Sei sulla pagina 1di 6

Results in Physics 12 (2019) 1834–1839

Contents lists available at ScienceDirect

Results in Physics
journal homepage: www.elsevier.com/locate/rinp

Microarticle

Numerical study of cavitation in centrifugal pump conveying different liquid T


materials

Song-Sheng Deng, Guo-Dong Li , Jin-Fa Guan, Xao-Chen Chen, Lu-Xing Liu
Department of Fuels, Army Logistics University of PLA, Chongqing 401331, China

A R T I C LE I N FO A B S T R A C T

Keywords: A numerical study on cavitation in the centrifugal pump during the transportation of pure water and diesel is
Pump carried out. An improved cavitation model based on the Zwart–Gerber–Belamri Model is employed in the si-
Cavitation mulation. A Cash–Karp fourth–fifth order Runge–Kutta method was used to solve the original Rayleigh-Plesset
Viscosity equation to get the numerical solution. The effects from viscosity and surface tension on bubbles’ expansion and
Surface tension
compression are taken into account in the improved model instead of the traditional simplification in which the
Simulation
bubble radius is just a function of pressure and density. The coefficients got by fitting the numerical solution are
added to the constant pre-multiplier of the Zwart-Gerber-Belrami cavitation model and finally enhance the
model. The vapor volume fraction was predicted to be larger in diesel compared to water. The distribution of
pressure in diesel is more homogenous than that in pure water.

Introduction that the first phase must be liquid phase and the second phase must be
vapor phase. It cannot be used in the simulations connected with LES.
Cavitation is a common hydraulic phenomenon that occurs when The Zwart–Gerber–Belamri model expresses the mass transfer rate per
the local pressure inside the liquid is below the local saturated vapor unit volume only as a function of vapor phase density, so the speed of
pressure. If cavitation occurs in pumps or on propellers, it will generate calculation stability and convergence is significantly improved. The
a lot of noise and vibration, further causing damage to components. Schnerr-Sauer model has a different description of mass transfer
In addition to experimental research, since CFD was came up with, equation in which expression αv(1−αv) is included (αv is the vapor
more and more researchers have used numerical simulation methods to volume fraction).
predict cavitation in the pump [1–5]. Since CFD is used for cavitation Actually, a lot of researches have been made on cavitation simula-
simulation, many scholars have tried different methods to study it tion associated with pumps. For example, Liu et al. [17] investigated
[6–11]. Commonly used methods are direct numerical simulation the sheet cavitation on INSEAN E779A propeller with the Singhal
(DNS), large eddy simulation (LES), and Reynolds-averaged Navier–- model. In their simulation, the influence of no-condensable gas was
Stokes equations (RANS). DNS can get all the details of the in- taken into consideration. Kanemaru and Ando [18] used a modeling
stantaneous solution of the turbulent flow field. However, its calcula- approach for the camber surfaces to capture the steady or unsteady
tion cost is very large. In terms of DNS, Giacomo Falcucci et al. [12–13] cavitation phenomenon and the mothed was finally validated by the
proposed a direct numerical simulation method based on the lattice experiment of two six-bladed propellers. Sato et al. [19] used SST k-ω
Boltzmann method (LBM) to study the hydrodynamic field in the ca- turbulence model and unstructured grids to simulate the cavitation
vitation injector nozzle. The simulation results agreed well with their phenomenon around the propeller. The fundamental characteristics of
experimental results. Compared with DNS and LES, RANS has ad- the sheet cavitation was predicted well.
vantages in saving computational cost and is therefore favored by many In the study of many scholars, the cavitation models used are all
scholars. For a cavitation simulation, an appropriate cavitation model based on the simplification of Rayleigh-Plesset equation in which
should firstly be chosen to describe the mass transfer mechanism be- bubble radius is only the function of pressure difference and liquid
tween vapor phase and liquid phase. The commonly used cavitation density. So, the effects from viscosity and surface tension are both ne-
models are Singhal model [14], the Schnerr-Sauer model [15] and the glected. Under this assumption, the above models are possible for the
Zwart–Gerber–Belamri Model [16]. The Singhal et al. model requires cavitation simulation in pure water, but they are not accurate enough


Corresponding author.
E-mail address: lgd341@163.com (G.-D. Li).

https://doi.org/10.1016/j.rinp.2019.02.009
Received 30 September 2018; Received in revised form 12 January 2019; Accepted 4 February 2019
Available online 08 February 2019
2211-3797/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
S.-S. Deng, et al. Results in Physics 12 (2019) 1834–1839

Fig. 3. Diagram of the geometry of the pump.


Fig. 1. Numerical solutions of primitive Rayleigh-Plesset equation in water and
diesel during bubbles’ compression process.

Fig. 4. The boundary conditions of computational domain.

Fig. 2. Numerical solutions of primitive Rayleigh-Plesset equation in water and


diesel during bubbles’ expansion process.

Table 1
Values of ke and kc in different conditions.
Media properties water diesel
μ = 0.001 Pa·s, μ = 0.0024 Pa·s,
σ = 0.072 N/m σ = 0.026 N/m

kc 0.121 0.028
ke 0.728 0.391
Fig. 5. Mesh of computational domain.

for viscous liquids such as diesel and gasoline. dR 2 |P∞ − Pb |


= (−1)n
In this article, we attempt to use our previously proposed cavitation dt 3 ρL (1)
model [20] to simulate the cavitation in centrifugal pump while
transporting fluids of water and diesel whose viscosities and surface Here R stands for the bubble radius, Pb stands for the bubble surface
tensions are different. pressure, P∞ denotes the local far-field pressure, ρL represents the liquid
density.
It is derived from the Rayleigh-Plesset equation as follows,
Cavitation model
Pb − P∞ d 2R 3 dR 2 4v dR 2S
In fact, we have proposed an improved cavitation model considering = R· 2 + ·⎛ ⎞ + L · +
ρL dt 2 ⎝ dt ⎠ R dt ρL R (2)
the effects from fluid’s viscosity and surface tension on bubbles’ growth
and collapse in our previous research and the presentation of our im-
νL is the liquid kinematic viscosity, S is the liquid surface tension
proved model has been elaborated in our recently published paper (Ref.
coefficient.
[20]). Here we just introduce some main points.
In our developed cavitation model, Eq. (2) was firstly converted into
In traditional cavitation models, the Rayleigh-Plesset equation is
Eq. (3) by introducing dimensionless variables into it.
usually simplified as follows.

1835
S.-S. Deng, et al. Results in Physics 12 (2019) 1834–1839

d 2β 3 dβ 2 C dβ 2D
β· 2
+ ⎛ ⎞ + · + ±1=0
dτ 2 ⎝ dτ ⎠ β dτ β (3)
where
R t |P∞ − Pb |
β= , τ = (−1)n
R0 R0 ρL (4)
4μL S
C= (−1)n , D= (−1)n ,
R 0 ρL |P∞ − Pb | R 0 ·|P∞ − Pb | (5)
By solving Eq. (3) with Runge–Kutta method, numerical solutions
were got and shown in Figs. 1 and 2. By combining Eq. (4) with Eq. (5),
it is easy to get the relationship between dR/dt and dβ/dτ,
dR dβ |P∞ − Pb |
= ·(−1)n
dt dτ ρL (6)
By fitting the numerical solution, the values of dβ/dτ can be ob-
tained, thus new expression of dR/dt will be got and it reflects the ef-
fects from fluid viscosity and surface tension to some extent.
Based on Zwart-Gerber-Belamri model, the new equations de-
scribing mass transfer mechanism are written as follows.
3αnuc (1 − α v ) ρv 2 (Pv − P )
R e = Fvap ∗ (P < Pv )
RB 3 ρl (7)
Fig. 6. Contours of vapor volume fraction in pure water in the pump.
3α v ρv 2 (P − Pv )
R c = Fcond ∗ (P > Pv )
RB 3 ρl (8)
Fvap* = 1.2247ke·Fvap, Fcond* = 1.2247kc ·Fcond. Fvap and Fcond are
original coefficients in Zwart-Gerber-Belamri model. ke is the slope of
the fitted line of bubble radius in expansion process and kc is the slope
of the fitted line of bubble radius in compression process. ke and kc are
obtained based on the numerical solutions of unsimplified Rayleigh-
Plesset equation, the numerical solutions are shown in Figs. 1 and 2.
Values of ke and kc are shown in Table 1.

Turbulence models and multiphase models

Generally, there are three multiphase models available in Ansys


Fluent, VOF model, mixture model and Eulerian model. The VOF model
is a surface tracking method under a fixed Euler grid, which has certain
advantages for capturing phase interfaces. The mixture model performs
better in solving the momentum equations, mass equations, energy
equations of the mixture, as well as the volume fractions and relative
velocity equations of the second phase. Typical applications of the
mixture model are low-mass load particle flow, bubble flow, sedi-
mentation, and Cyclone separator, etc. The Eulerian model is more
suitable for describing phase-to-phase interactions and cannot be used
for cavitation problems in evaporation and compression. Actually, the
VOF model is more suitable for simulating the evolution of a single
bubble, such as the growth and collapse of a single bubble. As the ca-
Fig. 7. Contours of vapor volume fraction in diesel in the pump. vitation region studied here is a cavitation cloud composed of a large
number of vacuoles, VOF model is not a good choice. Here the mixture
model is chosen as the multiphase model.
LES (Large Eddy Simulation) is undoubtedly a good choice for ca-
vitation simulation when selecting the turbulence model. However, the

Fig. 8. The comparison of vapor volume fraction in pure water and diesel.

1836
S.-S. Deng, et al. Results in Physics 12 (2019) 1834–1839

Fig. 9. The distributions of vapor volume fraction on the pump case conveying diesel.

Fig. 10. The distributions of vapor volume fraction on the pump case conveying pure water.

huge computational cost of LES makes it difficult to widely promote in of the viscous underlayer, and in the mainstream region it employs k-ε
engineering calculation. Therefore, the RANS method is still favored by model instead of k-ω model to reduce the sensitivity to the inlet tur-
many scholars. As one of the k-ε models, the Realizable k-ε model has bulent kinetic energy parameters. Just like many other researchers,
certain advantages in the simulation of water jets, so many scholars here we also choose the SST k-ω model as the turbulence model.
directly use the Realizable k-ε model for cavitation prediction, and to
some extent it meets the engineering needs. Coutier-Delgosha [21] Geometry, grid and boundary conditions
analyzed the effects of several RANS turbulence models on the cavita-
tion simulation and concluded that the compression effects must be The geometry of the centrifugal pump is shown in Fig. 3. The three
considered into the turbulence model to accurately simulate the self- diameters are 110 mm, 13 mm and 34 mm respectively. The boundary
oscillation phenomenon in cavitation. The SST k-ω model has the ro- conditions of pressure inlet and pressure outlet are shown in Fig. 4.
bustness of the k-ω model on the near wall which can capture the flow Pressure-based solver is chosen for this steady simulation. The Coupled

1837
S.-S. Deng, et al. Results in Physics 12 (2019) 1834–1839

distribution in pure water is more concentrated and its volume fraction


is higher.
Figs. 11 and 12 show the cavitation on the impellers. In order to
make a quantitative comparison, we represent the projection length of
the cavitation cloud as b, and the overall projection length of the im-
peller as a + b. By measurement, the ratio of b to a in pure water is
0.69, and in diesel, the value is 0.83. The above data gives the in-
formation that cavitation region in diesel extends longer than pure
water.
From the perspective of simulation, the liquid viscosity and surface
tension are taken into account in the cavitation model, resulting in the
differences in cavitation morphology. However, from a physics point of
view, liquid viscosity and surface tension have significant effects on
bubbles’ growth and collapse.
In the mixture of liquid and bubbles, the viscosity of fluid works to
decrease the overall energy of the oscillating bubble during the inter-
Fig. 11. Contours of vapor volume fraction on the impellers in pure water. actions between bubbles and liquid for the diffusion of linear mo-
mentum. It is obvious that the greater the viscosity of the liquid, the
stronger this effect. The surface tension force will slow down the bub-
ble’s expansion process and speed up the compression process. For the
viscosity of diesel is higher than that of water and the surface tension is
lower than that of water, the combination of viscosity and surface
tension causes the bubbles to change more slowly in diesel, which is
manifested in a macroscopically large size of cavitation region in diesel.

Conclusions

1. Cavitation in the centrifugal pump conveying pure water and diesel


is simulated based on an improved cavitation model. The effects
coming from the viscosity and surface tension in pure water and
diesel were taken into consideration.
2. When the pump delivers different liquid media, the scale of the
cavitation region and the vapor volume fraction in the pump are
different, which is related to the viscosity and surface tension of the
Fig. 12. Contours of vapor volume fraction on the impellers in diesel. media.

algorithm was used for pressure–velocity coupling, and a PRESTO! Acknowledgement


scheme was used for pressure interpolation. Discretization schemes for
momentum and volume fraction were selected as First Order Upwind. This work is supported by the Basic Science and Advanced
Convergence criteria were equal to 10−6 for all cases. The impeller Technology Research Special Fund of Chongqing, China
speed was set to 1200 RPM and the inlet pressure was set 600 kPa. The (cstc2017jcjAX0166).
outlet pressure was set 200 kPa. The size of grids was chosen tetra-
hedron and some details are shown in Fig. 5. The maximum velocity of References
pure water at outlet were 34.95449 m/s, 35.00383 m/s and
35.00391 m/s when the number of nodes is 556672, 605144 and [1] Hirschi R, Dupont P, Avellan F, et al. Centrifugal pump performance drop due to
leading edge cavitation: numerical predictions compared with model tests. J Fluids
635820, so we finally choose 605144 nodes to mesh the domain. Eng 1998;120(4):705–11.
[2] Wursthorn S, Schnerr GH. Numerical investigation of performance losses in a cen-
trifugal pump due to cavitation. Zeitschrift Angewandte Mathematik Und Mechanik
Results and analysis 2001;81(S3):579–80.
[3] Coutier-Delgosha O, Fortes-Patella R, Reboud JL, et al. Experimental and numerical
studies in a centrifugal pump with two-dimensional curved blades in cavitating
The comparison between our improved cavitation model and the condition. J Fluids Eng 2003;125(6):492–7.
traditional cavitation model has been clarified in our previous article [4] Lindau JW, Pena C, Baker WJ, et al. Modeling of cavitating flow through waterjet
propulsors. Int J Rotating Mach 2012;2012(8).
[20] and the improved cavitation model was proved to be more accu-
[5] Nohmi M, Goto A, Iga Y. et al.Cavitation CFD in a centrifugal pump. The
rate to simulate cavitation phenomenon in fluid of different viscosity Proceedings of the Fluids engineering conference. The Japan Society of Mechanical
and surface tension force. The contours of vapor volume fraction are Engineers; 2003.
[6] Zeidan Dia, Zhang Lucy, Goncalves Eric. Cavitating bubbly flow computations by
shown in Figs. 6 and 7. Fig. 6 shows the overall distribution of cavi-
means of mixture balance equations. 3rd Thermal and Fluids Engineering
tation in pure water and Fig. 7 shows that in diesel. It can be seen that Conference. 2018. p. 1681–7.
the cavitation areas in diesel are relatively larger than that in pure [7] Lebon GSB, Tzanakis I, Djambazov G, et al. Numerical modelling of ultrasonic
water. This phenomenon can also be seen in Fig. 8. The scale of cavi- waves in a bubbly Newtonian liquid using a high-order acoustic cavitation model.
Ultrason Sonochem 2017;37:660–8.
tation cloud in pure water (l1) is obviously smaller than that in diesel [8] Zeidan D, Goncalves E, Lewis R, et al. Numerical study of turbulent cavitating flows
(l2). in thermal regime. Int J Numer Meth Heat Fluid Flow 2017:1487–503.
The distributions of vapor volume fraction on the pump case are [9] Eric Goncalvès, Dia Z. Numerical simulation of unsteady cavitation in liquid hy-
drogen flows. Int J Eng Syst Model Simul 2017;9(1):41–51.
shown in Figs. 9 and 10. Here the x and y axes represent the x and y [10] Asnaghi A, Feymark A, Bensow RE. Improvement of cavitation mass transfer
coordinates, respectively, and the z axis represents the vapor volume modeling based on local flow properties. Int J Multiph Flow 2017;93:142–57.
fraction. The three-dimensional diagram of vapor distribution in pure [11] Zeidan D. Assessment of mixture two-phase flow equations for volcanic flows using
Godunov-type methods. Elsevier Science Inc.; 2016.
water is relatively higher and thinner, which means that the vapor

1838
S.-S. Deng, et al. Results in Physics 12 (2019) 1834–1839

[12] Falcucci G, Jannelli E, Ubertini S, et al. Direct numerical evidence of stress-induced steady and unsteady sheet cavitation on 2D and 3D hydrofoils. 2009.
cavitation. J Fluid Mech 2013;728:14. [18] Kanemaru T, Ando J. Calculation of steady cavitation on a marine propeller using a
[13] Giacomo F, Stefano U, Gino B, et al. Lattice Boltzmann simulation of cavitating simple surface panel method. J Jpn Soc Naval Arch Ocean Eng 2008;7:151–61.
flows. Comm Comput Phys 2013;13(3):685–95. [19] Sato R, Tasaki R, Nishiyama S. Observation of flow on a horizontal flat plate above a
[14] Singhal AK, Athavale MM, Li H, et al. Mathematical basis and validation of the full working propeller and physics of propeller-hull vortex cavitation. Proc. Int. Symp.
cavitation model. J Fluids Eng 2002;124(3):617–24. on Propeller-Hull Vortex Cavitation, Wuxi, China, April. 1986. p. 118–25.
[15] Schnerr GH, Sauer J. Physical and numerical modeling of unsteady cavitation dy- [20] Li Guo-Dong, Deng Song-Sheng, Guan Jin-Fa, Chen Yan. Numerical study on cavi-
namics. Fourth International Conference on Multiphase Flow, New Orleans, USA. tation generation in a water jet added with turbulent drag-reducing additives based
2001. on improved cavitation model and cross model. Ind Eng Chem Res
[16] Zwart PJ, Gerber AG, Belamri T. A two-phase flow model for predicting cavitation 2018;57:9665–73.
dynamics. Fifth International Conference on Multiphase Flow, Yokohama, Japan. [21] Coutier-Delgosha O, Reboud JL, Delannoy Y. Numerical simulation of the unsteady
2004. behaviour of cavitating flows. Int J Numer Meth Fluids 2003;42(5):519–30.
[17] Li D, Grekula M, Lindell P. A modified SST k-? turbulence model to predict the

1839

Potrebbero piacerti anche