You are on page 1of 9

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/267210609

CAVITATING FLOW SIMULATION WITH MESH


DEVELOPMENT USING SALOME OPEN SOURCE
SOFTWARE

Conference Paper · October 2014


DOI: 10.13140/2.1.2423.0402

CITATIONS READS

2 637

4 authors:

Victor Hugo HIDALGO DIAZ Xianwu Luo


Escuela Politécnica Nacional Tsinghua University
35 PUBLICATIONS 37 CITATIONS 104 PUBLICATIONS 1,066 CITATIONS

SEE PROFILE SEE PROFILE

An YU Luis Ricardo Soto


Tsinghua University Escuela Politécnica Nacional
15 PUBLICATIONS 36 CITATIONS 5 PUBLICATIONS 3 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Design optimization of hydraulic machinery View project

EPN PIJ 15-11 View project

All content following this page was uploaded by Victor Hugo HIDALGO DIAZ on 28 November 2014.

The user has requested enhancement of the downloaded file.


Proceedings of the 11th International Conference on
Hydrodynamics (ICHD 2014)
October 19 - 24, 2014, Singapore

CAVITATING FLOW SIMULATION WITH MESH DEVELOPMENT USING


SALOME OPEN SOURCE SOFTWARE

V.H. HIDALGO, X.W. LUO∗ and A. YU


State Key Laboratory of Hydro science & Engineering, Tsinghua University,
Beijing 100084, China
E-mail: luoxw@mail.tsinghua.edu.cn

R. SOTO
Escuela Politecnica Nacional University, Department of Mechanical Engineering,
Ladrn de Guevara E11 253, Quito, Pichincha 17-01-2759, Ecuador
E-mail: ricardo.soto@epn.edu.ec

Cavitating flow simulation is very helpful for understanding the mechanism of cavitation in turbomachin-
ery. However, the simulation accuracy may depend on the quality of mesh generation for computation
domain. Consequently this paper treats the development of an anisotropic structured mesh using open
source software Salome. The unsteady cavitating flow around a hydrofoil NACA 0015 is solved based
on OpenFOAM, the results show that grid/mesh has important effects on the cavitating flow charac-
teristics such as cavity revolution, vortex shedding. The comparison of cavitation flows among different
mesh generation methods is conducted and it is noted that the cavitating flows with the structured
mesh and well controlled node distribution can reproduce experimental results fairly well. Further,
mesh development using Salome much reduces calculation time of the flow simulation.

Keywords: Cavitation; Salome; OpenFOAM; Mesh.

1. Introduction
Studies about cavitation and unsteady cavitating flow around hydrofoils are basis of hydrody-
namic machinery, because a correctly analysis and prediction allows to prevent cyclical stress
and erosion [1]. Most of previous studies focused to predict the cavity sheet, vortex shedding,
pressure fluctuation and other important aspects of cavitation [2]. Kubota et al. [3] presented a
cavitation model called BTF (bubble two phase flow) as homogeneous flow. Based on this idea
other cavitation models have been developed [4–6].
The cavitation phenomena shows a high Reynolds number, in which the Reynolds Average
Navier Stokes (RANS) performed poorly according to Yejun, et al. [7], whereas the LES were
more consistent. Ji, et al.[8, 9] used large eddy simulation (LES) to simulate unsteady cavitating
flow around a twisted hydrofoil. The investigation shows that LES outcomes presented good
accuracy with experimental results [10]. However, The mesh for LES should be very especial to
capture the flow around hydrofoil [11], and the challenge of the present paper.
There are several commercial programs to get a structured mesh, e.g., ICEM, Gambit, and
open source programs as Gmsh, Salome, etc. In OpenFOAM [12], there are other options , e.g.,
blockMesh and snappyHexMesh, which can be used to generate the structured mesh by writing
the main code of mesh.

1
2

Looking for a better option for meshing, Salome shows friendly open source and useful for
complex shapes, as hydrofoil NACA 66 and NACA 0015 [13]. Then, the mesh is transformed
to OpenFOAM requirements for the numerical simulation. Therefore, the use of open source to
obtain a structured mesh and make a numerical simulation are challenges in this research.

2. Physical Description
The unsteady cavitating flow is considering like multiphase flow, so the Eq. (1) represent the
characteristic of mixture flow, density and dynamic viscosity are calculated by Eq. (2) and
Eq. (3).

∀L
γ= (1)

ρ = γρL + (1 − γ)ρV (2)

µ = γµL + (1 − γ)µV (3)


where γ is the volume fraction, ∀ is the volume, ρ is the density of fluid, µ is the dynamic
viscosity, and L,V are the subindex for liquid and vapor.

2.1. Equations
Favre-filtering operation is applied to continuity and momentum equations, which are observed
with over bar in Eq. (4) to Eq. (6). These equations are the basis of LES.
 
∂uj 1 1
= ṁ − (4)
∂xj ρL ρV

∂(γ) ∂(γuj ) ṁ
+ = (5)
∂t ∂xj ρL
 
∂(ρui ) ∂(ρui uj ) ∂p ∂
+ =− + ρ (Ra − G) (6)
∂t ∂xj ∂xi ∂xj
where u is velocity, x is variable of space, p is pressure, t is time, ṁ is mass rate, i and
j are the indexes of axes 1, 2 or 3. Ra is filtered viscous
  tensor 2νS ij , where ν is the
stress
1 ∂ui ∂uj
kinematic viscosity, S ij is the rate of stress tensor 2 ∂xj + ∂xi . G is the subgrid stress tensor
(ui uj − ui uj ).

3. Cavitation Model
The cavitation model of Zwart is based on Rayleigh-Plesset equation [14, 15], as shown in Eq. (7).
This cavitation model is not native in OpenFOAM, but it is selected for the numerical simulation.
 r
3rnuc (1 − α)ρV 2 pV − p
 ṁ+ = −CV if p < pV


ṁ = rRB 3 ρL (7)
 − 3αρV 2 p − pV
 ṁ = CC
 if p > pV
RB 3 ρL
3

where CV = 300 and CC = 0.03 are the calibration constants for evaporation and condensa-
tion, rnuc = 5.0 × 10−6 is the nucleation site volume fraction, RB = 1.9 × 10− 6 m is the typical
bubble size in water [15]. The vapor volume fraction α = 1 − γ.
Equation (7) shows that the change of phase is not symmetry process, because vaporisation
m+ dependents of γ and rnuc , but condensation m− dependents of α [14].

4. Computation Domain and Mesh Generation


4.1. Hydrofoil
In this study, the hydrofoil selected is NACA 0015, as shown in Fig. 1.

Chord line
upper surface
leading edge
B
A
trailing edge

lower surface max. thickness

Fig. 1. NACA 0015

where c is the length of chord line AB, and x is a distance that begins on leading edge A in
the direction of trailing edge B. The computation domain is shown in Fig. 2.

Fig. 2. Computation Domain

4.2. Mesh Generation


The research is focused to obtain and test a good grid/mesh for unsteady cavitating flow. There-
fore, two types of mesh are performed, as observed in Fig. 3 and Fig. 4
The unstructured meshing (UM) is shown in Fig. 3. This kind of mesh gives facilities for
drawing complex shapes as hydrofoils, and is based on experience from previous research of
VZLU [13] in 2D airfoils, and Salome tetrahedron (Netgen) with wire discretization in the wall
of hydrofoil.
Figure 4 gives information about a structured meshing (SM) around a NACA 0015, the special
characteristic is the scale distribution of elements and nodes based on Salome hexahedron (i,j,
4

mesh refinement

Fig. 3. Unstructured Mesh around NACA 0015

scale distribution

Fig. 4. Structured Mesh around NACA 0015

k) and wire discretization. Table 1 shows the distribution of elements and nodes for UM and
SM.

Table 1 Special features of the mesh

Object UM SM
Nodes 16541 50310
Elements Total 79890 78890

Two dimensionless numbers are important to evaluate the mesh quality. The first one is
Ω, which is the relationship between number of elements N E and number of nodes N D, it
is calculated using Ω = N E/N N and values from Table 1. The second one is Yplus number
y + = (uτ y)/ν to know whether the mesh matches the LES method, where, uτ is the friction
velocity , y is the distance from nearest hydrofoil wall and ν is the kinematic viscosity.

5. Solver
5.1. OpenFOAM
The OpenFOAM used in this study is the version 2.2x, which is a modified version from 2.2 for
CentOS operating system (OS) according to OpenFOAM web page [12].
The reason to use OpenFOAM 2.2x are explained in the following items:
(1) The Tsinghua university server used for this simulation is GNU/Linux RetHad 4.1, so this
OS does not allow to use the newest versions of OpenFOAM, so a modified version is needed.
(2) OpenFOAM 2.2x is focused for CentOS OS. However, CentOS is based on RedHat distribu-
tion, which is the official version of RedHat software company [16]. In fact, the version 2.2x
works well in Tsinghua university server.
5

(3) The version 2.2x is important due to the fact that PIMPLE algorithm, which is a hybrid
algorithm between PISO (Pressure Implicit with Splitting of Operators) and SIMPLE(Semi-
Implicit Method for Pressure-Linked Equations) algorithms, it can be used for larger time
scale and saving some computational time.
(4) In OpenFOAM 2.2x is easier to adapt the Zwart model than previous version of Open-
FOAM 2.0. Therefore, Zwart model is implemented, and a new solver is generated as
vInterPhaseChangeFoam. In a future research is going to explain the steps to adapt Zwart
model in OpenFOAM 2.2x.

Implicit LES is used in OPEN, considering the subgrid scale effects similar to a strictly
dissipation action, understanding that the truncation error of leading order emulate the energy
dissipation in the flux discretization [17].

5.2. Boundary and Operation Conditions


The numerical simulation beginning at 0 s and the end time is 6 × 10−1 s with step of 1 × 10−5 s
and every 25 steps a result is written. Table 2 gives information about boundary conditions.

Table 2. Boundary Conditions

BOUNDARY CONDITION
Inlet velocity in x axis U∞ = 5.4772 m
s
Outlet pressure p = pr = 20.3kP a
Top and Bot wall
Front and Back symmetry planes
Hydrofoil wall wall

5.3. Natural Cavity


The natural cavity is defined by cavitation number σ in Eq. (8), where pr is absolute pressure
of system at constant temperature.

pr − pV
σ= 1 2
(8)
2 ρU∞

For the present simulation, σ is equal to 1.2.

6. Results and Discussion


Looking at Fig. 5, the taper analysis concludes that the UM shows a better distribution of faces
in UM than SM. However, SM changes are not bigger than the fraction 0.032, as observed in
Fig. 5. In fact, these two meshes are correctly drawing.
Table 3 gives information of dimensionless numbers in the present study. It is noted that Ω
of UM is four times greater than SM, this means that UM has more elements per node than SM.
In fact, the UM use more computer resources than SM for calculation. The y + of hydrofoil wall
should be less than 10 to match LES method, then SM shows better accuracy than UM, due to
the maximum value is 6.53.
6

0.00 0.032 0.064 0.097 0.130

(a) (b)

Fig. 5. Taper analysis of mesh surfaces for (a) UM and (b) SM

Table 3. Dimensionless numbers

NUMBER UM SM
Ω 4.83 1.57
Max y + 53.27 6.53
Min y + 1.36 0.32

0.01 pr 0.01 pr
γ γ
0.001 0.001

1e-04 1e-04

1e-05 1e-05

1e-06 1e-06

1e-07 1e-07
0 0.04 0.08 0.12 0.16 0 0.2 0.4 0.6

(a) (b)

Fig. 6. Residual plot for (a) UM and (b) SM

Figure 6 is the residual plot obtained by pyFoam [12], this graph shows the residuals of pr
and γ in the vertical axes for UM and SM. Although the patterns are random, simulation results
are acceptable because the maximum value for these meshes is less than 0.001. It is observed
that UM finished before the end time established that is probably to a floating point.
Further results are the comparisons among UM, SM and experiment results, as observed in
Fig. 7. Looking at UM, is noted that the cavity sheet has not presented a re-entrant jet effects,
so the cavity never separate from hydrofoil wall. On the other hand, the SM gives equivalent
results to experiments and presents re-entrant jet effects. Therefore, SM shows a better accuracy
for the case of cavitation.
7

(a) UM (b) SM (c) Experiment

re-en
trant
Jet

1 0.81T

2 0.83T

Fig. 7. Cavitation results comparison among (a) UM, (b) SM and (c) experiments

7. Conclusions
The following milestones are the study conclusions:
(1) The adapted model of Zwart in OpenFOAM 2.2x presents similar results to experiments
with the application of structured mesh.
(2) The method of wire discretization in the open source software Salome permits to draw a
scale distribution of mesh nodes, which is suitable for LES.

Acknowledgement
This work was financially supported by the National Natural Science Foundation of China
(Project Nos. 51206087 and 51376100), and the Major National Scientific Instrument and Equip-
ment Development project (Grant No. 2011YQ07004901).

References
[1] M. Dular, B. Stoffel and et al., Wear 261, 642 (2006).
[2] J.-P. Franc and J.-M. Michel, Fundamentals of cavitation (Springer, 2006).
[3] A. Kubota, H. Kato and H. Yamaguchi, Journal of fluid Mechanics 240, 59 (1992).
[4] R. F. Kunz, D. A. Boger, D. R. Stinebring, T. S. Chyczewski, J. W. Lindau, H. J. Gibeling,
S. Venkateswaran and T. R. Govindan, Computers & Fluids 29, 849 (2000).
[5] A. K. Singhal, M. M. Athavale, H. Li and Y. Jiang, Journal of Fluids Engineering 124, 617
(2002).
[6] O. Coutier-Delgosha, J. L. Reboud and Y. Delannoy, International Journal for Numerical
Methods in Fluids 42, 527 (2003).
[7] Y. Gong and F. X. Tanner, Comparison of RANS and LES models in the laminar limit for
a flow over a backward-facing step using OpenFOAM, in Nineteenth International Multidi-
mensional Engine Modeling Meeting at the SAE Congress, 2009.
[8] B. Ji, X. Luo, X. Peng and Y. Wu, Journal of Hydrodynamics, Ser. B 25, 510 (2013).
[9] X. Luo, B. Ji, X. Peng, H. Xu and M. Nishi, Journal of Fluids Engineering 134, p. 041202
(2012).
[10] R. F. Kunz, J. W. Lindau, T. A. Kaday and L. J. Peltier, Unsteady RANS and detached
eddy simulations of cavitating flow over a hydrofoil, in Fifth international symposium on
cavitation, Osaka, Japan, 2003.
8

[11] P. R. Spalart, W. H. Jou, M. Strelets and S. R. Allmaras, Advances in DNS/LES 1, p. 48


(1997).
[12] OpenFOAM, The openfoam foundation documentation, www.openfoam.org (2014).
[13] VZLU, The Vyzkumny a zkusebni letecky ustav (VZLU) documentation, www.vzlu.cz
(2013).
[14] M. Morgut, E. Nobile and I. Bilu, International Journal of Multiphase Flow 37, 620 (July
2011).
[15] B. Ji, X. Luo, Y. Wu and H. Xu, Chinese Physics Letters 29, p. 076401 (2012).
[16] C. Negus, Linux Bible 2010 Edition: Boot Up to Ubuntu, Fedora, KNOPPIX, Debian,
openSUSE, and 13 Other Distributions (Wiley. com, 2010).
[17] N. X. Lu, Tutorial, SE 412, p. 96 (2008).

View publication stats

You might also like