Applied Mathematical Modelling 37 (2013) 64696488

Contents lists available at SciVerse ScienceDirect

Applied Mathematical Modelling

journal homepage: www.elsevier.com/locate/apm

Numerical simulation of cavitation around a two-dimensional

hydrofoil using VOF method and LES turbulence model
Ehsan Roohi , Amir Pouyan Zahiri, Mahmood Passandideh-Fard
Department of Mechanical Engineering, Faculty of Engineering, Ferdowsi University of Mashhad, P.O. Box 91775-1111, Mashhad, Iran

a r t i c l e

i n f o

Article history:
Received 3 May 2012
Received in revised form 16 July 2012
Accepted 7 September 2012
Available online 18 September 2012
Keywords:
Clark-Y hydrofoil
Cloud cavitation
Supercavitation
LES
VOF
Mass transfer model

a b s t r a c t
In this paper simulation of cavitating ow over the Clark-Y hydrofoil is reported using the
large eddy simulation (LES) turbulence model and volume of uid (VOF) technique. We
applied an incompressible LES modelling approach based on an implicit method for the
subgrid terms. To apply the cavitation model, the ow has been considered as a single uid,
two-phase mixture. A transport equation model for the local volume fraction of vapour is
solved and a nite rate mass transfer model is used for the vapourization and condensation
processes. A compressive volume of uid (VOF) method is applied to track the interface of
liquid and vapour phases. This simulation is performed using a nite volume, two phase
solver available in the framework of the OpenFOAM (Open Field Operation and Manipulation) software package. Simulation is performed for the cloud and super-cavitation
regimes, i.e., r = 0.8, 0.4, 0.28. We compared the results of two different mass transfer models, namely Kunz and Sauer models. The results of our simulation are compared for cavitation dynamics, starting point of cavitation, cavitys diameter and force coefcients with the
experimental data, where available. For both of steady state and transient conditions, suitable accuracy has been observed for cavitation dynamics and force coefcients.
2012 Elsevier Inc. All rights reserved.

1. Introduction
Formation of vapour bubbles within a liquid when its pressure is less than the saturated vapour pressure is called cavitation. Cavitation usually could appear over marine vehicles such as marine propeller blades. For efciency reasons, the propeller usually needs to operate in cavitating conditions but the negative effects of cavitation such as vibrations, noise and
erosion should be avoided. The radial section of these marine blades is a two-dimensional hydrofoil. Cavitation process is
P 1 Pv
characterized by a dimensionless number; i.e., r 0:5
called cavitation number, where pv is the vapour pressure, q is
qU 2
1

the liquid density, and P 1 ; U 1 are the free stream ow pressure and velocity, respectively. Five different cavitation regimes
are observed in the ow over a body: incipient, shear, cloud, partial, and supercavitation. Partial and cloud cavitation regimes
refer to the situation where vapour phase covers a subsection of the body. Alternatively, supercavitation refers to a long cavity that extends more than the body length and closes in the liquid. In all cavitation regimes, there is a constant movement of
a re-entrant liquid jet in the cavity closure section. In cloud cavitation regime, this backward movement of liquid results in
detachment of large vapour sections from the main body [1].
Numerical simulation of cavitating ows had shown a rapid progress during the last two decades. The key challenges in
numerical modelling of cavitating ows include sharp changes in the uid density, existence of a moving boundary and the
requirement of modelling phase change. Among different cavitation models, homogeneous equilibrium ow model had
Corresponding author. Tel.: +98 (511) 8805136; fax: +98 (511) 8763304.
E-mail address: e.roohi@ferdowsi.um.ac.ir (E. Roohi).
0307-904X/$ - see front matter 2012 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.apm.2012.09.002

6470

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

been widely employed [2]. This approach denes a single-uid model for both phases. Various categories in homogeneous
equilibrium ow model differ in the relation that denes the variable density eld. A barotropic water-vapour state law
could be applied to evaluate density eld. However, selection of an appropriate state law is a difcult task and needs enough
experimental data for any specic problem. Moreover, a typical barotropic state equation neglects vorticity production at the
cavity closure [3]. A more appropriate approach is to solve an advection equation for liquid or vapour volume fraction and
compute density as a weighted average of the volume fraction of the two phases. This approach, namely Transport Based
Equation Model (TEM), has extensively been used to simulate cavitating ows. The convective characteristic of the advection equation considers the effects of inertial forces like bubbles shedding from cavities [2]. Three key points should be considered regarding the TEM approach: (1) selection of an appropriate mass transfer model, (2) a solution strategy for the
advection equation, (3) appropriate turbulence model.
Sauer [4] and Yuan et al. [5] suggested cavitation models based on the classical Rayleigh equation with some improvements. Singhal et al. [6], Merkle et al. [7] and Kunz et al. [8] suggested alternative mass transfer models based on semianalytical equations. Senocak and Shyy [9] developed an analytical cavitation model based on the mass-momentum balance
around the cavity interface.
Volume of uid (VOF) technique could be utilized to solve the advection equation of the volume fraction and predict the
cavity interface accurately. The VOF technique is famous for its application in numerical simulation of free surface ows, e.g.
drop collision, liquid sloshing, uid jetting, and spray deposition [10]. This method is conservative, robust and capable of
treating both of large deformations of interface as well as small-scale interface topologies such as breakup and reconnection.
Consistently, the cavity interface can be tracked by VOF approach. Different VOF methods for tracking free surface interface
have been developed; the most known are Simple Line Interface Calculation (SLIC) [11], HirtNichols [12], Piecewise Linear Interface Calculation (PLIC) [13], and Compressive Interface Capturing Scheme for Arbitrary Meshes (CICSAM) [14,15].
In SLIC and HirtNichols approaches, the interface is reconstructed with piecewise constant and piecewise constant stairstepped line segments, respectively. However, in the PLIC method, piecewise linear segments are used to reconstruct the
interface. In contrast to the geometric reconstruction algorithms [1113], compressive scheme benets from a high resolution differencing schemes to calculate volume uxes [14]. Additionally, the implementation of compressive algorithms on
arbitrary unstructured meshes is quite straightforward. A review of the literature shows that VOF method is in accordance
with cavitation physics and can capture the cavity shape accurately. For example, Frobenius and Schilling [16], Wiesche [17],
and Bouziad et al. [18] used VOF technique to simulate cavitation over hydrofoils and pump impellers.
Since most of the cavitating ows performs at high Reynolds number and under unsteady condition, implementation of a
suitable turbulence model is of great importance for accurate prediction of cavitation. Different approaches such as standard
or modied two-equation turbulence models (ke, kx) have been utilized to implement turbulence effects on cavitating
ows [1923]. Use of large eddy simulation (LES) is another approach considered recently in numerical cavitation modelling [2427]. LES resolves large scales energy-containing eddies while it models small scale energy-dissipative one. The success of the LES approach in capturing the details of small-scale ow structures in cavitating ows demonstrates the
important role of turbulence modelling in the cavitation prediction.
As a continuation of our previous work [28], in this study we utilize a multi-phase ow solver of OpenFOAM package to
simulate cloud and supercavitation regime over two-dimensional Clark-Y hydrofoil whose experimental data is available [2].
Our simulation employs a compressive VOF technique [15] combined with two mass transfer models, namely Sauer model
[4] and Kunz model [8]. VOF model used in this work considers the effect of the surface tension force over the cavity surface.
Moreover, in order to capture unsteady features of cavitating ow accurately, we use an implicit large eddy simulation (LES)
turbulence approach. PISO (pressure implicit with splitting of operators) algorithm is used to solve the set of governing equations [29]. The results of our simulation are compared with the experimental data for cavitation dynamics, starting point of
cavitation and cavity diameter as well as lift and drag coefcients. Additionally, a comparison between two standard cavitation models, i.e., Sauer and Kunz models, and between the LES and standard ke turbulence model, will be reported.
2. Governing equations
2.1. Implicit LES model
Large eddy simulation (LES) is based on computing the large, energy-containing eddy structures which are resolved on
the computational grid, whereas the smaller, more isotropic, subgrid structures are modelled. Development of the LES
encounters a main obstacle of the strong coupling between subgrid scale (SGS) modelling and the truncation error of the
numerical discretization scheme. This link could be exploited by developing discretization methods where the truncation
error itself acts as an implicit SGS model. Therefore, the implicit LES expression is used to indicate approaches that merge
SGS model and numerical discretization [30]. Furthermore, the cell-averaging discretization of the ow variables can be
thought of as an implicit lter. In the other words, nite volume discretization provides top-hat-shaped-kernel ltered values as:

f
p

1
dV p

Z
Xp

fdV;

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6471

where over-bar denotes ltered quantity for cell Xp and Vp is the volume of the cell. In the implicit LES approach the truncation error of the discretization scheme acts as the subgrid modelling. In contrast to RANS approaches, which are based on
solving for an ensemble average of the ow properties, LES naturally allows for medium to small scale, transient ow structures. When simulating unsteady, cavitating ows, it is an important property in order to be able to capture the mechanisms
governing the dynamics of the formation and shedding of the cavity [2].
Starting from the incompressible NavierStokes equations, the governing ow equations consisting of the balance equations of mass and momentum are

@ t qv r  qv  v rp r  s;
@ t q r  qv 0;

where v is the velocity, p is the pressure, s = 2lD is the viscous stress tensor, where the rate-of-strain tensor is expressed as
D 12 rv rv T and l is the viscosity coefcient. The LES equations are theoretically derived, following Sagaut [31], from
Eq. (2) by applying a low-pass ltering G = G(x, D), using a pre-dened lter kernel function such that

@ t qv r  qv  v rp r  s  B;
@ t q r  qv 0:

As no explicit ltering is employed, commutation errors in the momentum equation have been neglected. Eq. (3) introduces one new term when compared to the unltered Eq. (2), i.e., the unresolved transport term B, which is the sub grid
stress tensor. B can be decomposed as [32]:

e
B q  v  v  v  v B;

e needs to be modelled. The most common subgrid modelling approaches utilizes an eddy or subgrid viswhere now only B
cosity, vSGS, similar to the turbulent viscosity approach in RANS, where vSGS can be computed in a wide variety of methods
[32]. In the current study, subgrid scale terms are modelled using one equation eddy viscosity model available in the
framework of OpenFOAM.
2.2. Multiphase ow modelling
To model cavitating ows, the two phases of liquid and vapour need to be specied as well as the phase transition mechanism between them. In this work, we consider a two-phase mixture method, which uses a local vapour volume fraction
transport equation together with source terms for the mass transfer rate between the two phases due to cavitation:

_
@ t c r  cv m:

The density and viscosity coefcient are assumed to vary linearly with the vapour fraction,

lm clv 1  cll ; lm;t lm lt ;

q cqv 1  cql :

The viscosity equation (Eq. (6)) consists of a continuous function of laminar viscosity coefcients of both phases (lm) as
well as turbulence viscosity (sub-gird scale viscosity in the LES approach). Since turbulence models employed for multiphase ows are typically the same as single-phase turbulence models, Eq. (6) does not consider two-phase ow effects
on the turbulence viscosity coefcient at sub-grid scales. However, Eq. (6) only used for the interfacial cells which are of
innitesimal width. Therefore, the introduced error is so small as the good treatment of the interface. This treatment of turbulent mixture viscosity coefcient is widely employed.
In this work, we had employed both of Sauer and Kunz models. The approach chosen by Sauer [4] is derived from a simplied Rayleigh equation as follows:

 2
3 dR
Pv  Pl

;
2 dt
ql

where water vapour nuclei with a radius R is assumed to grow when the liquid pressure drops below the vapour pressure, Pv.
The mass transfer rate could be derived from:

_ qv
m

 
3c dR
:
R dt

Substituting Eq. (8) in (9), we obtain the following relation for the mass ow rate [4,5]:

s
3c
2 jp  pv j
_ qv 1  c signp  pv
m
:
3
Rb
ql

10

6472

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

Sauer model expresses the vapour fraction as a function of the radius of the bubbles, Rb, which is assumed to be the same
for all the bubbles. It should be reminded that Rayleigh theory was based on the balance of forces over spherical bubbles. It
ignores bubble interactions, non-spherical bubble geometries and local mass-momentum transfer around the interface. It
has been reported that these characteristics can become important in predicting cavity region, especially in the case of supercavitation [19]. Another drawback of this method is that it requires estimation for the initial value of cavitation nuclei (n0)
and bubble radius (R). The amount of these values affects the predicted cavity length and diameter.
The Kunz approach is a semi-analytical model. This model is based on the conservation of mass-momentum around
cavity interface [19]. The exact analytical relation for cavitation mass transfer based on the local mass-momentum
conservation around cavity interface is [19]

@c ~
ql MinPl  Pv ; 0c MaxPl  Pv ; 01  c

;
r  cv
2
2
@t
V net
qv V net
I;n ql  qv t 1
I;n ql  qv t 1

11

where V net
I;n is the net interface velocity relative to the local ow eld and t1 is ow characteristic time which is dened as the
ratio of body diameter to free stream ow velocity (D/V1). The two terms in the right hand side of Eq. (11) are evaporation
and condensation terms, respectively. The evaporation term reduces the amount of liquid (function c decreases) when pressure drops below the vapour pressure, while the condensation term will add to liquid (function c increases) when the reverse
occurs. The main drawback of this method is on approximating the value of V net
I;n for which some suggestions are reported in
the literature [19]. Kunz et al. [5] proposed a semi-analytical model; whose condensation term has a different form, i.e., their
model reads:

@c ~
C dest qv MinPl  P v ; 0c C prod 1  cc2

;
r  cv
@t
ql t1
ql 0:5ql V 21 t1

12

where Cdest and Cprod are two empirical constants. The main difference between the Eqs. (11) and (12) is in the condensation
term which signicantly affects the ow near the cavity closure region. Due to condensation, there will be a continuous ow
of re-entrant liquid jet near the cavity closure which in turn causes small vapour structures to detach from the end of the
cavity continuously. To include this phenomenon more effectively, Kunzs model assumes a moderate rate of constant condensation. According to Senocak and Shyy [9], Kunzs model reconstructs the cavity region quite accurately especially in the
closure region.
2.3. VOF Model
OpenFOAM uses an improved version of The Compressive Interface Capturing Scheme for Arbitrary Meshes (CICSAM)
VOF technique, based on Ubbinks work [14]. CICSAM is implemented in OpenFOAM as an explicit scheme and could produce
an interface that is almost as sharp as the geometric reconstruction schemes such as PLIC [13]. In CICSAM approach, a supplementary interface-compression velocity (Uc) is dened in the vicinity of the interface in such a way that the local ow
steepens the gradient of the volume fraction function and the interface resolution is improved. This is incorporated in the
conservation equation for volume fraction (c) in the following form [33,34]:

@c
v r  ~
v c c1  c 0:
r  c~
@t

13

The last term on the left-hand side of the above equation is known as the articial compression term and it is non-zero
only at the interface. The compression term stands for the role to shrink the phase-interphase towards a sharper one [34]. The
compression term does not bias the solution in any way and only introduces the ow of c in the direction normal to the
interface. In order to ensure this procedure, Weller [35] suggested the compression velocity to be calculated as:

~
v c minC c jv j; maxjv j

rc
jrcj

14

In other words, the compression velocity is based on the maximum velocity at the interface. The limitation of vc is
achieved through applying the largest value of the velocity in the domain as the worst possible case [35]. The intensity of
the compression is controlled by a constant Cc, i.e., it yields no compression if it is zero, a conservative compression for
Cc = 1 and high compression for Cc > 1 [34]. Nevertheless, the CICSAM algorithm is far less costly to apply compared to PLIC.
Previous studies showed that OpenFOAM give accurate results for the interface position on moderate to high resolution
meshes [33,34]. The surface tension is evaluated per unit volume using the CSF model [36]:

fr rjrc;

15

where r is surface tension coefcient and curvature (j) of the free surface is determined as:

j r 

rc
:
jrcj

16

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6473

2.4. Pressurevelocity coupling

The employed transient multi-phase solver of OpenFOAM utilizes a cell-centre-based nite volume method and employs
the solution procedure based on the pressure implicit with splitting of operators (PISO) algorithm [29] for coupling between
pressure and velocity elds.
The PISO algorithm could be briey described as follows [29]:
(1) Momentum prediction: First, the momentum equations are solved with a guessed pressure eld, normally the pressure
eld of the previous time step. The solution of the momentum equations gives a new velocity eld which does not
satisfy the continuity condition. Additionally, the vapour volume fraction transport equation is solved in this step.
(2) Pressure solution: These predicted velocities are used to solve the pressure equation. The solution of the pressure equation gives rise to a new pressure eld. The mass transfer terms are incorporated into the pressure Poisson equation
through as a split source term.
(3) Explicit velocity correction: The new pressure eld is used to perform an explicit correction on the velocity. The new
velocity is now consistent with the new pressure eld. The velocity in a cell not only depends on the pressure gradient
but also on the contributions from the neighbouring cells. This iterative algorithm continues until a pre-dened tolerance is met.
3. Results and discussions
3.1. Simulation set-up
The computational domain and boundary conditions are given according to the experimental setup described in Ref. [2]
and are illustrated in Fig. 1. The Clark-Y hydrofoil is placed at the center of the water tunnel with the angle of attack equal to

No Slip Condition

10c

Velocity
Inlet = 10 m/s

2.7C

Hydrofoil

Pressure Outlet = 100 Kpa

No Slip Condition
Fig. 1. Computational domain and boundary conditions.

Fig. 2. Investigating the effect of different grid sizes on the average pressure prole over the hydrofoil surfaces, r = 0.8.

6474

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

8. The two important non-dimensional numbers used are the Reynolds number (Re) and cavitation number r. The properties of the inlet free stream are used in these numbers dened as follows:

Re

U1  c

p  p1
qU21

17

18

where p is the pressure, p1 is free stream vapour pressure and U1 is the free stream velocity which is imposed 10 m/s. With
the chord length equal to 7 cm, we have Re = 7  105. Time step is set small enough so that the Courant number is less than
0.45 in the domain.
3.2. Grid independency study
As the Clark-Y hydrofoil is not geometrically complex, we used structured quadrilateral meshes. Mesh size near the wall
has a key effect on the cavitation dynamics. Meshes are rened in both axial and normal directions to get a cavitation dynamic like the experimental data. The effect of using four different grid sizes on the average pressure prole over one period
of cavitation, on the upper and lower surfaces of the hydrofoil is shown in Fig. 2. Grids 1 to 4 have 65, 130, 270 and 420 cells
on the upper surface and 43, 87, 180 and 280 cells on the lower surface of the hydrofoil, respectively. It is observed that the
difference between the pressure curves becomes negligible as the number of surface cells increases. Additionally, this gure
shows that the grids 3 and 4 provide close solutions, especially for the upper surface where the cavitation occurs. Therefore,
we performed our simulations using grid 3. This grid had total of 126,480 cells in the entire computational domain. Fig. 3
illustrates the employed grid 3 close to the hydrofoil body as well as two close-up views from the mesh at the leading edge
and trailing edge of the hydrofoil. Concentration of the cells is ner at the trailing edge. We applied an expansion ratio of 1.15
for cells width adjacent to the hydrofoil walls, therefore, cells close to the surface are suitably ne. This grid had been used
for all test cases reported in this paper. Additionally, this grid corresponds to a value of y+ < 1 everywhere near the upper wall
of the hydrofoil.

Fig. 3. Top row: Employed structured mesh around the hydrofoil (270  180 cells around the hydrofoil). Bottom row: Close-up view of mesh near the
leading edge (left) and trailing edge (right) of the hydrofoil.

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6475

3.3. Cloud cavitation regime (r = 0.8)

At the rst step, we consider the details of cloud cavitation regime over Clark-Y hydrofoil at r = 0.8. In this regime, some
specic features including vapour cloud shedding at the end of cavity occurs. Therefore, a critical task of suitable turbulence
model is to capture dynamics of cavity growth and detachment correctly. We selected Clark-Y hydrofoil because experimental set of data is available in the literature [2].
Fig. 4 shows density distribution (averaged in one period) over the upper surface of the hydrofoil obtained from two
cavitation models, namely Kunz model and Sauer models. The coefcients of Kunz model are set as:
C dest 2:0  104 ; C prod 1:0  103 [30]. Density is computed from Eq. (7). As observed, Kunz model predicts that cavitation
starts a bit ahead in comparison with Sauer model, to be more precise, Kunz model predicts that cavitation starts at
x = 10 mm, while Sauer model gives a value of x = 14.4 mm. However, experimental data of Ref. [2] gives a value of
x = 9.8 mm. Therefore, Kunz model gives a more accurate prediction. As cavity extends along the hydrofoil, both models predict an increase in the density eld. Fig. 5 shows average pressure coefcient distribution over the upper and lower surfaces
of the cavity at r = 0.8. It is observed that both models predict close Cp distribution expect some deviations predicted in the
Sauer model solution. On the upper surface, Cp = r near the leading edge but it slightly decreases as the ow approaches the
trailing edge of the hydrofoil due to cavity detachments and vapour shedding.

Kunz Model
Sauer Model

900

Density (Kg/m )

800

700

600

500

400

300

0.02

0.04

0.06

X (m)
Fig. 4. Average density distribution over the upper surface of the hydrofoil obtained from two cavitation models, namely Kunz model and Sauer model,
r = 0.8.

-0.8

-0.4

Cp

0.4

0.8
Kunz Model
Sauer Model
1.2

0.02

0.04

0.06

0.08

X(m)
Fig. 5. Pressure distribution (averaged over one cavitation period) over the upper and lower surfaces of the hydrofoil obtained from two cavitation models,
namely Kunz model and Sauer model, r = 0.8.

6476

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

1.4
Kunz Model
Sauer Model
Experimental Data

1.2

CL

0.8

0.6

0.4

0.2
-0.2

0.2

0.4

0.6

0.8

1.2

t/T
Fig. 6. Variation of lift coefcient with time in cloud cavitation regime, current Kunz and Sauer models compared with the experimental data reported in
Ref. [22].

Table 1
Averages of lift and drag coefcients for cloud cavitation test case (r = 0.8).
CL
Sauer
Kunz
Experiment

0.70
0.78
0.76

Sauer
Kunz
Experiment

CD
0.140
0.140
0.120

Error %
7.90
3.00

Error %
16.66
16.66

Fig. 6 presents variation of lift coefcient on one cavitation cycle from two cavitation models in addition to experimental
data reported in Ref. [22]. Due to changes in cavity length and cavity detachment (cloud shedding), lift forces performs an
oscillatory behaviour with time. Maximum lift occurs once cavity is at maximum length while slight force oscillation refers
to small scale detachments stage of the cloud cavitation regime. In this gure solution of Kunz model is closer to the experimental data with less oscillatory peaks while Sauer model predicts higher peaks and hills for the lift. The averaged lift and
drag coefcients over one cavitation cycle is given in Table 1 for both cavitation models and compared with the average data
reported in Ref. [2]. As observed, the solution of Kunz model is quite close to the experimental data, with maximum of 3%
error in average lift coefcient. However, both models overpredict drag coefcient. It should be noted that changes in the
Kunz model coefcients is not inuential on the accuracy of the lift and drag forces, i.e., error in lift force increases to 5%
and drag force remains unchanged if we use the default setting of C dest 1:0  103 ; C prod 1:0  103 suggested in the OpenFOAM package.
Fig. 7 shows the temporal evolution of cloud cavitation over one cavitation cycle. Additionally, pressure contours are provided from our numerical solutions. These results correspond to LES turbulence model and both of Sauer and Kunz cavitation
models. Results of Kunz model had shown for fewer time steps to avoid lengthy gure. The experimental pictures from Ref.
[2] are also provided where available. In the cloud cavitation regime, the trailing edge of the hydrofoil becomes unsteady due
to substantial vapour shedding at the terminal section of the cavity. Additionally, cloud cavitation has a cyclic behaviour, i.e.,
as the re-entrant jet approaches the leading edge of the hydrofoil, the cavitating ow is pushed away from the wall and a
new cavity structure forms there [2]. Due to the large vapour shedding at the trailing edge, there is considerable growth
of the cavity thickness in the leading edge of the hydrofoil.
The frames in Fig. 7 show the cavitation cycle as follows:
Frame (a): Cavity starts its growth at the leading edge, shedding occurs at the trailing edge.
Frames (b and c): Cycle grows more, shed vapour moves downstream.
Frame (d): Cavity occupies most of the hydrofoil and is at its maximum extent. The peak observed in CL diagram (Fig. 6)
corresponds to this condition.
Frame (e): Breakdown of cavity starts at the trailing edge.
Frame (f): Breakdown continues, re-entrant jet moves further and enhances shedding.

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6477

(a) Cavity grows, detachment exists near the trailing edge, Top: solution of Sauer model, Middle: solution of
Kunz model, Bottom: Experiment [2], Left: contour of vapour, Right: contour of pressure

(b) cavity occupies most of the hydrofoil while vapour detachment and shedding exists at the trailing edge;
solution of Sauer model, Left: contour of vapour, Right: contour of pressure

(c) Cavity growth and shedding, Top: Sauer model, Middle: Kunz model, Bottom: Experiment [2], Left: contour
of vapour, Right: contour of pressure
Fig. 7. Cavitation dynamics (left) and pressure contours (right) from the current simulation (with Kunz and Sauer cavitation models, as indicated, and LES
turbulence model) for cloud cavitation regime, r = 0.8. Experimental pictures from Ref. [2] are provided where available.

6478

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

(d) Cavity occupies most of the hydrofoil, Top: Sauer model, Middle: Kunz model, Bottom: Experiment [2],
Left: contour of vapour, Right: contour of pressure

(e) Cavity starts breakdown at the trailing edge, Top: Sauer model, Middle: Kunz model, Bottom: Experiment
[2], Left: contour of vapour, Right: contour of pressure

(f) Cavity breakdown continues as re-entrant jet moves backward more, Top: Sauer model, Middle: Kunz model,
Bottom: Experiment [2], Left: contour of vapour, Right: contour of pressure

Fig. 7. (continued)

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

(g) Cavity breakdown/shedding magnifies, re-entrant jet reaches to middle of the hydrofoil, Top: Sauer model,
Middle: Kunz model, Bottom: Experiment [2], Left: contour of vapour, Right: contour of pressure

(h)

Cavity breakdown magnifies, solution of Sauer model, Left: contour of vapour, Right: contour of pressure

(i) Cavity breakdown magnifies, shedding from the leading edge of the hydrofoil; solution of Sauer model, Left:
contour of vapour, Right: contour of pressure

(j) Detached cavity mostly focused on the middle and trailing edge, Top: Sauer model, Middle: Kunz model,
Bottom: Experiment [2], Left: contour of vapour, Right: contour of pressure

Fig. 7. (continued)

6479

6480

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

(k) Cycle restarts, cavity restart growth and shedding at the trailing edge; solution of Sauer model compared with
experiment [2], Left: contour of vapour, Right: contour of pressure
Fig. 7. (continued)

Frame (g): Breakdown and shedding magnies, re-entrant jet moves further to upstream.
Frames (h and i): Cavity breakdown magnies, re-entrant jet reaches to the leading edge.
Frame (j): Cavity disappears at the leading edge while shed vapour accumulates mostly at the middle and trailing edge.
Frame (k): Cycle restarts, same as Frame (a).
As expected, cloud cavitation regime is accompanied with cavity breakdown and vortex shedding. As Fig. 7 shows, there
are good agreements between the current numerical solutions with those of experiments. This could be attributed to
employing complex turbulence model, i.e., LES, in addition to beneting from VOF technique in reconstructing the free surface
as well as suitable cavitation models. However, Kunz and Sauer models differ in their cavity prediction. Sauer model predicts
smaller detachments and stronger re-entrant jet compared to the Kunz model, i.e., the re-entrant jet is quite visible in Sauer
model solutions in Frames (ac). Stronger re-entrant jet could results in creation of smaller detachments. Pressure contours
show considerable oscillations, which could justify oscillations in the lift coefcient. These contours show that transportation
of the shed vapour structures towards the trailing edge is performed by the external ow, which is indicated as P = 100 kPa
region. Interestingly, condensation of these shed structures results in a local high pressure peak comparable to stagnation
pressure, see for example local pressure peak in Fig. 7 (bd). The videos illustrating temporal evolution of cloud cavity
dynamics (c function boundary, illustrated by blue colour) are labelled as Sauer Model r = 0.8 and Kunz Model r = 0.8
and could be observed in the web version of this paper. The videos show multiple cycles of cavitation and more clearly
llustrate the cloud cavitation dynamics, i.e., the backward movement of re-entrant jet and its role on the cavity shedding.
Fig. 8 shows the cavitation dynamics using the Sauer model and standard ke turbulence approach for three time steps.
Standard ke model predicts re-entrant jet while it could not predict vapour shedding at the trailing edge section at all.
Additionally, re-entrant jet is quite weak and could not reach the leading edge of the hydrofoil. Consequently, there is no

Fig. 8. Cavitation dynamics (left) and pressure contours (right) from the current simulation with Sauer model and standard ke turbulence model, r = 0.8.

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6481

0.04
0.035
0.03

d/c

0.025
0.02

0.015
0.01
Current
Experimental

0.005
0

0.2

0.4

0.6

0.8

t/T

(a) x/c=0.2
0.1
0.08

d/c

0.06
0.04
0.02
0

Current
Experiment

0.2

0.4

0.6

0.8

t/T

(b) x/c=0.4
0.16
0.14

d/c

0.12
0.1

0.08
0.06
0.04

Current
Experiment

0.2

0.4

0.6

0.8

t/T

(c) x/c=0.6
Fig. 9. Cavity thickness at different locations on the hydrofoil, comparison of the current numerical simulation from Kunz model with the experimental
data from Ref. [2], r = 0.8.

vapour shedding from the hydrofoil. Comparison of Figs. 7 and 8 indicates the importance of using a suitable turbulence
model for accurate prediction of unsteady cloud cavitation dynamics.
Comparison of the cavity thickness at different locations along the hydrofoil, i.e., x/c = 0.2, 0.4, and 0.6, where c is the
length of hydrofoil chord, from the current numerical simulation (Kunz model) with the experimental data from Ref. [2]
is provided in Fig. 9. Suitable agreement is observed between the transient predictions from the simulation and experiment
while the average values are quite close to each other, i.e., at x/c = 0.2, Kunz model predicts an average value of d/c = 0.26
while the experimental data gives d/c = 0.24, or at x/c = 0.4, our simulation predicts d/c = 0.78 while this value is 0.76 from
experiment.

6482

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

900

Kunz Model
Sauer Model

Density (Kg/m )

800

700

600

500

400

300
0

0.02

0.04

0.06

X (m)
Fig. 10. Density distribution over the upper surface of the hydrofoil obtained from two cavitation models, namely Kunz model and Sauer model, r = 0.4.

-0.4

Cp

0.4
Kunz Model
Sauer Model

0.8

1.2

0.02

0.04

0.06

X(m)
Fig. 11. Pressure distribution over the upper surface of the hydrofoil obtained from two cavitation models, namely Kunz model and Sauer model, r = 0.4.

3.4. Supercavitation regime

If we decrease the cavitation number further, supercavitation occurs. Supercavitation is the nal state of cavitation where
the cavity closes in the liquid instead of the bodys surface. In the supercavitation region, the pressure remains at a constant
value of vapour pressure and does not drop any further. Density distribution over the upper surface of the hydrofoil obtained
from two cavitation models, namely Kunz and Sauer models is shown in Fig. 10 for r = 0.4. Kunz model predicts the starting
point of cavity closer to the leading edge compared to the Sauer model while Sauer model predicts lower minimum density.
This behaviour is similar to our observation for cloud cavitation regime. Pressure coefcient distribution over the upper surface of the hydrofoil obtained from two cavitation models, Kunz and Sauer, is shown in Fig. 11. Unlike to cloud cavitation
case, pressure is almost constant over the cavitating surface of the hydrofoil due to the steady vapour formation at the supercavitation state. There are small oscillations on the pressure in the Sauer model solution. As will be discussed, Sauer model
performs worse than Kunz model for large scale supercavitating ows.
In Fig. 12, the shape of supercavitation from the current simulation and experiment [2] as well as pressure coefcient
distribution is shown. Compared to the cloud cavitation regime, the mechanism of cavitation growth and vortex shedding
is strongly decreased in supercavitation. It is observed that the shape of simulated supercavitation is in good agreement with
the experimental results while both numerical mass transfer models predict almost identical cavity shape. Very weak reentrant jet exists in supercavitation condition which results in small scale vapour shedding at the cavity closure point.

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6483

Fig. 12. Left: Supercavitation shape from the current simulation, (top: Sauer model, middle: Kunz model) and experiment [2] (bottom). Right: Pressure
coefcient contours (top: Sauer model, middle: Kunz model), r = 0.4.

Additionally, little uctuation of the interface of supercavity is observed. Fig. 13 compares the cavity thickness at different
locations on the hydrofoil, from the Kunz solution with the experimental data reported in Ref. [2]. This comparison is shown
with respect to the time from the early stage of supercavity formation until steady condition. As steady state condition approached, the current numerical simulation becomes closer to the experimental data, which conrms the accuracy of the
numerical simulations. We also extracted the average amounts of CL and CD for this case from two cavitation models, as reported in Table 2. There is suitable agreement between the numerical predictions with those of experimental data [2]. However, this table shows that Kunz model is closer to the experimental data for both of lift and drag forces.
Numerical simulation for a supercavitating ow at r = 0.28 is shown in Figs. 1416 for density and pressure coefcient
distributions as well as supercavity shape and pressure coefcient contours, respectively. Fig. 14 shows that Kunz model predicts a sharp decrease in the density after the starting point of the supercavity while Sauer model predicts peaks and hills in
the density proles. In fact, Sauer model predicts that there are very small regions on the surface where vapour bubbles condense and become liquid. This is a defect in the Sauer model simulation for low cavitation number ows. These oscillations
are also observable in the Cp of this model proles shown in Fig. 15 while Kunz model predicts a constant Cp = r, which is in
agreement with the theory. Therefore, we could recommend Kunz model for simulating supercavitating ows over hydrofoils. This recommendation further conrmed by the data provided for force coefcients for the previous test case, i.e.,
r = 0.4, see Table 2. Fig. 16 shows the supercavity dynamics from both cavitation models. In comparison to the previous case,
decreasing the cavitation number resulted in longer vapour region. As the supercavity length and diameter increases, the
effects of turbulence instabilities at the closure section decays and more stable cavities are observed.
3.5. Computational time
A note should be given about the computational time for investigated test cases. Table 3 provides computational expanse
for a simulation time of 235 ms for three investigated cavitation regimes. Simulation was performed in parallel using four
cores of Intel Core i7-2600K CPU equipped with 16 GB memory RAM. In the multi-phase solver of the OpenFOAM, time
step size was determined based on the estimates of the characteristic length and time scales of the eddies. Therefore, the
most time consuming case is the cloud cavity test case (r = 0.8), where small-scale/low-speed vapour eddy detachments increases computational costs to 25 h. Once cavitation number decreases, run time decreases, i.e., cases r = 0.4 and r = 0.28
need 9 and 6 h, respectively. This decrease could be attributed to higher ow speed and less vapour detachments at supercavitation regime.
3.6. A note on 2D LES simulation
The key reason for employing 2D LES in the current work is the large computational cost of 3D LES simulation.
3D LES simulation typically needs large computational resources [3739]. Therefore, like some other researches, i.e.,

6484

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

0.06
0.05

d/c

0.04
0.03
0.02
0.01
0

Current
Experiment

0.2

0.4

0.6

0.8

t (non-dimensionalized)

(a) x/c=0.2
0.09
0.08
0.07

d/c

0.06
0.05
0.04
0.03
0.02

Current
Experiment

0.01
0

0.2

0.4
0.6
0.8
t (non-dimensionalized)

(b) x/c=0.4
0.16
0.14

d/c

0.12
0.1
Current
Experiment

0.08
0.06
0.2

0.4
0.6
0.8
t (non-dimensionalized)

(c) x/c=0.6
Fig. 13. Cavity thickness at different locations on the hydrofoil, comparison of the current numerical simulation from Kunz model with the experimental
data from Ref. [2], r = 0.4.

Table 2
Averages of lift and drag coefcients for supercavitation test case (r = 0.4).
CL
Sauer
Kunz
Experiment

0.36
0.4
0.435

Sauer
Kunz
Experiment

CD
0.096
0.096
0.087

Error %
17.2
8

Error %
10
10

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6485

Fig. 14. Density distribution over the upper surface of the hydrofoil obtained from two cavitation models, namely Kunz model and Sauer model, r = 0.28.

Fig. 15. Pressure distribution over the upper surface of the hydrofoil obtained from two cavitation models, namely Kunz model and Sauer model, r = 0.28.

Refs. [38,4042], the current research was based on 2D LES. As we showed in Section 3.3, standard RANS based schemes
often gives physically unrealistic results when they are applied to highly unsteady cavitating ows. Therefore, 2D LES could
be considered as a great improvement over RANS for cavitation simulation. In other words, 2D LES is a superior approach
compared to RANS suitably affordable for cavitation analysis. Accuracy of 2D LES had also been reported in the literature.
For example, Kinzel et al. [37,38] provided a comparison between 2D and 3D LES simulations of cavitating ow over
hydrofoils. Their results show that for angle of attacks bellow the stall condition, 2D and 3D results agrees suitably for
the lift coefcient, general cavity development, stability, and cavity size. They showed that surface pressure distributions
of the 2D case are consistent with the three dimensional predictions, especially away from the tip [37]. Qin et al. [40,41]
and Arndt [42] considered 2D LES simulation of cavitating ow over hydrofoil and physical results were obtained for
classication of vortex shedding mechanism in sheet/cloud cavitation regimes. Therefore, for situation far from stall
condition, where there is not considerable vortical motion in the lateral direction, application of 2D LES could provide
accurate results. Evidently, suitable agreement between the current results and those of experiments is another conrmation
of the above statement.

6486

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

(a) Sauer Model

(b) Kunz model

Fig. 16. Supercavitation shape and Cp contours from the current simulation (r = 0.28).

Table 3
Details of computational cost for investigated test cases.

r = 0.8

r = 0.4

r = 0.28

25.0 h

9.0 h

6.0 h

4. Conclusion
In the present study, a nite volume solver beneting from the implicit LES turbulence model and accompanied with the
VOF interface capturing technique has been employed to capture unsteady cloud cavitation and steady supercavitating ows
over the Clark-Y hydrofoil. The simulation is performed under the framework of OpenFOAM. Effects of different mass
transfer models including Kunz and Sauer models had been investigated. Our simulation shows that combination of the
LES, VOF and Sauer or Kunz models could simulate the shape of cloud cavitation and its dynamics precisely. Also lift and
drag coefcients as well as cavity diameter and starting point are obtained close to the experimental data specially using
the Kunz model. We showed that substitution of LES model with the standard ke model results in poor accuracy for cloud

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

6487

cavitation dynamics due to weaker re-entrant jet prediction and no vapour shedding. We also observed that the solution of
the Kunz model is insensitive to the production constant, however, employing an extremely high value for this coefcient
make the simulation unstable. On the other hand, large Kunz destruction coefcient destroys the cavity soon. For steady
supercavitation regime, Sauer model gives higher error for force coefcients while it predicts oscillations in the density
and pressure eld over the hydrofoil. Therefore, we recommend using Kunz model for supercavitation regime.
Acknowledgements
The authors would like to sincerely thank Prof. Robert F. Kunz (Department of Aerospace Engineering, Pennsylvania State
University) and Dr. Santiago Marquez Damian (CIMEC, Santa Fe-Argentina) for fruitful discussions.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/
j.apm.2012.09.002.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]

J.P. Franc, J.M. Michel, Fundamentals of Cavitation, Kluwer Academic Publisher, Netherlands, 2004.
G. Wang, I. Senocak, W. Shyy, T. Ikohagi, S. Cao, Dynamics of attached turbulent cavitating ows, Prog. Aerosp. Sci. 37 (2001) 551581.
S. Gopalan, J. Katz, Flow structure and modelling issues in the closure region of attached cavitation, Phys. Fluids 12 (2000) 895911.
J. Sauer, Instationaren kaviterendeStromung Ein neues Modell, baserend auf Front Capturing (VoF) and Blasendynamik, Ph.D. Thesis, Universitat,
Karlsruhe, 2000.
W. Yuan, J. Sauer, G.H. Schnerr, Modelling and computation of unsteady cavitation ows in injection nozzles, J. Mech. Ind. 2 (2001) 383394.
N.H. Singhal, A.K. Athavale, M. Li, Y. Jiang, Mathematical basis and validation of the full cavitation model, J. Fluids Eng. 124 (2002) 18.
C.L. Merkle, J. Feng, P.E.O. Buelow, Computational modelling of the dynamics of sheet cavitation, in: Third International Symposium on Cavitation
(CAV1998), Grenoble, France, 1998.
R.F. Kunz, D.A. Boger, D.R. Stinebring, T.S. Chyczewski, J.W. Lindau, H.J. Gibeling, A preconditioned NavierStokes method for two-phase ows with
application to cavitation, Comput. Fluids 29 (2000) 849875.
I. Senocak, W. Shyy, Evaluation of cavitation models for NavierStokes computations, in: Proceedings of the FEDSM 02, ASME Fluid Engineering
Division Summer Meeting, Montreal, Canada, 2002.
M. Passandideh-Fard, E. Roohi, Coalescence collision of two droplets: bubble entrapment and the effects of important parameters, in: Proceedings of
the 14th Annual (International) Mechanical Engineering Conference, Isfahan, Iran, 2006.
W.F. Noh, P.R. Woodward, SLIC (Simple Line Interface Construction), Lect. Notes Phys. 59 (1976) 330.
F.H. Hirt, B.D. Nichols, Volume of uid (VOF) method for the dynamics of free boundaries, J. Comput. Phys. 39 (1981) 201.
D.L. Youngs, Time dependent multi material ow with large uid distortion, Numer. Methods Fluid Dyn. (1982) 273285.
O. Ubbink, Numerical Prediction of Two Fluid Systems with Sharp Interfaces, Ph.D. Thesis, Imperial College, University of London, 1997.
O. Ubbink, R.I. Issa, A method for capturing sharp uid interfaces on arbitrary meshes, J. Comput. Phys. 153 (1999) 2650.
M. Frobenius, R. Schilling, Three-dimensional unsteady cavitating effects on a single hydrofoil and in a radial pump-measurement and numerical
simulation, Cav03-GS-9-005, in: Proceedings of the Fifth International Symposium on Cavitation, Osaka, 2003.
S. Wiesche, Numerical simulation of cavitation effects behind obstacles and in an automotive fuel jet pump, Heat Mass Transfer 41 (2005) 615624.
A. Bouziad, M. Farhat, F. Guennoun, K. Miyagawa, Physical modelling and simulation of leading edge cavitation, application to an industrial inducer, in:
Proceedings of the Fifth International Symposium on Cavitation, Cav03-OS-6-014, Osaka, Japan, 2003.
J. Wu, G. Wang, W. Shyy, Time-dependent turbulent cavitating ow computations with interfacial transport and lter-based models, Int. J. Numer.
Methods Fluids 49 (2005) 739761.
D. Li, M. Grekula, P. Lindell, A modied SST kx turbulence model to predict the steady and unsteady sheet cavitation on 2D and 3D hydrofoils, in:
Proceedings of the Seventh International Symposium on Cavitation CAV2009, 2009.
D. Liu, F. Hong, F. Lu, The numerical and experimental research on unsteady cloud cavitating ow of 3D elliptical hydrofoil, J. Hydrodyn. B 22 (5) (2010)
759763.
B. Huang, G. Wang, Partially averaged NavierStokes method for time-dependent turbulent cavitating ows, J. Hydrodyn. B 23 (1) (2011) 2633.
S. Phoemsapthawee, J. Leroux, S. Kerampran, J. Laurens, Implementation of a transpiration velocity based cavitation model within a RANS solver, Eur. J.
Mech. B 32 (2012) 4551.
G. Wang, M. Ostoja-Starzewski, Large eddy simulation of a sheet/cloud cavitation on a NACA 0015 hydrofoil, Appl. Math. Model. 31 (3) (2007) 417
447.
T. Huuva, Large Eddy Simulation of Cavitating and Non-Cavitating Flow, Ph.D. Thesis, Chalmers University of Technology, Sweden, 2008.
D. Liu, S. Liu, Y. Wu, H. Xu, LES numerical simulation of cavitation bubble shedding on ALE 25 and ALE 15 hydrofoils, J. Hydrodyn. B 21 (6) (2009) 807
813.
N. Lu, R.E. Bensow, G. Bark, LES of unsteady cavitation on the delft twisted foil, J. Hydrodyn. B 22 (5) (2010) 784791.
M. Passandideh-Fard, E. Roohi, Transient simulations of cavitating ows using a modied volume-of-uid (VOF) technique, Int. J. Comput. Fluid Dyn.
22 (12) (2008) 97114.
R.I. Issa, Solution of the implicitly discretized uid ow equations by operator-splitting, J. Comput. Phys. 62 (1986) 4065.
R.E. Bensow, G. Bark, Simulating cavitating ows with LES in OpenFOAM, in: Fifth European Conference on Computational, Fluid Dynamics, 2010.
P. Sagaut, Large Eddy Simulation for Incompressible Flows, third ed., Springer, New York, 2006.
R.E. Bensow, C. Fureby, On the justication and extension of mixed methods in LES, J. Turbul. 8 (2007) N54.
E. Berberovic, N.P. van Hinsberg, S. Jakirlic, I.V. Roisman, C. Tropea, Drop impact onto a liquid layer of nite thickness: dynamics of the cavity evolution,
Phys. Rev. E 79 (2009) 036306.
E. Berberovic, Investigation of Free-surface Flow Associated with Drop Impact: Numerical Simulations and Theoretical Modelling, Technische
Universitat Darmstadt, Darmstadt, Germany, 2010.
H.G. Weller, A new approach to VOF-based interface capturing methods for incompressible and compressible ow, Technical Report TR/HGW/04,
OpenCFD Ltd., 2008.
J.U. Brackbill, D.B. Kothe, C. Zemach, A continuum method for modelling surface tension, J. Comput. Phys. 100 (1992) 335354.
M.P. Kinzel, J.W. Lindau, R.F. Kunz, Free-surface proximity effects in developed and super-cavitation, in: DoD HPCMP Users Group Conference, IEEE
Computer Society, 2008.
M.P. Kinzel, Computational Techniques and Analysis of Cavitating-Fluid Flows, Ph.D. Thesis, Pennsylvania State University, 2008.

6488

E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 64696488

[39] S.E.A. Kim, Numerical study of unsteady cavitation on a hydrofoil, in: Proceedings of the Seventh International Symposium on Cavitation, CAV2009,
Ann-Arbor, Michigan, USA, 2009.
[40] Q. Qin, C.C.S. Song, R.E.A. Arndt, Numerical study of an unsteady turbulent wake behind a cavitating hydrofoil, in: Fifth International Symposium on
Cavitation (CAV03-GS-9-001), Osaka, Japan, 2003.
[41] Q. Qin, C.C.S. Song, R.E.A. Arndt, Incondensable gas effect on turbulent wake behind a cavitating hydrofoil, in: Fifth International Symposium on
Cavitation (CAV03-GS-9-001), Osaka, Japan, 2003.
[42] R.E.A. Arndt, From Wageningen to Minnesota and back: perspectives on cavitation research, in: Sixth International Symposium on Cavitation,
CAV2006, Wageningen, The Netherlands, September 2006.