Applied Mathematical Modelling 37 (2013) 6469–6488

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 a b s t r a c t

Article history: In this paper simulation of cavitating flow over the Clark-Y hydrofoil is reported using the
Received 3 May 2012 large eddy simulation (LES) turbulence model and volume of fluid (VOF) technique. We
Received in revised form 16 July 2012 applied an incompressible LES modelling approach based on an implicit method for the
Accepted 7 September 2012
subgrid terms. To apply the cavitation model, the flow has been considered as a single fluid,
Available online 18 September 2012
two-phase mixture. A transport equation model for the local volume fraction of vapour is
solved and a finite rate mass transfer model is used for the vapourization and condensation
Keywords:
processes. A compressive volume of fluid (VOF) method is applied to track the interface of
Clark-Y hydrofoil
Cloud cavitation
liquid and vapour phases. This simulation is performed using a finite volume, two phase
Supercavitation solver available in the framework of the OpenFOAM (Open Field Operation and Manipula-
LES tion) software package. Simulation is performed for the cloud and super-cavitation
VOF regimes, i.e., r = 0.8, 0.4, 0.28. We compared the results of two different mass transfer mod-
Mass transfer model els, namely Kunz and Sauer models. The results of our simulation are compared for cavita-
tion dynamics, starting point of cavitation, cavity’s diameter and force coefficients with the
experimental data, where available. For both of steady state and transient conditions, suit-
able accuracy has been observed for cavitation dynamics and force coefficients.
Ó 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 cav-
itation. Cavitation usually could appear over marine vehicles such as marine propeller blades. For efficiency reasons, the pro-
peller 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 ¼ 0:5 qU 2
called cavitation number, where pv is the vapour pressure, q is
1

the liquid density, and P 1 ; U 1 are the free stream flow pressure and velocity, respectively. Five different cavitation regimes
are observed in the flow 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 cav-
ity 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 flows had shown a rapid progress during the last two decades. The key challenges in
numerical modelling of cavitating flows include sharp changes in the fluid density, existence of a moving boundary and the
requirement of modelling phase change. Among different cavitation models, ‘‘homogeneous equilibrium flow 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 Ó 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) 6469–6488

been widely employed [2]. This approach defines a single-fluid model for both phases. Various categories in ‘‘homogeneous
equilibrium flow model’’ differ in the relation that defines the variable density field. A barotropic water-vapour state law
could be applied to evaluate density field. However, selection of an appropriate state law is a difficult task and needs enough
experimental data for any specific 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 flows. The convective characteristic of the advec-
tion equation considers the effects of inertial forces like bubbles shedding from cavities [2]. Three key points should be con-
sidered 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 improve-
ments. Singhal et al. [6], Merkle et al. [7] and Kunz et al. [8] suggested alternative mass transfer models based on semi-
analytical equations. Senocak and Shyy [9] developed an analytical cavitation model based on the mass-momentum balance
around the cavity interface.
Volume of fluid (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 flows, e.g.
drop collision, liquid sloshing, fluid 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], Hirt–Nichols [12], ‘‘Piecewise Lin-
ear Interface Calculation’’ (PLIC) [13], and ‘‘Compressive Interface Capturing Scheme for Arbitrary Meshes’’ (CICSAM) [14,15].
In SLIC and Hirt–Nichols approaches, the interface is reconstructed with piecewise constant and piecewise constant stair-
stepped line segments, respectively. However, in the PLIC method, piecewise linear segments are used to reconstruct the
interface. In contrast to the geometric reconstruction algorithms [11–13], compressive scheme benefits from a high resolu-
tion differencing schemes to calculate volume fluxes [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 flows 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 modified two-equation turbulence models (k–e, k–x) have been utilized to implement turbulence effects on cavitating
flows [19–23]. Use of ‘‘large eddy simulation (LES)’’ is another approach considered recently in numerical cavitation model-
ling [24–27]. LES resolves large scales energy-containing eddies while it models small scale energy-dissipative one. The suc-
cess of the LES approach in capturing the details of small-scale flow structures in cavitating flows 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 flow 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 flow 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 equa-
tions [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 coefficients. Additionally, a comparison between two standard cavi-
tation models, i.e., Sauer and Kunz models, and between the LES and standard k–e 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 flow variables can be
thought of as an implicit filter. In the other words, finite volume discretization provides top-hat-shaped-kernel filtered val-
ues as:
Z
f ¼ 1
p fdV; ð1Þ
ðdV p Þ Xp
 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488 6471

where over-bar denotes filtered quantity for cell Xp and Vp is the volume of the cell. In the implicit LES approach the trun-
cation 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 flow properties, LES naturally allows for medium to small scale, transient flow struc-
tures. When simulating unsteady, cavitating flows, 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 Navier–Stokes equations, the governing flow equations consisting of the balance equa-
tions of mass and momentum are
@ t ðqv Þ þ r  ðqv  v Þ ¼ rp þ r  s;
ð2Þ
@ t q þ r  ðqv Þ ¼ 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 ¼ 12 ðrv þ rv T Þ and l is the viscosity coefficient. The LES equations are theoretically derived, following Sagaut [31], from
Eq. (2) by applying a low-pass filtering G = G(x, D), using a pre-defined filter kernel function such that
@ t ðqv Þ þ r  ðqv  v Þ ¼ rp þ r  ðs  BÞ;
ð3Þ
@ t q þ r  ðqv Þ ¼ 0:
As no explicit filtering is employed, commutation errors in the momentum equation have been neglected. Eq. (3) intro-
duces one new term when compared to the unfiltered Eq. (2), i.e., the unresolved transport term B, which is the sub grid
stress tensor. B can be decomposed as [32]:
e
B ¼ q  ðv  v  v  v þ BÞ; ð4Þ
where now only B e needs to be modelled. The most common subgrid modelling approaches utilizes an eddy or subgrid vis-
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 flow modelling

To model cavitating flows, the two phases of liquid and vapour need to be specified as well as the phase transition mech-
anism 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 þ r  ðcv Þ ¼ m:
_ ð5Þ
The density and viscosity coefficient are assumed to vary linearly with the vapour fraction,
lm ¼ clv þ ð1  cÞll ; lm;t ¼ lm þ lt ; ð6Þ

q ¼ cqv þ ð1  cÞql : ð7Þ

The viscosity equation (Eq. (6)) consists of a continuous function of laminar viscosity coefficients of both phases (lm) as
well as turbulence viscosity (sub-gird scale viscosity in the LES approach). Since turbulence models employed for multi-
phase flows are typically the same as single-phase turbulence models, Eq. (6) does not consider two-phase flow effects
on the turbulence viscosity coefficient at sub-grid scales. However, Eq. (6) only used for the interfacial cells which are of
infinitesimal width. Therefore, the introduced error is so small as the good treatment of the interface. This treatment of tur-
bulent mixture viscosity coefficient 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 sim-
plified Rayleigh equation as follows:
 2
3 dR Pv  Pl
¼ ; ð8Þ
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:
 
3c dR
_ ¼ qv
m : ð9Þ
R dt
Substituting Eq. (8) in (9), we obtain the following relation for the mass flow rate [4,5]:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
3c 2 jp  pv j
_ ¼ qv ð1  cÞ signðp  pv Þ
m : ð10Þ
Rb 3 ql
6472 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488

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 sup-
ercavitation [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 MinðPl  Pv ; 0Þc MaxðPl  Pv ; 0Þð1  cÞ
þ r  ðcv Þ ¼ 2
þ 2
; ð11Þ
@t qv ðV net
I;n Þ ðql  qv Þt 1 ðV net
I;n Þ ðql  qv Þt 1

where V net
I;n is the net interface velocity relative to the local flow field and t1 is flow characteristic time which is defined as the
ratio of body diameter to free stream flow 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 pres-
sure 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 MinðPl  P v ; 0Þc C prod ð1  cÞc2
þ r  ðcv Þ ¼ þ ; ð12Þ
@t ql ð0:5ql V 21 Þt1 ql t1
where Cdest and Cprod are two empirical constants. The main difference between the Eqs. (11) and (12) is in the condensation
term which significantly affects the flow near the cavity closure region. Due to condensation, there will be a continuous flow
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, Kunz’s model assumes a moderate rate of constant con-
densation. According to Senocak and Shyy [9], Kunz’s 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 Ubbink’s 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 sup-
plementary ‘‘interface-compression velocity (Uc)’’ is defined in the vicinity of the interface in such a way that the local flow
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 Þ þ r  ½~
þ r  ðc~ v c cð1  cÞ ¼ 0: ð13Þ
@t
The last term on the left-hand side of the above equation is known as the artificial 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 flow 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:
rc
v c ¼ min½C c jv j; maxðjv jÞ
~ : ð14Þ
jrcj
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 ¼ rjrc; ð15Þ
where r is surface tension coefficient and curvature (j) of the free surface is determined as:
 
rc
j ¼ r  : ð16Þ
jrcj
 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488 6473

2.4. Pressure–velocity coupling

The employed transient multi-phase solver of OpenFOAM utilizes a cell-centre-based finite 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 fields.
The PISO algorithm could be briefly described as follows [29]:

(1) Momentum prediction: First, the momentum equations are solved with a guessed pressure field, normally the pressure
field of the previous time step. The solution of the momentum equations gives a new velocity field 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 equa-
tion gives rise to a new pressure field. The mass transfer terms are incorporated into the pressure Poisson equation
through as a split source term.
(3) Explicit velocity correction: The new pressure field is used to perform an explicit correction on the velocity. The new
velocity is now consistent with the new pressure field. 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-defined tol-
erance 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

Hydrofoil Pressure Outlet = 100 Kpa

Velocity 2.7C
Inlet = 10 m/s

No Slip Condition

Fig. 1. Computational domain and boundary conditions.

Fig. 2. Investigating the effect of different grid sizes on the average pressure profile over the hydrofoil surfaces, r = 0.8.
6474 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488

8°. The two important non-dimensional numbers used are the Reynolds number (Re) and cavitation number r. The proper-
ties of the inlet free stream are used in these numbers defined as follows:
U1  c
Re ¼ ; ð17Þ
v
p  p1
r¼ qU21
; ð18Þ
2

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 refined in both axial and normal directions to get a cavitation dy-
namic like the experimental data. The effect of using four different grid sizes on the average pressure profile 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 figure
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 finer 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 fine. 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) 6469–6488 6475

3.3. Cloud cavitation regime (r = 0.8)

At the first step, we consider the details of cloud cavitation regime over Clark-Y hydrofoil at r = 0.8. In this regime, some
specific 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 experimen-
tal 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 coefficients of Kunz model are set as:
C dest ¼ 2:0  104 ; C prod ¼ 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 pre-
dict an increase in the density field. Fig. 5 shows average pressure coefficient 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 flow approaches the
trailing edge of the hydrofoil due to cavity detachments and vapour shedding.

Kunz Model
900
Sauer Model

800
Density (Kg/m )
3

700

600

500

400

300
0 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

0
Cp

0.4

0.8
Kunz Model
Sauer Model
1.2

0 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) 6469–6488

1.4
Kunz Model
Sauer Model
1.2 Experimental Data

CL
0.8

0.6

0.4

0.2
-0.2 0 0.2 0.4 0.6 0.8 1 1.2
t/T

Fig. 6. Variation of lift coefficient 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 coefficients for cloud cavitation test case (r = 0.8).

CL Error %
Sauer 0.70 7.90
Kunz 0.78 3.00
Experiment 0.76 –
CD Error %
Sauer 0.140 16.66
Kunz 0.140 16.66
Experiment 0.120 –

Fig. 6 presents variation of lift coefficient 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 figure solution of Kunz model is closer to the exper-
imental data with less oscillatory peaks while Sauer model predicts higher peaks and hills for the lift. The averaged lift and
drag coefficients 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 coefficient. However, both models overpredict drag coefficient. It should be noted that changes in the
Kunz model coefficients is not influential 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 ¼ 1:0  103 ; C prod ¼ 1:0  103 suggested in the Open-
FOAM package.
Fig. 7 shows the temporal evolution of cloud cavitation over one cavitation cycle. Additionally, pressure contours are pro-
vided 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 figure. 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 flow 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) 6469–6488 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) 6469–6488

(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) 6469–6488 6479

(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)
6480 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488

(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 magnifies, re-entrant jet moves further to upstream.
Frames (h and i): Cavity breakdown magnifies, 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 benefiting 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 (a–c). Stronger re-entrant jet could results in creation of smaller detachments. Pressure contours
show considerable oscillations, which could justify oscillations in the lift coefficient. These contours show that transportation
of the shed vapour structures towards the trailing edge is performed by the external flow, 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 (b–d). 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 k–e turbulence approach for three time steps.
Standard k–e 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 k–e turbulence model, r = 0.8.
 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488 6481

0.04

0.035

0.03

0.025

0.02

d/c
0.015

0.01
Current
0.005 Experimental

0
0.2 0.4 0.6 0.8
t/T
(a) x/c=0.2
0.1

0.08

0.06
d/c

0.04

0.02
Current
Experiment

0
0.2 0.4 0.6 0.8
t/T
(b) x/c=0.4
0.16

0.14

0.12
d/c

0.1

0.08

0.06 Current
Experiment

0.04
0.2 0.4 0.6 0.8 1
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) 6469–6488

900 Kunz Model

Sauer Model

800

Density (Kg/m )
3
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

0.4
Cp

Kunz Model
Sauer Model

0.8

1.2

0 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 final state of cavitation where
the cavity closes in the liquid instead of the body’s 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 coefficient distribution over the upper sur-
face 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 sup-
ercavitation 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 flows.
In Fig. 12, the shape of supercavitation from the current simulation and experiment [2] as well as pressure coefficient
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 re-
entrant 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) 6469–6488 6483

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

Additionally, little fluctuation 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 ap-
proached, the current numerical simulation becomes closer to the experimental data, which confirms the accuracy of the
numerical simulations. We also extracted the average amounts of CL and CD for this case from two cavitation models, as re-
ported in Table 2. There is suitable agreement between the numerical predictions with those of experimental data [2]. How-
ever, this table shows that Kunz model is closer to the experimental data for both of lift and drag forces.
Numerical simulation for a supercavitating flow at r = 0.28 is shown in Figs. 14–16 for density and pressure coefficient
distributions as well as supercavity shape and pressure coefficient contours, respectively. Fig. 14 shows that Kunz model pre-
dicts a sharp decrease in the density after the starting point of the supercavity while Sauer model predicts peaks and hills in
the density profiles. In fact, Sauer model predicts that there are very small regions on the surface where vapour bubbles con-
dense and become liquid. This is a defect in the Sauer model simulation for low cavitation number flows. These oscillations
are also observable in the Cp of this model profiles 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 flows over hydro-
foils. This recommendation further confirmed by the data provided for force coefficients 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 IntelÒ 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 in-
creases 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 flow speed and less vapour detachments at superca-
vitation 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 [37–39]. Therefore, like some other researches, i.e.,
6484 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488

0.06

0.05

0.04

d/c
0.03

0.02

0.01 Current
Experiment

0
0.2 0.4 0.6 0.8 1
t (non-dimensionalized)

(a) x/c=0.2
0.09
0.08
0.07
0.06
0.05
d/c

0.04
0.03
0.02 Current
Experiment
0.01
0
0.2 0.4 0.6 0.8 1
t (non-dimensionalized)
(b) x/c=0.4

0.16

0.14

0.12
d/c

0.1

Current
0.08 Experiment

0.06
0.2 0.4 0.6 0.8 1
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 coefficients for supercavitation test case (r = 0.4).

CL Error %
Sauer 0.36 17.2
Kunz 0.4 8
Experiment 0.435 –
CD Error %
Sauer 0.096 10
Kunz 0.096 10
Experiment 0.087 –
 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488 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. [38,40–42], 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 flows. 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 flow over
hydrofoils. Their results show that for angle of attacks bellow the stall condition, 2D and 3D results agrees suitably for
the lift coefficient, 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 flow over hydrofoil and physical results were obtained for
classification 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 confirmation
of the above statement.
6486 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488

(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 finite volume solver benefiting 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 flows
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 coefficients 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 k–e model results in poor accuracy for cloud
 E. Roohi et al. / Applied Mathematical Modelling 37 (2013) 6469–6488 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 coefficient
make the simulation unstable. On the other hand, large Kunz destruction coefficient destroys the cavity soon. For steady
supercavitation regime, Sauer model gives higher error for force coefficients while it predicts oscillations in the density
and pressure field 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] J.P. Franc, J.M. Michel, Fundamentals of Cavitation, Kluwer Academic Publisher, Netherlands, 2004.
[2] G. Wang, I. Senocak, W. Shyy, T. Ikohagi, S. Cao, Dynamics of attached turbulent cavitating flows, Prog. Aerosp. Sci. 37 (2001) 551–581.
[3] S. Gopalan, J. Katz, Flow structure and modelling issues in the closure region of attached cavitation, Phys. Fluids 12 (2000) 895–911.
[4] J. Sauer, Instationaren kaviterendeStromung – Ein neues Modell, baserend auf Front Capturing (VoF) and Blasendynamik, Ph.D. Thesis, Universitat,
Karlsruhe, 2000.
[5] W. Yuan, J. Sauer, G.H. Schnerr, Modelling and computation of unsteady cavitation flows in injection nozzles, J. Mech. Ind. 2 (2001) 383–394.
[6] N.H. Singhal, A.K. Athavale, M. Li, Y. Jiang, Mathematical basis and validation of the full cavitation model, J. Fluids Eng. 124 (2002) 1–8.
[7] 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.
[8] R.F. Kunz, D.A. Boger, D.R. Stinebring, T.S. Chyczewski, J.W. Lindau, H.J. Gibeling, A preconditioned Navier–Stokes method for two-phase flows with
application to cavitation, Comput. Fluids 29 (2000) 849–875.
[9] I. Senocak, W. Shyy, Evaluation of cavitation models for Navier–Stokes computations, in: Proceedings of the FEDSM 02, ASME Fluid Engineering
Division Summer Meeting, Montreal, Canada, 2002.
[10] 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.
[11] W.F. Noh, P.R. Woodward, SLIC (Simple Line Interface Construction), Lect. Notes Phys. 59 (1976) 330.
[12] F.H. Hirt, B.D. Nichols, Volume of fluid (VOF) method for the dynamics of free boundaries, J. Comput. Phys. 39 (1981) 201.
[13] D.L. Youngs, Time dependent multi material flow with large fluid distortion, Numer. Methods Fluid Dyn. (1982) 273–285.
[14] O. Ubbink, Numerical Prediction of Two Fluid Systems with Sharp Interfaces, Ph.D. Thesis, Imperial College, University of London, 1997.
[15] O. Ubbink, R.I. Issa, A method for capturing sharp fluid interfaces on arbitrary meshes, J. Comput. Phys. 153 (1999) 26–50.
[16] 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.
[17] S. Wiesche, Numerical simulation of cavitation effects behind obstacles and in an automotive fuel jet pump, Heat Mass Transfer 41 (2005) 615–624.
[18] 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.
[19] J. Wu, G. Wang, W. Shyy, Time-dependent turbulent cavitating flow computations with interfacial transport and filter-based models, Int. J. Numer.
Methods Fluids 49 (2005) 739–761.
[20] D. Li, M. Grekula, P. Lindell, A modified SST k–x 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.
[21] D. Liu, F. Hong, F. Lu, The numerical and experimental research on unsteady cloud cavitating flow of 3D elliptical hydrofoil, J. Hydrodyn. B 22 (5) (2010)
759–763.
[22] B. Huang, G. Wang, Partially averaged Navier–Stokes method for time-dependent turbulent cavitating flows, J. Hydrodyn. B 23 (1) (2011) 26–33.
[23] 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) 45–51.
[24] 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.
[25] T. Huuva, Large Eddy Simulation of Cavitating and Non-Cavitating Flow, Ph.D. Thesis, Chalmers University of Technology, Sweden, 2008.
[26] 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.
[27] N. Lu, R.E. Bensow, G. Bark, LES of unsteady cavitation on the delft twisted foil, J. Hydrodyn. B 22 (5) (2010) 784–791.
[28] M. Passandideh-Fard, E. Roohi, Transient simulations of cavitating flows using a modified volume-of-fluid (VOF) technique, Int. J. Comput. Fluid Dyn.
22 (1–2) (2008) 97–114.
[29] R.I. Issa, Solution of the implicitly discretized fluid flow equations by operator-splitting, J. Comput. Phys. 62 (1986) 40–65.
[30] R.E. Bensow, G. Bark, Simulating cavitating flows with LES in OpenFOAM, in: Fifth European Conference on Computational, Fluid Dynamics, 2010.
[31] P. Sagaut, Large Eddy Simulation for Incompressible Flows, third ed., Springer, New York, 2006.
[32] R.E. Bensow, C. Fureby, On the justification and extension of mixed methods in LES, J. Turbul. 8 (2007) N54.
[33] E. Berberović, N.P. van Hinsberg, S. Jakirlić, I.V. Roisman, C. Tropea, Drop impact onto a liquid layer of finite thickness: dynamics of the cavity evolution,
Phys. Rev. E 79 (2009) 036306.
[34] E. Berberović, Investigation of Free-surface Flow Associated with Drop Impact: Numerical Simulations and Theoretical Modelling, Technische
Universitat Darmstadt, Darmstadt, Germany, 2010.
[35] H.G. Weller, A new approach to VOF-based interface capturing methods for incompressible and compressible flow, Technical Report TR/HGW/04,
OpenCFD Ltd., 2008.
[36] J.U. Brackbill, D.B. Kothe, C. Zemach, A continuum method for modelling surface tension, J. Comput. Phys. 100 (1992) 335–354.
[37] 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.
[38] 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) 6469–6488

[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.