You are on page 1of 9

WEAR

ELSEVIER Wear 199 (1996) 36-44

Measurement of erosion due to particle impingement and numerical


prediction of wear in pump casing
Yuan Zhong, Kiyoshi Minemura
Nagoya University, Furo-cho,Chikusa-ku,Nagoya 464.01, Japan
Received17 August1995;accepted7 March 1996

Abstract

Erosion damage of wear-resistant materials due m sand particle impingement is measured and correlated based on Bitter's erosion model
to clarify the effects of particle impinging velocity and angle, particle size and concentration on the wear. Using the empirical formula for the
correlation and calculating impinging velocities of sand particles on a casing wall of a pump, successive erosion of the wall is numerically
calculated to demonstrate the viability of the prediction method proposed in our previous study.

Keywords: Erosiondamage;Particleimpingement;Numericalprediction;Pumpcasing;Trajectory;Turbomachinery

1. I n t r o d u c t i o n wear is thought to be caused by direct impingements, random


impingements,and friction of the sliding bed. When applying
Turbomachines are increasingly being applied to extensive all these models, however, dependence of the wear on elastic
industrial fields and tend to be operated at a higher rotational and plastic characteristics of the particle and wall materials,
speed and under multiphase flow conditions. When the called erosion coefficients, must be determined in advance
machines transport fluid containing solid particles, their ero- by corresponding experiments. The existing coefficients,
sion becomes serious. Therefore, to maintain high perform- however, are limited either to soft and ductile metals (copper,
ance over a long time, it is desirable to establish a definite aluminum) or to hard and fragile materials (glass, annealed
design method by paying attention to the wear. steel); that is, there are no available data for the metals usually
Erosion of turbomachines due to entrained particles had employed in turbomachinery. Moreover, the wearing effects
been experimentally investigated many times over the past of particle velocity, concentration, size, and shape as well as
years, and the results obtained until the early 1970s were local flow behavior have scarcely been investigated so far.
surveyed by Truscott [ 1 ]. Since then there have been many Since these factors are indispensable for predicting erosion
studies [2,3]: erosion by solid particles with higher imping- of turbomachinery, the wear is first measured in this study
ing speeds (over 100 m s - i) in a gas turbine has been inves- using the same method as Shimizu et al. [ 13]. Based on the
tigated by Tabakoff [4]; compound erosion of sand and measured values, Bitter's model is confirmed, and an empir-
cavitation [ 5,6], and wear due to the sand particles from the ical formula for the erosion coefficients is proposed.
Yellow River [7,8] have been investigated. With regard to the numerical method of predicting erosion
With regard to the erosion of a wall surface due to particle of tarbomachinery, Roeo and Addie have proposed a method
impingements, Bitter [9] has theoretically analyzed elastic using their own erosion model [15]. In this method, the
and plastic behaviors of the contacting surfaces and con- terminal velocity of a particle is used for the impinging veloc-
firmed experimentally the relationship between the particle ity and a diffusion-convection equation is solved to obtain
impinging velocity and angle. More simplified relationships particle concentrations necessary for the model. They also
have been proposed by Neilson and Gilchrist [ 10] and Finnie compared the calculated wear with the values measured by
[11]. All these have been reviewed by Hashimoto [ 12]. detecting the reduction in resin painted on a casing wall [ 16].
Bitter's model has also been checked recently by Shimizn et But the advisability of using the terminal velocity and the
al. [ 13], who measured the erosion of copper due to the diffusion equation is in question. A novel and consistent
impingements of glass spheres. On the other hand, Roco et approach has been proposed by the authors [ 17] based on
al. [ 14] have proposed a unique erosion model in which the Bitter's erosion model and calculating impinging velocities
0043-1648/96/$15.00 1996Publishedby ElsevierScienceS.A.All fightsreserved
Pll S0043-1648 (96) 06974-8
E Zhong, g MinemuraI Wear 199 (1996) 36--44 37

of particles on the wall by considering the spinning caused resin through which particle trajectories can be observed with
by collision with the wall and other flying particles. The stroboscopic lighting.
calculated results agree well with Roco's experiment. In this
paper, this method is extended to simulate a progressive 2.2. Procedure
change in the wall surface gradually worn down by the par-
ticle impingement, and to demonstrate the effects of particle The test pieces were square bars 1 0 X l 0 X , - 3 0 m m in
diameter and flow capacity on the wear rate of a radial-flow size with one end cut off at varying angles (/3) of - 15, 0 ,
pump casing. 15, 30 and 57 . This angular surface alone was exposed to
particle impingement, while the other exposed surfaces were
protected by rubber fittings. A group of test pieces made of
2. Experimental details cast iron (FC20) or wear-resistant stainless steel (SUS403)
and was fastened with holders around the rotating disk so as
2.1. Apparatus to be eroded at the same time. Sand, used for concrete from
downstream of the Kiso River, served as particles for this
experiment. After putting the sand through sieves with 2 0 -
Fig. I shows the experimental apparatus, which is basically
and 30 - mesh screens, the resultant particle diameter ranged
the same as that employed by Shimizu et al. [13]. Solid
from 0.6 to I mm. The sand was found to be constituted of
particles stored in a feeding tank are released through an
quartz including chert (75%), amphibolite (19%) and feld-
electromagnetic slide-valve on the bottom of the tank into the
spar (6%). Since chert is a kind of quartz and quartz is the
center of a horizontally rotating disk (diameter: 119 ram).
hardest material, the physical properties of quartz were taken
Centrifugal force causes the particles to fly radially along two
as representative of the sand, and the specific weight of the
radial guide vanes on the disk. Thus they impinge on a set of
sand was taken as 2.65. To verify the effect of particle size
test pieces fixed with holders on a standing frame around the
on erosion, steel balls used for shot blasting (specific weight:
disk. The velocity of particles flying from the disk can be
7.2 with diameters of 0.4, 0.6 and 1.0 ram) were employed
controlled by altering the rotational speed of an invertermotor
in this experiment instead of sand.
driving the disk. To determine the effect of particle concen-
Each test piece was weighed before testing with an elec-
tration on erosion, the flow rate of the particles is altered with
trical scale (accuracy + 0.1 rag). The feeding tank was filled
a series of nozzles installed just behind the slide-valve. By
every time with a fixed amount of particles and the total
alternating nozzles having different opening sizes, we meas-
erosion time of the particles was measured. The pieces were
ure the effect of particle concentration on erosion. The feed-
then removed from the holders and weighed to obtain ~ e loss
ing tank and the rotating disk are installed in an airtight
in weight, i.e. the degree of erosion. This procedure was
container which is evacuated (6 cm Hg) in order to eliminate
repeated several times. Since the weight loss became nearly
the effect of air resistance on the flying particles. Part of the
constant after one or two runs, the values in this range were
upper window of the container is made of transparent acrylic
averaged to obtain the wear.
ol~ervation window In this experiment, impinging velocity Vand particle con-
centration Cv varied in the range of 10-29 m s - t and 0.05-
0.20, respectively. Their methods of evaluation are described
below.

2.3. Impinging velocity and angle

Impinging velocity V is geometrically determined based


on photographs taken of particle trajectories, in much the
same way as Shimizu et al. [ 13] did. Let the trajectory CE
be known, where the particle is supposed to leave the disk
(radius ra) at C and impinge on the test piece at E (radius
rE), as shown schematically in Fig. 2. By observing a rectan-
gle ABED with the diagonal_~ corresponding t___othe particle
path in a rotating system (AD = r ~ o s 0 - rd, A B = re sin 0),
we obtain the relation -A-D/V,=-KB/Vo, where Vr and Vo
( = oJra) are the radial and circumferential components oftbe
flying velocity. Since the flying velocity is constant due to
the evacuated space, the impinging velocity V is thus deter-
lnver~r ~ Exhaustv~lw mined as a function of angle 0, by
far p m t c ~
V= (cord/sin 0)~/1 + (fairs) 2 - 2(rdlrE)cos 0 ( I)
Fig. I. Testdevice.
38 Y.Zhong.K. MinemuraI Wear199 (1996)36--44

Teat piece Mean path 2.4. Bitter's erosion coe.Oicients


_ I l:.>.-.i-ff~a~l.
Let A 0 be the angle enveloping a test piece from the center
of rotation on a rotating surface. The mass of the particles
impinging on the piece Mp can be obtained from the mass of
the particles charged in the feeding tank My, as Mp =
Co~/v( A 0/2~r). Since the particles scatter in a vertical direc-
tion beyond the thickness of the pieces, the ratio of the par-
ticles impinging on the test piece to the total number of
particles released is taken into consideration as (7= in the
aboveequation. Under the assumption ofanormal probability
distribution of particles in a vertical direction, the ratio C=
can be evaluated with the vertical length eroded on the test
piece holder Al and with the height of the test piece.
When a test piece reduces its mass by Am due to erosion,
its wear amount per unit mass by particle impingement Wis
thus obtained as W= Am/Mp.
Fig. 2. Velocityrelations. According to Bitter [9], the amount of wear Wdue to the
particle impingement is formulated as the sum of two factors:
the deformation wear Wdowing to fatigue damage from plas-
tic deformation repeated with the normal component of
where 0is the angle between radii vectors OA and OE around impinging velocity V,; and the cutting wear Wc doe to the
the center of rotation O. cutting off with the tangential component of impinging veloc-
While ap_.__~icle is in flight, point A on the disk moves to ity V,, that is,
C, so that A B = AC, or re sin 0_~..rd(~:+_~0), in which g is the
angle between the radii vectors 0(2 and OE. Since ~:becomes W = Wd"l-Wc (2)
definite from the particle trajectory, 0can be found, hence V
where
can be determined.
The pictures show that the particles scatter along a mean w~ = (I/2)Mp(V.~- K,)21ed
trajectory within a confined region, as schematically shown
by broken lines in Fig. 2. Since the mean trajectory is quite
similar to a logarithmic spiral ( r = raka, where k is a constant), f 2MDB( V.~- K, ) 2{ V~ - B ( V.~- K, ) 2e J ~~.~ ) / V~.~
such particle trajectories are approximately expressed with a /for a<ao
constant m as r = (rdk)% which reduces to O=a log r - b ,
(a,b constant). The coordinates of the scattered particles are Wc=I (I/2)Mp{V2-K2(Vni-KI)3/2}/,sc
substituted in this equation after a polar coordinate system is ~ f o r ~z>%
constructed with the origin O at the axis of the rotating disk,
Thus the constants a and b can be determined using the least The terms B, Kj and K2 are constants depending on the mafe-
squares method. These constants obviously depend on the rial properties of the particle and wail, such as density, Pois-
rotational speed of the disk. son's ratio, Yonng's modulus and elastic yield limit. This
Since the velocity V in Eq. (1) indicates only the value at formulation assumes that the wear occurs only when V.
the radius rE, i.e. on the center of the surface of the test piece, exceeds the critica! ve!ocit7 Ki, corresponding to a wall defor-
V varies from the leading to the traiiing edge of the surface mation velocity for the elastic yield limit. It further assumes
and is also dependent upon the trajectory displaced due to that the tangential component of the rebounding velocity
particle scattering. But the length of the edge is rather small, becomes zero when the impinging angle a becomes greater
the radius rE is relatively large, and the scattering is relatively than critical angle Oto.The terms ~d and e= are called erosion
small. Thus, the values at the center of each piece are repre- coefficients and their values should be determined by exper-
sented as the impinging velocity. By assuming that the par- iment, though ab can be calculated from ~=.
tiele scattering to the mean trajectory is a normal probability
distribution, the margin for error of the representative value 2.5. Particle concentration
will be expressed with a three-fold standard deviation and a
dependence on the length of the surface. Let A T be the time of erosion for the total amount of the
After the flying angle 0 is obtained, the impinging angle a particles in the feeding tank My. Since the impinging area of
on the test piece having cutting off angle fl, can be calculated the particles ~(A/)2/4 is sweeping with a speed of re~a, the
as a = f l + O + K , in which K is the departure angle of the volumetric concentration of the particles on the area, Cv, is
particle from the disk, or K= tan- ~{cos 0 - (to/rE)/sin 0}. evaluated as
E Zhong, K. Minemura I Wear 199 (1996) 36--44 39

MT/ pp figures, the deformation wear and cutting wear are also drawn
C~ = ('n A/Z/4)rEo AT as a thin solid line and a broken line, respectively. Wdaccounts
for a major portion of the wear for FC20, while Wc does
likewise for SUS403. The wear is seen m occur when the
impinging angle a becomes larger than about 10 and reaches
3. Measured wear rate and discussion a maximum at about 50. These characteristics hold for a
variant velocity V as well as for particle concentration ,.
3.L Relation between impinging velocity and angle Since ductile metals such as copper and aluminum reach their
maximum at ao--30 , while hard and brittle materials such
When the test pieces made of cast iron (FC20) and stain- as glass reach their maximum at ao = 90 [ 1 ], the cast iron
less steel (SUS403) are eroded by the impingement of sand and stainless steel used are found to have an intermediate
particles (dp=0.6-1.0 mm), the reduction in the wall mass property between those two extremes.
per unit mass of the impinging particles, W ( m g k g - I ) , Fig. 3(c) shows the results when the cast iron (FC20) is
depends on the impinging angle a as shown in Fig. 3(a) and eroded by shot-blast steel balls (dp=0.4 ram), instead of
Fig. 3(b), respectively. In the figure, the @ symbol indicates sand particles, at an impinging velocity of V= 14.2 m s - i.
the data for the averaged values at the representative radius Although the wall material remains consistent, the maximum
of the piece (rE), and the thick vertical solid lines denote the wear alters to peak at a = 9 0 , and the portion of Wd on the
scattering range of the data. The range of probability error, wear is much increased due to the increased density of the
as described in Section 2.3, is denoted by quadrilaterals panicles, showing that the resultant characteristics are
enclosing each data item. changed into those of hard and brittle materials.
When substituting the measured values into Eq. (2) and
determining the empirical coefficients ed and e~ with the least 3.2. Effect of impinging velocity
squares method, the resultant relation of Eq. (2) is drawn as
a thick solid line in the figures, showing that the measured In all cases, evaluated wear coefficients ed and ec are
values are well expressed by Bitter's erosion model. In the decreased with an increase in V, as shown in Fig. 4. This

FC20 vs. Sand I SUS403 w.Sand


fd= I.I,X l012j/m3 ~..6.03 X 10ttJ/m3 ftp=4.lgX 1012j/las ec =I.U X t012j]m3
dp ::0.6-1.0 nun ~6.C d p=0.6-l.Omm
V=22n ~ , ~ v==~/s z'J I
a0-57.1"
~15 c~=o.n

I0

5.(

% 15 30 45 60 7.5 90
(a) a (deg) (b) a (deg)
@ FLY20vs. Steelball

I ~-1.88 1012j/m3 ~-4.63 I01=J/m~


d p :=0.4mm
Oe~4' V=14.2n~s
~, ao -26.4~ ~ I~

o. 15 30 ,
45 ' >60
- / , - . ,75
90
(c) a (deg)
Fig. 3. Dependenceof wearon impingingangle. (a) FC20vs. sand. (b) SUS403vs. sand. (c) FC20vs. steel ball.
40 It'. Zhong, K. Minemura/ Wear 199 (1996) 36--44

d 4,
xtd~
45,
~4J~
~ ~ .
,
dp=0.4m m , SUS403 vs.Sma
dp=0.6"1.0 ram V=14~m~,~
Cv ~=0"2 Cv=0.i l
1.9~

~ d ~ vs. ~_n.d
0.6~ dp =0.6-1.0 mm 0.~ ~ ~ c O~~.6.-1.0 ~
, , ~, Cv=0.1,1
10 15 ~0 25 30
v (m/s) t I
0.05 0.1 01.2' Cv
Fig. 4. Effect of impinging velocity.
Fig. 6. Effect of panicle concentration.
X 10z:
62 FC20 vs. St~l l~licle According to existing reports, the relationships between W
V= 14.2m/s and dp vary greatly [ 1 ]. Hashimoto [ 12] has summarized
the results; W increases with an increase in dp, but there is a
critical value for the particle diameter; i.e. with the increase
o
~ 3.! in dp beyond this critical value, the wear will increase,
Ii1
decrease or remain constant depending on the elastic and
2.6 plastic characteristics of the materials and the wearing envi-
I ronment. Since this critical value is considered to be much
1.3 smaller than the cases dealt with in this experiment, the results
~d presented in Fig. 5 are not inconsistent with the existing data.
0.4 0.6 0.8 1.0
dp (==) 3.4. Effect of particle concentration
Fig. 5, Effect of particle size.
Fig. 6 shows the measured relationship between the ero-
means that the higher the particle velocity V the more the sion coefficients and particle concentration Cv. Both 8d and
materials are eroded. The curves drawn in the figure denote e~ are seen to increase as Cv increases.
the relation given by the empirical formula described later According to experiments with a sand pump [1] (Cv<
(Eq. (3)), and are found to be well in agreement with the 0.15), the wear is reported as WccC~s2 independent of the
experiment. The resultant wear coefficients are expressed as material. On the other hand, it is reported that, with an
~d,Oc (~ V --0.45 increa3c in Cv, the wear increases within a dilute Cv and then
According to the wear of hydranlic machines [ l ], the wear decreases when Cv rises beyond acritical value. Experimental
is expressed as Wcc V 2"2- -3.o (V: circumferential velocity results on pumps by De Bree et al. [2] are similar, showing
of impeller). The wear is also expressed as WR V 2's- -3.o a critical value of about 0.05. Since Cv in this experiment is
for erosion due to fluid jet on the material of $35C or SUS40 greater than this value, the increase in ed and e~ along with
[8]. Since Bitter's formula is approximately expressed as the increase in C~, i.e. wear decreases as C v increases, is
WctVele, the relationship in Fig. 4 can be expressed as consistant with De Bree's results.
Wc V z4s, being very close to the existing data mentioned
above. 3.5. Empirical formulation

As seen in Figs. 4--6, it is clear that the relations among the


3.3. Effect of particle diameter erosion coefficients for impinging velocity V, particle diam-
eter dj, and volumetric concentration of the particles Cv are
When the test pieces of FC20 are eroded by the steel balls almost the same irrespective of differences in the wall and
at a velocity V = 14.2 m s - 1, the resultant erosion coefficients particle materials and in the sand particles and steel balls.
depend on particle size dp, as shown in Fig. 5. The value 8d From these results, the relationships might be unified into an
remains almost unaltered by dp hut 8 increases with an empirical formula, as
increase in dp. This means that the larger the particle diameter,
ed, e,: = a( 1 + bCv) [ 1 + c(dp/dro ) e } ( VIK 1)f (3)
the smaller the degree of cutting wear and the greater the
degree of defotanation wear becomes. The solid lines in the where dpo is taken to be 0.8 mm as a mean value of the sand
figure indicate the relation by the empirical formula. particles, the terms a, b, c, e and f a r e material constants and
Y. Zhong. K. Minemara / Wear 199 (1996) 36-.44 41

Table I
Empiricalconstantsin F.q.(3)
Input Data, t=O [
Wall and particle a (J m - 3) b c e f
Mesh Gener~ion [
SUS403and sand ra I 1.5x 10~2 -0A5 t
Cal. of Water Flow Field [
ec 5.23X 10~z -0.45
FC20and sand ~'a 3.1 x 1012 2.52 -0.45 Cal, of particle T r a j e ~ l
and I m p i n g i n g Velocity- [
,% 1.86X10'2 2.52 -0,45
FC20and steelball ea 4.6 X 10~2 2.52 0 0 - 0A5
~ 6.16x10'z 2.52 1.18 0.55 -0,45

K~ denotes the term in Eq. (2). When using the values of the
constants listed in Table !, Eq. (3) is in good agreement with
the measured values as denoted by the curves in Figs. 4-6,
with V between 10.8 and 28.8 m s -m, dp between 0.4 and
1.0 mm and C,, between 0.05 and 0.20.
The listed data are obtained under evacuated conditions to Fig. 7. Flow chart of calculation.
exclude the effects of drag and virtual mass forces. Also,
Bitter's model describes how the wear rate is related to
impinging velocity and material properties of both particles
Section B"~-"x
and wall. Thus, the data can be applied to gas-sand as well as
water-sand conditions if the particle impinging velocity is
evaluated from the equation of motion of the particles (see
Eq. (5)).

4. N u m e r i c a l prediction o f w e a r in casing
- TF
4.1. Method of predicting erosion

The numerical procedure to predict erosion is shown as a


flowchart in Fig. 7. In our previous paper [ 17], the progres- -,t rB.
sion of wear on a pump casing was simply estimated by
assuming the wear rate to be unaltered. Fig. 8. Geometryof casingand coordinatesystem.
This paper reports on a special case of a spiral casing in a
radial-flow impeller pump in which the wear has already been ible and inviscid, and that it remains unaltered by the presence
investigated experimentally and numerically by Roco et al. of the particles due to their small size and low concentration,
[14] and calculated by us [17]. Though the procedure as was the case in Roco's study and our previous study. In
described here is a two-dimensional problem, it is easily the present case, the water velocity w has a potential ~b
extended to a three-dimensional one. (w=V~b) and can be determined by solving the Laplace
equation (V2~b=0) [ 17,18].
4.1.1, Mesh generation
The shape and dimensions of the casing are shown in
4.1.3. Trajectory of a solid particle
Fig. 8, where the finEe elements (simplex triangular ele-
The governing equation of motion of a solid particle with
ments) are drawn. Since the radial coordinate of each node
mass mp and velocity Vpcan be expressed as [ 16,17]
is located in equal segments, the shape of the casing wall can
easily be changed as i! undergoes erosion. The shape of the nlp dvp/dt = Fp + F d + F v + F L (4)
wall is represented b~ a cubic spline-fit curve. Using this
curve, it can be deterrained whether a particle impinges on where Fp denotes the force caused by the pressure gradient,
the wall or not, and where the node coordinate of the impinge- Fd the drag force determined by the particle Reynolds num-
ment or that after repartitioning is located. ber, Fv the virtual mass force, and FL the lift force due to the
velocity gradient and particle spinning.
4.1.2. Waterflow fieM BY uumerieaily integrating Eq. (4) under the assumption
Although the water flow inside the casing is turbulent, the that the initial vd,-,z.ity is the ~ame as the water velocity in
movement of the particles is predominantly affected by iner- the initial se~.'.tion, the distributions of particle impinging
tia. Thus, we may assume that the flow is steady, incompress- velocity and angle on the casing wall can be evaluated [ 17].
42 Y. Zhong, K. Minemara/ Wear199 (1996) 36--44

4.1.4. Erosion rate 0.8


Let Np denote the m~mber of impinging particles on the Q=egt # / h C,= 0.069 /-~f~.
e - 2.~s ~ J . J
wall per unit area and unit time. By replacing M v in Eq. (2)
06
with ~ = l, wear rate, that is, wear .per unit area W, can be
converted to wear rate pet- unit area W, as
#=#d+#o (5)
~0.4 I "f~'~'-=:='~"
After recording the calculated impinging velocity and g l~ ~ dp=l.Omm
angle on every control surface, the distribution of the wear 0.2 ~ L "-'.] ;'-'3 dp=3.0mm
rate I~"can then be evaluated by the erosion coefficients, 6d
and ~c, estimated from Eq. (3). 0.L I .

4.1.5. Deformation due to erosion


As the wear proceeds, the shape oftbe casing wall ( density: ~ 0.2 .......... .~.~-%
p~,) is also changed. After a time interval of A t, the volumetric
wear of a control surface j (area Sj).is (VjAtSJpw. Thus, the
mean depth of the worn surface hj is W~At/p~. Iftbe integrated
(deg)
depth of hj exceeds a critical value ht, the wall surface is
displaced by ht to reflect the changes in the flow field and Fig. 9. Distribution of impinging velocity (effect of particle size).
impinging velocity. In this case, the shape of the boundary eter dp. This is shown in Fig. 9, where O is the angle counter-
expressed by a spline-fit curve is rewritten by the way dis- clockwise from the tongue, and the normal component Vn
placement takes place at the centroid of the control surface and circumferential component Vt are expressed dimension-
by h l, and the coordinates of corresponding nodes are recal- lessly by the peripheral speed of the impeller u2. As described
culated. In this calculation, hi is taken to be 1/10 of each in our previous paper, Vngenerally takes the maximum at the
element height. first impingement and decreases every time the particle
impinges, finally diminishing to zero. On the other hand, V,
4.2. Calculated results and discussion also takes the maximum at the first impingement and
decreases with repetition of the impingement. When the
In our previous paper [ 17], the movement of spherical rebounding distance of the particles diminishes, however,
particles having a diameter of dp= 0.27 mm and the specific their speed increases again to almost the initial value. Distri-
weight corresponding to sand ( = 2.65) was analyzed under butions of the impinging velocities for dp= 1.0 :nm are
the rated flow capacity ( Q = 681 m 3 b - l ), and wear rate shaded and enclosed with a thick solid line.
was evaluated with the resultant distributions of particle The largest values, that is, the upper periphery of V, for
impinging velocity V. The results obtained coincided fairly dp=0.27 and 1.0mm, approximately coincide with each
well with the values m~asured by Roco. In the calculation, other. However, the values for dp= 3.0 mm increase greatly
the particle movement was analyzed based on three models: over the whole range except in the vicinity of the tongue
model I, in which the particles move without spinning and ( 0 < 4 5 ) . This is attributable to the fact that the inertia of
mutual collisions; model II involves spinning only; and model the particle increases in proportion to the cube of do. On the
III involves both spinning and mutual collisions. From the other hand, Vt is remarkably small in the tongue area, while
results, it is found that these additional effects scarcely affect in other areas Vt is large and increases with an increase of 0.
the erosion of the casing. To discover whether or not this The values of Vt for dp= 0.27 and 1.0 mm are also the same,
result is right regardless of the particle diameter, the calcu- and decrease with the increase of dp, owing to the increased
lation for three kinds of spherical particles having diameters impinging angle due to greater inertia, whereas the values of
of alp=0.27, 1.0, 3.0 ram, respectively, are carried out in this Vt and V, in the tongue .area remain unaltered for different dp.
study. The results calculated by model III are superimposed using
In each case, it is assumed that the inlet value of particle thin lines in Fig. 9. The resultant differences using this model
concentration C,o is 0.069, and that the particles distribute lessen when dp increases. However, the differences are trifling
uniformly in a circumferential direction. It is also assumed in the range of dp> 0.27 mm, suggesting the entire applica-
that the easing is to be made of cast iron (FC20) and eroded bility of model I is independent ofdp. This is related to the
by sand (specific weight e = 2.65), and that the wear coeffi- fact that the particles having greater diameter encounter less
cie.n_ts~re gi,,en by Eq. (3). mutual collision because of the fewer particles included in
the medium with the particle concentration unaltered.
4.2.1. Distributionofimpinging velocity Fig. 10 shows the effects of alteration in flow capacity Q
When applying model I to calculate particle trajectories, on the impinging velocity, when model I is applied to the
the resultant impinging velocity Vis distributed in a confined particle dp= 1 mm and at capacities of 4-35% of the rated
region, enclosed by various lines depending on particle diam- one (Q=681 m 3 h - J ) . V, is seen to be almost the same
Y, Zhong, K. Minemura/Wear199(1996) 36-44 43

0.8 I'~ Q.~lOmS/la ~'.]Q~Y3Om3/h/e.


~
--- q-cl01Om
m q--~t,,'m ~ , ; C~68t ml/li clp-LOmm

g o.4

o.~

C,- 0.069
I
..r .....
b ~ tm 27o 36o

Fig. 12. Effectof flowrateon wearrate.


C: 6 ~o l~o ~o 360
o (deg) - - t=3ree~ - - t~12muetht[ ~ " ~
Fig. 10. Distributionof impingingvelocity(effectof flowrate).

0'21 modell modelm


[ . . . . . . dp-0.27 mm f.
L ~ -- dp=i.O mm
...... ,,=,.o_ I

"t
}o} / /

Fig. 13. Deformation of c~ing due to wear.

O (dee) no overall meaningful difference between them for all dp.


Fig. I I. Effectof particlesizeon wearrate (differencebetweenmodel! and Since model llI consumes much greater computing time than
modelIll). model I, it may be concluded that model I is sufficient to
predict erosion by particles of the usual size (dp > 0.27 ram).
except in the vicinity of the tongue. With an increase of O, Fig. 12 shows the effect of flow capacity on the wear rate
V~decreases in the range of 65 < O < 180, while it increases IT' when calculated by model I for dp = 1.0 mm. With an
in the range of O > 180. increase in Q, if" increases in the vicinity of the tongue area
due to the increased value of V,, and in the exit region of the
4.2.2. Distributions of wear rate casing due to the increased value of Vt.
When the impinging velocity of sand particles is calculated
4.2.3. Deformation of casing due to wear
by model I, the resultant distributions of wear rates I~'for east
When calculating the impinging velocity of particles cor-
iron (FC) are approximately the same when alp_<1.0 ram,
responding to sand with a diameter ofdp ffi 1.0 mm by model
and reach a maximum at the tongue area and exit region of
I, the resultant deformation of the casing wall made of cast
the casing (O--320), as shown in Fig. 11. When dp is
iron (FC20) progresses hourly as shown in Fig. 13. The
increased to 3 ram, however, W is almost doubled in the
tongue area is seen to be locally eroded on a large scale and,
whole region except for the tongue area. This insensitivity in
except for this area, the erosion depth progressively increases
the tongue area is attributable to the fact that the larger par-
in the direction of flow. These conditions appear quite similar
ticles have less opportunity for impinging when the particle
to those observed by Roco and Addie[ 16].
concentration remains unaltered.
The wear rate if" based on the impinging velocities calcu-
5. Conclusions
lated by model III is also superimposed using thin lines in
Fig. ! i. It can be seen that the difference between models I Erosion damage by particle impingements is measured to
and III lessen as dp increases. However, there appears to be formulate the wear coefficients based on Bitter's model and,
44 Y. Zhong, K. Minemura / Wear 199 (1996) 36-44

using this formula, the erosion d a m a g e to a p u m p casing is [5] Z. Wang, Several new materials for anti-wear, anti-cavitation use on
calculated. The results obtained are summarized as follows: surface of watar turbine pump, IAHR 13th Syrup., 2 (1986) 84-1.
[6] J. Sarah, K. Usami and T. Okamara, Basic study of coupled damage
1. The erosion data of east iron and stainless steel eroded by caused by silt abrasion and cavitationerosion ( I st report: Experiments
sand particles confirm Bitter's erosion model. with submerged water jets ), Trans. JSME, 56-523 B (1990--3) 696
2. Erosion coefficients e d and e o f Bitter's model increase (in Japanese).
with a decrease in impinging velocity, and with an [7] B.-E. Chert and J. Ye, Experimental investigation on effects of silt-
increase in particle diameter and volumetric concentra- laden flow in Huanghe River upon abrasion resistance of materials for
hydromachines,Alternative Energy Sources, 7 (3) (1987) 77.
tion. These effects are expressed, within the range meas- [ 8] N. Mankawa,Q.G. Yuan, Y. lwai and T. Okada, Slun'y wear test using
ured, by Eq. ( 3 ) . liquidjet containing sand (in the case of sand from the Yellow River),
3. The effects o f spinning and mutual collisions of the par- Trans. JSME, 59-560A (1993--4) 262 GinJapanese).
ticles on the w e a r rate decreases with an increase in par- [9] J.G.A. Bitter, A study of erosion phenomena, Wear. 6 (1963) 5, 169.
ticle diameter, though the difference is trifling. [ 101 J.H. Nailson and A. Gilchrist, Erosion by a stream of solid particles,
Wear, I1 41968) I11.
4. The larger the particle diameter, the greater the erosion [ 11] I. Finale, Some observations on the erosion of ductile metals, Wear.
rate becomes; though the effect of particle diameter on the 19(1972) 81.
w e a r rate is fairly small for diameters less than 1.0 mm. [12] K. Hashimoto, Wear due to Particle Impingements, Industrial
With an increase in flow capacity, the wear in the vicinity Technology Center, Tokyo, 1974. p. 75 (in Japanese).
[13[ A. Shimizu, Y. Yagi, H. Yoshida and T. Yokomime, Erosion of
o f the tongue and the casing exit progressively increases. gaseous suspension flow duct due to particle collision (!). J. Nncl. Sci.
TeehnoL, 30 (9) (1993) 881.
[ 14] M.C. Roco, P. Nair, G.R. Addie and J. Dennis,Erosion of concentrated
slurries in turbulent flow, J. Pipelines, 4 (1984) 213.
References [ 15] M.C. Rocoand G.R. Addie, Analyticalmodel and experimentalstudies
on slurry flow and erosion flow and erosion in pump casings, Slurry
[ I ] G.F. Tmscott, A literature survey on abrasive wear in hydraulic Transportation, 8 (1983) 263.
machinery, Wear. 20 (1970) 29. [ 16] M. C. Roco and G.R. Addie, Erosion wear in slurry pumps and pipes,
[2l S.E.M. De Bree, A. Begalinger and A.WJ. De Gee. On the erosion Powder TechnoL, 50 (1987) 35.
resistance of steels in water-sand mixtures for application in dredge [ 17] K. Minemuraand Y. Zhong, Numerical prediction of erosion wear on
pumps, Proc, 9th World Dredging Conference. )980, p. 621. pump casing under selid-watar two-phase flow, in A. Serizawa, T.
[ 3l M.A. Rayan and M. Shawky,Evaluation of wear in a centrifugalslurry Fukano and J. Bataille (eds.), Advances in Muhiphase Flow, Elsevier,
pump. Proc. Inst. Mech. Eng., 203-A (1989) 19. New York, 1995,p. 561.
[4] W. Tabakoff. Review-Turbomachinery performance deterioration [ 18] K. Minemura and T. Uchiyama, Calculation of the three-dimensianal
exposed to solid particulatesenvironment, Trans. ASME. J. FluidEng.. behaviorof spherical solid particlesentrained in a radial-flowimpeller
106 (1984-6) 125. pump, Proc. Inst. Mech. Eng., Part C, 204 (1990) 159.

You might also like