Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Proceedings of ASME-JSME-KSME Joint Fluids Engineering Conference 2011

AJK2011-FED
July 24-29, 2011, Hamamatsu, Shizuoka, JAPAN

AJK2011-06084

INDUSTRY-UNIVERSITY COLLABORATIVE PROJECT ON NUMERICAL


PREDICTIONS OF CAVITATING FLOWS IN HYDRAULIC MACHINERY
- Part 1: Benchmark test on cavitating hydrofoils -

Chisachi Kato
Institute of Industrial Science, The University of Tokyo
Meguro-ku, Tokyo, Japan

ABSTRACT INTRODUCTION
Through an industry-university collaborative project, Prediction and control of cavitating flows are one of the
extensive benchmark studies have been made for numerical crucial issues in the design of hydraulic machinery such as a
prediction of cavitating flows around two-dimensional pump and a turbine. Various cavitation models have been
Hydrofoils: Clark-Y 11.7% and NACA0015. The emphases are proposed and implemented in commercial CFD software and
placed on the ability of present cavitation models to predict the they are, at least to some extent, used in the design of such
breakdown characteristics for these hydrofoils. The machinery (see [1] for example). The prediction capability of
benchmarking was done for a light and a moderate loading present cavitation models, however, has not yet been clarified
condition of these hydrofoils at a chord-based Reynolds and their intrinsic limitations have not fully been understood.
number in the order of 106. Four commercial CFD flow solvers, Although present cavitation models perform quite well for the
ANSYS CFX, ANSYS Fluent, and STAR-CCM+, and prediction of head-down characteristics of hydraulic machinery
SCRYU/Tetra, along with four open-source or in-house flow in some cases, they miserably fail to do so in other cases. Along
solvers in universities participated in this benchmark. All the with relatively high computational cost (i.e. long CPU time)
cavitation models, except one, implemented in these flow generally required for the prediction of a breakdown
solvers are based on an assumption of homogenous media of characteristics, this is why cavitation CFD has not yet been
one fluid, for which inception, growth, decay and destruction of used so widely in the design of hydraulic machinery as non-
cavitation are expressed by density change of the mixture fluid cavitating flow CFD is used for performance prediction of such
composed of liquid and gas phases. They differ with each other machinery. Accumulation of know-hows by which most
in how they determine the mixture fluid density and can be reliable prediction of cavitating flows can be achieved seems
categorized into of barotropic type or of source-sink type. also necessary in order for an engineer to apply cavitation CFD
Despites these differences in the cavitation models to the design of hydraulic machinery with confidence.
themselves and differences in the Navier-Stokes solvers, With this situation regarding cavitating flow computations
turbulence models and computational grids, the results of the in mind, we recently initiated an industry-university
benchmark show a consistent trend of discrepancy between the collaborative research project that is aimed at benchmarking
predicted and measured breakdown characteristics. Namely, various cavitation models for their prediction capability. The
none of the cavitation models is able to predict sudden drop of specific objectives of this project are, through extensive
the lift coefficient near the breakdown point confirmed in the validation studies, to understand the real capability of the
measured characteristics. The lift coefficients predicted by all present cavitation models to predict cavitating flows around a
the cavitation models show a gradual decrease with decreasing simple hydrofoil and breakdown characteristics of typical types
cavitation number. This discrepancy between the predicted and of pumps, to understand the intrinsic limitations (if any)
measured breakdown characteristics is most prominent at the associated with the present cavitation models, and finally, based
higher loading condition for NACA0015. But, it is consistently on the knowledge acquired in the above-mentioned benchmark
confirmed for the other cases investigated in this benchmark. studies, to improve the present cavitation models and/or to
The difference seems to be the results of under prediction of the propose a new cavitation model for a better prediction of
cavity length, which probably comes from an intrinsic cavitating flows.
limitation associated with a cavitation model based on an Cavitation in hydraulic machinery is also concerned with
assumption of homogeneous media of one fluid. structural vibration, noise, surging, material erosion, and

Copyright © 2011 by JSME


thermal effects. But, prediction of these phenomena is out of water temperature was set from 12.1 ℃ to 13. 8 ℃ and the
scope of this project although some of them are most likely to volumetric air content ratio was approximately 1 %. The data
become the main theme of a succeeding project, which will were taken at a water speed of 11.0 m/s, resulting in a chord-
probably be started in October, 2011. based Reynolds number of 7.7 × 105 . The angle of attack was
This paper describes the organization of the project and set to 1.97 degrees and 7.97 degrees. This data set contains
results of benchmark studies for basic cavitating flows around variations of lift and drag coefficients with respect to cavitation
the hydrofoils. It also discusses limitations intrinsically number, and time-averaged distributions of the static pressure
associated with the present cavitation models that are becoming on the suction and pressure surfaces of the hydrofoil at the non
apparent through these benchmark studies. cavitating condition.
The benchmark test for the NACA0015 hydrofoil refers to
experimental data measured in Marine Propeller Cavitation
PROJECT ORGANIZATION Tunnel in The University of Tokyo. A hydrofoil with a chord
This is a two-year long project started in April, 2009 and is length of 150 mm was installed in the test section of a water
led by the present author of The University of Tokyo with tunnel, which was 600 mm high and 1,000 mm long. The water
cooperation of participating groups. One research institute temperature was set from 13 ℃ to 15 ℃ and the volumetric
(National Agency), seven laboratories from five universities, air content ratio was approximately 1.1%. The water speed was
five CFD software venders and eight turbomachinery set to 8 m/s, resulting in a chord-based Reynolds number of
manufactures, listed at the end of this paper, are participating in 1.2 × 106 . The angle of attack was set to 4.0 degrees and 8.0
this initiative. The project is organized by three work groups, degrees. This data set contains variations of lift and drag
each of which has a specific goal associated with the project’s coefficients and cavity length with respect to cavitation number,
overall objectives. and time-averaged distributions of the static pressure on the
The first work group is responsible for extensive survey of suction and pressure surfaces of the hydrofoil for typical
the literature regarding CFD predictions of cavitating flows, the cavitation numbers. On the side wall of the tunnel, fluctuations
second work group for benchmark for basic cavitating flows in static pressure were also measured and cavitation was
around hydrofoils, and the third for benchmark for actual visualized by a high-speed camera at a maximum frame rate of
products of pumps. 10 kHz.
The literature search has been made in order to classify the
cavitation models presented so far, to identify the leading
institutions in this field of research, and to know cutting-edge
applications of cavitation computations both to basic flows and
to turbomachinery flows. The benchmark of the cavitation
models for actual hydraulic machinery is currently underway
for a total of six turbo pumps: three centrifugal pumps and three
mixed-flow pumps, results of which will be presented
elsewhere.

BENCHMARK PROBLEMS

Test Hydrofoils (a) Clark Y-11.7%


Benchmark tests of various cavitation models for basic
types of flows have been done for two types of hydrofoils:
Clark Y-11.7% and NACA0015. The former is a chambered
hydrofoil and has moderate loading (suction peak) near its
leading edge while the latter is a symmetric hydrofoil and has
strong loading (suction peak) near its leading edge. The section
of these hydrofoil is presented in Fig. 1.

Reference Data Sets


Experimental data measured in former Institute of High
Speed Mechanics at Tohoku Imperial University (Institute of
Fluid Science, Tohoku University at present) are used as
reference data set for the Clark Y-11.7% hydrofoil. The data (b) NACA0015
were measured for a hydrofoil with a chord and spanwise
length both of 70 mm installed in the test section of a cavitation FIGURE 1. SECTION OF TEST HYDROFOILS
tunnel with a height of 190 mm and length of 700 mm. The

Copyright © 2011 by JSME


Benchmark Rules CAVITATION MODELS AND FLOW SOLVERS
Standard computational domain and boundary conditions Four commercial flow solvers and four in-house or open-
were set for these benchmark tests as shown in Fig. 2 and their source flow solvers participated in this benchmark for the basic
use was recommended. But, choice of Navier-Stokes solver (i.e. cavitating flows. Except one cavitation model, all the other
incompressible or compressible, two or three dimensional, and cavitation models implemented in these flow solvers and thus
steady or unsteady computations), turbulence model, and tested in this benchmark are all based on an assumption of
numerical scheme and generation of the computational grid homogeneous media of one fluid. In this type of cavitation
were all entrusted to the individual participating group. The model, inception, growth, decay and destruction of cavitation
reference data set were distributed to the participants prior to are expressed by density change of the mixture fluid composed
their computations. Under these conditions, the participating of liquid and vapor phases. These cavitation models differ with
groups were requested to present the best possible predictions each other in how they determine the mixture fluid density and
for the benchmark cases described above. can be categorized into two groups. The first group, generally
called as a barotropic model, determines the density of the
mixture fluid by an equation of state for the mixture fluid while
the second group, generally called as a source-sink model,
solves a transport equation for the liquid or vapor fractions with
source and sink terms of the vapor phase, which can be
determined either by a simplified Rayleigh-Plesset equation
(bubble dynamic model) or by the Henry’s law (phase change
model). The flow solvers and their cavitation models are
summarized in Tab. 1 along with turbulence model, Navier-
Stokes formulations, and numerical schemes.
FIGURE 2. RECOMMENDED COMPUTATIONAL DOMAIN
AND BOUNDARY CONDITIONS

TABLE 1. FLOW SOLVERS AND CAVIATION MODELS

Discretized
software cavitation model NS solver 2 D/3 D Turbulence model numerical scheme
by
Simplified Rayleigh- 2 D and Finite Hybrid upwind+2nd
ANSYS CFX incompressible k-ω SST
Plesset 3D Volume order Euler Backward
Finite QUICK
ANSYS Fluent Full Cavitation Model incompressible 2D k-ω SST
Volume +Euler Backward
Simplified Rayleigh- Finite SIMPLE
STAR-CCM+ incompressible 2D DES (S-A model)
Plesset Volume +2nd order upwind
Eq. of state +full Finite
SCRYU/Tetra compressible 2D RNG k-ω 2nd order MUSCL TVD
cavitation model Volume
FrontFlow/blue Simplified Rayleigh- LES (Dynamic Finite Fractional step+
incompressible 3D
(Univ. Tokyo) [2]-[4] Plesset Smagorinsky Model) Element Crank-Nicolson
in-house Simplified Rayleigh- Finite QUICK
incompressible 2D k-ω SST
(Univ. Tokyo) [5] Plesset+Level Set Volume +Euler Backward
in-house Eq. of state + phase Modified Baldwin- Finite ADI+3rd order MUSCL
compressible 2D
(Tohoku Univ.) [6]-[8] change model Lomax Model Volume TVD
in-house Simplified Rayleigh- LES (Standard Finite Fractional step+2nd
incompressible 3D
(Osaka Univ.) [9], [10] Plesset Smagorinsky Model) Volume order Adams-Bashforce

𝑝∞ − 𝑝𝑣
σ≡
RESULTS 1
𝜌 𝑢2
2 𝑙 ∞
Breakdown Characteristics
where 𝑝∞ and 𝑢∞ are respectively the static pressure and
Figures 3 and 4 respectively show variation of lift (CL) and
velocity of the upstream uniform water flow while pv and 𝜌𝑙
drag (CD) coefficients with respect to cavitation number, σ.
are respectively the saturated vapor pressure and density of
Cavitation number is defined as:
water.

Copyright © 2011 by JSME


(a) Clark Y-11.7%, α = 1.97° (a) Clark Y-11.7%, α = 1.97°

(b) Clark Y-11.7%, α = 7.97° (b) Clark Y-11.7%, α = 7.97°

(c) NACA0015, α = 4.0° (c) NACA0015, α = 4.0°

(d) NACA0015, α = 8.0° (d) NACA0015, α = 8.0°

FIGURE 3. VARIATIONS OF LIFT COEFFICIENTS FIGURE 4. VARIATIONS OF DRAG COEFFICIENTS

Copyright © 2011 by JSME


None of the cavitation models tested in this benchmark and measured static pressure on the suction surface of the
was able to predict the breakdown characteristics of these hydrofoil at around 𝑥 ⁄𝐶 = 0.3~0.7 where the loading of the
hydrofoils correctly. When the cavitation number is decreased hydrofoil is remarkably under-predicted. A similar trend of
from a non-cavitating condition, all the predicted lift difference is confirmed for other cavitating flow conditions
coefficients show a general trend of gradual decrease while the although they are not shown in this paper. This is most likely to
measured ones show a sudden drop near the cavitation number be related to under-prediction of the cavity length, for which
where the breakdown takes place. Although this discrepancy detailed discussion will be given in the next section.
between the predicted and measured breakdown characteristics
is most prominent at the higher angle of attack (α = 8.0°) for
NACA0015 hydrofoil, it is consistently confirmed for other
cases, too. It is surprising to note that despite many differences
in the cavitation models as well as the Navier-Stokes solvers
and computational grids, all the cavitation models show an
essentially similar trend for variation of the lift coefficients.
It may have been reported in the literature that these
cavitation models have successfully predicted breakdown
characteristics of hydraulic machinery. From the results
presented here, however, we cannot fully rely on the present
cavitation models for prediction of the breakdown
characteristics of hydraulic machinery, which is primarily
determined by the lifting of the impeller’s blades. A similar
trend of discrepancy is confirmed for the drag coefficients (a) Clark Y-11.7%, α = 1.97°
where the predicted drag coefficients begin to increase at a
cavitation number much greater than the measurements.

Distributions of Static Pressure around Hydrofoils


Figure 5 compares, for some typical cases, the predicted
and measured distributions of the static pressure coefficients,
CP, on the hydrofoil surfaces at the non-cavitating condition,
where CP is defined as:

𝑝 − 𝑝∞
𝐶𝑃 ≡
1
𝜌 𝑢2
2 𝑙 ∞

Agreement between the predictions and measurements is (b) NACA0015, α = 8.0°


satisfactory for these cases. The computed distributions of the
static pressure show a slight over-prediction of loading for the FIGURE 5. COMPARISONS OF STATIC PRESSURE AROUND
Clark Y-11.7% hydrofoil. But, the author believes that this HYDROFOIL AT NON-CAVITATING CONDITION
discrepancy is presumably attributed to the measurement errors.
Likewise, the static pressure on the pressure side of the
NACA0015 hydrofoil is slightly over-predicted as compared to
the measured one, which is most likely to be the result of
difference in the reference pressure in the measurement and
computations. Although not shown in this paper, other cases
also show the same level of agreement at the non-cavitating
flow condition.
For cavitating-flow conditions, none of the models was
able to predict static pressure distributions accurately, as shown
in Fig. 6 where static pressure distributions are compared at
α = 8.0° , σ = 1.4 for the NACA0015 hydrofoil. At this
cavitation number, the measured lift coefficient is almost
unchanged from the non-cavitating condition while the
NACA0015, α = 8.0°, σ = 1.4
predicted lift coefficients are already considerably smaller than
the non-cavitating conditions (see the bottom figure of Fig. 3). FIGURE 6. COMARISON OF STATIC PRESSURE AROUND
This discrepancy is attributed to the difference in the predicted HYDROFOIL AT A CAVITATING CONDITION

Copyright © 2011 by JSME


DISCUSSIONS As is expected, the measured static pressure takes
Hereafter, we focus on the case of NACA0015 hydrofoil at approximately the same value as the saturated vapor pressure
α = 8.0° since for this case the discrepancy between the (i.e. 𝐶𝑃 = −𝜎 ) in the portion on the suction side of the
predicted and measured breakdown characteristics is most hydrofoil where it was lower than the vapor pressure (i.e.
prominent and also detailed measured data are available for 𝐶𝑃 < −𝜎) at the non-cavitating flow conditions. This is also the
comparisons. case for the predicted static pressures. In a short region
Cavitation inception generally takes place when and where downstream of the leading edge of the hydrofoil, the predicted
static pressure decreases to the saturated vapor pressure. This static pressures have not reached to, and are lower than the
inception condition is expressed as: saturated vapor pressure. This means that in this region
cavitation has not yet been developed such that it alters the flow
𝑝𝑚𝑖𝑛 = 𝑝𝑣 , field pressure. Since experimental data are not available in this
or, region, one does not know if this predicted trend of delay in
𝐶𝑃 𝑚𝑖𝑛 = −𝜎 fact occurs in the experiments.
where Except in the above-mentioned short region downstream of
𝑝𝑚𝑖𝑛 − 𝑝∞ the leading edge, the predicted static pressure is similar to the
𝐶𝑃 𝑚𝑖𝑛 ≡
1 measured one until 𝑥 ⁄𝐶 reaches to the point where it was
𝜌 𝑢2
2 𝑙 ∞ equal to the vapor pressure at the non-cavitating flow condition.
For further reference in this paper, we call this point as
As is seen from the lower figure of Fig. 5, the minimum “matching point”. From this matching point, however, the
value for the static pressure coefficient is about -3.0, which predicted static pressure starts to deviate from the measured one
means the cavitation inception is likely to take place at σ = 3.0 and tends to obey the distribution at the non-cavitating flow
on the suction surface of the hydrofoil near the leading edge condition. On the other hand, the measured static pressure
(i.e. near suction peak point) for this angle of attack. In region gradually increases beyond this matching point, but, it is still
cavitation is generated, the surface pressure on the hydrofoil lower than it was at the non-cavitating flow condition. The
tends to remain as the vapor pressure, which acts to reduce the loading of the hydrofoil, therefore, decreases upstream of this
loading of the hydrofoil. However, the measured lift coefficient matching point, but it increases downstream of the matching
is kept almost constant until around σ = 1.2 when the point. This is why the measured lift coefficient, 𝐶𝐿 , remains
cavitation number, σ, is decreased, as is seen in the bottom almost the same value as it was at the non-cavitation condition
figure of Fig. 3. Figure 7 plots the measured variation of the until the ambient pressure becomes close to the vapor pressure
static pressure distributions around the hydrofoil with respect to where sudden drop in 𝐶𝐿 takes place.
cavitation number, σ , together with predicted variations To achieve an accurate prediction of the breakdown
computed by FrontFlow/blue[2]-[4] (see also TABLE 1). The characteristics of a hydrofoil, the above-mentioned distribution
cavitation model implemented in this flow solver is of source of the static pressure beyond the matching point has to be
and sink type based on a simplified Rayleigh-Plesset equation. reproduced. In order for the static pressure to decrease and for
FrontFlow/blue solves all the active eddies in the turbulent the cavity to be elongated, the streamline has to have a strong
boundary layers, and thus its solution is least affected by the curvature at the trailing edge of the cavity. In the present
turbulence model, giving most reliable prediction for the flow computations this does not occurs, and as a result, the cavity
fields themself. length is considerably under-predicted as shown in Fig. 8.

FIGURE 7. COMPARISONS OF STATIC PRESSURE COEFFICIENT FIGURE 8. COMPARISON OF CAVITY LENGH


(Predicted by FFB for NACA0015 at 𝛂 = 𝟖. 𝟎°) (Predicted by FFB for NACA0015 at 𝛂 = 𝟖. 𝟎°)

Copyright © 2011 by JSME


computtations, Dr. Iga investigated effects of grid resolutions
on non-cavitating flows, and found that the non-cavitating flow
fields do not change by grid refinement beyond 200 × 120
grid. However, this is not the case for cavitating flow
computations. With grid refinements, the computed cavity
shape becomes thinner and longer, which gives a better
agreement with the cavitation visualized by the measurements.
One also notes that the finer the computational grid becomes,
the spatial gradient of the void fraction distribution becomes
FIGURE 9. COMPUTED DISTRIBUTION OF VOID FRACTION steeper at the “interface” region between the liquid and vapor
(Predicted by FFB for NACA0015 at 𝛂 = 𝟖. 𝟎°, 𝛔 = 𝟏. 𝟎) phases although cavitation models based on homogenous media
of one fluid do not really acknowledge the interface between
the phases.

(a) computed by 200 × 120 grid

FIGURE 10. EXPERIMENTALLY VISUALIZED CAVITATION


(NACA0015 at 𝛂 = 𝟖. 𝟎°, 𝛔 = 𝟏. 𝟎)

Figure 9 and 10 respectively show predicted void fraction


and cavitation visualized by a high-speed camera. Apparently,
the predicted cavity region is considerably shortened and
thickened as compared to the measurements, which implies that
some of the important aspects of cavitating flows cannot be (b) computed by 800 × 480 grid
reproduced by the present cavitation model.
Before trying to improve the present cavitation model, or
developing a new cavitation model, the work group on the
benchmark for the basic cavitating flows is investigating the
“true limit” of the present cavitation models since the
benchmark results obtained so far may have been affected by
other aspects of computations such as turbulence models,
numerical methods, and computational grid. Among the efforts
of this kind, Dr. Iga of Tohoku University has investigated the
effects of the grid resolution on the predicted cavitating flows
for the same benchmark case (NACA0015, α = 8°, σ = 1.4).
She solved the preconditioned compressible Navier-Stokes
(c) computed by 1,600 × 960 grid
equations in a two-dimensional domain with Baldwin-Lomax
zero-equation turbulence model. Cavitation is predicted by a
barotropic type of model detailed in [6]-[8]. FIGURE 11. INSTANTANEOUS DISTRIBUTIONS OF VOID
Figure 11 shows instantaneous distributions of void FRACTION AND VELOCITY VECTORS OF MIXURE FLUID
fraction and flow velocity of the mixture fluid around the (Predicted by in-house code developed in Tohoku University for
leading edge of the hydrofoil. Prior to cavitating flow NACA0015 at 𝛂 = 𝟖. 𝟎°, 𝛔 = 𝟏. 𝟒)

Copyright © 2011 by JSME


The effects of grid refinement are also confirmed by this benchmark have been carefully evaluated for breakdown
comparisons of static pressure distributions and lift and drag characteristics of the lift and drag forces, time-averaged
coefficients, respectively shown in Fig 12 and Tab. 2. With the distributions of the static pressure on the hydrofoil, and cavity
grid refinement, both the static pressure distribution and lift and length.
drag coefficients become closer to the measured equivalents. The benchmarking has been made for a light and a
However, with the finest computational grid, the predicted moderate loading condition of these hydrofoils at a chord-based
characteristics still has some essential difference to the Reynolds numbers in the order of 106. Four commercial flow
measured one, which we believe is presumably attributed to the solvers, ANSYS CFX, ANSYS Fluent, STAR-CCM+, and
intrinsic limitations associated with the present cavitation ScrewTetra, along with four open-source or in-house flow
model. solvers participated in this benchmark. All the cavitation
models, except one, implemented in those flow solvers and
tested in this benchmark are based on an assumption of
homogenous media of one fluid, for which inception, growth,
decay and destruction of cavitation are expressed by density
change of the mixture fluid composed of liquid and gas phases.
These models do not have a concept of “interface” between the
liquid and gas phases.
Some groups performed Large-eddy simulation (LES) for
solving the fluid flows, but most of the groups solved the two
dimensional compressible or incompressible Navier-Stokes
equations with a Reynolds-Averaged Navier-Stokes (RANS)
model.
The cavitation models tested in this benchmark differ with
each other mainly in how they determine density of the mixture
fluid and can be categorized into two groups. Barotropic
models determine the density of the mixture fluid by an
equation of state for the mixture fluid while bubble dynamic
models determine the void fraction by looking at the difference
FIGURE 12. COMPARISON OF STATIC PRESSURE COEFFICIENT of the fluid pressure and the saturated vapor pressure. The latter
(Predicted by in-house code developed in Tohoku University for NACA0015 types of models are based either on simplified (reduced)
at 𝛂 = 𝟖. 𝟎°, 𝛔 = 𝟏. 𝟒) Rayleigh-Plesset equation or Henry’s law.
Despites all these differences in the cavitation models
themselves and differences in the flow solvers and
CL CD computational grids, the results of the benchmark show a
Measured 0.859 0.0596
universal trend of difference between the predicted and
measured breakdown characteristics. Namely, the predicted lift
200 × 120 0.650 0.0638 coefficients gradually decrease from the one corresponding to
the non-cavitating flow condition while the measured lift
400 × 240 0.722 0.0797
coefficient suddenly drops near the breakdown point. This trend
800 × 480 0.714 0.0857 of discrepancy between the computations and measurements is
most prominent at the higher loading condition for the
1,600 × 960 0.736 0.0876
NACA0015 hydrofoil. But, it is also consistently confirmed for
the other cases investigated in this benchmark. It seems that the
difference is most likely to be attributed to under prediction of
TABLE 2. COMPARISON OF LIFT AND DRAG COEFFICIENTS the cavity length, which is also seen for all the cavitation
(Predicted by in-house code developed in Tohoku University for NACA0015 models tested in this benchmark. A grid resolution study shows
at 𝛂 = 𝟖. 𝟎°, 𝛔 = 𝟏. 𝟒)
that cavitaing flows are much more strongly affected by the
grid resolutions, and with grid refinement, the shape of the
predicted cavity becomes thinner and longer, which gives a
CONCLUSIONS better agreement with the cavitaion visualized by the
Through an industry-university collaborative project, measurements. But essential difference still remains between
extensive validation studies have been made for the prediction the predicted and measured characteristics of cavitation.
of basic cavitating flows around different-loading-type These discrepancies are presumably attributed to some
hydrofoils: Clark-Y 11.7% and NACA0015 in order to clarify intrinsic limitations associated with the present cavitation
the ability of present cavitation models to predict cavitating models, which are based on an assumption of homogeneous
flows. Prediction capabilities of each of the models tested in media of one fluid.

Copyright © 2011 by JSME


ACKNOWLEDGMENTS AND LIST OF PARTICIPANTS [6] Iga, Y., et al., 2003, “Numerical Study of Sheet Cavitation
Those engineers and researchers who are participating in Break-off Phenomenon on a Cascade Hydrofoil”, Journal
this project are listed below, to whom the author would like to of Fluids Engineering, Trans. ASME, Vol. 125, No. 4, pp.
convey his utmost thanks for their dedicated and indispensable 643-651.
cooperation for promotion of this project. The author also likes [7] Iga, Y., et al., 2004, “Numerical Analysis of Cavitation
to thank Turbo machinery Society of Japan, under which this Instabilities Arising in The Three-Blade Cascade”,
project has been organized, for its indispensable support for this Journal of Fluids Engineering, Trans. ASME, Vol. 126,
project. No. 3, pp. 419-429.
[8] Iga, Y., Yoshida, Y., and Ikohagi, T., 2009, “Numerical
Universities Analysis of Occurrence Mechanism of Rotating Cavitation
Dr. Y. Iga of Tohoku University, Profs. T. Kajishima and H. in Cascade”, 6th International Symposium on Pumping
Horiguchi of Osaka University, Prof. S. Watanabe of Kyushu Machinery, Colorado, USA, FEDSM2009.
University, Dr. S. Tsuda of Shinshu University, Dr. N. Tani of [9] An, B-J., Kajishima, T., and Okabayashi, K., 2009,
Japan Aerospace Exploration Agency (JAXA), and former Prof. “Generality in the unsteady cavitating flows in two-
T. Kawamura of The University of Tokyo dimensional cascades”, 7th International Symposium on
Cavitation (CAV2009), Michigan, IL, USA, August, 2009,
Software Venders Paper No. 90.
Mrs. K. Tsuboi and T. Kuwayama of ANSYS Japan, K. K., [10] Okita, K., et al., 2003, “Numerical Analysis for Unsteady
Mrs. M. Ishikawa and J. Mukai of CD-adapco JAPAN Co., Cavitating Flow in a Pump Inducer”, 5th International
Ltd., Mrs. M. Kuba and D. Hitomi of Software Cradle Co., Symposium on Cavitation (CAV2003), November 1-4,
Ltd., Mr. Y. Fujikawa and Dr. Y. Takekoshi of VINAS Co. Ltd., 2003, Osaka, Japan.
and Mr. Yoshinobu Yamade of Mizuho Information & Research
Institute, Inc.

Turbomahinery Manufactures
Drs. A. Goto and M. Nohmi of Ebara Corporation, Mr. R.
Hashi of Shin Nippon Machinery Co.,Ltd., Mrs. T. Fujii, M.
Hayakawa and H. Takuno of Teral Inc., Dr. S. Tomimatsu of
DMW Corporation, Mrs. M. Miyabe and T. Miura of Torishima
Pump Mfg. Co., Ltd., Dr. T. Nagahara of Hitachi Plant
Technologies, Ltd., Dr. M. Fukaya of Hitachi, Ltd., Mr. K.
Miyagawa of Mitsubishi Heavy Industries, Ltd., Dr. H. Tomaru
and Dr. K. Yada of IHI Corporation

REFERENCES
[1] Nohmi, M., et al., 2003, “Experimental and Numerical
Study of Cavitation Breakdown in a Centrifugal Pump”,
ASME-JSME Joint Fluids Engineering Summer
Conference, July 6-10, 2003, Honolulu, FEDSM2003-
45409.
[2] Kato, C., Kaiho, M., and Manabe, A., 2003, “An Overset
Finite-Element Large-Eddy Simulation Method With
Applications to Turbomachinery and Aeroacoustics”,
Journal of Applied Mechanics, Trans. ASME, Vol. 70, pp.
32-43.
[3] Kato, C., et al., 2007, “Numerical Prediction of Sound
Generated from Flows with a Low Mach Number”,
Computers & Fluids, Vol. 36, No. 1, pp. 53-68.
[4] Yamanishi, N., et al., 2007, “LES Simulation of Backflow
Vortex Structure at the Inlet of an Inducer”, Journal of
Fluids Engineering, Trans. ASME, Vol. 129, pp. 587-594.
[5] Kawamura, T., and Sakoda, M., 2003, “Comparision of
Bubble and Sheet Cavitation Models for Simulation of
Cavitating Flow over a Hydrofoil”, 5th International
Symposisum on Cavitaion (CAV2003), November 1-4,
2003, Osaka, Japan, CAV03-OS-1-008.

Copyright © 2011 by JSME

You might also like