rankings in
both fluids except for the carbon steel which ranked seventh in mercury
and eighth hi water. This could have been due to the additional corrosion
suffered by the carbon steel in water.
2. Damage in mercury was 3 to 20 times greater than in water. The
stainless steels, in particular, suffered about three times as much damage
in mercury as in water. Note that the comparison is on the basis of equal
TABLE 8—Fluid properties data at various temperatures (taken from Refs [23] and [24]).
Fluid
Fluid Property
Water, 70 F Mercury, 70 F Mercury, 500 F Pb-Bi, 500 F Pb-Pi, 1500 F
6
Acoustic impedance (AI), lbm/ft2 sec 0,,299 X 10" 4. 03 X 10 3..85 X 106 3 .08 X 10" 2, 86 X 10"
Density (o) . g/cm3 . 1,,0 13,.55 12.,98 10 .38 9. 64
Surface tension (<r), dynes/cm 72.,8 465. 0 419.,0 397 .0 367. 0
Net positive suction head (NPSH) , f t 36.,5 2.,7 2.,8 3 .5 3.,8
Bulk modulus (B), psi 0..31 X 10« 4,,11 X 10« 3.,94 X 10« 3 .16 X 10« 2.,92 X 10«
Kinematic viscosity (»»), ft 2 /hr 0..039 0,,0044 0,,0030 0 .0064 0..0047
Specific heat (Cp), cal/g/deg C) 1,,00 0 .033 0.,032 0 .035 0,,035
Thermal conductivity (k) , cal/sec cm deg C . 1,.41 X 10~3 0,,021 0,,030 0 .025
Heat of vaporization (HV) , cal/g 585 69.,7 69.,7
Vapor pressure (Pv)> psi 0.,36 0 1,,93 0 0
TABLE 9—Mechanical properties data at 70 F from Pratt & Whitney Aircraft (CANEL).
Tensile Yield Engineering True Strain DPH Elongation, Area Elastic
Material Strength, psi Strength, Strain Energy, Energy, Hardness, Reduction, Modulus,
psi psi psi 1kg % % psi
304 SS 94 500 64 700 57 300 41 300 47 500 237 63.8 77.9 29..0 X 10«
316 SS 87 200 63 600 48 850 38 200 49 500 227 57.8 80.3 29..0
T-lll 131 600 124 900 16 750 16000 68 600 308 14.8 80.4 28. 0
T-222 154 200 154 200 15 250 16 050 70 350 338 10.6 55.6 28. 0
T-222(A) 108 900 91 100 23 950 22 180 52 350 288 23.1 61.1 28.,0
Mo-^Ti 165 800 150 400 21 300 14 570 11 600 295 9.3 7.9 45.,0
Cb-lZr 59 200 59 000 6 650 6 300 29 600 151 14.3 88.4 15,,0
Cb-lZr(A) 36 300 19 200 13 200 7 050 12 110 99 41.9 91.4 15,.0
static suppression pressures rather than equal NPSH.8 For a comparison

with equal NPSH, it is likely that the mercury damage would be propor-
tionately increased, since the static suppression pressure for mercury
would then be 13.6 times as great.
3. The stainless steels were the most cavitation-resistant materials in
mercury, while the tantalum-base alloys, T-lll and T-222, were far su-
perior in water. This may be due to the change in character of damage in-
curred by the stainless steel between the two fluids, as previously discussed.
The Plexiglas was the least resistant in both fluids, as opposed to the
venturi tests where it was quite resistant in water but poor in mercury.
This may indicate that materials which rely to some extent on a superior
yield deflection range for their protection (as rubberized coatings and also
Plexiglas in the present tests) are suitable in relatively low-intensity cavita-
tion fields, but fail under more intense attack. This observation is at least
consistent with much field experience.
Presumably, the primary cause of the greater damage suffered by all
the materials in mercury than in water is the much greater density of
mercury. The pressures generated by bubble collapse are theoretically
roughly proportional to fluid density, if the suppression heads seen by the
bubbles and the kinetic behavior of the horn were the same in both fluids.
This may be approximately the case in the present tests since the major
contribution to the suppression head at the start of bubble collapse is the
dynamic portion of the head caused by the horn motion. This portion
would be the same for all tests, although the static suppression heads differ
by the density ratio, since constant static suppression pressure was main-
tained.
Comparison of Mercury and Lead-Bismuth Results at 500 F

A further comparison of fluid effects can be obtained between the mer-
cury and lead-bismuth 500 F tests (see Tables 2 and 4). For this case, the
applicable properties of the fluids do not differ widely (Table 8). Again
the comparison is made on the basis of average MDP rate.
1. For both the lead-bismuth and the mercury tests, the materials in-
vestigated had almost identical comparative ratings (as opposed to the
water-mercury comparison). The only exception was Mo-^Ti (P & W)
which ranked third in lead-bismuth and fifth in mercury.
2. Any given material tested in both fluids suffered damage which was
of the same order of magnitude. This is not surprising considering the
similarity of the fluids.
3. The T-lll, T-222(A), 316 stainless steel, and 304 stainless steel
all suffered less damage in the mercury than in the lead-bismuth alloy,
8
The facility limitations are such that it is not possible to maintain a constant
NPSH (net positive suction head) at the horn tip.
TABLE 10—Mechanical properties data at 500 F from Pratt & Whitney Aircraft (CANEL).
Tensile Yield Engineering True Strain DPH Area Elastic
Material Strength, Strength, Strain Energy, Energy, Hardness, Elongation,
%
Reduction, Modulus,
psi psi psi psi l.lkg % psi
304 SS 92 500 56 700 16 150 18 200 37 200 154 30.8 72.9 26.0 X 106
316 SS 72 400 52 300 18 050 17 700 38 000 203 30.4 78.2 26.0
T-lll 101 800 100 800 15 100 10 700 50 900 218 13.8 86.2 27.0
T-222 133 800 133 800 12 850 12 900 67 800 286 10.9 71.5 27.0
T-222(A) 92 300 63 400 20 650 33 800 42 200 209 23.6 66.9 27.0
Mo-^Ti 84 100 79 700 10 700 11 000 44 400 207 15.0 75.9 43.0
Cb-lZr 54 700 54 700 6 450 5 185 27 700 133 12.7 88.7 14.5
Cb-lZr(A) 25 000 11 600 8 100 3 780 7 890 71 35.9 92.2 14.5
TABLE 11—Mechanical properties data at 70 F from University of Michigan Laboratories.

Tensile Yield Engineering True DPH Area Elastic
Material Strength, Strength, Strain Energy, Strain Energy, Hardness, Elongation, ^™ion, Modulus,
psi psi psi psi l.lkg %
% psi
1100-0 Al 12 250 7 600 4 950 4 320 22 600 27 44.5 85.5 10.0 X 106
2024 Al 72 000 57 900 13 300 13 600 31 050 171 20.0 34.5 10.0
6061 Al 45 300 40 000 25 800 9 840 38 620 127 19.4 56.7 10.0
Carbon steel, 70 F 45 300 41 600 18 440 30 530 111 000 193 46.3 76.1 29.0
Carbon steel, 500 F 62 510 18 400 19 225 20 900 66 650 125 37.2 63.6 28.0
Plexiglas 10 445 1 600 320 320 320 9 4.0 0.0 0.4
Cu 53 400 49 500 3 100 11 800 11 800 133 6.2 19.8 17.0
Cu 900 F 31 500 9 500 13 900 26 900 26 900 51 51.3 48.5 17.0
Cu 1500 F 30 700 5 000 6 100 11 800 11 800 41 32.5 33.2 17.0
Cu-Zn 93 900 82 000 4 700 55 400 55 400 197 5.3 40.7 16.0
Cu-Zn, 850 F 47 600 20 000 28 600 57 000 57 000 71 62.6 60.9 16.0
Cu-Zn, 1400 F 40 400 11 000 15 300 33 000 33 000 48 58.9 51.7 16.0
Cu-Ni 87 300 77 000 6 100 13 200 13 200 197 4.5 15.4 22.0
Cu-Ni, 1300 F 57 900 20 000 3 100 36 200 36 200 96 34.9 43.5 22.0
Cu-Ni, 1800 F 53 300 18 000 16 300 21 800 21 800 77 34.4 34.4 22.0
Ni 93 100 82 000 3 200 8 300 8 300 206 3.9 10.2 30.0
Ni HOOF 50 500 13 000 18 300 48 300 48 300 67 43.8 51.6 30.0
Ni 1600 F 48 700 7 000 16 100 40 500 40 500 59 41.8 49.7 30.0
while the Mo-V^Ti, Cb-lZr, and Cb-lZr(A) suffered more in mercury.

However, the differences were not great in any case.
Mechanical Properties Data

In order to obtain a meaningful correlation between the cavitation
resistance of the various materials tested, their mechanical properties, and
suitable fluid coupling parameters, it is absolutely essential that the appli-
cable mechanical properties such as tensile strength, yield strength, engi-
neering strain energy, true strain energy, hardness, elongation, reduction
in area, and elastic modulus be measured at the test temperatures using
tension bars machined from the same stock as were the cavitation speci-
mens. Otherwise the variations between material lots due to differences
in heat treat, cold work, etc., are too large to allow useful results. Accord-
ingly, for a given material all cavitation test specimens, tensile bars, and
special hot hardness specimens were machined from the same piece of bar
stock. In the case of the copper, copper-zinc, copper-nickel, and nickel
materials that were available only in sheet stock, flat tension specimens
were fabricated and tested.
The mechanical properties data for the stainless steels and refractory
materials that were available only in sheet stock, flat tension specimens
(CANEL). All the refractory materials tested in this program were sup-
plied by Pratt & Whitney Aircraft (CANEL), whereas the stainless steels
were supplied by this laboratory. Tbe results of the portion of the me-
chanical properties determination program conducted at Pratt & Whitney
were supplied to this laboratory by private communication [16] and are
listed in Tables 9 and 10.
The mechanical properties data for the aluminum alloys, carbon steel,
Plexiglas, and the copper, copper-zinc, copper-nickel, and nickel ma-
terials were determined at room temperature in the Department of Chemi-
cal and Metallurgical Engineering laboratories at the University of Michi-
gan and were reported earlier [17,18]. The mechanical properties for
carbon steel were also determined at 500 F. The data obtained from these
tests are listed in Table 11.
The mechanical properties data determined by Pratt & Whitney Air-
craft and the University of Michigan consist of values of tensile strength
(TS), yield strength (YS), engineering strain energy (ESE), true strain
energy (TSE), hardness (H), per cent elongation (ELON), per cent reduc-
tion in area (RA), and elastic modulus (E). Most of these have been sug-
gested at one time or another as possible correlating parameters for cavi-
tation damage. Most recently, both abroad [19] and in the United States
[20], strain-energy concepts have been advanced.
The "engineering strain energy" is based on the "approximate" or
engineering stress-strain curve and is equal to the area under this curve as
it is produced by a standard tensile machine [18]. Two values of "true

strain energy" were used in our correlations (and listed in Tables 9, 10,
and 11) and are based on approximations to the true stress-strain curve
[18]. The first value of true strain energy, denoted by TSEE, takes into
account elongation of the test specimen as a whole in computing the
strain as well as reduction of area in computing the breaking stress, while
the second value, denoted by TSER, takes into account the "necking" of
the specimen in computing the local strain in the failure region, and reduc-
tion in area to compute the local true breaking stress. The large discrepan-
cies that exist between these three strain-energy values for highly ductile
materials indicate the difficulties and uncertainties incurred in using energy
as a correlating parameter. If it occurs that the engineering strain energy
(ESE) proves to be a better correlating parameter than the true strain
energy based on reduction in area (TSER), then this may indicate that
brittle rather than ductile failures are typical of cavitation damage.
Correlations of Cavitation Data With Mechanical and Fluid Properties

Data
General
To investigate the dependence of cavitation resistance on mechanical
material properties and fluid properties, and to obtain a better understand-
ing of the damage mechanisms, it is desirable to subject the damage data
and the appropriate mechanical and fluid properties data to a least-mean-
squares-fit correlation procedure. For these studies, a quite sophisticated
least-mean-squares stepwise regression program [21,22] was utilized. For
the first-order interaction form of the program, the problem can be simply
stated: it is required to determine the appropriate coefficients and expo-
nents in a predicting equation of the form:
where: C0 , Ci, C2 , C3 , C4 , . . . , Cn are constant coefficients; a, b, c, d,

... , q are constant integer or reciprocal integer exponents; Xt are the in-
dependent variables, in this case the mechanical properties of the materials
and the fluid properties; and Y is the dependent variable, the average MDP
rate. The independent variables are allowed to appear in the predicting
equation any number of times, each time raised to a different value of
exponent and multiplied by an appropriate coefficient. The general pro-
gram allows great latitude in the possible exponents. However, that form
of the program used here allows any or all of the independent variables to
be raised to the following exponents: ±1, ±2, ±l/2,±3,±ys.
A predicting equation of the type of Eq 1 would be obtained by allow-
ing only a "first-order interaction" of the possible terms, that is, terms
involving products of the independent variables would not be allowed.

The general program, however, also allows the option of a "second-order
interaction."
In the present analysis, nine mechanical properties (independent varia-
bles) were considered, that is, TS, YS, ESE, TSEE, TSER, H, ELON, RA,
and E (symbols are previously explained). In addition, one fluid property
was included among the independent variables for each of the correlations
as a coupling parameter to attempt to account for differences jn cavitation
resistance of a given material in different fluids. The fluid coupling parame-
ters that were investigated included the ratio of acoustic impedances of
test fluid and specimen material (AI), density of fluid (p), surface tension
(<r), net positive suction head (NPSH), bulk modulus (B), and kinematic
viscosity (v). Hence, in a given correlation, there were a total of ten inde-
pendent variables, and ten possible exponents were allowed for each
independent variable. As a result, a total of 100 terms are possible candi-
dates for inclusion in the predicting equation, plus an additive constant.
From a physical point of view, it is desired that a good statistical correla-
tion be obtained with a minimum number of terms so that the predicting
equation may hopefully be justifiable on physical grounds. The interested
reader is referred to the literature previously cited for further details on the
program.
Lead-Bismuth Correlations
General—The lead-bismuth cavitation data obtained at 500 F and at
1500 F was submitted to the least-mean-squares regression program to
obtain a first-order interaction correlation applicable at both temperatures,
and, hence, having the greatest generality allowed by the limited data.
The nine mechanical properties (Tables 9, 10, and 11) and one fluid
property (Table 8) were taken to be the independent variables with the
average MDP rate being the dependent variable. The fluid coupling
parameter employed in this case was the ratio of the acoustic impedances
of the test fluid and specimen material.9 These properties were selected
since previous investigators had attempted correlations with them or be-
cause many of the properties had been involved in hypothesized damage
mechanisms.
Single-Property Correlations—As a first step in the analysis, an attempt
was made to correlate the damage data with each mechanical property
individually. True strain energy based either on the reduction in area
(TSER) or elongation (TSEE) was found quite successful as a single cor-
relating parameter for all of the lead-bismuth data, although, as will be
discussed, this was not the case for the other data subsets. The tensile
9
This is one of several quantities that have been chosen as coupling parameters
between the fluid and material, and is related to the ratio of reflected to transmitted
energy as liquid shock waves or jets impinge on the solid.
strength, hardness, and engineering strain energy, although having much

lower values of coefficient of determination,10 are also fairly successful in
this regard, while the other mechanical properties are much less so. It is
further noted that the average MDP rate is inversely proportional to vari-
ous powers of true strain energy, engineering strain energy, tensile strength,
or hardness in this analysis, as would be expected. The statistically best
predicting equation obtained from the single-property correlations was:
Average MDP rate = 0.233 + 2.57 X lO^TSER)-1
Coefficient of determination = 0.986 (2)
Average absolute per cent deviation11 = 0.1 per cent
Multiple-Property Correlations—Further attempts at complete cor-
relations of the experimental data were conducted in which all nine me-
chanical properties and one fluid property noted previously, each allowed
to be raised to ten separate exponents, were possible terms in the predict-
ing equation. The statistically best predicting equation then obtained was:
Average MDP rate = 0.713 + 3.12 X lO^TSER)-1
- 6.55 X lO^TS)-1 + 2.97 X 1(P(E)-3
Note that the true strain energy based on reduction in area (TSER), tensile
strength (TS), and elastic modulus (E) all enter the predicting equation in
an inverse manner, as was the case with the single-property correlations.
Also, the true strain energy based on reduction in area and the tensile
strength, which were both successful as single correlating parameters, are
prominent in the multiple-property correlating equation.
Mercury Correlations
Single-Property Correlations—True strain energy based on elongation
(TSEE) and hardness (H) are quite successful as single correlating parame-
ters for all of the mercury data. Tensile strength, yield strength, and elastic
modulus are considerably less successful. The statistically best single-
property predicting equation was:
Average MDP rate = 0.338 + 4.90 X 107(TSEE)-2
Note that a different form of strain energy is involved than that used in the
10
The coefficient of determination is a statistical quantity that can be interpreted
as the proportion of the total variation in the dependent variable that is explained
by the predicting equation. Its values range from 0 (no prediction) to 1.0 (perfect
prediction).
11
The average absolute per cent deviation is the average of the algebraic devia-
tions existing between individual experimental and predicted values of MPD rate.
best lead-bismuth single-property correlation, Eq 2, and that the exponent

is —2 rather than — 1.
Multiple-Property Correlations—Multiple correlations using all nine
mechanical properties and one fluid property were conducted. The sta-
tistically best predicting equation was:
Average MDP rate = -0.577 + 1.39 X 10n(TSEE)-;
+ 16.49(H)-
(5)
Note that the true strain energy based on elongation (TSEE) and the hard-
ness (H) enter the predicting equation in an inverse manner, as was the
case with the single-property correlations. However, the form of strain
energy involved differs from that in the lead-bismuth correlation, as does
the exponent on this term. These same properties were also successful as
single correlating parameters.
Water Correlations
Single-Property Correlations—Hardness, tensile strength, and yield
strength were most successful as correlating parameters among the ten
properties for the complete set of water data. However, only hardness was
reasonably successful from a statistical point of view. For this data set,
both forms of true strain energy were quite unsuccessful. The statistically
best predicting equation obtained in the single-property correlations was:
Average MDP rate = -6.023 + 1.30 X 104(H)-2 + 53.63(H)~1/a
- 6.17 X lOW-1 - 8.00 X 104(H)~3
Coefficient of determination = 0.946'' ^ ' (6)
The fact that the correlations for water were not as good as for the liquid
metals may indicate that other factors not considered hi this analysis, for
example, corrosion, may be more important for the water tests.
Multiple-Property Correlations—When all nine mechanical properties
and one fluid property are allowed to enter the predicting equation, the
statistically best predicting equation is:
Average MDP rate = -0.068 + 3.07 X lO^TS)-2
- 8.32 X 10-7(RA)3 - 2.03 X lO^H)-3
+ 1.49 X lO^TS)-1/2 (7)
Summary—Hardness, tensile strength, and yield strength adequately
predict the experimental water data on a single-property basis, either for
subset one, subsets two and three combined, or the full water data set to
which Eq 6 applies. In addition, the elastic modulus is successful as a
single correlating parameter for subset one. These same properties are
the most prominent in the ten-property water correlations. In fact, gen-
erally those properties most successful as single correlating parameters
in a given fluid are the most prominent in the ten-property correlations.
Whereas energy properties were quite important in the correlation of
the data from the tests with high-density liquid metals, they are almost
completely insignificant in the water tests. On the other hand, the strength
properties (including hardness in this category) are predominant in the
water tests.
Intuitive arguments can be advanced to show that a correlation would
be expected to involve both energy and strength terms. For example, if all
cavitation stresses were less than the fatigue limit of a very strong but
brittle material, it would not be damaged at all even if the strain energy
were zero. On the other hand, a highly ductile material with low strength
properties but high strain energy also might not suffer material removal
even though considerable surface distortion occurred. It would be ex-
pected that the first condition would apply more closely to the water tests
(relatively low-density fluid) than to the liquid-metal tests. Thus these
arguments are consistent with the present experimental data.
Fluid Coupling Parameters

Thus far, the only property included in the regression analysis that is
a function of the fluid has been the ratio of acoustic impedances of the
test fluid and specimen material. This quantity was not particularly suc-
cessful either for the lead-bismuth or mercury data, but was somewhat
better for the water data.
It would be desirable to obtain a predicting equation of high statistical
accuracy valid for all the lead-bismuth, mercury, and water data com-
bined. Since the dynamics of bubble growth and collapse are controlled
by the physical properties of the fluid, it is necessary to consider the varia-
tions in fluid properties for this case. Since the rankings of the materials
differed between these various test fluid conditions, it is clear that fluid
properties must be considered hi some way. This can be done by including
one or more suitable fluid coupling parameters among the mechanical
properties allowed to enter the predicting equation. The ratio of acoustic
impedances is one possible fluid coupling parameter. Other possibilities
are fluid density, surface tension, net positive suction head, compressibility
or bulk modulus, and kinematic viscosity. The values of each of these
properties for the fluid-temperature combinations investigated are listed
in Table 8.
Comprehensive Lead-Bismuth, Mercury, and Water Correlations

The combined lead-bismuth, mercury, and water data were submitted
to the regression program in an attempt to find a single correlation for all
the experimental data in terms of mechanical and fluid properties. The
mechanical properties considered were those previously used. In addition,

each of the six fluid coupling parameters discussed previously was com-
bined separately with the group of nine mechanical properties. Hence, six
comprehensive correlations were attempted, allowing a total of nine me-
chanical properties and one fluid property in each. It was hoped that such
a procedure would indicate those fluid properties that were most success-
ful as coupling parameters. However, it was found that all fluid properties
considered in this way were fairly successful as coupling parameters. In
retrospect, this is not surprising since all are smooth functions of each
other, and the power series type of correlating equation used has sufficient
flexibility to accommodate any one of them. For example, with density
as fluid coupling parameter, the statistically best predicting equation is:
Average MDP rate = 2.44 - \.62(p}~^ + 4.82 X lOÊ)-1/3'
+ 2.07 X K^CTSEE)-1 - 1.99 X 1013(E)-2
+ 4.26 X lO^TSER)-1 - 7.35 X 10-^H)1'2
- 5.48 X lO^TSER)-1/2 + 2.36 X lO^TS)"2 (8)
Three of the terms present in the equation involve the true strain energy
and two terms are functions of the elastic modulus. The tensile strength
and hardness are also present, each indicating that a decrease in damage
rate would be expected with increasing tensile strength or hardness. Even
though the density term is raised to a negative exponent, increasing density
still would result in increased damage, since the whole term has a negative
coefficient.
It will be recalled that strain energy, tensile strength, and hardness
were the most successful properties in the correlations of lead-bismuth,
mercury, and water separately. These mechanical properties along with
any of the fluid coupling parameters are also the most prominent in the
comprehensive lead-bismuth, mercury, and water correlations.
Summary and Conclusions

Ultrasonic-induced cavitation studies have been conducted in lead-
bismuth alloy at 500 and 1500 F, in mercury at 70 and 500 F, and in
water at 70 F for a wide variety of materials, including refractory alloys,
steels, brasses, copper, nickel, plastics, and so on. The detailed results
for the various fluid-material-temperature combinations are listed in the
appropriate sections of the report. Various salient features include the
following:
(a) The tantalum-base alloys, T-lll and T-222, were the most cavita-
tion resistant of the materials tested in lead-bismuth alloy at 500 and 1500
F, in mercury at 500 F, and in water at 70 F. The stainless steels were the
most resistant in mercury at 70 F.
(b) As would be expected, all materials tested in lead-bismuth sus-

tained greater damage at 1500 F than at 500 F. All materials tested in
mercury sustained greater damage at 500 F than at 70 F with the excep-
tion of the hot-rolled carbon steel. This can be explained by the superior
mechanical properties of this material at the higher temperature.
(c) Plexiglas was relatively less cavitation resistant in mercury than in
water, and in fact there are numerous substantial differences in rankings
between materials in different fluids at the same temperature. Thus,
clearly, any equation to predict cavitation damage must consider fluid
as well as material properties. These observations are consistent with
previous venturi test results from this laboratory.
(d) Detailed examination of the wetted but noncavitated parts of the
equipment and test specimens indicates that corrosion effects in the ab-
sence of cavitation were insignificant in these investigations.
(e) For these tests wherein the applied static pressure above vapor
pressure was held constant for the different tests, damage rates with mer-
cury were 3 to 20 times greater than for water. The damage noted in
lead-bismuth and mercury at 500 F was about the same.
(/) In general, the damage rate was constant throughout the tests for
the liquid metals and for most of the water investigations. However, in
some of the water tests, the damage rate decreased markedly during the
test. It is felt that this is a result of the very deep, isolated pitting encoun-
tered in these tests. Generally, water tests differed from the liquid-metal
tests in that quite uniform and relatively fine damage was encountered in
the liquid metal tests as opposed to the water tests on the same materials.
This may be a result of improper modeling of the fluid flow regime be-
tween the two fluids. The test durations were so adjusted that the total ac-
cumulative mean depth of penetration was always less than about 10 mils.
It is within the range between zero and 10 mils MDP (ignoring the very
early part of the test below perhaps and MDP of 1 mil) that the damage
rates were substantially constant.
(g) Direct comparison of venturi and vibratory results from this labo-
ratory shows that the ranking of materials for cavitation damage is quite
similar. The damage rates in the vibratory tests are of the order of 103
times those in the venturi, so that corrosive effects are much less impor-
tant in the vibratory tests.
Computer correlations of the cavitation damage data with applicable
mechanical and fluid properties indicate the following important con-
clusions:
(a) There is no single material mechanical property which can be used
to correlate the damage, even if coupling parameters to account for fluid
property changes are included in the correlation.
(b) In general, the best correlations include one or more energy-type
mechanical properties, one or more strength-type properties, and one or
more fluid coupling parameters if the fluid properties are varied in the
data set.
(c) The energy-type properties are more predominant in the tests with
the high-density liquid metals, while the strength-type properties pre-
dominate for the water tests. This is consistent with theoretical expecta-
tions.
(d) No relatively simple single correlating equation applies well to all
the data. If sufficient terms are allowed, of course, any degree of statistical
fit can be obtained. This lack of a single simple correlating equation may
indicate that all important mechanisms in cavitation damage have not
been considered. For example, it may not be possible to explain cavitation
damage in terms of properties which are determined under semistatic
conditions. Final conclusions in this regard must await the obtaining of
additional data and more comprehensive correlations.
A cknowledgment
Financial support for this investigation was provided by a grant from
the National Science Foundation. Mechanical properties data supplied
by Pratt & Whitney Aircraft (CANEL) and The University of Michigan
Department of Chemical and Metallurgical Engineering is also gratefully
acknowledged. Special thanks are also due C. A. Siebert, M. J. Robinson,
R. L. Crandall, and A. R. Schaedel, The University of Michigan, and
Henry Leeper and Glenn Wood, Pratt & Whitney Aircraft (CANEL),
for many helpful suggestions and continuing interest in this project.
References
[1] Royal Society Discussion on Deformation of Solids Due to Liquid Impact,
London, May 27, 1965, Philosophical Transactions, Royal Soc. of London, A,
Vol 260, 1966.
[2] O. Decker, "Cavitation Erosion Experience in Liquid Mercury Lubricated
Journal Bearings," First Annual Mercury Symposium, Atomics International,
Canoga Park, Calif., November, 1965, p. 14.
[3] A. A. Shoudy and R. J. Allis, "Materials Selection for Fast Reactor Applica-
tions," Proceedings, Michigan ANS Fast Reactor Topical Meeting, Detroit,
Mich., April, 1965.
[4] G. M. Wood, R. S. Kulp, and J. V. Altieri, "Cavitation Damage Investigations
in Mixed-Flow Liquid Metal Pumps," Cavitation in Fluid Machinery, Am. Soc.
Mechanical Engrs., November, 1965, pp. 196-214.
[5] P. G. Smith, J. H. DeVan, and A. G. Grindell, "Cavitation Damage to Cen-
trifugal Pump Impellers During Operation with Liquid Metals and Molten Salt
at 1050-1400°F," Journal of Basic Engineering, Transactions, Am. Soc. Me-
chanical Engrs., September, 1963, pp. 329-337.
[6] F. G. Hammitt, "Cavitation Damage and Performance Research Facilities,"
Symposium on Cavitation Research Facilities and Techniques, Am. Soc. Me-
chanical Engrs., Fluids Engineering Division, May, 1964, pp. 175-184. See
also, ORA Technical Report 03424-12-T, Department of Nuclear Engineering,
The University of Michigan, Ann Arbor, Mich., November, 1963.
[7] R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation Studies,"
ORA Technical Report 05031-1-T, Department of Nuclear Engineering, the
University of Michigan, Ann Arbor, Mich., October, 1964.
[8] R. Garcia and F. G. Hammitt, "Amplitude Determination of an Ultrasonic

Transducer By Means of an Accelerometer Assembly," ORA Internal Report
05031-7-1, Department of Nuclear Engineering, The University of Michigan,
Ann Arbor, Mich., December, 1965.
[9] R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation in Liquid
Metals at 1500°F," Internal Report 05031-1-1, Department of Nuclear Engi-
neering, The University of Michigan, Ann Arbor, Mich., February, 1965; also
Transactions, Nuclear Soc., Vol 8, No. 1, June, 1965, pp. 18-19.
[10] R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation in Liquid
Metals at 500°F," Internal Report 05031-3-1, Department of Nuclear Engineer-
ing, The University of Michigan, Ann Arbor, Mich., April, 1965.
[11] R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation Studies in
Lead-Bismuth Alloy at Elevated Temperatures," ORA Technical Report
05031-2-T, Department of Nuclear Engineering, The University of Michigan,
Ann Arbor, Mich., June, 1965.
[12] R. Garcia and F. G. Hammitt, "Ultrasonic-Induced Cavitation Studies in
Lead-Bismuth Alloy at Elevated Temperatures," Corrosion, Vol 22, No. 6,
June, 1966, p. 157.
[13] R. Garcia, R. E. Nystrom, and F. G. Hammitt, "Ultrasonic-Induced Cavita-
tion Studies in Mercury and Water," ORA Technical Report 05031-3-T, De-
partment of Nuclear Engineering, The University of Michigan, Ann Arbor,
Mich., December, 1965.
[14] R. E. Nystrom, "Ultrasonic-Induced Cavitation Study in Mercury at 70°F,"
M.S.E. thesis, Department of Nuclear Engineering, The University of Michi-
gan, Ann Arbor, Mich., October, 1965. See also, "Ultrasonic-Induced Cavita-
tion Study in Mercury at 70°F," ORA Internal Report 05031-6-1, Department
of Nuclear Engineering, The University of Michigan, Ann Arbor, Mich.,
October, 1965.
[75] M. S. Plesset and R. E. Devine, "Effect of Exposure Time on Cavitation
Damage," Paper 65-WA/FE-23, Am. Soc. Mechanical Engrs.; to be published
in Journal of Basic Engineering, Transactions, Am. Soc. Mechanical Engrs.
[16] Personal communication from Henry P. Leeper, Project Metallurgist, Pratt
& Whitney Aircraft (CANEL), to F. G. Hammitt, February 26, 1965 and May
13, 1965.
[17] M. J. Robinson, "On the Detailed Flow Structure and the Corresponding
Damage to Test Specimens in a Cavitating Venturi," Ph.D. thesis and ORA
Technical Report 03424-16-T, Department of Nuclear Engineering, The Uni-
versity of Michigan, Ann Arbor, Mich., August, 1965.
[18] C. A. Harrison, M. J. Robinson, C. A. Siebert, F. G. Hammitt, and J.
Lawrence, "Complete Mechanical Properties Specifications for Materials as
Used in Venturi Cavitation Damage Tests," ORA internal Report 03424-29-1,
Department of Nuclear Engineering, The University of Michigan, Ann Arbor,
Mich., August, 1965.
[19] K. K. Shalnev, 'The Energetics Parameter and Scale Effect in Cavitation
Erosion," Akademiya nauk SSSR, Izvestiya Otdeleniye teknicheskikh nauk.
Mekhanika i machinomostroyeniye, No. 5, 1961, pp. 3-10.
]20] A. Thiruvengadam, "A Unified Theory of Cavitation Damage," Journal of
Basic Engineering, Transactions, Am. Soc. Mechanical Engrs., Vol 85, 1963,
pp. 365-376.
[21] F. H. Westervelt, "Automatic System Simulation Programming," Ph.D.
thesis, College of Engineering, The University of Michigan, Ann Arbor,
Mich., November, 1960.
[22] R. L. Crandall, 'The Mathematical and Logical Procedure of the Stepwise
Regression Program with Learning," University of Michigan Computing
Center Internal Report, 1965.
[23] Liquid Metals Handbook, R. N. Lyon, editor-in-chief, 2nd edition, June,
1952.
[24] International Critical Tables, compiled by C. J. West and Callie Hull, Mc-
Graw-Hill Book Company, Inc., New York, 1933.
DISCUSSION
/. Z. Lichtman1' 2 (written discussion)—The authors have, conducted a

investigation of the cavitation erosion resistance of metallic and plastic
materials in liquid metals and in water, and of relationships between the
erosion resistance and quasi-static mechanical properties of these mate-
rials. This discussion will touch on two topics noted in the paper: (a)
corrosion effects and (b) comparisons between Plexiglas and rubber coat-
ings.
Detailed examination of the wetted metal parts not exposed to cavita-
tion indicated that corrosion effects in the absence of cavitation were
insignificant. However, no comment was made concerning the possibility
that corrosion may have contributed to the erosion of the specimen sur-
faces exposed to cavitation. In view of the severe localized erosion
stresses acting on the specimen faces as shown by the localized pits, and
the resultant possibility of occurrence of stress corrosion, the contribu-
tion of corrosion may be significant and should be determined. High-
frequency fatigue tests such as described by Thiruvengadam et al3 may
be applicable to this study, and this study should be coordinated with
ASTM Committee G-l on Corrosion of Metals.
The low erosion resistance of Plexiglas in the magnetostriction tests
as compared to its high resistance in the low-intensity venturi tests was
considered to be dependent on its yield deflection range. The authors
stated that rubberized (rubber?) coatings have a similar yield deflection
range and also fail under high erosion intensities. Many elastomeric
materials have shown very high erosion resistance in high-intensity rotat-
ing-disk tests,4 while Plexiglas showed very low erosion resistance in
such tests. Although the yield strength of Plexiglas (1600 psi in Table
11) is in the range of ultimate tensile strengths of many elastomeric mate-
rials, the ultimate elongation values differ drastically (4 per cent and
several hundred per cent, respectively). Field experience4 with elasto-
meric materials selected on the basis of their high erosion resistance in
the laboratory have shown performance deficiencies (separation and
erosion) because of application problems.
1
U. S. Naval Applied Science Laboratory, Brooklyn, N. Y.
2
The opinions or assertions contained in this discussion are the private ones of
the discusser and are not to be construed as official or reflecting the views of the
Naval Services at large.
3
A. Thiruvengadam, H. S. Preiser, and S. L. Rudy, "Cavitation Damage in
Liquid Metals," Report TPR 467-3, Hydronautics, Inc., June 30, 1965.
* Cavitation Damage Design Handbook, AD 460-524, NASL Project 9300-17,
Final Report, Sept. 30,1964.
DISCUSSION ON CORRELATION OF CAVITATION DAMAGE 281
R. Garcia, F. G. Hammitt, and R. E. Nystrom (authors)—There is

a definite possibility that corrosion may have contributed to the total
damage exhibited by the various specimens tested. As mentioned pre-
viously, detailed examination of the wetted surfaces not exposed to
cavitation indicated that corrosion effects in these areas were negligible.
In addition, however, photomicrographs of several of the specimen sur-
faces (316 stainless steel, T-lll, and Cb-lZr (A)) subjected to cavita-
tion damage in lead-bismuth alloy at 1500 F were obtained and indicated
no trace of corrosion. It is still possible that the cleansing action of the
cavitation field rapidly removed any corrosion products from the speci-
men surface as they were formed. Solution corrosion is another possi-
bility. Additional tests are planned in an effort to determine the relative
contribution of cavitation erosion and corrosion to the total damage
mechanism. Plesset5 has suggested a pulsed-type experiment for this
determination.
It is apparently true that many elastomeric coatings exhibit ultimate
elongation values much greater than that of Plexiglas. In another study
we subjected Teflon-coated type 304 stainless steel to cavitation in our
ultrasonic rig in mercury at 350 F. The Teflon apparently offered no
protection in this high-intensity cavitation field. As suggested by Licht-
man, there may have been an application problem here also.
Allen Smith6 (written discussion)—Can the ring patterns on some of
the eroded specimen surfaces (Fig. 5) be explained by the outer edge
bending under the vibration, or are they caused by the radial flow of the
fluid across the surface?
Messrs. Garcia, Hammitt, and Nystrom—The origin of the ring
patterns on some of the eroded specimen surfaces (Fig. 5) have been the
subject of much speculation. In some cases (Fig. 15, for example) this
outer undamaged ring is much wider than that shown in Fig. 5 and
appears to be a function of the test fluid, becoming larger as the fluid
density is decreased. We feel that the existence of these undamaged
rings is due to so-called "edge effects," as described by Nyborg and
Jackson,7- 8 which result in vortex action near the outer edge of the
specimen and a corresponding region of high pressure and decreased
bubble population.
B
M. S. Plesset, "The Pulsation Method for Generating Cavitation Damage,"
Journal Basic Engineering, Transactions, Am. Soc. Mechanical Engrs., Vol 85,
Series D, No. 3, 1963, pp. 360-364. See also "Pulsing Technique for Studying
Cavitation Erosion of Metals," Corrosion, Vol 18, No. 5, May, 1962, pp. 181-
188.
6
C. A. Parsons & Co. Ltd., Heaton Works, Newcastle upon Tyne, England.
7
F. J. Jackson, and W. L. Nyborg, "Sonically-Induced Microstreaming Near
a Plane Boundary. I. The Sonic Generator and Associated Acoustic Field," Journal,
Acoustical Society of Am., Vol 32, No. 10, October, 1960, pp. 1243-1250.
8
F. J. Jackson, "Sonically-Induced Microstreaming Near a Plane Boundary. II.
Acoustic Streaming Field," Journal, Acoustical Society of Am., Vol 32, No. 11,
November, 1960, pp. 1387-1395.
/. M. Hobbs9 (written discussion)—The authors are to be commended

on presenting such a comprehensive amount of cavitation erosion data for
a variety of liquid and metal combinations. Their correlation of data by
the method of least squares has led to some very informative conclusions.
Further developments in this work will be awaited with great interest,
and it is hoped that the gap between a statistical fit to the data and a
physical explanation of the behavior may soon be bridged.
From an examination of the tabulated data and the predicting equa-
tions presented in the paper, and also drawing on some personal ex-
perience, three points emerge:
1 Some properties have little or no apparent significance for example,
yield strength, elongation or reduction in area.
2 Some properties are quite well related, for example, tensile strength
and hardness.
3 Some properties are derived from others, for example, engineering
strain energy.
It therefore follows that some reduction in the number of variables
used in the analysis might be possible in the future. However, allowance
would have to be made for second-order interaction between variables in
the predicting equation to compensate for this. Similarly, the number of
fluid properties used might also be reduced, as it is stated that "all are
smooth functions of each other."
This simplification is desirable not only on the grounds of physical
justification but also on those of engineering application. For the latter
purpose, a simple, approximate, but widely applicable, empirical formula
for determining relative erosion rates from mechanical properties would
be more suitable than an exact, but complicated, multiterm power series
having a limited field of application. Although, as concluded by the
author, "it might not be possible to explain cavitation damage hi terms of
properties which are determined under semistatic conditions," there is
promise of some guidance for the designer. If it should be found that
cavitation erosion data correlates with some type of dynamic material
test, this test may offer little or no advantage over the accepted erosion
tests now in use.
Messrs. Garcia, Hammitt, and Nystrom—We fully agree in all respects
with the various points raised by Hobbs. Some of the mechanical
properties we have considered in our computer correlations have little
'Properties of Fluids Div., National Engineering Laboratory, East Kilbride,

Glasgow, Scotland.
10
R. Garcia, "Comprehensive Cavitation Damage Data for Water and Various
Liquid Metals Including Correlations with Material and Fluid Properties," Ph.D.
thesis and ORA Technical Report No. 05031-6-T, Department of Nuclear Engi-
neering, The University of Michigan, Ann Arbor, Mich., August, 1966.
DISCUSSION ON CORRELATION OF CAVITATION DAMAGE 283
or no apparent significance, and in our more recent work10' n these

properties have not been considered. As a result we have been able to
correlate a large body of cavitation data involving many materials and
several fluid-temperature combinations with the most significant mechani-
cal and fluid properties. This has led to simple correlating (predicting)
equations involving only one mechanical and one fluid property. The
most significant equation of this type is as follows:10' u
Average MDP Rate = d-C2[0.142 + 8.918CUR)-1/2]
Here Ci is a correction for variation in static NPSH and can be expressed
as: Ci = 1.29(NPSH)-°-266. C2 is a correction for the "thermodynamic
effect".10' n UR denotes the "ultimate resilience," first suggested by
Hobbs.12 While the statistical precision of the above equation is not as
good as might be hoped, it is recommended by its simplicity as an ap-
proximate empirical formula for determining relative erosion rates by
the designer of equipment susceptible to cavitation damage.
(Authors' reply to discussion by R. W. Wilson, see p. 217)—Proce-
dures suggested by Plesset5 and Thiruvengadam et al3 can be employed
to determine the corrosion component of the total damage suffered by a
material in a short-term cavitation test such as those conducted in our
laboratory. Future tests include such a determination.
We feel that the various damage patterns observed are a function of
the static net positive suction head. Hence, in our studies in heavy liquid
metals such as mercury and lead-bismuth alloy (large density and small
NPSH), the damage pattern covers essentially the complete face of the
specimen, whereas in water and lithium10 (small density and large NPSH)
the damage pattern is quite irregular. As the NPSH increases, the damage
tends to concentrate at the center of the specimen.
The point raised by Mr. Wilson concerning the acoustic impedance is
well taken, and, in fact, our computer correlations have involved the
ratio of acoustic impedances of fluid and test specimen material. How-
ever, it has not proven to be an effective correlating parameter in our
vibratory tests, whereas it was important in our venturi tests.
11
R. Garcia and F. G. Hammitt, "Cavitation Damage and Correlations with
Material and Fluid Properties," ORA Internal Report No. 05031-11-1, Department
of Nuclear Engineering, The University of Michigan, Ann Arbor, Mich., September,
1966.
13
See p. 159.
This page intentionally left blank
TfflS PUBLICATION is one of many
issued by the American Society for Testing and Materials
in connection with its work of promoting knowledge
of the properties of materials and developing standard
specifications and tests for materials. Much of the data
result from the voluntary contributions of many of the
country's leading technical authorities from industry,
scientific agencies, and government.
Over the years the Society has published many tech-
nical symposiums, reports, and special books. These may
consist of a series of technical papers, reports by the
ASTM technical committees, or compilations of data
developed in special Society groups with many organiza-
tions cooperating. A list of ASTM publications and
information on the work of the Society will be furnished
on request.

ASTM - STP 408 - Erosion by Cavitation or Impingement

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ASTM - STP 408 - Erosion by Cavitation or Impingement

Uploaded by

Copyright:

Available Formats

EROSION BY CAVITATION

ASTM SPECIAL TECHNICAL PUBLICATION NO. 408

List price $20.00; 30 per cent discount to members

The Society is not responsible, as a body,

Printed in Baltimore, Md.

The papers in the Symposium on Erosion by Cavitation or Impringe-

Erosion and Captation, STP 307 (1961), $4.00

ASTM is interested in the properties of materials and the ways of

A Test Rig for Studying Impingement and

REFERENCE: J. F. Ripken, "A Test Rig for Studying Impingement and

For nearly forty years various investigators have attempted to develop

solids at stationary drops of liquid. For hydraulic devices and ship

The foregoing methods were not considered directly applicable to

FIG. 1—The rotor assembly.

FIG. 2—A mounted test specimen at the rotor tip.

The resulting experimental equipment described in this paper produces

The aluminum-alloy rotor as shown in Fig. 1 consists of two central

FIG. 3—The test rig—housing chamber closed.

is infinitely adjustable down to the minimum values of interest of about

FIG. 4—The test rig—housing chamber open.

represents the limit of the available vacuum system. A lower value is

(a) Velocity = 500 ft/sec, N about 145,000 cycles.

General Radio strobotac which is synchronized with the motion of the

FIG. 7—Impact damage, Type 430 stainless steel annealed.

larger outer end of this hole as shown in Fig. Ib is approximately the

FIG. 8—Damage-rate values for 304 and 430 stainless steels.

mens are not included but have a considerable resemblance to those of

Zone 1. An incubation or no-loss zone. In cavitation tests with a

FIG. 9—A comparison of impact failure with failure by a standard fatigue

test data may perhaps be fortuitous but is nevertheless encouraging in

[9] J. A. Brunton, "Deformation of Solids by Impact of Liquids at High Speeds,"

W. S. Owen1 (written discussion)—It is clear that the strain rate im-

be a very useful research facility, but it will need further refinement if it

that obtained using resonant-cavity erosion test devices. In these tests,

The Concept of Erosion Strength

REFERENCE: A. Thiruvengadam, "The Concept of Erosion Strength,"

Em = Young's modulus of material

FIG. 1—Definition sketch for the material response to erosive forces.

zm = Acoustic impedance of material

The word "erosion" from the Greek "erodere"—to gnaw away—

ment of a liquid on a solid surface or of a solid on another solid surface.

where a is of the order of unity at high velocities; the factor Y% is due to

Em = Young's modulus of material

Experiments by Brunton [6,8] show that the measured values of the

where YD is the dynamic yielcfWength of the material. However, in

where <je is the endurance limit.

Since (from [75])

if we assume the Poisson's ratio, v, as nearly Vs for most metals. Term

The threshold criterion would become

From Eqs 5 and 9,

where KI is the bulk modulus of the liquid.

For water, ^ = 300,000 psi, and for steel, Em = 30 X 106 psi.

Necessity for the Concept of Erosion Strength

is introduced. Just as yield strength, ultimate strength, fatigue strength,

FIG. 2—Correlation between strain energy and reciprocal of rate of volume

If we can devise a method by which we can accurately determine the

Determination of Erosion Strength

NaCl up to 9 per cent, and it is the same as in distilled water. If we use

TABLE 2—Erosion strength of 1020 steel in NaCl solution.°

FIG. 4—High-frequency corrosion fatigue of SAE 1020 steel (Ref 18).

[12] S. M. DeCarso and R. E. Kothman, "Erosion by Liquid Impact," Erosion

W. C. Leith1 (written discussion)—The erosion strength postulated in

in writing a general threshold criterion, the pressure exerted by a liquid-

FIG. 5—Strain energy related to reciprocal of rate of volume loss.