23

The 23th National Computational Fluid Dynamics Conference

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

Investigation on dynamic behaviors of the turbulent cavitating flows and the lift
coefficient based on the Euler viewpoints
1, 2

Ping-Ben Liu1*, Chien-Chou Tseng2

Department of Mechanical Engineering and Electro- Mechanical, National Sun
Yat-Sen University

Abstract
The objective of this study is to investigate the relationship between the lift coefficient and dynamic
behaviors of cavitating flow around a two-dimensional ClarkY hydrofoil at 8 angle of attack, cavitation
number of 0.8, and Reynolds number of 7105. The flow field is investigated numerically by using a vapor
transfer equation and a modified turbulence mode which applies the filter and local density correction. The
results including time-averaged lift/drag coefficient and shedding frequency agree well with experimental
observations, which confirmed the reliability of this simulation. According to the variation of lift coefficient,
the cycle which consists of growth and shedding of cavitation can be divided into three stages, and the lift
coefficient at each stage behaves similarly due to the formation and shedding of the cavity around the trailing
edge.

KeywordsCFD, Cavitation, Turbulence, lift coefficient

1*, 2
1,2

ClarkY ( 8 0.8
7105)
/
ClarkY

 23
The 23th National Computational Fluid Dynamics Conference

1.

Introduction

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

2.

Governing equations and numerical

settings
2.1 Navier-Stokes equations

In liquid flow, Cavitation is a phase-change

phenomenon which is triggered when the local
pressure drops below the vapor pressure and it can
be observed in a variety of hydraulic machine. The
intensity of cavitation is controlled by cavitation
number
Ca,
which
is
defined
as
Ca ( P Pv ) / 0.5lU where P is local fluid

In this study, Favre-averaged Navier-Stokes

turbulence treatment is adopted due to the
compressibility effect.
The continuity and
momentum equations are listed below in Equation
(1) and (2):
m ( mu j )
(1)

0
t
x j
m ui ( m u j ui )

pressure, Pv is vapor pressure, l is liquid phase

density, and U is free stream velocity. When
cavitation occurs, heat transfer from the liquid
phase to vapor phase conquers the latent heat to
keep the cavity, cooling the liquid around the cavity.
However, this thermal effect is usually neglected
due to the large liquid-to-vapor density ratio (order
of 105 ) of water at room temperature [1-3].
In cavitating flow modeling, mixture model is
widely used [4-8]. Under this framework, the liquid
phase and the vapor phase are treated as a single
phase and the slip velocity between two phases is
neglected. Therefore, the governing equations of
two phases are simplified into one set of equations
for mixture phase. Furthermore, mixture model is
often coupled with the transport equation-based
model (or so-called cavitation model). Transport
equations can base on either vapor or mass fraction
with appropriate source terms to regulate the mass
between phases [9-11].
Besides, some numerical simulations about
cavitating flow around hydrofoil were investigated
during past decades. Huang et al. [12] shows that
cavity shedding can induce self-oscillatory
behaviors of the hydrofoil. Wang et al. [13] presents
that the adverse pressure gradient at the end of
cavity is the reason for shedding of cloud cavitation.
Luo et al. [14] applies different sizes of hydrofoils
and shows that cavitation is suppressed over a
smaller hydrofoil. Saito et al. [15] indicates that the
walls on the both sides of hydrofoil hugely change
the structure of cavitation compared to the original
hydrofoil.
This study presents unsteady numerical simulations
of turbulent cavitating flow over a two-dimensional
ClarkY hydrofoil through a commercial CFD code
Ansys-Fluent. The calculations involve Zwart
cavitation model [10] and a modified standard
k turbulence model by a hybrid function
between
filter-based
model
(FBM)
and
density-based model (DCM). The objective of this
study is to investigate the relationship among the
averaged lift coefficient, the structure of cavitation,
and the variations of pressure on the upper and
bottom hydrofoil surface.

xi x j

x j

ui u j
2 u

) ij k
( t )(
x j xi
3 xk

(2)

In Equation (1) and (2), x is the coordinate, t is

time, u is velocity, m is density of mixture
phase, p is pressure, is fluid viscosity, and

t is turbulent viscosity. The subscript i , j , and

k represent for Einstein index. ij is Kronecker

delta.
The density of mixture phase and other mixture
fluid properties are defined by homogeneous
theory:
(3)
m ll v (1 l )
In Equation (3), is a property of mixture phase,
is volume fraction. The subscript m , l , and
v represent for mixture, liquid, and vapor phase.

2.2 Turbulence model

Standard k turbulence model is used to solve
t . k is turbulent kinetic energy, is turbulent
dissipation rate. The equations are listed below in
Equation (4) and (5):
k
( m k ) ( mu j k )

Pt m
[( t )
]
t
x j
x j
k x j

(4)

( m ) ( mu j )

t
x j

2

C 1 Pt C 1 m
[( t )
]
k
k x j
x j

(5)

In Equation (4) and (5), Pt and t are defined by

the equations below:
u
u u j
2 u
(6)
Pt i [ t ( i
) ij k ]
x j
x j xi
3 xk
T

m C k 2

(7)

 23
The 23th National Computational Fluid Dynamics Conference

In Equation (4) ~ (7), C 1 =1.44 C 2 =1.92

=1.3 k =1.0 C 0.09 .
However, since standard k turbulence model
often leads to an over predicted turbulent viscosity,
a hybrid-type limiter is introduced to reduce
turbulent viscosity. The filter-based model f FBM
and density-based model f DCM are listed in
Equation (8) and (9):

(8)
f FBM min(1, 1.5 )
k
(1 v )n ( l v )
(9)
f DCM v
v (1 v )( l v )

transport-based cavitation model listed in Equation

(13):
v v
t

v vu j

Re Rc

x j

The source term

represent

for

Re

(13)
and the sink term

evaporation

and

Rc

condensation

processes. The Zwart cavitation model utilized in

present

study

is

derived

from

simplified

RayLeigh-Plesset equation where the second-order

derivative of the bubble radius and surface tension
effect are not considered [10]. Accordingly, the

In Equation (8), the turbulence length scale k 1.5

is compared with the filter size based on the
resolution. If k 1.5 is larger than , the term

source and sink terms of Zwart cavitation model

can be expressed as follows:

k 1.5 is smaller the 1, and the excessive

Re

turbulent viscosity is reduced by f FBM ; If k

is smaller than , f FBM is equal to 1, and the
turbulent viscosity remains the same value as
calculated.
In Equation (9), f DCM is used to capture the large
drop of density on the interface from liquid phase to
vapor phase. The turbulent viscosity is reduced
aggressively around the location of cavity.
Consequently, the T in Equation (7) is modified
by Equation (10):
k2
(10)
t 0.09 m
f hybrid
1.5

Rc
RB

3Fvap nuc (1 v ) v
RB
3Fcond v v
RB

2 max(0, Pv P)
3
l

(14)

2 max(0, P Pv )
3
l

(15)

is bubble radius. The triggering condition of

evaporation and condensation processes is the value

of local pressure. If the local pressure is lower than
the vapor pressure, the term Re in Equation (14) is
activated. Otherwise, the term Rc in Equation (15)
is activated. Furthermore, in evaporation process,

f hybrid f FBM (1 ) f DCM

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

the nucleation site density must decrease due to

(11)

increase of vapor volume fraction. To descript the

C (0.6 m l C2 )
tanh[ 1
]
0.2(1 2C2 ) C2
0.5
(12)
2 tanh(C1 )
The hybrid function shown in Fig.1 is
controlled by m l . If m l is close to 0 at
certain location, it means that cavitation bubble
exists, and f DCM is triggered. On the other hand,
If m l is close to 1, it means that there is no
phase-change phenomena, and f FBM is enabled.
In Equation (8) ~ (12), is set to be 4.4% of
characteristic length scale, the value of n in f DCM
is 4. C1 and C2 are chosen as 4 and 0.2.

reaction,

evaporation

is
term,

replaced
and

by

nuc

nuc (1 v )

in

defined

as

is

non-condensable gas in liquid phase. The empirical

constants Fvap and Fcond are designed to control
the rate of evaporation and condensation. Here,
is 106 m , nuc is 5 104 , Fvap and

Fcond

RB

are 90

and 4 103 .

2.4 Geometry, grid, and boundary conditions

The hydrofoil used in the study is ClarkY with 8
degree angle of attack and 0.07m of cord length
C. The Reynolds number is 7 105 and the
cavitation number Ca is 0.8. The computational
domain shown in Fig.2 is chosen to be the same as
experimental one [16]. After a series of tests, an
orthogonal mesh shown in Fig.3 with 61000 cells is

2.3 Cavitation model

Under the framework of mixture model and
homogeneous theory, the entire cavitating flow field
is treated as a single phase field. The mass transfer
between liquid and vapor phases is governed by a
3

 23
The 23th National Computational Fluid Dynamics Conference

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

adopted. The non-dimensional distance from solid

wall y is given between 30 and 100 to fit the
requirement of turbulence model. Table.1 shows the
time-averaged lift/drag coefficient and frequency
based on lift coefficient. It reveals that the
numerical results with the chosen mesh are able to
capture the lift/drag force and close to real flow
field.
At the inlet boundary, the value of inflow velocity
comes from Reynolds number, the vapor volume
fraction is 0, and the turbulent quantities are given
by the eddy-to-laminar viscosity ratio of 1000 and
turbulence intensity of 0.02. The pressure is
extrapolated. At outlet boundary, the pressure is
based on the cavitation number Ca while other
variables are extrapolated. A no-slip boundary
conditions is applied on the walls and hydrofoil
surface. The turbulent quantities are calculated by
the wall function. The working fluid is water at
room temperature. The vapor pressure is 3141.69 Pa
and the density is 998.2kg m3 for liquid phase and

shortest, which connects to the newly-formed

0.0228kg m3

6(b), the adverse pressure gradient at the cavity

attached cavity in Fig. 5(b). Moreover, the pressure

coefficient yields down at trailing edge, which
means the pressure on the bottom surface is affected
by the lower pressure of trailing edge cavity. The
lift coefficient is 0.41, which is the lowest in whole
cycle due to the small size of cavity and lower
pressure on the bottom surface as shown in Fig. 8.
At 13.2% of the cycle, Fig. 6(b) illustrates that the
cavity which formed at leading edge grows rapidly
and occupies more region of the top surface. On the
other hand, the cavity around the trailing edge
disappears, making no influence on the bottom
surface, so the pressure of the bottom surface
gradually recovers as Fig. 8 shown. Besides, in Fig.

for vapor phase. The laminar

closure region becomes unneglectable, slows down

viscosity is 0.001003kg m s .

the incoming fluid near the closure region, and

3.

Results and discussion

induces a vortex. The lift coefficient is 0.63 and

A typical cycle of lift coefficient is chosen as shown

because of the cavity growth and pressure recovery

in Fig.4, and time of calculation is divided by time

of the bottom surface, the lift coefficient is rising at

of a cycle T to become dimensionless. The cycle is

this stage.

taken apart into three parts by 49% and 79% of the

At 27.5% of the cycle, the lift coefficient comes to

cycle where shedding of trailing edge cavity

the maximum value 1.04 in the entire cycle in Fig. 4.

happens. In each part, there is a local maximum

In Fig. 7(b), the size of cavity is larger than one in

value and the tendency is the same which rises first

Fig. 6(b) at 13.2% of the cycle. It is worth to be

and then decreases. The phenomenon in the first

mentioned that the rate of growth of attached cavity

part before 49% of the cycle will be discussed in

declines due to the development of the vortex. The

following article.

pressure of the center of the vertex decreases to

In Fig. 5(a) and (b), the liquid volume fraction l

keep the vertex, which pushes the lower-pressure

of numerical results is compared with experimental

region closer to the end of hydrofoil illustrated in

observations at 0% of the cycle, showing that

Fig. 8. Furthermore, the adverse pressure gradient

numerical results are highly consistent with

on the hydrofoil surface can trigger the re-entrant

experimental observations. The new attached cavity

jet to stem the cavity back to the leading edge. The

forms at leading edge and is growing. The detached

pressure coefficient in Fig. 8 also illustrates that the

cavity around the trailing edge comes from the

longest cavity length and higher pressure of the

previous cycle and is about to leave with incoming

bottom surface occur at this moment, which refers

flow. The pressure coefficient distribution in Fig. 8

to the highest lift coefficient mentioned above.

reveals at this moment, the cavity length is the

4

 23
The 23th National Computational Fluid Dynamics Conference

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

At the next moment, 28.6% of the cycle, the vertex

surface yields down and drops to a lower value due

moves away from the hydrofoil. Therefore, the fluid

to the effect of trailing edge cavity.

along the bottom surface is attracted to the top side

Last but not least, the variation of the lift coefficient

as Fig. 9(a) and (b) shown. The pressure of the

in Fig. 4 shows that there are three local maximum

bottom side is no longer to keep at that high value

values which connect to 27.5%, 54%, and 83.6% of

because the fluid along the bottom surface can

the cycle. However, the peak is lower and lower as

travels around the lower-pressure region at the

times goes by. This can be explained in Fig. 13 that

trailing edge. The pressure drop is shown in Fig.

the cavity length is pushed back continuously,

9(c).

which causes a higher-pressure region on the top

At 36% of the cycle, in order to make the fluid turn

surface.

around the end of trailing edge, the pressure must

4.

decrease to a much lower value to create required

Conclusion

centrifugal force. Therefore, the lower-pressure area

This investigation focus on the relationship between

at the end of trailing edge encourages the formation

the lift coefficient and dynamic behaviors of the

of trailing cavity and leads to the reduction of

cavitating flow. The increase of the lift coefficient

pressure of the bottom surface, as Fig. 10 and Fig.

mainly depends on the rising pressure on the bottom

12 shown. In addition, the cavity is forming at the

surface of the hydrofoil. However, when the fluid

center of the vertex on the top side. The reason is

turns up to the top surface, the pressure on the

that in order to maintain such a vertex, the pressure

bottom surface suddenly falls down, leading to the

of the center must decrease, and cavity forms if the

drop of the lift coefficient. As the trailing edge

pressure is lower than the vapor pressure. In Fig. 12,

cavity leaves from the end of hydrofoil, the

although the lower-pressure region on the top side

lower-pressure region makes less and less influence

is larger than one at 27.5% of the cycle, the lift

on the bottom side, and the lift coefficient can rise

coefficient falls down to 0.8 due to the declining

up again.

pressure on the bottom surface. There is no

referring observation in experiment at this moment.
The lift coefficient keeps decreasing to a local
minimum value 0.75 at 49% of the cycle. The
attached cavity is pushed back toward the leading
edge by the re-entrant jet, as illustrated in Fig. 11.
The newly-formed cavity structure above the
hydrofoil grows up due to the low pressure of the

Acknowledgements

vertex center. Similar to the occasion at 0% of the

The financial support was provided by Ministry of

cycle, the trailing edge cavity is about to move

Science and Technology under the grant number

away from the hydrofoil, and the impact of the

MOST 104-2221-E-110-049 and communication

trailing edge cavity on the bottom surface weakens,

platform

making the lift coefficient rise up at next moment

Astronautical Society of the Republic of China

shown in Fig. 4. Also, the phenomenon can be

(A.A.S.R.C.) and National Pingtung University of

described in Fig. 12 that the pressure of the bottom

supported

by

Aeronautical

Science and Technology (NPUST).

5

and

 23
The 23th National Computational Fluid Dynamics Conference

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

[12] Huang, B. Zhao, Y. Wang, G. Y., Large

Eddy
Simulation
of
Turbulent
Vortex-Cavitation Interactions in Transient
Sheet/Cloud Cavitating Flows,
[13] Wang, G. Y., Zhang, B., Huang, B., Zhang,
M. D. Unsteady Dynamics of Cloud
Cavitating Flows around a Hydrofoil, 7th
International Symposium on Cavitation, Aug.
17-22, Ann Arbor, Michigan, USA, 2009.
[14] Luo, Xi. W., Bin, J., Zhang. Y., Xu H. Y.,
Cavitating Flow over a Mini Hydrofoil
CHIN. PHYS. LETT. Vol. 29, No. 1, 2012.
[15] Saito. Y., Takami. R., Nakamori. I., Ikohagi. T.,
Numerical Analysis of Unsteady Behavior of
Cloud Cavitation around NACA0015 Foil,
Computational Mechanics, Vol. 40, No. 1, pp.
85-96, June 2007.
[16] Wang, G. Y., Senocak, I., Shyy, W. Ikohagi, T.
Cao, S., Daynamics of Attached Turbulent
Cavitating Flows, Aerospace Sci, Vol. 37, NO.
6, 2001, pp. 551-581.

References
Franc, J. P., Claude, R., and Alain, C., An
Experimental Investigation of Thermal Effects
in a Cavitating Inducer, J. Fluids Eng.
Vol.126, No. 5, 2004, pp.716-723.
[2] Tseng, C. C., Shyy, W., Modeling for
Isothermal and Cryogenic Cavitation, Int. J.
Heat and Mass Transfer, Vol.53, No. 1, 2010,
pp. 513-525.
[3] Tseng, C. C., Wei, Y. J., Wang, G. Y., Shyy, W.,
Modeling of Turbulent, Isothermal and
Cryogenic
Cavitation under
Attached
Conditions, Acta Mechanica Sinica, Vol. 26,
No. 3, 2010, pp. 325-353.
[4] Dular, M., Bachert, R., Stoffel, B., irok, B,
Experimental Evaluation of Numerical
Simulation of Cavitating Flow around
Hydrofoil, European Journal of Mechanics
B/Fluids, Vol. 24, 2005, pp. 522-538
[5] Ji, B., Luo, X. W., Wua, Y. L., Peng, X. X.
Duan, Y. L., Numerical Analysis of Unsteady
Cavitating Turbulent Flow and Shedding
Horse-Shoe Vortex Structure around a Twisted
Hydrofoil, Int. J. Multiphase Flow, Vol. 51,
2013, pp. 33-43.
[6] Bensow, R. E., Bark, G., Simulating
Cavitating Flows with LES in OPENFOAM,
European Conference on Computational Fluid
Dynamics, Lisbon, Portugal, 2010.
[7] Li, D. Q., Grekula, M., Lindell, P., Towards
Numerical Prediction of Unsteady Sheet
Cavitation on Hydrofoils, 9th International
Conference on Hydrodynamics, Oct. 11-15,
Shanghai, China, 2010.
[8] Ducoin, A., Huang, B., Young, Y. L.,
Numerical Modeling of Unsteady Cavitating
Flows around a Stationary Hydrofoil.
International Journal of Rotating Machinery,
Vol. 2012, Article ID 215678.
[9] Singhal, A. K., Athavale, M. M., Li, H. Y.,
Jiang, Y., Mathematical Basis and Validation
of the Full Cavitation Model, J. Fluids Eng.
Vol. 124, 2002, pp. 617-624.
[10] Zwart, P. J., Gerber, A. G., Belamri T., A
Two-Phase Flow Model for Predicting
Cavitation Dynamics, ICMF International
Conference on Multiphase Flow, Paper
No.152, 2004.
[11] Yuan, W. X., Sauer, J. G., Schnerr, H.
Modeling and Computation of Unsteady
Cavitation Flows in Injection Nozzles, 1st
Intl Colloq. Microhydrodynamics, Paris, Vol.
2, 2001, pp. 383-394.
[1]

Fig. 1. The function for the hybrid turbulence

model.

Fig. 2. The geometry and boundary conditions of

two-dimensional ClarkY hydrofoil.

 23
The 23th National Computational Fluid Dynamics Conference

Fig. 3. The grid near the ClarkY hydrofoil.

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

Fig. 6. The flow structure over the ClarkY hydrofoil

with streamlines at 13.2% of the cycle. The
schematic in (a) shows the experimental

0.79T

0.49T

observation [16] and in (b) shows the liquid volume

fraction l of numerical results.

0.9

Cl

0.8

0.7

0.6

0.5

0.4

0.2

0.4

0.6

0.8

time/T

Fig. 4. The variation of lift coefficient in the chosen

Fig. 7. The flow structure over the ClarkY hydrofoil

cycle versus dimensionless time. The cycle is

with streamlines at 27.5% of the cycle. The

divided into three parts by 0.49T and 0.79T.

schematic in (a) shows the experimental

observation [16] and in (b) shows the liquid volume
fraction l of numerical results.

27.5%T
13.2%T
0%T
1

Fig. 5. The flow structure over the ClarkY hydrofoil

Cp along the bottom surface

0.5

at 0% of the cycle. The schematic in (a) shows the

Cp

experimental observation [16] and in (b) shows the

liquid volume fraction l of numerical results.

leading
edge

trailing
edge
Cp along the top surface

-0.5

cavity length

-1
0

0.2

0.4

0.6

0.8

x/c

Fig. 8. The pressure coefficient along the

ClarkY hydrofoil surface at 0%, 13.2%T, 27.5% of
the cycle.
7

 23
The 23th National Computational Fluid Dynamics Conference

105 8 18-20
Kaohsiung City, August 18-20, 2016
XX-XX

27.5%T
36%T
49%T
1

Cp along the bottom surface

Cp

0.5

leading
edge

trailing
edge

Cp along the top surface

-0.5

cavity length

-1
0

0.2

0.4

0.6

0.8

x/c

Fig. 12. The pressure coefficient along the

ClarkY hydrofoil surface at 27.5%, 36%T, 49% of
the cycle.
Fig. 9 The schematic in (a) shows the flow structure
at the trailing edge of the ClarkY hydrofoil with

1.4

streamlines at 27.5% and in (b) shows the flow

27.5%T
54%T
83.6%T

1.2
1

structure at the trailing edge of the ClarkY hydrofoil

0.8
0.6

shows the pressure coefficient along the

0.4

Cp along the bottom surface

Cp

with streamlines at 28.6%. The schematic in (c)

0.2

ClarkY hydrofoil surface at 27.5% and 28.6% of the

cycle.

leading
edge
trailing
edge

-0.2
-0.4
Cp along the top surface

-0.6
-0.8
-1

0.2

0.4

0.6

0.8

x/c

Fig. 10. The schematic shows the numerical flow

Fig. 13. The pressure coefficient along the

structure over the ClarkY hydrofoil with

ClarkY hydrofoil surface at 27.5%, 54%T, 83.6% of

streamlines at 36% of the cycle

the cycle.
Table 1. Validation between the numerical and
experimental results [16] by Cd, Cl, and flift
(AOA=8, Re=7105, =0.8).

Fig. 11. The schematic shows the numerical flow

flift (Hz)

Cd

Cl

structure over the ClarkY hydrofoil with

Exp.

Num.

Exp.

Num.

Exp.

Num.

streamlines at 49% of the cycle.

24.1

23.3

0.119

0.123

0.760

0.660