Proceedings of FEDSM2005

2005 ASME Fluids Engineering Division Summer Meeting and Exhibition

June 19-23, 2005, Houston, TX, USA

FEDSM2005-77240
FEDSM2005-77240

PREDICTION OF SHEET CAVITATION IN A CENTRIFUGAL PUMP IMPELLER WITH THE

THREE-DIMENSIONAL POTENTIAL-FLOW MODEL
R.J.H. Dijkers B. Fumex
Flowserve Corporation, Pump Division Department of Mechanical Engineering
Hengelo, The Netherlands University of Twente, Enschede, The Netherlands
[email protected] [email protected]

J.G.H. Op de Woerd N.P. Kruyt H.W.M. Hoeijmakers

Flowserve Corporation Department of Mechanical Department of Mechanical
Pump Division Engineering, University of Twente Engineering, University of Twente
Hengelo, The Netherlands Enschede, The Netherlands Enschede, The Netherlands
[email protected] [email protected] [email protected]

ABSTRACT and/or pump (impeller) life is reduced. A good economic pump

The occurrence of cavitation is one of the main limiting has optimum size and speed combined and should be running
factors in the operation and design of centrifugal pumps. In this very close to the physical limitations imposed by cavitation.
paper a model for the prediction of sheet cavitation is Cavitation occurs in regions in the pump where the
described. This model has been implemented in a three- static pressure is below the vapor pressure of the fluid due to
dimensional finite-element package, employing the potential- high (local) fluid velocities. The highest velocities generally
flow approximation of the governing flow equations. At the occur at the impeller inlet, mostly due to incidence of the flow
interface between vapor and liquid, pressure equilibrium is relative to the leading edge of the impeller blades. These inlet
required. The closure region of the cavity is modeled as the velocities can be limited by decreasing the rotational speed of
collapse of a bubble, whose motion is described by the the pump and/or increasing the dimensions of the impeller.
Rayleigh-Plesset equation. The effect of displacement of the Furthermore, the occurrence of cavitation can be steered by
flow due to presence of the sheet cavity is incorporated by the control over the incidence angle through proper design of the
transpiration technique. This is a linearised approach which is shape of the leading edge of the impeller blade.
well-known from techniques for coupling inviscid-flow The occurrence of cavitation in a centrifugal pump during
methods to boundary-layer methods. The model gives the operation is admitted when the performance of the pump is not
location of the sheet cavity (if present); its length is thus also significantly influenced and/or the lifetime of the pump is
predicted. sufficiently guaranteed. The lifetime decrease due to cavitation
The model has been validated by comparing sheet erosion has been studied by Gülich (1999). It was found that
cavitation at the blades of a centrifugal pump impeller, obtained cavitation erosion is mainly dependent on the length of the
from CFD-computations and from visual observations in a cavitation sheet and on the circumferential speed, besides
model test. properties of the impeller material. The erosion rate is
proportional to the length of the cavitation sheet l2.6-2.83 and the
circumferential speed U6. From this point of view it seems very
Keywords: Centrifugal Pumps, Cavitation, NPSH, Potential logical to design pump impellers with respect to the length of
Flow, CFD Analysis the cavitation sheets that occur in the field.
From the authors’ experience with analysis and design
INTRODUCTION purposes of centrifugal pumps, the use of a three-dimensional
In the design process of a centrifugal pump the occurrence CFD-method that is based on the potential-flow approximation
of cavitation is one of the most important aspects that need to (augmented by loss models) is very efficient and effective
be considered. Cavitation is a major contributing factor to (Dijkers et al., 2000; van Esch et al., 1997). Using COMPASS,
failure and/or inefficiency in these rotating fluid-handling a three-dimensional finite-element package for solving the
machines. Often when cavitation occurs, performance is lost potential-flow field, it has already been possible to predict

1 Copyright © 2005 by ASME

cavitation-inception curves of centrifugal pump impellers. The along the wall. The Rayleigh-Plesset part of the interface is
use of this potential-flow model, extended with a model that started at the location s0 where the sheet thickness is maximal.

conditions R(s0)=αmax and dR/ds(s0)=0.

predicts the occurrence of sheet cavitation, makes it possible to The Rayleigh-Plesset equation is solved numerically with initial
obtain impeller designs that are both economically attractive
and technologically reliable.

NOMENCLATURE
CP Pressure Coefficient (p-pv)/(0.5ρU2)
D1 Impeller inlet diameter
D2 Impeller outlet diameter
g Gravitational acceleration

Specific speed ΩQ0.5/(gH)0.75

H Head
nω
NPSH Net Positive Suction Head
p Static pressure Figure 1 Cavitation sheet with closure point condition.
pv Vapour pressure

sheet α (see figure 1), are unknown, we have a free-boundary

Q Flow rate Since the location of the interface and the thickness of the
R Bubble radius
s Length coordinate problem. At the interface between liquid and vapour two
t Time conditions hold, the kinematic condition and the dynamic
U Circumferential speed condition. In the developed numerical method, the kinematic
v Absolute velocity vector condition is applied in determining the flow field in the channel
W Relative velocity formed by two blades. The dynamic condition is used to update
z Elevation the location and thickness of the cavitation sheet.
Greek symbols The effect of displacement of the flow due to the presence
α Cavitation sheet thickness of the sheet cavity is incorporated by the transpiration
φ Flow coefficient: Q/(ΩD23)
ϕ
approach, instead of actual adjustment of the blade geometry
Velocity potential and remeshing after each iteration. The transpiration approach
κ Cavitation coefficient gNPSH/(ΩD1)2 is a linearized approach which is well-known from techniques
υ
The complete model gives the location and thickness α of
for coupling inviscid-flow methods to boundary-layer methods.
ρ
Kinematic viscosity

ψ
Fluid density
the sheet cavity (if present); its length is thus also predicted.
Head coefficient: gH/(ΩD2)2
Ω
The sheet cavity model has been implemented in a
Angular velocity of impeller
numerical method based on the potential-flow approximation of
SHEET CAVITATION MODEL the governing equations. The potential-flow approximation

incompressible, ( ∇ ⋅ v = 0 ), and irrotational ( ∇ × v = 0 ),

For inlet stagnation pressures below the cavitation inception means (mathematically) that the core flow is considered as
value, a cavitation sheet will form. A model to determine the
location and shape (length and thickness in particular) is where v is the fluid velocity. This latter assumption is justified
described here. A full description is given by Fumex (2003). as long as the importance of the viscous forces compared to the
In the present model the pressure inside the sheet equals the non-viscous (inertia) forces acting upon the fluid is small (see
vapour pressure, so at the interface between the sheet and the Batchelor, 1967). An additional assumption is that the fluid
liquid the pressure is also equal to the vapour pressure. This is enters the machine free from vorticity. For pumping machinery
the so-called dynamic condition. Furthermore, the interface is a these assumptions usually hold.

can then be described by a velocity potential ϕ obeying the

stream surface, hence the normal velocity is zero. This is the The (core) flow field (outside of boundary layers and wakes)
kinematic condition.
The rear part of the cavitation sheet is described separately,

v = ∇ϕ
since it resembles the collapse of a bubble cloud, see also figure relation
1. This motion is described (in simplified form) by the
Rayleigh-Plesset equation for the bubble’s radius R (Batchelor, The equation of continuity then reduces to the Laplace

∇ 2ϕ = 0
1967) that governs the motion of a single bubble in an infinite equation, i.e.

pv − p ∞
medium
&& + R =
ρ
3 &2 The pressure distribution is determined from the (unsteady)
RR
∂ϕ 1
2 Bernoulli equation,

+ 2 v ⋅ v + + g z = c(t )
∂t ρ
Here R& denotes the time-derivative of R, pv is the vapour p
pressure, p∞ is the inlet stagnation pressure and ρ is the fluid
density. The time derivatives are transformed to spatial where t is time, p is thermodynamic pressure, z is elevation,
derivatives by dR/dt = Ws dR/ds, where s is the coordinate
along the wall and Ws is the component of the relative velocity and c( t ) is a time-dependent constant.

2 Copyright © 2005 by ASME

 SHEET CAVITATION RESULTS
Since pumps most often operate at duties with a certain
level of cavitation, the occurrence of sheet cavitation for both

different flow coefficients and cavitation coefficients κ, the

impeller designs has been further investigated. For a number of

method described before. For a flow coefficient of φ = 0.05, at

occurrence of sheet cavitation has been predicted using the

72% of BCP, figures 5 and 6 show the cavitation volumes and

maximum lengths and thicknesses. Although the value for
inception cavitation coefficient is twice as high for the
symmetrical vane thickness profile as for the asymmetric vane
thickness profile, it can be seen that with a decrease in inlet
pressure especially the length of the sheet cavitation grows
Computed Asymmetric Profile Computed Symmetric Profile
Figure 2 Mixed-flow impeller. Test Asymmetric Profile Test Symmetric Profile
0.3

DESIGN VARIATIONS
0.25
Predictions and measurements of sheet cavitation will be
presented of a mixed-flow impeller with a specific speed of 0.2
nω=1.6, as shown in figure 2. This impeller has been analyzed

κ (-)
before in the works of van Os et al. (1997) and van Esch et al. 0.15 BCP

(1997). In these studies the capabilities of predicting the head, 0.1

power and NPSH inception curves by means of three-
dimensional potential-flow computations were demonstrated. 0.05
The performance of the impeller with respect to cavitation
0
inception has been optimized, using a profiled blade thickness
φ (-)
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
distribution with an asymmetric elliptical nose (instead of a
symmetric elliptical nose) followed by a uniform blade Figure 4 Influence of vane profile on cavitation inception.
thickness, as shown in figure 3. The performance with respect
to cavitation inception of the two designs, with the same more rapidly for the asymmetric vane profile. For cavitation
impeller hydraulic layout but different vane thickness profiles, coefficients just below the inception cavitation coefficient, the
has been determined using the computed three-dimensional cavitation sheet is longer for the asymmetric vane profile. Both
potential-flow field. In figure 4 the computed values of the designs show the same tendency: an increase of sheet cavitation
cavitation coefficients κ with respect to cavitation inception are
plotted versus the flow coefficient φ for both designs. With the
with a decrease in cavitation coefficient. Although the sheet
cavitation is smaller in height and length for the symmetric
asymmetric nose profile the cavitation inception values of the
for cavitation coefficients smaller than κ = 0.024, while the
vane profile, the computations of the cavitation sheet diverge
design have been reduced significantly at lower flow rates,
while the best cavitation point (BCP) remained approximately
cavitation coefficients smaller than κ = 0.016. For both vane
computations for the asymmetric vane profile diverge for
constant. At flow rates higher than BCP, the inception
cavitation coefficient slightly increased.
profiles, the non-dimensional pressure distributions along the

for a cavitation coefficient of κ = 0.031. Comparison of these

blade (at mid span between hub and shroud) are shown figure 7

pressure distributions shows that the impeller with the

symmetric vane profile shows a deeper drop in pressure, but
also a steeper rise to a value higher than the vapor pressure. The
presence of the cavitation sheet results in a less steep drop in
pressure in the sheet cavitation zone at the suction side of the
impeller vanes, compared to the computations without the sheet
cavitation model. The corresponding cavitation sheets are
shown in the figures 8 and 9. It is clear that in case of the
asymmetric vane profile, the cavitation sheet is larger than with
the symmetric vane profile. Due to the added thickness at the
leading edge, the sheet cavitation is moved slightly away from
the leading edge. It is also seen that the small cavitation sheet
on the pressure side of the impeller with the symmetric vane
profile is not present on the other impeller.

Figure 3 Descriptions of the vane profile thickness.

3 Copyright © 2005 by ASME

 SS Symmetric PS Symmetric SS Asymmetric PS Asymmetric
25

20
Volume (cm )
3

15

10

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 Figure 9 Sheet cavitation at the suction (left) and pressure
κ (−)
for φ = 0.05, κ = 0.031.
(right) side of the impeller vane with symmetric profiled vanes

Figure 5 Volume of cavitation sheets for φ = 0.05.

Length Symmetric Length Asymmetric For flow coefficients at the best cavitation point (BCP) and
Height Symmetric Height Asymmetric higher, cavitation inception occurs at the pressure side of the
0.1 0.01
vanes. The position and shape of the sheet cavitation have
shifted as well with respect to the suction side cavitation, as the
0.08 0.005
pressure side cavitation sheet is positioned near the shroud for
Thickness (m)

For a flow coefficient of φ = 0.074, at 8% over BCP, the growth

higher flow rates.
Length (m)

0.06 0

function of the cavitation coefficient κ is shown in figures 10

0.04 -0.005 of the lengths, heights and volumes of the sheet cavitation as

0.02 -0.01 and 11. With the asymmetric vane profile, cavitation starts at a
higher inlet pressure and the sheet profile remains larger than

At a cavitation coefficient of κ = 0.030 a sheet cavitation

0 -0.015 with the symmetric vane profile.
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
κ (−)
occurs at the suction side of the vanes with the asymmetric

κ = 0.028 for both designs a second cavitation sheet at the

vane profile just after the swelling at the leading edge, while at
φ = 0.05.
Figure 6 Length and thickness of cavitation sheet for

cavitation sheets at suction and pressure side for κ = 0.026 are

0.8 0.8 suction side of the vanes starts in the middle at the shroud. The
0.6 0.6

0.4 0.4 shown in figures 13 and 14. The corresponding pressures along
the shroud position of the blade are shown in figure 12.
C P (-)

C P (-)

0.2 0.2

0 0

-0.2 -0.2
PS Symmetric PS Asymmetric SS Symmetric PS Asymmetric
-0.4 -0.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 25
Position (-) Position (-)

Figure 7 Pressure coefficient along the mid span of the vane 20

sheet cavitation at φ = 0.05, κ = 0.031 for symmetric (left) and

with (dotted line) and without (straight line) the occurrence of
Volume (cm )
3

15

asymmetric (right) profiled vane.

10

0
0 0.01 0.02 0.03 0.04 0.05 0.06
κ (−)

Figure 10 Volume of cavitation sheets for φ = 0.074.

Figure 8 Sheet cavitation at the suction (left) and pressure

for φ = 0.05, κ = 0.031.

(right) side of the impeller vane with asymmetric profiled vanes

4 Copyright © 2005 by ASME

 Length Symmetric Length Asymmetric Height Symmetric Height Asymmetric
0.1 0.006

0.08 0.004

Thickness (m)
Length (m)

0.06 0.002

0.04 0

0.02 -0.002

0 -0.004
0 0.01 0.02 0.03 0.04 0.05 0.06
κ (−)

at φ = 0.074.
Figure 11 Length and thickness of pressure side cavitation sheet Figure 15 Lay-out of the test loop.

0.6 0.6
COMPARISON WITH EXPERIMENT
0.5 0.5
In order to validate the sheet cavitation model,
0.4 0.4
computations executed on the impeller with asymmetric vane
0.3 0.3
C P (-)

C P (-)

0.2 0.2
profile at the leading edge have been compared with visual
0.1 0.1
observations obtained from a model test. In order to keep the
0 0
deviations between the geometry of the actual impeller and the
-0.1 -0.1 CFD geometry to a minimum, the impeller of the model pump
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Position (-) Position (-)
has been fabricated by means of numerical-controlled milling.
Figure 12 Pressure coefficient along the shroud of the vane The actual test has been executed in a test loop, specially built
for executing scaled model tests, see figure 15, at the Flowserve
sheet cavitation at φ = 0.074, κ = 0.026, for symmetric (left)
with (dotted line) and without (straight line) the occurrence of
site in Hengelo. The test loop is a closed system filled with
water, connected to various auxiliary systems. The model pump
and asymmetric (right) profiled vane.
in the test loop is driven by a 300 kW speed-controlled electric
motor and has special arrangements to control the inlet pressure
and flow rate. Transparent Plexiglas windows are used at the
suction pipe and volute for visual study of cavitation and flow
phenomena. A cross-sectional view of the model pump is
shown in figure 16.

range from φ = 0.035 to 0.08. Cavitation inception

Tests have been executed for flow coefficients varying in the

measurements have been performed both visually by means of

stroboscopic illumination and acoustically by means of
hydrophones. For the impeller with the symmetric vane profile
the inception NPSH values have been determined from
averaging measurement values of different tests using precision
Figure 13 Sheet cavitation at the suction (left) and pressure cast impellers. For the design with the asymmetric vane profile,
a single measurement has been used. Figure 4 shows computed
for φ = 0.074, κ = 0.026.
(right) side of the impeller vane with asymmetric profiled vanes
and measured cavitation inception values for the mixed-flow
impeller with both symmetric and asymmetric vane thickness.

Figure 14 Sheet cavitation at the suction (left) and pressure

for φ = 0.074, κ = 0.026.

(right) side of the impeller vane with symmetric profiled vanes
Figure 16 Cross section of model test pump.

5 Copyright © 2005 by ASME

 For both impeller designs the agreement between
computations and measurements is similar. The position of the CONCLUDING REMARKS
BCP is fairly well predicted, as are the cavitation inception For two designs of a mixed-flow pump impeller three-
values around the BCP. For flow rates further away from BCP dimensional potential-flow analysis has been used for
the computed values become more unreliable: they could be predicting both cavitation inception and the occurrence of sheet
both too low and too high. cavitation. It is shown that the NPSH inception values obtained
Tests results have also been used to compare the computed with computations show fairly good agreement with

flow coefficient of φ = 0.035, 50% of BCP, and an cavitation

cavitation sheets with those in the actual pump impeller. At a measurements.

coefficient of κ = 0.063, figure 17 shows a computed cavitation

The numerical method used to take the effect of
displacement of the flow into account due to the presence of

respect to size and position. For flow coefficients of φ = 0.0517

sheet in accordance with the measured cavitation sheet with sheet cavity shows good qualitative agreement with respect to
size and location of the sheet when compared to measurements.
(73 % of BCP) and φ = 0.0655 (95% of BCP) similar results are The effect of the use of an impeller with the exact
shown in figures 18 and 19. hydraulic layout, but an asymmetric vane thickness distribution
at the leading edge, gives lower values of the NPSH-inception
values at flow rates lower than BCP. However, with the
cavitation sheet analysis it is shown that for operating
conditions where cavitation occurs, the asymmetric design in
most cases shows larger cavitation sheets than the design with
the symmetric vane profile. This applies for flow conditions
both above and below BCP. This implies that the design with
the asymmetric vane profile is more sensitive to cavitation
erosion than the impeller with the symmetric vane profile.

ACKNOWLEDGMENTS
The authors wish to thank the management of Flowserve
for their support and permission to publish this paper.
The research into the development of the numerical
method for computing sheet cavitation has been supported by
φ = 0.035, κ = 0.063.
Figure 17 Comparison of the sheet cavitation at
NOVEM.

REFERENCES
Batchelor, G.K., 1967, An introduction to fluid dynamics,
Cambridge University Press, Cambridge, U.K.
Dijkers, R. J. H., Visser, F. C., and Op de Woerd, J. G. H.,
2000, Redesign of high-energy centrifugal pump first stage
impeller, I. A. H. R., Charlotte, North Carolina, U.S.A.
Fumex, B., 2003, Modeling and Numerical Simulations of
Sheet Cavitation in Centrifugal Pumps, PhD thesis, Enschede,
The Netherlands.
Gülich J.F., 1999, Kreiselpumpen: ein Handbuch für
Entwicklung, Anlageplanung und Betrieb, Springer-Verlag,
Berlin, Germany.
φ = 0.0517, κ = 0.0264.
Figure 18 Comparison of the sheet cavitation at
Van Esch, B. P. M., Kruyt, N. P. and Jonker, J. B., 1997,
An inviscid-viscous coupling method for computing unsteady
flows entire pump configurations. FEDSM97-3373. In
Proceedings of the 1997 ASME Fluids Engineering Division
Summer Meeting, Vancouver, Canada.
Van Os, M.J., Op de Woerd, J.G.H. and Jonker, J.B., 1997,
A parametric study of the cavitation inception behaviour of a
mixed-flow pump impeller using a three-dimensional potential-
flow model. FEDSM97-3374. In Proceedings of the 1997
ASME Fluids Engineering Division Summer Meeting,
Vancouver, Canada.

φ = 0.0655, κ = 0.025.
Figure 19 Comparison of the sheet cavitation at

6 Copyright © 2005 by ASME