You are on page 1of 10

A COMPARISON STUDY OF DIFFERENT CAVITATION MODELS ON THE

DEVELOPMENT OF 2-D CAVITATING FLOWS

S.-T. Chiang and J.-H. Chen


Department of Systems Engineering and Naval Architecture
National Taiwan Ocean University
2 Pei-Ning Road, Keelung 202, Taiwan
positiveman03@yahoo.com.tw, b0105@mail.ntou.edu.tw

ABSTRACT: In the present study, three transport models for liquid/mass fraction developed
by Singhal et al. [1], Zwart et al. [2], and Schnerr & Sauer[3], respectively, were employed
for computing the cavitating flow phenomena in several 2-D hydrofoils. For each case, the
computational results were compared to each other for different models. It is interesting to
find that in some cases, the results agree with each other quite well; however, in other cases,
they differ significantly. Furthermore, the existence of non-condensible gas in water can play
an important role in the development of cavitating flow. In this paper, the cavitation
characteristics, such as its size, location, and physical properties, are presented and discussed.
Finally, some features on the cavitation simulation using these models are concluded.

INTRODUCTION

Flow cavitation is a vital, yet complicated, phenomenon in many engineering applications.


For some examples, cavitation is an important factor in the design of propeller and impeller,
as it could induce many detrimental effects, such as noise, structural damage, and severe
efficiency reduction. In the area of biochemical engineering/biotechnology, it is known that
cavitation results in the generation of hot spots, highly reactive free radicals, and turbulence
associated with liquid circulation currents, which can result in the intensification of various
physical/chemical operations [4]. Cavitation is also related to valves fractures of a mechanical
heart cavitation and could result in damage of blood cells and increase the risk of
thromboembolic complications found in patients implanted with a mechanical heart [5].
Cavitation has been one of the most important issues which need be considered in designing
and operating a pump, especially when the rotating speed of pump increases because it could
cause severe damage on the surface of flow channels and affect stability of system [6].

With the advancement of computational capability, rapid progress has been made it possible
to reveal cavitation physics in various forms by computational fluid dynamics. Various
cavitation models have been proposed in the past few decades. Roughly speaking, they can be
divided into three main categories: interface tracking models, density-pressure coupling
models, and transport models for liquid/mass fraction [7]. Among them, the type of models in
the last category appears to be the most popular one in numerical simulations since the first
one is difficult to take care of the detached cavity and the second one does not have the
capability to couple with the transport phenomenon.

The transport models are most suitable for large-scale cavitation, such as sheet cavitation that
often occurs on hydrofoils and propellers. These models include two-way interactions
between the phases. The homogeneous mixture concept has been the popular focus in the
development of two-phase modelling. In this concept, the contents of individual cells are
assumed to be uniform. Recently, the mixture model has been advanced rapidly with various
mass transfer formulae. It is commonly assumed in the mixture model that the diffusion
balance is maintained for both phases and no velocity/temperature differences between two
phases for the entire computing domain. These two phases are thus regarded as one mixed
pseudo-fluid. Because of this fact, the mixture model is also referred as the diffusion model.

Under the premise of diffusion balance, the mixture model often incorporates with various
mass transfer mechanisms derived from either bubble dynamics or vaporization-condensation
process. Singhal et al. [1] proposed a mixture model based on the reduced Rayleigh-Plesset
equation. They considered various first-order effects including phase change, bubble
dynamics, turbulent pressure fluctuations, and non-condensable gases and thus named it the
full cavitation model. Later, Huuva derived a similar formulation in which the vapor mass
fraction was replaced by the vapor volume fraction [8]. The model proposed by Zwart et al. [2]
(Zwart-Gerber-Belamri model) is based on the simplified Rayleigh-Plesset equation. This
cavitation model assumes large vapour bubbles of equal size, which do not affect each other.
However, the expressions for condensation and evaporation are not symmetric. In particular,
the evaporation expression takes into account that, as the vapour volume fraction increases,
the nucleation site density must decrease accordingly. They proposed that the total interphase
mass transfer rate per unit volume was determined by the bubble density numbers and the
mass change rate of a single bubble. The study conducted by Schnerr and Sauer [3] (Schnerr-
Sauer model) appears to be the first which leads to the conclusion that the mass transfer is
mainly driven by the pressure differential. Their formulation of mass transfer rate is able to
simulate the cyclic formation of cavitation cloud, the formation of re-entrant jet, and so on.

In all models, empirical coefficients must be given before the cavitation physics can be
determined. In fact, a common problem of mass transfer models is that they rely on empirical
parameters which need to be tuned for specific flow problems and are not known a priori.
Recently, some studies on the influence of these coefficients on predicting cavitating flows
have been conducted. Employing the Zwart-Gerber-Belamri model [2] as an example, Liu et
al. [9] investigated their influence on the pump cavitation performance prediction and carried
out experiments to validate the computational results.

For the mass transfer model using the vaporization-condensation mechanism, it is assumed
that the pressure in the cavity keeps unchanged and the mass change rate is proportional in
some way to the difference of liquid and vapor pressure. Based on these concepts, several
models have been proposed in the literature. Saito et al. [10] proposed a model according to
the theory of evaporation/condensation on a plane surface. Okita and Kajishima [11]
formulated a mass transfer in the form of volume fraction. The model proposed by Kunz et al.
[12] is somewhat complicated. The derived the mass transfer which was based on two
different strategies for creation and destruction of liquid. The transformation of liquid to
vapor was assumed to be proportional to the amount by which the pressure was below the
vapor pressure. The transformation of vapor to liquid, however, was based on a third order
polynomial function of volume fraction.

In the present study, three mixture models were employed for computing the cavitating flow
phenomena in several 2-D hydrofoils. They are the full cavitation model, the Zwart-Gerber-
Belamri model, and the Schnerr-Sauer model. The commercial code FLUENT was employed
for the study and all these three models are available in the code. Some particular turbulence
models were used. The details are described in the next section.
THEORETICAL FORMULATION AND COMPUTATIONAL STRATEGY

Physical problem under consideration

We consider two-dimensional viscous liquid


flow past various hydrofoils. Figure 1 depicts
the typical flow physics. The angle of attack of
the incident flow is . A cavitation bubble is
formed somewhere on the airfoil.

Considering the cavitation physics, we can


express the two-dimensional governing
equations as follows, Figure 1 Schematic of the problem.

m
( mu m ) 0 (1)
t

and


t

2
( mu m ) ( mu mu m ) p m u m uTm u dr, u dr, (2)
1

where t denotes time, udr, is the drift velocity of phase which is equal to 1 for the liquid
phase and 2 for the gas phase, and m and u m are the mixed density and velocity defined as

2
m , (3)
1

and

u
um 1
, (4)
m

respectively. In the above two equations, represents the volume fraction of phase .
Furthermore, p denotes the pressure and m the mixed viscosity
2
m . (5)
1

To find the fluid velocities and densities of liquid and gas phases, we must employ a proper
cavitation model.
For the boundary conditions, an inlet uniform velocity is specified with 1% turbulent intensity;
no slip condition is specified on the surface of the hydrofoil; a pressure outlet condition with
zero gauge pressure is specified at the downstream exit.

Mass transfer models

In a cavitating flow, the liquid-vapour mass transfer (evaporation and condensation) is


governed by the vapour transport equation


( 2 2 ) ( 2 2u 2 ) Re Rc (6)
t

For the full cavitation model, if p < pv,

k 2 pv p
Re Ce 2 1 (1 f 2 f g ) , (6)
3 1

and, if p > pv,

k 2 p pv
Rc Cc 2 1 f2 . (7)
3 1

The symbols in the above expressions are defined as follows: Ce and Cc are empirical
constants for evaporation and condensation, respectively, and take the values of 0.02 and 0.01,
respectively; is the surface tension of liquid; pv the phase-change threshold pressure; k is
the turbulent kinetic energy; f2 and fg are the mass fraction of vapour and non-condensable
gases. In this model, the liquid-vapor mixture is assumed to be compressible.

For the Zwart-Gerber-Belamri cavitation model, if p < pv,

3 nuc (1 2 ) 2 p p
Re Ce 2 v , (8)
RB 3 1
and, if p > pv,

3 nuc 2 2 p pv
Rc Cc 2 . (9)
RB 3 1

In the above two expressions, Ce and Cc take the values of 50 and 0.01, respectively; RB is the
radius of bubble and is assumed to be 106 m; nuc represents the nucleation site volume
fraction and its default value is assumed to be 5 104.

Finally, for the Schnerr-Sauer cavitation model, if p < pv,


2 1 3 2 pv p
Re (1 ) , (10)
m RB 3 1

and, if p > pv,

2 1 3 2 p pv
Re (1 ) , (11)
m RB 3 1

where the volume fraction is defined as

4
nb RB3
3 . (12)
4 3
1 nb RB
3
with nb being the number of bubbles per volume of liquid which is assumed to be 1013.

It is noted that for the Zwart-Gerber-Belamri model and the Schnerr-Sauer model, the non-
condensable permanent gas is not accounted for.

Turbulence models

Two different turbulence models were employed for the series of computations. The first one
is the standard k- model which is one of the most common turbulence models. It has been
shown that this model is suitable for free-shear layer flows with relatively small pressure
gradients.

The second model employed in the present study is the shear-stress transport (SST) k- model.
This model takes advantages of two worlds. In the viscous sub-layer region of a boundary
layer, it uses a k- formulation and in the free stream, it switches to a k- behaviour in the
free-stream to cope with the free-stream turbulence properties. It is suitable for flows with
adverse pressure gradients and separation.

Computational approach

The commercial code FLUENT (version 15) was employed for the present study. In the code,
the finite volume method was employed to discretize all differential equations and the
algorithm of pressure implicit with splitting of operators (PISO) was adopted for nonlinear
iterations of pressure and velocity solutions. The PISO algorithm is efficient and more accu-
rate to solve the Navier-Stokes equations in unsteady problems because the momentum cor-
rector step is performed more than once, compared to the traditional SIMPLE algorithm. For
unsteady flows, the second-order upwind scheme was employed asr the time-marching
strategy.

For the grid generation, the commercial codes Pointwise (version 17.2) and ANSYS (version
15) were used. We used the former to generate structured grids in the following case 1 and the
latter unstructured grids in case 2.
RESULTS AND DISCUSSIONS FOR SOME STUDY CASES

Case 1: Supercavitation with a NACA0015 section

A uniform flow past the NACA 0015 section at an


angle of attack = 8. The local mesh near the
hydrofoil is shown in Figure 2. The total grid number
is about 40,000. The standard standard k- model was
employed for the computations. The cavitation number
for the present study is = 0.1 which evidently results
in a supercavitating cavity. The Reynolds number
based on the incoming flow speed and the chord length Figure 2 Local mesh for Case 1.
is Re = 4.4 107. For the full cavitation model, we
specified fg = 1.5 105. We also compared the results with and without non-condensable gas
content for the full cavitation model. The evaporation pressure for water is 3540Pa.

The computations show that the parameters in each model may have significant influence on
the development of the flow and cavitation bubble. However, if we adjusted the default values
set in a proper way, then we may obtain results which agree very well with one another for
different models. The computational results shown in Figure 3 were obtained with a change of
nucleation site volume fraction from its default value 5 104 to a new one 4 104. The
adjustment is purely a trial-and-error process. We may find that the results coincide.

The cavitation bubble is shown in Figure 4. The shape with 2 = 0.5 is also shown in the plot.
The shape is comparable to the results in the literature.

Case 2: Partial cavitation with a plano-convex hydrofoil section

We first define the hydrofoil section: its upper side is plane and its lower side circular (radius

Figure 3 Pressure coefficients for different cavitation models.

(a) Present study.

(b) Cavitation bubble in [13].

Figure 4 Cavitation bubble for Case 1.


(a) Whole domain. (b) Local enlargement.
Figure 5 Mesh for Case 2.

(a) Zwart-Gerber-Belamri model

(b) Schnerr-Sauer model

Figure 6 Time variations of the drag and lift coefficients for two models at = 2.5.

1.327c) with a maximum thickness of 0.102c; the leading edge is rounded with a radius of
0.0051c mm. For details, see [14].

Shown in Figure 5, an unstructured mesh was generated for this study. Around the hydrofoil
section, the mesh is structured. The total grid number is 311,253. The innermost layer of the
grid has an average height y+ = 2 from the solid boundary. The SST k- model was employed
to simulate the turbulent flow. The flows at = 2.5 and 3.5 were computed for three
different models. The cavitation number for the present study is = 0.55 which evidently
results in a supercavitating cavity. The Reynolds number based on the incoming flow speed
and the chord length is Re = 1.96 106.

The computation shows that convergent results were obtained only for the Zwart-Gerber-
Belamri model and the Schnerr-Sauer model. The employment of the full cavitation model
did not reach convergence in the computation. Therefore, in the following, only the solutions
for the two models which led to convergence are discussed.

Figure 6 shows the time-variations of lift and drag coefficients at = 2.5. It can be observed
that the periods with different models are somewhat different but the fluctuation amplitudes
and the time variations are similar. Shown in Figures 7 and 8 are the cavitation bubble
variation with equal time intervals in a period. The results are similar to those in [14] but the
(a) t = 0 (b) t = 0.125T

(c) t = 0.25T (d) t = 0.375T

(e) t = 0.5T (f) t = 0.625T

(g) t = 0.75T (h) t = 0.875T

Figure 7 Bubble variation in a period for Zwart-Gerber-Belamri model at = 2.5.

(a) t = 0 (b) t = 0.125T

(c) t = 0.25T (d) t = 0.375T

(e) t = 0.5T (f) t = 0.625T

(g) t = 0.75T (h) t = 0.875T

Figure 8 Bubble variation in a period for Schnerr-Sauer model at = 2.5.


bubble size on the upper side is smaller than that in [14]. Figure 9 shows the corresponding
pressure coefficient distributions on the lower surface for the eight time points. A strong
variation in time is observed. Furthermore, at t = 0.25T, the pressure coefficient curve by the
Zwart-Gerber-Belamri model agrees quite well with the result in [14] which is the only
available data for comparison.

Figures 10-11 shows the cavitating bubble evolution in a period for = 3.5. It is evident that
the cavitating bubble is much larger on the upper face, compared to that for = 2.5. And the
result shows that the cavitation computed by the Zwart-Gerber-Belamri model is weaker than
that by the Schnerr-Sauer model.

CONCLUSIONS

A comparison study of different cavitation models on the development of 2-D cavitating


flows has been conducted in the present study. Three mass/vapor transport models for, i.e. the
full cavitation model, the Zwart-Gerber-Belamri model, and the Schnerr-Sauer model, were
employed for study. All the three mixture models are derived from the simplified bubble
dynamics, based on the diffusion balance. Two study cases consist of one steady
supercavitation and one unsteady partial cavitation. For the case of supercavitation, all three
model exhibit good convergence and the solution agree with one another very well. However,
for the partial cavitation case, the full cavitation model led to divergent computations while
the other two models led to convergence in computations. The convergent solution exhibits a
periodic pattern. Even though they are comparable to each other, the deviation between the
two solutions is significant. We believe more tests are required before meaningful conclusions
can be reached.

(a) Zwart-Gerber-Belamri model (b) Schnerr-Sauer model

Figure 9 Pressure coefficient distributions at different time for = 2.5.

REFERENCES

1. Singhal, A.K., Athavale, M.M., Li, H., and Jiang, Y. (2002) Mathematical basis and
validation of the full cavitation model. J. Fluids. Eng., 124, 617-624.
2. Zwart, P.J., Gerber, A.G., and Belamri, T. (2004) A two-phase flow model for predicting
cavitation dynamics. 5th Int. Conf. Multiphase Flow, Yokohama, Japan.
3. Schnerr, G.H. and Sauer, J. (2001) Physical and numerical modelling of unsteady
cavitation dynamics, 4th Int. Conf. on Multiphase Flow, New Orleans, USA.
4. Gogate, P.R. and Kabadi, A.M. (2009) A review of applications of cavitation in
biochemical engineering/biotechnology. Biochem. Eng. J. 44, 60-72.
5. Huang, C.K. and Chen, J.-H. (2013) A comparison study of cavitating flow in a
ventricular assist device using laminar and turbulent model. J. Flow Control,
Measurement and Visualization. 1, 33-48.
6. Christopher S. and Kumaraswamy S. (2013) Identification of critical net positive suction
head from noise and vibration in a radial flow pump for different leading edge profiles of
the vane. J Fluids Eng. 135, 121301-1-15.
7. Tseng, C.-C. (2010) Modelling of Turbulent Cavitating Flows, Ph.D. Dissertation,
University of Michigan, Ann Arbor, MI, USA.
8. Huuva, T. (2008) Large Eddy Simulation of Cavitating and Non-cavitating Flow. Ph.D.
thesis, Chalmers University of Technology, Gothenburg, Sweden.
9. Liu, H.L., Wang, J., Wang, Y., Zhang, H., and Huang, H. (2014) Influence of the
empirical coefficients of cavitation model on predicting cavitating flow in the centrifugal
pump. Int. J. Archit. Ocean. Eng. 6, 119-131.
10. Saito, Y., Takami, R., Nakamori, I., and Ikohagi, T. (2007) Numerical analysis of
unsteady behavior of cloud cavitation around a NACA0015 foil. Comput. Mech. 40, 1-12.
11. Okita, K. and Kajishima, T. (2002) Numerical simulation of unsteady cavitating flow
around a hydrofoil. Trans. Jpn. Soc. Mech. Eng., Ser. B. 68, 637-644.
12. Kunz, R.F., Boger, D.A., Stinebring, D.R., Chyczewski, T.S., Lindau, J.W., Gibeling, H.J.,
Venkateswaran, S., and Govindan, T.R. (2000) A preconditioned NavierStokes method
for two-phase flows with application to cavitation prediction. Comput. Fluids 29, 849-875.
13. Wu, P.C. and Chen, J.-H. (2016) Numerical study on cavitating flow due to a hydrofoil
near a free surface. J. Ocean Eng. Sci. http://dx.doi.org/10.1016/j.joes.2016.02.002.
14. Frank, T., Lifante, C., Jebauer, S., Kuntz, M. and Rieck, K. (2007) CFD simulation of
cloud and tip vortex cavitation on hydrofoils. 6th Int. Conf. Multiphase Flow, Leipzig,
Germany.

(a) t = 0 (b) t = 0.125T

(c) t = 0.25T (d) t = 0.375T

(e) t = 0.5T (f) t = 0.625T

(g) t = 0.75T (h) t = 0.875T

Figure 10 Bubble variation in a period for Zwart-Gerber-Belamri model at = 3.5.

(a) t = 0 (b) t = 0.125T

(c) t = 0.25T (d) t = 0.375T

(e) t = 0.5T (f) t = 0.625T

(g) t = 0.75T (h) t = 0.875T

Figure 11 Bubble variation in a period for Schnerr-Sauer model at = 3.5.

You might also like