You are on page 1of 11

Sixth International Symposium on Cavitation

CAV2006, Wageningen, The Netherlands, September 2006

EXPERIMENTAL AND NUMERICAL MODELING OF CAVITATION EROSION

Matevž Dular, Brane Širok


Laboratory for Water Turbine Machines, University of Ljubljana, Slovenia
matevz.dular@fs.uni-lj.si

Bernd Stoffel
Chair of Turbomachinery and Fluid Power, Darmstadt University of Technology; Germany

ABSTRACT k1 coefficient
A study of visual and erosion effects of cavitation on k2 coefficient
simple single hydrofoil configurations in a cavitation p pressure
tunnel was made. A two-dimensional hydrofoil with pv vapour pressure
circular leading edge was used for the experiments. In P(mj) micro-jet occurrence probability
addition, the hydrofoil geometry was modified to obtain Pwave pressure wave power
some three-dimensional cavitation effects. A thin copper Re vapour generation source term
foil, applied to the surface of the hydrofoil, was used as an Rc vapour condensation source term
erosion sensor. The cavitation phenomenon on hydrofoils v velocity
at different flow conditions (system pressure, water gas vcrit critical velocity
content, flow velocity) was observed. Images of vapour vjet jet velocity
cavities from above and from a side view were taken. vm mixture velocity
Plausible results that showed a significant relationship t time
between cavitation erosion and the visual effects of x distance from cloud implosion
cavitation made it possible to use these information to α void fraction
develop a cavitation erosion model. The model is based on φ water gas content
the physical description of different phenomena, which are γ liquid surface tension
involved in the process of pit formation. The model is µ viscosity
capable to predict the influences of significant parameters µm mixture viscosity
as flow velocity and gas content of water. ρ density
The final goal of the study is numerical prediction of
ρl liquid density
cavitation erosion. Therefore a CFD simulation of a
ρm mixture density
developed cavitating flow was pursued. Numerical images
of cavitation structures were generated and used as an ρv vapour density
input to the cavitation erosion model. Numerically σ cavitation number
predicted distribution and magnitude of cavitation erosion τ time of exposure to cavitation
are in a relative good agreement to the measured values. ω angular velocity
Results promise a good possibility of accurate prediction of
cavitation erosion in real hydraulic machines in the future. INTRODUCTION
Though there exist many empirical rules developed to help
NOMENCLATURE engineers to evaluate the potential cavitation damage rate
Apit pit area in a given application (Hammit 1979), there remain a
Aref reference area number of basic questions regarding the fundamental
Arel part of eroded surface mechanisms involved. Recently there were many attempts
c sonic velocity to predict the magnitude of the cavitation erosion. For
Cc empirical constant example Pereira et al. (1998) found a relation between the
Ce empirical constant volume of transient cavities and its rate of production to
ES integral part of eroded surface the material deformation energy. Fortes-Patella et al.
f shedding frequency (2004) suggested that the damage of the solid surface is a
fg gas mass fraction consequence of a sequence of events – from cavitation
fv vapour mass fraction cloud collapse to the spherical implosion of a single bubble
k turbulence kinetic energy that causes the damage. Present authors Dular et al. (2004)

1
and Širok et al. (2002) suggested that a strong correlation Water quality, which can be measured by its content of
between erosion of the surface and visual cavitation dissolved and undissolved gasses φ, was controlled using
structures exists. However an attempt to include the whole the bubble generator system. For the majority of tests the
sequence of events, that lead to the appearance of damage gas content was kept at a minimum possible level -
on the surface, into a method of damage prediction, has not approximately 14.3 ± 0.5 mgg/lw (milligram of gas per liter
yet been successfully made. of water). Additionally three measurements were made at
The present paper is a conclusion of a systematical higher water gas contents (25.9 mgg/lw, 36.7 mgg/lw and
research of cavitation erosion at Darmstadt University of 48.9 mgg/lw) to study the influence of this parameter on the
Technology. A number of experimental and numerical aggressiveness of cavitation erosion.
studies were made by present authors (Dular et al. 2004a,
Širok et al. 2002, Dular et al. 2005). Some ideas can also Cavitation image capturing
be found in the PhD theses of Boehm (1998), Hofmann A CCD camera SensiCam with sensor CCD-Interline
(2001), Lohrberg (2001) and Bachert (2004). Progressive Scan was used. Images were captured at 8 bit
The model gives a new option of cavitation erosion resolution in m-jpeg format.
prediction and more importantly embraces the theories of Fig. 2 shows shedding of the cavitation structures on the
cavitation cloud collapse (Shimada et al. 1999, Brennen ALE25 hydrofoil at low gas content and at a cavitation
1995), attenuation of the pressure wave (Beranek 1996), number σ = 2.3 from top view. Flow is from left to right. It
micro-jet formation (Plesset & Chapmann 1971) and is obvious that the cavitation zone is asymmetrical and that
finally pit formation (Lush 1983). cavitation cloud separation occurs only in the region where
The commercial CFD code Fluent 6.1.18 was used for 3D the length of the hydrofoil is the shortest while it is steady
transient simulations of cavitating flow. A cavitation on the other side.
model, based on bubble dynamics equations, described in
Singhal et al. (2002) is included in the code and it was used
to describe the unsteady behaviour of cavitation. A very
good prediction of different features (velocity distribution,
cavitation cloud shedding frequency, pressure distribution
etc.) of cavitating flow was achieved. Images of
numerically predicted cavitation structure shapes were
generated and were used as an input to the cavitation
erosion model.
The results of model prediction are compared to the
experimental results of pitcount measurements on single
hydrofoils.

EXPERIMENTAL INVESTIGATIONS Fig. 2: Cavitation cloud separation on ALE25 hydrofoil at


Three simple hydrofoils were used. The basic geometry is
cavitation number σ = 2.3
a 50 mm wide, 107.9 mm long symmetric hydrofoil with a
circular leading edge and constant thickness of 16 mm,
Mean value and standard deviation of gray level in the
having a wedge shape near the trailing edge (CLE –
images was calculated (50 images of each cavitation
Circular Leading Edge hydrofoil). In order to obtain three-
condition were used for statistical evaluation).
dimensional cavitation effects the basic geometry was
modified by sweeping back the leading edge at an angle of
15 and 25 degrees (ALE15 and ALE25, respectively; ALE
– Asymmetric Leading Edge hydrofoil) (Fig. 1).

Fig. 3: Mean value and standard deviation of gray level on


ALE25 hydrofoil at cavitation number σ = 2.3

Results of statistically evaluated cavitation condition for


Fig. 1: Copper coated CLE, ALE15 and ALE25 hydrofoils ALE25 hydrofoil at a cavitation number σ = 2.3, flow
velocity v = 13 m/s and low gas content φ = 14.7 mgg/lw in
The velocity in the reference plane upstream of the Fig. 3. Left images represent the mean value (scaled to 0 –
hydrofoil was held constant at 13 m/s, except for two black, 250 - white) while the right ones represent standard
measurements at 10 and 16 m/s. Developed cavitating flow deviation of gray level (scaled to 0 – white, 50 - black).
was observed at 5° incidence angle and at different values The upper two images show the top view and the bottom
of cavitation number (2.5, 2.3, 2.0). two images show the side view.

2
If one compares Fig. 3 to Fig. 2 a distinctive relation
between the region of cavitation cloud separation and the
maximum of the standard deviation of gray level can be
seen.

Cavitation erosion tests


Due to the time limitation of the experiment only damage
in the incubation period was studied (where damage is
already present but there is no material loss).
To get the information about the erosion on the whole
surface of the hydrofoil, a polished copper foil, 0.2 mm Fig. 5: Cavitation damage on ALE25 hydrofoil at
thick, was fixed to its surface using an adhesive film. The cavitation numbers σ = 2.0, 2.3 and 2.5
hardness of the copper coating was approximately 40 HV.
A sufficient number of pits was obtained after 1 hour Fig. 5 confirms our hypothesis that the cavitation erosion is
exposure to the cavitating flow (the exposure time was conditioned by the separation of cavitation clouds. The
constant for all operating conditions). damaged surface in predominantly in the region near the
Pits have a diameter in order of magnitude 10-5 m, and can edge where the length of the hydrofoil is the shortest – in
be distinguished only by sufficient magnification. Images the same region where unsteady cavitation (with cloud
of the pitted surface were acquired using an Olympus BX- separation) is present.
40 microscope and a CCD camera.
The intensity of cavitation erosion was determined with the
pit-count method. The method is based on the assumption
that the area of the pitted surface and the number of pits
that are created by bubble implosions (micro-jet impacts)
in a certain time of exposure to cavitating flow give a
quantitative measure of the intensity of cavitation erosion.
The pit-count method gives a distribution of the number
and the area of the pits and consequently the distribution of
the magnitude of cavitation erosion on the surface.
Results of measurements are presented in Figs. 4-7. Fig. 6: Cavitation damage on CLE hydrofoil at cavitation
Contour diagrams are a result of an interpolation of pit- number σ = 2.3 at flow velocities v= 10, 13 and 16 m/s
count measurements at approximately 925 positions on the
hydrofoil surface. The flow is from bottom to top. The Tests with variable flow velocity revealed an obvious
results of surface damage are scaled from 0 % surface relation between flow velocity and erosion rate. The
damage...white to 10 % surface damage...black (in case of cavitation is much more aggressive at higher flow
variable flow velocity the scale reaches to 12 % surface velocities. The distribution of the pits and the position of
damage). The value of eroded surface (ES) represents the the maximum magnitude of damage are similar for all the
part of the whole surface that is damaged (covered by pits). tests since cavitation number was held constant (σ = 2.3) -
the topology of cavitation structures practically does not
change when the gas content is altered.

Fig. 4: Cavitation damage on CLE hydrofoil at cavitation


Fig. 7: Cavitation damage on CLE hydrofoil at cavitation
numbers σ = 2.0, 2.3 and 2.5
number σ = 2.3 at gas contents α = 13.8, 25.9,
Fig. 4 reveals that, as expected, the region of maximum 36.7 and 48.9 mg/l
cavitation damage moves away from the leading edge
when cavitation number is decreased. Interestingly, the One can see that erosive aggressiveness decreases
maximum value of total damaged surface (ES) occurs at significantly when the gas content rises. The surface
cavitation number 2.3 and not at 2.0 as expected. The sustains almost 50 times less damage in the case with high
reason lies in the fact that the region of cloud separation gas content (α = 48.9 mgg/lw) than in cases with low gas
lies downstream of the copper coated surface and also content (α = 13.8 mgg/lw). Similarly to the tests with
further away from the hydrofoil surface. variable flow velocity, the position of maximum erosion
magnitude and the distribution of pits on the hydrofoil

3
remain almost constant for all the cases, since the ⎛ dV ⎞
cavitation number was constant. Pwave = ∆p⎜ ⎟, (1)
⎝ dt ⎠
The previous study (Dular et al. 2004a) showed that the
magnitude of cavitation damage is correlated with where ∆p is the difference between the surrounding
unsteadiness of cavitation structures. For example, the pressure and vapour pressure (psur – pv) and dV/dt is the
effect can be seen if we compare Fig. 2, Fig 3 and Fig. 5 change of the vapour cloud volume in time t.
(middle hydrofoil - σ = 2.3). Fig. 2 shows a sequence of From the acoustics the magnitude of the emitted pressure
cavitation images on ALE25 hydrofoil. Statistical wave is proportional to the square root of its power (p0 ∝
evaluation (Fig. 3) reveals that the standard deviation has
Pwave ). If we consider the pressure difference ∆p to
its maximum in the region where cavitation cloud
separation occurs (in the region where the hydrofoil length remain approximately constant, we can write that the
is the shortest). It is obvious (from Fig. 5 - middle distribution of the mean change in cavitation cloud volume
hydrofoil - σ = 2.3) that most of the damage also occurs in on the hydrofoil reveals the mean distribution of amplitude
this region. Results point out that the standard deviation of of the pressure wave that is emitted by the cavitation cloud
gray level is a variable that could be included in a collapse.
cavitation erosion model as a measure of unsteadiness, Since we have no possibility of measuring the
which is indirectly linked to the magnitude of cavitation instantaneous change of the cavitation cloud volume a
erosion. standard deviation of gray level σ was used as the
parameter, which is related to the power of the emitted
THEORETICAL MODEL pressure wave. Standard deviation can be used in this
A successful prediction of cavitation erosion still remains manner since it is a function of the change of the gray level
one of the big goals in this field of research. As shown in the image, which is a function of the cavitation cloud
above one could use visualization techniques as an input volume:
for cavitation erosion model to determine the region and
possibly the magnitude of cavitation damage. ⎛ dV ⎞
gray level = f (V ) ⇒ σ ∝ ⎜ ⎟. (2)
The process of pit formation is very complex. The ⎝ dt ⎠
presented theory explains the pit formation in the following
way (Fig. 8): The hypothesis is that the relation between the time
• Collapse of the cavitation cloud causes a shock wave derivative of vapour cloud volume and the distributions of
that spreads in the fluid. standard deviation of gray level from side and top view
• The magnitude of the shock wave is attenuated as it exists. A possible measure of the emitted pressure wave
travels toward the solid surface. power Pwave can be formulated in the following way:
• Single bubbles are present near the solid surface. As the
⎡⎛ 1 M
⎞ ⎤
∑σ
shock wave reaches them, they begin to oscillate and a
micro-jet phenomenon can occur. Pwave (n, l ) ∝ ∆p ⋅ ⎢⎜ m ⎟ ⋅σ n ⎥ , (3)
• The damage (single pit) is caused by high velocity liquid ⎣⎝ M m ⎠ ⎦ l =const .
jet impact to the solid surface.
where ∆p is the mean pressure difference, M is the height
of the matrix (see Fig. 9), l, m and n are direction vectors
(see Fig. 9) and σm and σn are standard deviations of gray
level in distribution matrix of side and top view,
respectively (see Fig. 9).
Fig. 8: Cavitation erosion model idea

If we take into account only the incubation period (surface


is plastically deformed but no material loss is present) we
can say that the eroded surface of a submerged body is a
result of repetition of abovementioned processes.

Amplitude of the emitted pressure wave


The power and consequently the magnitude of the emitted
pressure wave are closely related to the velocity of the
change of the vapour cloud volume (velocity of cavitation
cloud collapse) and to the surrounding pressure (Fortes-
Patella et al. 2004). We can write the following relation:

Fig. 9: The distribution of standard deviation of gray


level in the coordinate system.

4
compressive stress is reached and then behaves as a
It was found that a linear function shows the best perfectly plastic solid, for which the stress will remain
correlation to the experimental results (the relation was constant. The deformation only occurs if the water hammer
obtained by comparison of model predictions and results of pressure is higher than the limit pressure at which the
experimental measurements from Hofmann (2001): plastic flow of the material occurs. In other words: “the
liquid jet velocity must exceed some critical velocity at
p 0 (n, l ) = k1 ∆p ⋅ ξ (n, l ) ; (4) which the damage occurs". The critical liquid jet velocity
at which the stress, high enough to produce a plastic flow
of the material py, is reached is:
Distance of the cloud implosion
The position of the cloud collapse and consequently the
distance of cloud collapse from the hydrofoil surface x can ⎛ − ⎞
1
py ⎜ ⎛ py ⎞ n ⎟
vcrit = ⎜1 − ⎜1 + ⎟ ⎟,
be determined from the standard deviation of gray level
ρ l ⎜ ⎜⎝ B ⎟⎠ ⎟
(8)
from the side view (see Fig. 9). An equilibrium function
was used: ⎝ ⎠

∑ (σ ⋅ m) − m
M
where py is the yield stress of the material.
x(l ) = m m
h, (5) The duration of the water hammer stress is as long as the
∑σ
M
m
time for the impact signal to traverse the radius of the jet
m
(rjet):
where mh is the position of the hydrofoil surface at a
specified l. The distance of cloud implosion had to be r jet
considered constant for each value of l, what represents a
t def = . (9)
c
considerable simplification especially for the cases with
asymmetrical hydrofoils. . After that time a stagnation pressure 1/2ρv2 is established.
It is unlikely that any damage occurs in this period since
Attenuation of the pressure wave the stagnation pressure is an order of magnitude smaller
As the pressure wave travels away from its source its than the water hammer pressure (unless exceptionally high
energy is gradually converted into heat. For our problem impact velocities are encountered).
the main energy loss mechanism are the viscous losses If we consider only the center of the impact where only
generated from the friction within the fluid itself (Beranek motion (plastic flow) normal to the surface is present the
1996). The pressure wave attenuation with the distance x maximum depth of the pit can be calculated:
can then be written as:

2 ω2
d pit = v def t def . (10)
− µ⋅x
3 ρ0c3
p = p0 ⋅ e −δ ⋅ x = p0 ⋅ e , (6)
Pit geometry
where ω is the pressure wave frequency, µ is the viscosity The ratio between the pit radius and pit depth is not
and c is the sonic velocity. µ and c are functions of the constant. Previous investigations showed that it usually lies
local vapour volume fraction αv. between 15 and 30 but can also be as low as 2 and as high
as 500 (Fortes-Patella et al. 2000). The mean ratio between
Formation of micro-jet the pit radius and pit depth (26.7) was determined from the
A well know theory developed by Plesset & Chapmann laser profilometry measurements of the pitted surface of a
(1971) determines the jet velocity under the following copper specimen from the hydrofoil from the same
assumptions: fluid inside the jet behaves as a solid cavitation tunnel done at LEGI – Grenoble (Reboud et al.
material, the potential energy of the bubble is fully 1999).
transformed into kinetic energy and the Kelvin impulse is The measure of the damage caused by cavitating flow (the
converted to the jet impulse. Micro-jet velocity is then: part of the damaged surface after a certain amount of time
τ) can now be written:
p − pv
v jet = k 2 , (7) A pit ⋅ τ ⋅ f ⋅ P(mj )
ρ Arel = , (11)
Aref
Pit formation
The water hammer stress applied to the material at the where τ is the duration of the exposition to the cavitating
impact of the liquid micro-jet can be considered as the flow, f is the frequency of cavitation cloud implosion,
main mechanism of damage to the solid surface (Plesset & P(mj) is the probability of micro-jet occurrence and Aref is
Chapmann 1971). In the present approach (Lush 1983) the the reference area. The Eqn. 11 is only valid in the
surface responds as a perfectly rigid solid until a certain

5
incubation period of cavitation erosion where no mass loss The values of densities and viscosities of water and water
is present. vapour (ρl, ρv, νl, νv) and the sonic velocity (cl) and water
vapour pressure (pv) correspond to the ambient temperature
Influences of fluid properties of 20 °C.
Experimental results show obvious influences of fluid A value of f = 0.75 MHz was chosen for the frequency of
properties on cavitation erosion. Is was shown that when the pressure wave that is emitted by cavitation cloud
the experiment was conducted in water with high gas collapse on the basis of theoretical and experimental
content cavitation erosion was less aggressive. The main studies (Shimada et al. 1999, Lohrberg 2001) and also
reason lies in the fact that the sonic velocity is lower in based on measurements of pressure waves on similar
water with high gas content. Consequently compressibility geometries done by Hofmann (2001).
and pressure wave attenuation are higher. Since the local gas volume fraction α was not a subject of
The density and viscosity of the water considering the measurements, it was estimated that (right after the cloud
presence of gases are: collapse) it is in the same order as the volume fraction of
the initial gas content of the water φ. The values from α ≈
ρ = αρ g + (1 − α ) ρ l (12) 0.01 for low gas content to α ≈ 0.04 for high gas content
relate well to the results of experimental measurements of
and void fraction within cavitation by Stutz and Reboud
(2000).
It is possible to compare results of experimental
ν = αν g + (1 − α )ν l . (13) measurements and model prediction on CLE geometry at a
cavitation number 2.3, flow velocity 13 m/s and low gas
Influences of flow velocity content φ = 13.8 mgg/lw. The histograms could reveal if the
Previous investigations (Dular et al. 2005, Knapp et al. model predictions fall within the expected (still physical)
1970) have shown that the flow velocity has an enormous range of values. Measurements of pressure on the hydrofoil
influence on the aggressiveness of cavitation. A power law surface (Hofmann 2001) were compared to model
with coefficient n = 5…8 was found. Here we state some prediction. Additionally results of present pit-count
physical explanations of the velocity influence that are measurements (local relative damaged surface Arel) were
included in the cavitation erosion model. compared to the model results.
• When the cavitation number is constant the pressure
difference has to increase with the power of 2 when
velocity is increased. This means that the pressure
emitted at bubble cloud collapse will also rise with the
power of 2.
• When the cavitation number is constant the pressure
difference has to increase with the power of 2 when the
flow velocity is increased. Higher system pressure acts
on the compressibility of the fluid. Fig. 10: Histograms of experimentally measured and
predicted pressure and local relative damaged surface
• Past experiments have shown that the Strouhal number
increases linearly when the velocity is increased (for the One can see that the model correctly predicts the histogram
same cavitation number) (Böhm 1998), hence shedding of pressures on the surface of the hydrofoil (Fig. 10). It
frequency will rise with the power of 2 when the flow also relates well to the histogram of values of the local
velocity is increased. relative damaged surface from the pit-count measurements.
• There is a finite number of bubbles that have a potential The model was used for prediction of surface damage
to form a micro-jet in the flow. Since the time of bubble caused by cavitation on 3 different hydrofoils. The
implosion (a few µs) is much smaller then the time diagrams have the same form as the experimental ones
needed for transition of a bubble through the control (Figs. 4-7). The flow is from bottom to top. The results of
volume (a few ms) the probability for a bubble to prediction of surface damage are scaled to 0 % surface
implode in micro-jet form does not alter with velocity damage...white and 10 % surface damage...black (in case
(this hypothesis was also confirmed by past experiments of variable flow velocity the scale reaches to 12 % surface
(Knapp et al. 1970) Hence when the velocity is damage). The value of eroded surface (ES) represents the
increased, more bubbles implode in the form of micro-jet part of the whole surface that is damaged (covered by pits).
in a certain time period.

RESULTS – SINGLE HYDROFOILS


Setting up the problem needs some iterative work. Guide
lines about the values of the parameters can be found in
past studies of Shimada et al. (1999), Brennen (1995),
Plesset & Chapman (1971), Hofmann (2001) and Lush
(1983).

6
Fig. 11:Predicted damage on CLE hydrofoil at cavitation Fig. 13:Predicted damage on CLE hydrofoil at cavitation
numbers σ = 2.0, 2.3 and 2.5 number σ = 2.3 at flow velocities v= 10, 13 and 16 m/s

We see that the predicted surface damage for the CLE Finally the ability of the model to predict the influence of
hydrofoil relates very well to the experimental results water gas contents was put in question.
(comparing Figs. 4 and 11). If we compare model
prediction and experimental measurements for the case of
cavitation number 2.0, we see that the model predicts the
damage to be concentrated in the region near the end of the
plane part of the hydrofoil, while the experimentally
measured damage stretches towards the leading edge and is
not so concentrated. The predicted distributions at
cavitation numbers 2.3 and 2.5 relate even better to the
experimental results. The model also correctly predicts the
Fig. 14: Predicted damage on CLE hydrofoil at cavitation
cavitation at cavitation number σ = 2.3 to be the most
number σ = 2.3 at gas contents α = 13.8, 25.9,
aggressive one.
36.7 and 48.9 mg/l

The diagrams in Fig. 14 show that the model responds very


well to the change of water gas content. When the gas
content increases the aggressiveness decreases
dramatically.

SIMULATION OF CAVITATING FLOW


CFD simulation of cavitation was performed. A program
package Fluent 6.1.18 was used to calculate the flow. The
accuracy of the simulations was previously evaluated
Fig. 12:Predicted damage on ALE25 hydrofoil at (Dular et al. 2004b) and the results were in a good
cavitation numbers σ = 2.0, 2.3 and 2.5 agreement to the measurements.

As a consequence of the hypothesis that cavitation damage Multiphase model


is conditioned by cavitation cloud separation (which is A single fluid (mixture phase) approach was used. The
characterized by higher values of standard deviation of basic approach consists of using the standard (Navier-
gray level), the model correctly predicts the damage to Stokes) viscous flow equations and a conventional
occur predominantly in the region where the hydrofoil turbulence model (RNG k-ε model). The mass (eq. 4) and
length is the shortest, for ALE25 (Fig. 12). The predicted momentum (eq. 5) conservation equations together with
damage is somewhat more evenly distributed over the the transport equation (eq. 6) and the equations of the
hydrofoil than experiments have shown (comparing Figs. 5 turbulence model form the set of equations from which the
and 12). The model wrongly predicts a significant stretch fluid density (which is a function of vapour mass fraction
of damage reaching from the “main damage cluster” fv) is computed. The ρm - fv (mixture density – vapour mass
toward the region where the hydrofoil length is the longest. fraction) relationship is:
The stretch cannot be seen in experimental results (Fig. 5)
and is a result of small oscillations of quasi-steady
1 fv 1− fv
cavitation in this region that contribute to higher values of = + (14)
standard deviation of gray level (seen also in standard ρm ρv ρl
deviation diagram in Fig. 3) but have nothing to do with
cloud separation.
The volume fraction of the vapour phase (αv) is related to
We can see that the model correctly predicts a significant
of the mass fraction of the vapour phase with:
decrease of aggressiveness of cavitation when the flow
velocity is decreased (Fig. 13).

7
ρm leads to a quasi steady behaviour of the cavitation sheet,
αv = fv . (15) which becomes stabile. Also the overall length of the
ρv predicted cavity structure is about 50 % too short
The mass conservation equation for the mixture is: compared to experimental results. The problem seems to
lie in the overprediction of the turbulent viscosity in the

(ρ m ) + ∇ ⋅ (ρ m vrm ) = 0 .
region of cavity closure. To improve the simulation a
(16) modification of the turbulent viscosity was applied as it
∂t was proposed and successfully proven by Reboud et al.
(1998). In regions with higher vapour volume fractions
The momentum conservation equation for the mixture is: (lower mixture densities) a modification of the RNG k-ε
turbulence model was made by artificially reducing the

(ρ m vrm ) + ∇ ⋅ (ρ m vrm vrm ) = turbulent viscosity of the mixture.
∂t (17)
[ ( r r r v
= −∇p + ∇ ⋅ µ m ∇v m + ∇v mT + ρ m g + F )] Simulation
The computational domain stretched from 10 chord lengths
in front of the hydrofoil to 10 chord lengths behind the
And the transport equation for the vapour is: hydrofoil. The cross sectional dimensions of the domain
were the same as the test section dimensions (50 mm wide

(ρ m f v ) + ∇ ⋅ (ρ m vrm f v ) = Re − Rc .
and 100 mm high).
(18) A C-type structured meshes with about 360000 nodes (Fig.
∂t 15) and time step 2*10-5 s were used. Standard wall
functions were applied, hence the y+ value lies between 30
Cavitation model and 80.
Source terms Re and Rc that are included in the transport
equation define vapour generation (liquid evaporation) and
vapour condensation, respectively. Source terms are
functions of local flow conditions (static pressure, velocity)
and fluid properties (liquid and vapour phase densities,
saturation pressure and liquid vapour surface tension). The
source terms are derived from the Rayleigh-Plesset
equation, where high order terms and viscosity term have
been left out. The derivation of the source terms can be Fig. 15: Computational domain around ALE15 and ALE25
found in Singhal et al. (2002). hydrofoil (the front wall is not represented)

They are given by: Approximately 40 iterations per time step were needed to
obtain a converged solution.
2 ( pv − p)
⋅ (1 − f v − f g ) , (19)
k Conditions applied for the simulation were the following:
Re = C e ⋅ ⋅ ρl ⋅ ρv ⋅ ⋅ • Boundary condition: Imposed velocity at inlet and static
γ 3 ρl
pressure at outlet. The channel walls are also considered
so that the possible wall effects are taken into account.
when pv > p, and by:
• Initial transient treatment: A low velocity was initially
applied to the flow field, for which no vapour appears.
k 2 ( p − pv ) The velocity was then increased until the desired
Rc = C c ⋅ ⋅ ρl ⋅ ρl ⋅ ⋅ ⋅ fv , (20)
γ 3 ρl operating point (cavitation number) was reached. The
turbulence level at inlet boundary condition was set to 3
when pv < p, where Ce and Cc are empirical constants, k is %.
the local kinetic energy, γ surface tension, fv vapour mass • Although vapour-fluid mixture flow can reach sound
fraction and fg mass fraction of noncondensable (dissolved) speed at relatively low velocities, compressibility effects
gases. Ce and Cc were determined by comparing were neglected since Fluent does not allow compressible
experimental and numerical results at different multiphase flow computation.
combinations of initial conditions and geometries, reported
by Singhal et al. (2002); their values are 0.02 and 0.01 It was shown by Coutier-Delgosha et al. (2003) that
respectively. plausible results can be obtained without consideration of
compressibility effects if a modification of turbulent
Turbulence model viscosity is applied. Despite the compressibility effects
The RNG k-ε turbulence model was applied for solving the were neglected is the simulation capable to predict many
transport equations of the turbulent kinetic energy and its features of the developed cavitating flow (like cloud
dissipation rate. The model itself is unable to correctly separation and re-entrant jet) but for example cannot
simulate the unsteady behaviour of cavitation. After an predict the shock waves that are emitted at cavitation cloud
initial fluctuation of the cavity volume, the calculation collapse.

8
Sequence of images of numerical simulation of cavitation The numerically predicted damage distribution shows good
around ALE25 hydrofoil is presented. Isosurfaces of 10 % correlation to the experimentally measured one (Fig. 4).
vapour volume fractions are shown (the value of 10 % When the cavitation number is decreased the model
vapour volume fraction was chosen on the basis of correctly predicts the maximum of damage to move away
previous experience that shows that it relates best to the from the hydrofoil leading edge towards the back of the
real cavity shapes) (Okita & Kajishima 2002). The time hydrofoil. The predicted damage is more evenly distributed
delay between the images is 0.6 ms. As in the experiment along the hydrofoil surface (no significant location of
(Fig. 2) a steady attached cavitation can be seen in the extremely aggressive cavitation can be seen). The model
region where the hydrofoil is the longest. The cloud correctly predicts the cavitation in the case with cavitation
separation occurs only in the region where the hydrofoil is number σ = 2.3 to be the most aggressive one.
the shortest. The cavitation structure firstly grows. The re- Numerically predicted pit distributions for the ALE25
entrant jet (not presented in the sequence) causes the hydrofoils at cavitation number σ = 2.3, low gas content
cavitation cloud separation in the region near the front and flow velocity v = 13 m/s can be seen in Fig. 18.
wall. The separated cloud travels with the flow and
implodes downstream in a higher pressure region. The
implosion of the cloud forms a new re-entrant jet that
causes the next cavitation cloud separation.

Fig. 18: CFD predicted damage on ALE15 and ALE25


hydrofoils at cavitation number σ = 2.3

Similar to the model prediction from the experimental


visualization (Fig. 12) the numerical visualization model
predicts the damage to occur in a extension reaching from
the ``main damage cluster'' towards the region where the
hydrofoil length is the longest. The reason for the error is
the same as before - the small quasi steady oscillations of
cavitation cloud contribute to higher values of standard
deviation in this region. The pit distribution prediction is
Fig. 16: Numerically predicted time evolution of cavitation nevertheless still plausible. The ES values of the model
structure on ALE25 hydrofoil at σ = 2.3 prediction are comparable to the experimentally measured
ones.
CFD PREDICTION OF CAVITATION EROSION In general the cavitation erosion model works better for the
Finally, the results of numerical simulations were used as cases with low gas content. The average discrepancy
an input into the cavitation erosion model. This way, the between model predictions and experimental results was
cavitation erosion was predicted solely on the basis of estimated to less than 20 % for these cases.
numerical simulations and the cavitation erosion model.
CONCLUSIONS
Fig. 17 shows the numerically predicted pit distribution for In previous investigations (Dular et al. 2004a, Širok et al.
the CLE hydrofoil for cavitation numbers σ = 2.0, 2.3 and 2002, Dular et al. 2005, Bachert et al. 2003) erosion on
2.5, low gas content and flow velocity v = 13 m/s. single hydrofoils and its relationship to visual cavitation
structures was studied. It was found that the value of
standard deviation of grey level could be used as the
parameter for describing the unsteadiness of cavitation
and that this parameter could be used as an input for the
cavitation erosion model. Effects of parameters as the
flow velocity and the water gas content were also studied
and some physical explanations were given.
This paper presents the development of the cavitation
erosion model, which is based on the knowledge that was
gained during past measurements. The model is capable
of predicting the magnitude and distribution of damage
Fig. 17: CFD predicted damage on CLE hydrofoil at caused by cavitation erosion. It is also capable to
cavitation numbers σ = 2.0, 2.3 and 2.5

9
reproduce the influences of flow velocity and water gas Flow Agrssiveness in Cavitation Erosion’, Workshop on
content. Cavitation Erosion, Bassin d'essais des carenes, Val de
A CFD simulation of cavitating flow was performed and Reuil, France, 2004
the numerical results were used as an input data to the
cavitation erosion model. The numerically predicted Hammitt F. G., ‘Cavitation Erosion, The State of The Art
cavitation erosion distribution and magnitude agree And Predicting Capability’, Appl. Mech. Rev., 32 (6), pp.
relatively well to the measurements. 665–675, 1979
The presented results promise a good possibility of
prediction and control of damage caused by cavitation Hofmann M., ‘Ein Beitrag zur Verminderung des erosiven
erosion in real hydraulic machines. Potentials kavitierender Stömungen - PhD Thesis’,
Technische Universität Darmstadt, 2001
REFERENCES
Bachert B., ‘Zusammenhang zwichen visueller Knapp R.T., Daily J.W., Hammitt F.G., ‘Cavitation’,
Erscheinung und erosiver Agressivität kavitierenden McGraw-Hill Book Company, London, 1970
Strömungen - PhD Thesis’, Technische Universität
Darmstadt, 2004 Lohrberg H., ‘Messung und aktive Kontrolle der erosiven
Aggressivität der Kavitation in Turbomaschinen - PhD
Bachert B., Ludwig G., Stoffel B., Širok B., Novak M., Thesis’, Technische Universität Darmstadt, 2001
‘Experimental Investigations Concerning Erosive
Aggressiveness of Cavitation in a Radial Test Pump With Lush P.A., ‘Impact of a liquid mass on a perfectly plastic
the Aid of Adhesive Copper Films’, Procedings of the solid’, Journal of Fluid Mechanics, Vol. 135, pp. 373-387,
Fifth International Symposium on Cavitation, Osaka, 1983
Japan, 2003
Okita K., Kajishima T., ‘Three-Dimensional Computation
Beranek L.L., ‘Acoustics’, Acoustical society of America, of Unsteady Cavitating Flow in a Cascade’, The 9th of
New York, 1996 International Symposium on Transport Phenomena and
Dynamics of Rotating Machinery Honolulu, Hawaii,
Böhm R., ‘Erfassung und hydrodynamische 2002
Beeinflussung fortgeschrittener Kavitationsustände und
ihrer Aggressivität - PhD Thesis’, Technische Universität Pereira F., Avellan F., Dupont Ph., ‘Prediction of
Darmstadt, 1998 Cavitation Erosion: An Energy Approach’, Journal of
Fluids Engineering, Vol. 120, pp. 719-727, 1998
Brennen C.E., ‘Cavitation and Bubble Dynamics’, Oxford
University Press, New York, Oxford, 1995 Plesset M.S., Chapman R.B., ‘Collapse of an Innitially
Spherical Vapour Cavity in the Neighbourhood of a Solid
Coutier-Delgosha O., Fortes-Patella R., Reboud J.L., Boundary’, Journal of Fluid Mechanics, vol. 47, pp. 283-
‘Evaluation of turbulence model influence on the 290, 1971
numerical simulations on unsteady cavitation’, Journal of
Fluids Engineering, 125, pp. 38-45, 2003. Reboud J.L., Fortes-Patella R., Archer A., ‘Analysis of
damaged surfaces: Part I: Cavitation mark measurements
Dular M., Bachert B., Stoffel B., Širok B., ‘Relationship by 3D laser profilometry’, Proceedings of the 3rd ASME /
between cavitation structures and cavitation damage’, JSME Joint Fluids Engineering Conference, San
Wear 257, pp. 1176–1184, 2004 Francisco CA, 1999

Dular M., Bachert R., Stoffel B., Širok B., ‘Experimental Reboud J.L., Stutz B., Coutier O., ‘Two-phase flow
evaluation of numerical simulation of cavitating flow structure of cavitation: experiment and modelling of
around hydrofoil’, European Journal of Mechanics - unsteady effects’, Third International Symposium on
B/Fluids, Vol. 24 (4), pp. 522-538, 2005 Cavitation, Grenoble, France, 1998

Dular M., Širok B., Stoffel B., ‘Influence of gas content Shimada M., Kobayashi T., Matsumuto Y., ‘Dynamics of
in water and flow velocity on cavitation erosion Cloud Cavitation and Cavitation Erosion’, Proceedings of
aggressiveness’, Journal of Mechanical Engineering, the ASME/JSME Fluids Engineering Division Summer
2005 Meeting, San Francisco CA, 1999

Fortes-Patella R., Reboud J.L., Archer A., ‘Cavitation Singhal A.K., Li H., Atahavale M.M., Jiang Y.,
Damage Measurement by 3D Laser Profilometry’, Wear ‘Mathematical Basis and Validation of the Full Cavitation
246, 2000 Model’, Journal of Fluids Engineering, 124, pp. 617-624,
2002
Fortes-Patella R., Reboud J.L., Briancon-Marjollet L., ‘A
Phenomenological and Numerical Model for Scaling the

10
Stutz B., Reboud J.L., ‘Measurements within unsteady
cavitation’, Experiments in Fluids 29, 2000

Širok B., Dular M., Novak M., Hocevar M., Stoffel B.,
Ludwig G., Bachert B., ‘The Influence of Cavitation
Structures on the Erosion of a Symmetrical Hydrofoil in a
Cavitation Tunnel’, Journal of Mechanical Engineering,
Vol. 48, No. 7, p. 368-378, 2002

11

You might also like