You are on page 1of 5

International Renewable Energy Congress

November 5-7, 2010 – Sousse, Tunisia

A new model to simulate a cavitating flow

Hatem KANFOUDI1, Hedi LAMLOUMI2 and Ridha ZGOLLI3

1
LMHE, ENIT, Hydraulic engineering
B.P. 37, Tunis, 1002, Tunisia
e-mail: hatem.kanfoudi@enit.tn
2 LMHE, ENIT, Hydraulic engineering
B.P. 37, Tunis, 1002, Tunisia
e-mail: lamloumihedi@yahoo.fr
3 LMHE, ENIT, Hydraulic engineering
B.P. 37, Tunis, 1002, Tunisia
e-mail: ridha.zgolli@enit.tn

Abstract - For numerical simulation of cavitating other numerical methods.


flows, many numerical models currently proposed
use some assumptions or/and empirical 2. Mathematical formulation
formulations that must limit their performance. A
new model based on the void fraction transport 2.1. Governing equations
equation solved with the source term evaluating
vaporization and condensation processes. The The flow with possible coexistence of liquid and
model is coupled with a CFD code solving the vapour (and /or gas) is treated as a homogeneous
Reynolds-averaged Navier-Stokes equations for mixture, and the governing equations are the
the mixture (liquid and/or vapour) to approach continuity (1) and the momentum (2) equations;

 m    mu j 
the cavitating flow. To test the validation of the
numerical simulation, the results obtained with a  0 (1)
2D approach for the flow around a NACA0009 t x j
hydrofoil for steady flow.
  mui    mu j ui  p
Keywords – vaporization, condensation,  
turbulence,hydrofoil t x j xi
(2)
   u u j  
 m  t   i 
1. Introduction
  
The phenomenon of cavitation that occurs within the x j   x j xi  
flow of a liquid can be searched for specific
industrial applications as it should be avoided in The mixture is a hypothetical fluid with variable
order not to suffer adverse consequences in other density  m ;
applications. In all cases we must learn to predict,
and in this regard there is more research work. Our
contribute here with the presentation of our
m  v  1    l (3)
development with the aim to develop a numerical
Where α is the vapour fraction (α =1 for pure vapour
method to simulate the cavitating flow. The model
and α =0 for liquid). The density of liquid and vapour
presented here is developed in an attempt to predict
are assumed to be constant.
the onset of cavitation as a result of pressure drop
The mixture viscosity is defined for given a constant
and also the changes in the flow. The model is based
vapour viscosity and a constant liquid viscosity;
on the source term of the transport equation
computing the vapour volume fraction which has the m  v  1    l (4)
special permit to reflect the quality of the liquid and
also its tension surface. To validate the method we 2.2. Cavitation Source Terms
consider the flow around hydrofoils that have been
the subject of experimental measurements and also The vapour volume fraction α is an additional

ID111/ ©IREC2010 310


unknown of the problem. And to close the system of 3k
R 
equations, an additional relation is needed. In this  behavior of the gas pg  pg 0  0 
numerical approach we use the dynamics of spherical in the bubble:  R
bubbles to simulate the presence of the cavity into the
liquid flow.  local liquid pressure p
The vapour volume fraction can be written as; :
4
n0  R3 2
Vvap 3 
pst 
  (5) surface tension : R
Vtot 4
1  n0  R3
3 4 R
 viscosity: pvi 
where Vvap and Vtot are the volume of vapour and the R
total volume of cell from the computed flow’s
domain, R is the bubble radius and n0 is defined as   3 R 2  1 d  R 2 R3 
RR
 2 dt  
Noting that
nuclei concentration per unit volume of pure liquid 2 2 RR
(n0≈1012 B/m3 for water Fujimoto [1]).For the model Equation (10) can be integrated, after neglect the
proposed, the given parmeter n0 will be used to treat viscosity and the behavior of the gas terms :
the quality of liquid and to calibrate with
1
experimental results.  2 pv  p  R03  2  R02 
2

Using this assumption you can calculate the vapour R   1  3   1  2  (11)
production rate with the following equation:  3 l   R  l R  R 
d 3 1    Rewritten the equation (9), the new model can be
 (6) expressed in this form :
dR R
Using the assumption of mixture considered as S  C pd  f  g  , p   sign  g  , p   (12)
homogenous fluid, the equation for the additionnal
variable , to be solved with the contunuity and with :
momuentum equations, is the vapour volume fraction  v l
C pd  3.3  n0  3
1
transport equation that can be easely written as; (13)
m
  v    v u j 
  S  m (7)
t x j f    1  n0 h    3
2
2
(14)
where
  d
1

 2 pv  p   4 R03 
2
m  v l (8)
m dt   1  
 3  l   h   
g  , p   (15)
it is the interphase mass flow rate per unit volume,
3  2.5 R 2 
that can be written, using the relation (6) :  1  0

 l h    h   3 
1 2

  R 3
 
m  3 1    v l (9)
m R

Where R is the bubble vapour-liquid interface
h    (16)
n0 1   
 , we use the Rayleigh-
velocity. To evaluate the m
Plesset equation (10) to compute R .
  3 R 2  1  p  p  p  p  p  (10)
RR In this study, the effect of surface tension ( = 0.0
2 l  v g st vi 
N/m) is disable. The surface tension pressure will be
with the terms on the right-hand side are defined as : the focus of future study further.

pv A comparative study between the differnet


 saturation pressure: vaporization and condensation terms propsed for the
the void fraction transport equation. Most of the
terms depend mainly on the diffence between the
local pressure and the vapour pressure p-pv. Thus,

ID111/ ©IREC2010 311


the following comparaison between the models is 3. Numerical Result
based on the expression of the sources terms as a
function of p-pv. However, the void fraction α usually 3.1. Numerical study
also appears in the expression of the source terms. To
expres them as a function of p-pv only, the barotropic To validate the new model, we confronted to an
state law of Delannoy is used (Fig. 1). experimental measures of pressure coefficient and
velocity distribution, in addition we compare with a
numerical model (Schnerr and Sauer model [3]) for
the same flow condition. The application is a
NACA0009 hydrofoil, truncated at 90% of the
original chord length. It has the final dimensions of
100 mm of chord length. The 3D test section is
modeled by a quasi 2D domain, with five rows of
cells in spanwise direction for the numerical domain.
Papers should be formatted for standard A4 size (210 The same mesh and numerical setup is used for the
× 297 mm) paper. All printed material, including computations with the two models. The density of
text, illustrations, and charts, must be kept within a bubbles and the initial radius has been adjusted and
print area of 172 mm (6.75 inches) wide by 247 mm fixed for all this study, n0=1018 B/m3, R0=10-6m.
(9.7 inches) high.
Do not write or print anything outside the print area. The domain is 9 blocks C-type grid of (5×88 000
The top margin must be 25 mm (1 inch), except for mesh cells (Fig. 2). The steady state RANS
the title page where top margin is 35 mm. The left simulations with SST turbulent model. The boundary
margin must be 19 mm (0.75 inch). All text must be conditions are set using a velocity inlet (Cref=20 m/s)
and a pressure at the outlet (the parameter which
fixes the cavitation number).

in a two-column format. Columns are to be 82 mm


(3.25 inches) wide, with 8 mm (0.31 inch) space
between them. Text must be fully justified.
Figure 2 : NACA0009 domain grid.

3.2. Pressure Distribution


Figure 1 : Comparison between the sources terms,
the Upper the vaporization process and the Lower the To validated the new model for a non-cavitating
condensation process. flow, a comparison with an experimental
measurements for a =1.38 has been performed; the
In literature, empirical factors are determined through figure Fig. 3 illustrates this comparison.
numerical/experimental results and are adjusted for
different geometries and different flow conditions.
To make this comparison possible, the empirical
factors (production/destruction coefficients) are
adjusted to obtain the same maximum value for the
source terms. The empirical factors have the
following values:

Cp=10. , Cd=0.7 for the Kunz model, Cp=410-4. ,


Cd=1.4 for the Singhal model, Cp=50. , Cd=0.005 for
the Schnerr model and n0=1010 B/m3 for the Yuan and
New models.

Figure 3 : Pressure distribution on NACA0009,


i=2.5°, =1.38[2]

ID111/ ©IREC2010 312


We remark the good agreement of the numerical
model compare to experimental data for a non-
cavitating flow.

After validation for a low regime of cavitation, we


proceed with a high regime of cavitation for a small
value of . Experimental data concerning the
NACA0009 hydrofoil are reported and compared
with the results of the computations (New Model and
Schnerr and Sauer model [3]) in Figs. 4 and Fig. 5.
We have presented in the same figure the vapour
volume fraction computed by both models for
different .

If we compare New Model with the model of Schnerr


and Sauer, we note that the model of Schnerr and New Model
Sauer model has failed to modeling the transition
from vapour to liquid, however the New Model has a
tendency to approach the experimental
measurements.

The Fig. 6 shows the impact of the growth of the Schnerr and Sauer
cavity pocket to the pressure distribution.

Clearly the New Model can predict the cavitation


flow for a high and low regime.
α

0 0.5 1

Figure 5 : Pressure distribution and cavity shape on


NACA0009, i=2.5°, =0.80 [4].

New Model

Schnerr and Sauer


Figure 6 : Volume fraction of vapour distribution,
Pressure distribution (New Model) and cavity shape
on NACA0009, i=2.5°, =0.75.
α
3.3. Velocity Distribution
0 0.5 1
Figure 4 : Pressure distribution and cavity shape on In this section, the velocity distribution around a
NACA0009, i=2.5°, =0.75 [4]. developed leading edge cavity is analyzed. The case
study concerns the same NACA0009 hydrofoil with a

ID111/ ©IREC2010 313


cavitation number =0.81 where the experimental =0.81. Profiles 20% of chord length [4].
data are reported by Dupont [7]. This regime is
characterized by an attached cavity (x/c = 33 %). The
pressure distribution and velocity distribution for the We remark the good agreement with the
two used models are reported in Fig. 7, 8 and 9 experimental measurements for x/c =10% and 20%
respectively. for the New Model compared to Schnerr and Sauer
model [3].

4. Conclusion
This study present a numerical method approaching
cavitating flows that uses a CFD code solving the
Navier-Stokes equations with homogeneous mixture
consideration. This method is based on the
introduction of a model of source term of transport
equation coupling the pressure calculation with the
volume fraction distribution. The model presented
here has shown the ease with which one can calibrate
to suit different qualities of the liquid considered.
Previously we gave some insights on how to validate
Figure 7 : Pressure distribution on NACA0009,
and especially the opportunities to rely on such
i=2.5°, =0.81[4].
models to monitor changes in length and also the
shape of the cavity which occurs in cavitating flow.

5. References
[1] FUJIMOTO, H.G., T. NISHIKORIi, Y.
HOJYO, T. TZUMAKOTO, et J. SENDA.
"Modelling of atomization and vaporization process
in flash boiling spray." France, 1994.
[2] Sauer, G. H. Schnerr and J. «Physical and
numerical modeling of unsteady cavitation
dynamics.» in Proceedings of the 4th International
Conference on Multiphase Flow (IMCF’01). New
Orleand,La, USA, 2001.
[3] Dupont, P. Etude de la Dynamique d’une
Poche de Cavitation Partielle en Vue de la Prédiction
Figure 8 : Computed and measured averaged de l’ Erosion dans les Turbomachines Hydrauliques.
dimensionless velocity profiles Cx/Cref and Cy/Cref Ecole Polytechnique Fédérale de Lausanne, 1991.
on the NACA009 hydrofoil suction-side at i=2.5°, [4] Youcef, AIT BOUZIAD. Physical
=0.81. Profiles at 10% of chord length [4]. modelling of leading edge cavitation: computational
methodologies and application to hydraulic
machinery. Fensh: Thesis, 2006.

Figure 9 : Computed and measured averaged


dimensionless velocity profiles Cx/Cref and Cy/Cref
on the NACA009 hydrofoil suction-side at i=2.5°,

ID111/ ©IREC2010 314

You might also like