Wear 268 (2010) 946–954

Contents lists available at ScienceDirect

Wear
journal homepage: www.elsevier.com/locate/wear

Cavitation enhancement of silt erosion: Numerical studies

P.J. Dunstan a , S.C. Li a,b,c,∗
a
University of Warwick, Coventry CV4 7AL, UK
b
State Key Lab of Hydroscience & Engineering, Tsinghua University, China
c
Hebei University of Engineering, China

a r t i c l e i n f o a b s t r a c t

Article history: Cavitation and silt erosion often co-exist causing severe damage. The effect that the silt erosion is
Received 9 March 2009 worsened by the presence of cavitation is here referred as cavitation enhancement of silt erosion. A
Received in revised form micro-model proposed by Li (2003, 2006) [1,2] describes how the silt particles gain extra-damaging
14 December 2009
potential from the collapsing bubbles. In order to verify this envisaged model, as suggested in Li (2006)
Accepted 17 December 2009
[2] we have performed initial numerical studies by employing a simplified model which considers a single
Available online 24 December 2009
silt particle entrained in the micro-jet of a collapsing cavitation bubble. Initial CFD simulations have been
conducted by using both commercial software Star CCM+ and our simplified analytical approach (written
Keywords:
Synergistic erosion model
in MATLAB). Results from both methods support the mechanism of the micro-model that the silt particle
Calculation of particles trapped in micro-jet damage potential to nearby solid boundary is largely enhanced through the acceleration of particle by the
Cavitation erosion micro-jet of collapsing bubble in which the particle entrained. Our calculations further show that given
Silt erosion the size of bubble the enhancement is mainly correlating with the mass (Its shape also plays an important
Cavitation enhancement role in gaining the drag force from the micro-jet flow, which will be discussed in detail later.) of particle
Bubble–particle interaction and the ratio between the distance from the collapsing bubble to a solid boundary and its radius upon
collapse. Though there are more attributions of the enhancement need to be further investigated such as
the influence of presence of the particle on the micro-jet flow, etc. (by using a real coupled calculation
approach) these initial results do reveal the major features of this phenomenon quantitatively for the
first time.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction cles and the cavitation events. In order to make a breakthrough, we

must understand the essential issue, that is, the driving force that
Cavitation erosion has been studied for over a century. Extensive the particles gained from fluid (in this case, cavitation bubbles in
books and articles are available about its mechanism and damage particular). Without such knowledge, we cannot answer ‘why some
effect, etc. It is generally agreed that the damage power is generated types of cavitation or some part of a cavitation do not enhance the
by collapsing bubbles and in most cases the damage to commonly silt erosion, and even retard the erosion processes’. Li [2] has pos-
used materials is a fatigue process. However, we have little knowl- tulated a potential micro-model1 for this interaction. That is, silt
edge of the synergism of cavitation and silt erosion, in particular the particles entrained in the resulting micro-jet of a cavitation bubble
enhancement role of cavitation in the process of silt erosion. That is, collapsing near a solid surface are projected at high velocity toward
the least effort has been made for the most severe type of damage. the surface resulting in significant impact damage. This theory is
This problem remains as a challenging in many areas such as the evidenced in experimental studies by Jin et al. [3].
synergistic damage to turbines by concert action of both cavitation
and silt erosion.
Moreover, controversial reports from different laboratories
1
around world make this phenomenon even more mysterious. The co-existence of cavitation and silt erosion does not always enhance each
Whether or not the cavitation enhances the process of silt erosion is other. Their synergism is governed by the characteristics of flow/cavitation (bub-
bles), solid particles and materials and their interplays. Li’s model is intended to
solely determined by the interaction between the entrained parti- reveal the mechanism underlying the enhancement of silt erosion owing to the
presence of cavitation, which is an essential aspect of the synergism. The model rep-
resents the situation where the characteristics of flow field and cavitation (bubbles)
are virtually not affected by the presence of solid particles. That is, the concentration
∗ Corresponding author at: Fluid Dynamics Research Centre, School of Engineer- of suspended particles is low and the size of particles is also small compared with
ing, Gibbet Hill, Coventry CV4 7AL, UK. Tel.: +44 024 76528027; the bubble. Therefore, Li’s model focuses rather on how the driving forces acting
fax: +44 024 76418922. upon the particle are altered owing to the presence of bubbles. For details, readers
E-mail address: [email protected] (S.C. Li). are referred to [2].

0043-1648/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.wear.2009.12.036
 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954 947

A simplified approximation is considered where a single silt par-

ticle becomes entrained in the micro-jet of a collapsing bubble and
is consequently projected at high velocity toward a solid surface.
The independent variables considered for comparison are the dis-
tance from the bubble centre to the solid boundary S, the maximum
radius of the bubble before collapse Rmax , the ratio  = S/Rmax and
the geometry and mass mp of the particle. A silt particle can take
any shape. Two geometries that can be easily scaled have been con-
sidered here: a cuboid and an extruded two-dimensional Joukovski
foil. The dependant variables of interest are the velocity and kinetic
energy (considered to be a measure of damage potential) of the par-
ticle at the point where it makes contact with the solid surface. In
reality, particle size varies, for instance in silt-laden rivers, from
much smaller than 10–60 ␮m to much larger than 500 ␮m. The
bubble size also varies from case to case, in most cases ranging from
fractions of mm to several mm. For this numerical study, we have
Fig. 1. Coordinate system.
chosen the particles varying from 0.1 mm to 0.5 mm and bubbles
varying from 0.5 mm to 2 mm which are commonly encountered
cases. Experimental results and Dynaflow simulations have shown 3. Basic equations of motion
micro-jet diameters in the region of 25% of the bubble diameter.
It is important that the initial position of the particle is constant In general, the equations of motion for a particle in a 3D flow are
throughout all of the simulations to ensure that similar situations consisted of translations and rotations. The equations of translation
are being compared to each other. From initial simulations it was motions are as follows:
found that for all bubble sizes, all particles within the test range d vx
would reach their maximum velocities (subject to fluctuations in Fx = mp (1)
dt
the micro-jet velocity) within 0.75 mm. It was thus decided to set
dvy
the translation distance required by a particle before striking the Fy = mp (2)
dt
wall to be 1 mm. The position of the bubble centre was determined
by the value chosen for . dvz
Fz = mp (3)
Two calculation methods are used. The first makes use of two dt
computational fluid dynamics packages. 2DynaFS is used to model Here, the subscripts x, y and z refer to the three dimensions, v is the
the collapse and subsequent micro-jet of a cavitation bubble. The particle velocity, F is the sum of the forces acting on the particle
data from these simulations are then used in Star CCM+ to model and mp is the particle’s mass. For our 2D problem as configured in
the motion of a silt particle entrained in the micro-jet. The second Section 2, Eq. (3) disappears.
is a simplified analytic approach using MATLAB code written by The equations of rotation motions are as follows:
Dunstan [4].
dωx
Ix − ωy ωz (Iy − Iz ) = Tx (4)
dt
2. Coordinate system dωy
Iy − ωz ωx (Iz − Ix ) = Ty (5)
dt
The initial position for the particle in this study is directly dωz
between the collapsing bubble and the solid boundary wall. Iz − ωx ωy (Ix − Iy ) = Tz (6)
dt
The particle’s movement has been restricted to two translational
Here, I is the moment of inertia of the particle, ω is the angular
degrees of freedom and a single rotational degree of freedom.
velocity and T is the sum of the torques applied to the particle. Val-
Therefore, it reduces to a 2D problem.
ues for I are specific to each individual particle. The subscripts with
The problem has been split into two separate coordinate sys-
primes are referred to a frame of reference moving with the particle
tems. The first has origin O at the solid boundary with y-axis along
on which the rotational terms are expressed while the translation
the wall and x-axis parallel to the direction of the micro-jet induced
terms are expressed in the inertial reference-frame fixed on the
flow. The second coordinate system has its origin O at the centre of
earth. For our 2D problem, since ωy = ωz = 0 and Tx = Ty = 0, the
the particle and with axes x and y parallel to x and y, respectively.
equations of rotation motions reduce to
This system is displayed in Fig. 1, in which
dωz
Iz = Tz  (6a)
dt
• x and y are the coordinate axes fixed on the boundary wall;
• x and y are the coordinate axes fixed on the centre of the particle 4. Particle motion in flow
and parallel to x and y;
• Both z and z axes are normal to the page; Maxey and Riley’s equation [5], which describes the motion of a
• u is the flow velocity vector; spherical particle with mass mp moving with a velocity v in a flow
• v is the particle velocity vector; with velocity u, has been further simplified by Yin et al. [6] as
• w is the relative velocity vector u − v;
dv
• ω is the rotational velocity of the particle; mp = FD + FL + Fg + FVM + FPG (7)
•  is the angle of the particle compared to x axis (anticlockwise dt
from positive x is positive ); Here, mp is the particle’s mass, FD is the drag force upon the particle,
•  is the angle of the resultant velocity compared to the x axis FL is the lift force acting on the particle, Fg is the force acting on the
(anticlockwise from negative x is positive ); particle due to gravity, FPG is the force due to a pressure gradient
• ˛ is the aerodynamic angle of attack which is equal to  − . and FVM is a force due to the virtual mass.
948 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954

4.1. Drag force

Let A1 be the particle area in the direction normal to the direction

of the drag force and L the density of the fluid, the hydrodynamic
drag on the particle can be expressed as
1  
FD = CD L A1 w
  (w)
 (8)
2
Here, w is the relative velocity (u − v). The drag coefficient CD is
evaluated in two ways in this paper. The first involves using the
CFD package Star CCM+ to calculate the coefficient and the second
is based on analytic formula. Chabbra et al. [7] collected around
1900 pieces of experimental data on the drag coefficient of non- Fig. 2. Torque due to the resistance [6].
spherical particles from 19 studies. These studies covered a wide
range of particle sizes and shapes, Reynolds numbers and kinematic
conditions. The data were then used to test five commonly used
the acceleration of the displaced fluid and is defined as mp (Du/Dt).
methods for calculating CD . It was found that the best method was
According to Yin et al. [6], the force due to virtual mass, FVM , is
that proposed by Ganser [8]. The mean error of Ganser’s study was
around 16%. Variations between the model and actual data suggest 1 d
that for solid objects the model is suitable but this is not true for 
L V (w) (16)
2 dt
hollow bodies.
CD 24 0.4305
= (1 + 0.1118(ReK1 K2 )0.6567 ) + (9) 4.4. Rotational motion
K2 ReK1 K2 3305
1+
ReK1 K2 The particle will also be subject to rotations since at most inci-
where dence angles the pressure centre does not coincide with the mass
1 d 2
−1 centre. As the hydrodynamic forces described above act at the pres-
K1 =
n
+ −0.5
(10) sure centre, it generates a torque around the mass centre. In order
3 dv 3 to evaluate the torque on a particle the distance, xcp , of the pressure
)0.5743
centre from the mass centre will need to be found. The position of
K2 = 101.8148(− log (11) the centre of pressure will vary depending on the attack angle so a
dn is the diameter of a sphere with equivalent projected area to the relationship will need to be found connecting the angle of attack ˛
object in question. to xcp . This value will clearly be different for all particles. Yin et al.
 [6] suggest the following relation, which can be used for thin rod
4A1 like particles with uniform cross-section.
dn = (12)

xcp = 0.25b(1 − e3(1−ˇ) )| cos3 ˛| (17)
dv is the diameter of a sphere with equivalent volume to the object
in question. is the sphericity of the particle defined as the ratio Here ˇ is the particle aspect ratio.
between the surface area of an object and the surface area of a At this point the hydrodynamic torque, T1 , due to the forces
sphere with equivalent volume. acting upon the particle is
1 2
As  3 (6VP ) 3 T1 = xcp × (FL + FD + FVM + FPG ) (18)
= = (13)
Ap Ap
Note that Fg is excluded here as that always acts at the mass centre
Re is the Reynolds number.
instead of at the pressure centre not only as negligible. The torque
T1 will give the particle a rotational velocity ω at which balanced
4.2. Lift force by a resistive torque from the liquid. Fig. 2 shows how the resistant
torque is generated by the rotating particle.
The lift force will act orthogonally to the drag and follow the This resistance contribution can be calculated by integration
relation given by Yin et al. [6] as along the length of the particle. If we consider that the relative
1 z · w
 velocity of the fluid due to particle rotation is (ωf − ωp )l, where ωf
FL = CL L A2   [z × w
 ] × w
 (14) is used to count for the undisturbed fluid vorticity at the point occu-
2  
w
pied by the particles centre of mass and l is the distance from the
Here, A2 is the particle area normal to the lift force. Yin recommends centre of mass then the resistant torque is
the use of the relationship given by Hoerner [9] as  B
 
FL  T2 = FD dl (19)
  = | sin2 (˛) cos(˛)| (15) A
FD 
Here, L is the particle length and D is the height of the projected
4.3. Other forces face at each point along the length L, referring to the side view of
the particle profile shown in Fig. 3.
The gravity force is simply Fg = Vp (p − L )g with Vp being the Substituting expression for FD into Eq. (19),
particle volume, p the particle density and g the gravity accelera-  B
tion. However, for this study, the gravity force Fg is considered to T2 = ω2 l2 DCD L ldl (20)
be negligible. The force due to pressure gradient, FPG , accounts for A
 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954 949

2. Solve the flow (Navier–Stokes) equations using the CFD package

Star CCM+.
3. Use Star CCM+ to calculate the lift and drag forces acting upon
the particle.
4. Use Star CCM+ to calculate the torques acting upon the particle.
5. Solve the equations of motion and update the particle’s position,
orientation and relative velocity for a given time step.
6. Repeat steps 2–5 until the particle has reached the boundary.
Fig. 3. Side view of a particle.

5.2. Analytic solution

If the particle is split into two segments either side of the centre
of mass then Approximation that the velocity of a collapsing micro-jet does
 0  L−xcm  not vary with time, i.e. constant u, has been made for this approach.
1 A custom written program in MATLAB has been employed for cal-
T2 = CD L ωp 2 D l3 dl + l3 dl
2 culations. Though it was not expected that the results from such a
−xcm 0

0 1
L−xcm 
simplified approach would be identical with those obtained from
1 1 4 CFD approach they do display surprisingly the same trends. For this
= CD L ωp 2 D l + l4
2 4 −xcm 4 0 approach, rather than allowing CFD software to calculate CD and CL
these values are determined directly by Eqs. (9) and (15). So do the
1 torques T1 and T2 by Eqs. (18) and (21). Under a given set of ini-
= CD L ωp 2 D(L4 − 4xcm L3 + 6xcm 2 L2 − 4xcm 3 L + 2xcm 4 ) (21)
8 tial conditions the forces on a particle are calculated allowing its
When the centre of mass is equidistant from either end of the par- motion in a given time period to be calculated. The program then
ticle, this reduces as loops with the new position, orientation and velocity of the parti-
cle as initial conditions. This process is repeated until the particle
1
T2 = CD L ωp 2 DL4 (22) has traveled a set distance determined to be the distance from its
64 initial position to the position where the solid boundary would be
located.
5. Simulation procedure
6. Results and discussion
CFD and analytical approaches have been used to calculate the
motion of a particle and their results are compared. The analyti-
6.1. CFD approach
cal approach is thus tested to see if this simple approach can still
provide sensible solutions for this type of problems. For the CFD
6.1.1. Single bubble and range of particles
approach, the solid boundary wall and its effect on the micro-jet
As the mass of a particle increases it is expected that its velocity
flow which in turn alter the motion of particle are included in the
upon impact will decrease as a consequence. Should the decrease
calculation. This method essentially follows a series of repeated
in v be significantly large then it may be found that there is a point
steps. For the analytical approach, the boundary wall is removed
at which the impact kinetic energy 0.5mp v2 reaches a maximum
and the micro-jet flow is extended to infinite on x direction with
beyond which the increase in mass and size will yield reduced dam-
analytical expressions for the lift and drag coefficients.
age potentials. In order to investigate this correlation, a range of
particles are considered with a single bubble with  = S/Rmax = 1.
5.1. CFD approach The extruded foil particle is considered in this case with chord
lengths ranging between 0.025Rmax and 0.25Rmax . In each case the
The input velocities for the micro-jet flow are determined by the particles initial position is set such that its mass centre is equidis-
results of 2DynaFS simulations of bubbles collapsing near a solid tant from the solid boundary and the bubble center. All distance
boundary. dimensions are measured relative to Rmax , all velocities are non-
dimensional relative to the impact velocity of a particle with chord
1. Set the initial conditions for the particle position, orientation and length 0.1Rmax and the particle density is unity such that the mass
relative velocity at the beginning of one time step. and volume are identical. Plots of volume (i.e. mass) against impact

Fig. 4. Variation of impact velocity against foil particle volume in a micro-jet flow.
950 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954

Fig. 5. Variation of impact kinetic energy against foil particle volume in a micro-jet flow.

Fig. 6. Variation of impact velocity against cuboid particle volume in a micro-jet flow.

velocity and impact kinetic energy are shown in Figs. 4 and 5, in Fig. 5, resulting a mono-ascending curve of kinetic energy. In
respectively. order to further clarify this phenomenon, a more accurate numer-
First the impact velocity reduces sharply then becomes mildly. ical model taking into account of this fluid–particle interaction is
Whereas, the kinetic energy increases with the increase of the par- needed in our future studies. From current results, the relationship
ticle volume. But the results above do not show a maximum kinetic of impact velocity with the particle size is approximately inversely
energy reached as expected presumably owing to the assump- proportional. This relationship is likely useful (i.e. true) for parti-
tions made in this numerical model. In reality, as the particle size cles with their size an order smaller than that of micro-jet (i.e. the
increases to the order comparable with the size of micro-jet which micro-jet flow can be regarded as infinite flow around the particle).
is around 20% of bubble size, the flow field of micro-jet can no longer This also indicates that the increase in mp overwhelms the decrease
be considered as an infinite flow field for the particle and conse- in v2 , thus resulting in the increase of the kinetic energy 0.5mp v2
quently the driving force accelerating the particle motion is affected for the cases of those particles at least an order smaller than that of
(reduced). But this is not reflected in current calculations. That is, the micro-jet.
the impact velocities might be over estimated towards the large For particles shaped as cuboids, similar simulations were con-
end of particle volume in Fig. 4 and so the impact kinetic energy ducted. Initially a particle shaped as a long thin rod was chosen

Fig. 7. Variation of impact kinetic energy against cuboid particle volume in a micro-jet flow.
 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954 951

Fig. 8. Variation of particle impact velocity against the relative distance of bubble to wall.

Fig. 9. Variation of particle impact kinetic energy against the relative distance of bubble to wall.

with dimensions a × b × c where b and c are equal. For each set of [10] and Dular et al. [11] further suggests that vjet is related to
simulations a remains constant while b and c vary but remain equal  2 . The variations of impact velocity and impact kinetic energy
to each other until all three dimensions are equal. The variations against  (and  2 ) are shown by Figs. 8 and 9 in which variables
of impact velocity and kinetic energy against particle volume are are normalized as before.
shown in Figs. 6 and 7, respectively. All values are normalized in an As expected the particle’s impact velocity, and hence kinetic
identical manner to the data for foil particles. energy, increase with . However, it is not clear, from these data
The behaviour of these particles is similar to that of the foil whether the relationship is linear or quadratic.2 There are many
shaped examples and as such the analysis is much the same. It once factors that possibly alter final velocities of the particle. The out-
again appears that the change in mass of a particle is the dominant put from Dynaflow does not suggest that the micro-jet velocity
factor affecting the impact energy. from a bubble continually and smoothly increases during collapse.
Although the relationships shown by both geometries are very There are fluctuations, some of which are sharp. The impact veloc-
similar, their damage potential would be quite different owing to ity of the particle will be certainly affected by this fluctuation of
their geometry. The square particles travel faster and hence have a the micro-jet velocity. Furthermore, the particle is free to rotate
larger damage potential than a foil of equal mass. This is due to the as well thus for each bubble not only will the translational motion
fact that the drag (i.e. driving) force on the foils is lower. be different but also the rotational motion. When the particle is
positioned such that it has a large surface area orthogonal to the
6.1.2. Single particle and range of bubbles flow it will have a larger drag coefficient than when it is positioned
In order to reveal how the distance of a collapsing bubble to the with a small surface area normal to the flow. Thus the rotational
boundary wall affects the particle behavior, calculations have been motion of a particle will alter its translation motion. Develop-
performed for the case of a single particle being exposed to a range ing any exact relationships for the motion of a particle directed
of collapsing bubbles with different relative distance  to the wall.
For this study a foil particle with chord length 0.2 mm was cho-
2
sen. Bubble conditions were selected with values of  between 0.5 This was something confusing from our simulation results against the relation-
and 1. Preliminary investigation with the 2DynaFS software sug- ship suggested firstly by Plesset and Chapman [10] and the relation of vjet being
related to  2 further suggested by Dular et al. [11]. Apart from the reasons men-
gested that an increase in  of a collapsing bubble should lead to tioned in the paper, this might also contribute to the lack of enough datum points in
an increase in the velocity of subsequent micro-jet vjet . This rela- the plots to determine fitting curves properly such that they look like straight lines
tionship has been suggested previously by Plesset and Chapman against both  and  2 .
952 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954

Fig. 10. Comparison of flow velocity effect by both CFD and analytical approaches.

toward a solid boundary by a collapsing micro-jet is clearly not 6.2.1. Change in flow velocity
possible. For this analysis a single particle was subjected to a range of
flows with constant velocities. Fig. 10 shows the variation of the
6.2. Analytic method particle velocity and impact kinetic energy against the flow veloc-
ity. Though the results from both approaches are not identical they
The analytic method has been performed using the MATLAB do present the same trends. Considering the simplifications, the
scripts in order to see if this simpler approach can still provide sen- analytical approach does provide significant insight of this phe-
sible results. For this method, rather than the data from 2DynaFlow nomenon.
to define the flow, a constant flow velocity was used for each sim-
ulation. It has been assumed that the micro-jet flow is not affected 6.2.2. Change in particle mass
by the boundary wall (i.e. no boundary wall as the micro-jet flow The same analytical approach has been used as before but each
field concerns). The cuboid particles were employed in these sim- loop increases one or more dimensions of the particle by a pre-set
ulations. percentage. The first condition considers the effect of scaling all of

Fig. 11. Comparison of particle volume effect by both CFD and analytical approaches.
 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954 953

Fig. 12. Effect of particle volume on impact velocity of a particle (variable dimensions shown in legend).

the dimensions of a cuboid particle such that it maintains its overall X refers to the depth of the particle into the page, L is its length as
shape. This is similar to the CFD simulations performed in Section viewed from the top down and D is its depth as viewed from the top
6.1.1 where a foil particle was scaled and retested. Initially a particle down. The labels in the legends refer to the dimensions which are
with dimensions 0.1 mm × 0.02 mm × 0.02 mm is tested. The pro- altering. For example, “DX” refers to the case where the dimensions
gram then loops back to the beginning, increasing all three of the D and X increase by 2% each loop but L remains constant and “DFX”
particle’s dimensions by 2% and then runs again from the beginning. refers to the case where all three dimensions alter.
Comparing the CFD and analytical approaches in Fig. 11 shows very All previous tests with a varying volume had shown a trend
similar trends, however, in this instance the discrepancies are more that an increase in particle volume would lead to a decrease in the
pronounced. This enhanced overestimation is presumably caused impact velocity. However, Fig. 12 shows that this is not always the
by the negligence of the effect of larger particles on the micro-jet case and is highly dependent on the shape variation which leads
flow (i.e. particle-flow coupling). to an alteration in the velocity. In the case marked ‘DLX’ where all
three parameters increase at the same rate the particle velocity
6.2.3. Change in particle shape decreased in a similar manner to that shown in Fig. 6. The same
The analytical approach has also been used to investigate the is true for the cases marked ‘DX’3 and ‘LX’. However, when only
effects of shape as well as mass on the impact velocity and kinetic one dimension is altered and when D and L change together an
energy. Seven simulations were performed each starting with the increase in volume leads to the increase of impact velocity. The pos-
same 0.1 mm × 0.02 mm × 0.02 mm particle as before. One simula- sible explanation can be offered as follows. In previously considered
tion was performed with all three dimensions increasing by 2% for cases, the increase of particle volume is achieved by scaling up all of
each loop; three simulations allowed two dimensions to increase its dimensions proportionally, remaining its shape unchanged and
by 2% while the third remained constant and the final three simula- so its drag coefficient CD . However, for the cases where only one
tions allowed one dimension to alter while the other two remained dimension changes, there is a dramatic change in shape. In particu-
constant. The results are shown in Figs. 12 and 13. The dimension lar for the cases where the increased particle surface is orthogonal
to the flow direction, particle’s drag coefficient CD increases dra-
matically and leads to a overwhelmingly larger drag force FD that
makes the impact velocity of the larger particle increases.

6.3. Further discussion

In reality, particle size varies, for instance in silt-laden rivers,

from smaller than 10–60 ␮m to larger than 500 ␮m. For flows
without cavitation, the nature of silt erosion is dominated by the
interaction between the particle and the flow structure subject to
particle size as briefly described below, for details referring to Li
and Carpenter [12].
For particles smaller than 10–60 ␮m, it is almost impossible to
eliminate them from entering the turbines, because they are always
present in water as suspended load no matter how good the scheme
design is. These fine particles, owing to their light nature, respond
strongly to turbulences and thus gain the erosive energy, causing

3
Fig. 13. Effect of particle volume on impact energy of a particle (variable dimensions The ‘DX’ curve shows in general a trend of reduction though it is not
shown in legend). monotonously decreasing.
954 P.J. Dunstan, S.C. Li / Wear 268 (2010) 946–954

damage to the material. After striking the material each time, the be true in reality. However, we think it worth further investigation
particle will regain energy when it is brought to the outer layer by by using more vigorous approach. As we pointed out before the
the turbulence and thus hits repeatedly the surface. The particles infinite condition of the flow around the particle is no longer valid
may gain highly rotational speed from highly rotating small vor- when the particle size increases to the order comparable with the
tices near the wall, and thus cause abrasive damage to the material. micro-jet’s diameter which is just the case for larger particles in our
Therefore, turbulence structure in the boundary layer dominates calculation. This may lead to overestimations of the impact veloci-
the erosion process. ties for large particles and thus their impact energies, shading the
For large particles sized above 500 ␮m, the inertia force dom- peaking feature of the impact energy variation with particle size.
inates their behaviour. They virtually do not respond to small Some calculations show that mass increase has led to a subse-
turbulent structures. Their erosive power comes from the main flow quent rise in velocity. This occurs only in those cases where the
where they are entrained. The centrifugal forces encountered in the volume change is combined with a significant alteration in parti-
flows such as flow over concave surface and secondary vortex flows cle profile and as such it has been postulated that the difference in
etc may shift these particles towards the wall material. Owing to the drag coefficient due to the shape change has increased the velocity.
significant momentum of these particles they usually cause severe As demonstrated by Fig. 13, the relationship between a particle’s
damage to the material. mass/volume and its impact kinetic energy is strongly affected by
For particles with size between these two extremes cases, a com- its shape. In general, if particles with longer and thinner profile
bined mode may exist subject to flow structure encountered, or one are likely to have more damage potential than their more spherical
mode is more dominant than the other. counterparts.
However, if cavitation presents in the flow and the particle Clearly a bubble which produces a faster micro-jet will make
is entrained in the micro-jet generated from the collapsing bub- an entrained particle possess increased damage potential. Work by
ble, the damaging power mainly stems from the high velocity jet Plesset and Chapman [10] has shown that an important param-
which accelerates the particle to a remarkably high velocity, hit- eter in determining a micro-jet’s velocity is the ratio  between
ting the already indented pit (caused by micro-jet) violently. It is the distance S from the bubble centre to a solid boundary and its
thus enhanced the damage. For details, readers are referred to Li maximum radius Rmax . Data obtained from the software package
[2] which should be regarded as the part 1 of this subject. Dynaflow have tended to agree with this. As a result it has been
found that an increase in  results in a larger damage potential.
7. Conclusions Though the number of our calculated cases is limited, the damage
potential does show the trend that the impact kinetic energy is
Using the simplified numerical approach has proven the hypoth- linearly related to  2 .
esis [1,2] that the particles suspended in the micro-jet generated
from collapsing bubbles can be accelerated to extremely high speed Acknowledgement
compatible with the micro-jet itself within short distance and time
duration subject mainly to the particle shape and mass. The parti- The financial support for the EPSRCs WIMRC Phases I & II projects
cle possessing such high velocity and kinetic energy further attack (R.ESCM.9001, 9219 and 9217) is acknowledged.
the micro-jet damaged area on the nearby boundary walls and
thus results in an enhancement of silt particle on cavitation ero- References
sion. It is also demonstrated that in most cases in our calculation
the sharp-edge is turned forward which creates a most effective [1] S.C. Li, Cavitation enhancement in silt erosion: obstacles & way forward, in:
Proceedings of Fifth International Symposium on Cavitation, Osaka, Japan,
cutting.4 November 1–4, 2003.
Apart from the simplified CFD approach, an even simpler ana- [2] S.C. Li, Cavitation enhancement of silt erosion: an envisaged micro model, Wear
lytical approach has been tested. The results show the same 260 (2006) 1145–1150.
[3] H. Jin, F. Zheng, S. Li, C. Hang, The role of sand particles on the rapid destruction
trends as those revealed by the CFD approach in particular about of the cavitation zone of hydraulic turbines, Wear 112 (2) (1986) 199–205.
the kinetic energy (damage potential). Though the magnitudes [4] P.J. Dunstan, Research thesis (Master of Science), School of Engineering, Uni-
of the analytical results show slightly discrepancies it offers a versity of Warwick, September 19, 2007.
[5] M.R. Maxey, J.J. Riley, Equation of motion for a small rigid sphere in a non-
very simple but effective method for evaluating the dynamics and
uniform flow, Physics of Fluids 26 (4) (1983) 883–889.
damaging potential of suspended particles in cavitation if con- [6] C. Yin, L. Rosendahl, S.N. Kaer, H. Sorenson, Modelling the motion of cylindri-
sider such a simplification has been made that no boundary wall cal particles in a non uniform flow, Chemical Engineering Science 58 (2003)
exits. 3489–3498.
[7] R.P. Chabbra, L. Agarwal, N.K. Sinha, Drag on non-spherical particles: an evalu-
As the impact kinetic energy is proportional to the mass (i.e. vol- ation of available methods, Powder Technology 101 (1999) 288–295.
ume in our case) of the particle and the square of velocity, it was [8] G.H. Ganser, A rational approach to drag prediction of spherical and nonspher-
expected that with the increase of mass (volume) the increase of ical particles, Powder Technology 77 (1993) 143–152.
[9] S.F. Hoerner, Fluid Mechanics Drag, Hoerner Fluid Dynamics (Published by the
impact kinetic energy due to the mass increase will be offset by the author), New York, 1965.
decrease of impact velocity at some point. However, the calculation [10] M.S. Plesset, R.B. Chapman, Collapse of an initially spherical vapour cavity in
results though show the decreasing trends of the impact velocity the neighbourhood of a solid boundary, Journal of Fluid Mechanics 47 (1971)
283–290.
with the increase of mass (volume); the deduction of the impact [11] M. Dular, B. Stoffel, B. Sirok, Development of a cavitation erosion model, Wear
velocity does not grow fast enough at some stage to offset the 261 (2006) 642–655.
kinetic energy increase contributed by the mass (volume) increase. [12] S.C. Li, P.W. Carpenter, 4.1.2 models in anti-erosion turbine technology, Techni-
cal Report (Revised Version, 30 January 2004), Fluid Dynamics Research Centre,
The calculated curves render monotonous increases instead of a Warwick University, UK, 2004, pp. 10–12.
peak (maximum) at some point. This unexceptional result might

4
There are a few cases if the initial angle of attack is in the stall region that
may results in funny oscillating rotating movement of particle with sharp-edge
backwards while travelling towards the targeted wall. This feature is being studied
further and will be reported in due course.