The 1997 ASME Fluids Engineering Division Summer Meeting FEDSM97 June 2226, 1997

FEDSM97-3374
A PARAMETRIC STUDY OF THE CAVITATION INCEPTION BEHAVIOR OF A MIXED-FLOW PUMP IMPELLER USING A THREE-DIMENSIONAL POTENTIAL FLOW MODEL
M. J. van Os Department of Mechanical Engineering University of Twente P.O. Box 217 7500 AE Enschede The Netherlands J. G.H. Op de woerd BW/IP International B.V. Stork Engineered Pumps P.O. Box 55 7550 AB Hengelo(O) The Netherlands

J.B. Jonker Department of Mechanical Engineering University of Twente P.O. Box 217 7500 AE Enschede The Netherlands

ABSTRACT The cavitation inception behavior of mixed-ow impellers has been investigated parametrically using a geometric impeller model and a three-dimensional potential ow model. The computed cavitation inception is validated by cavitation inception measurements. Computed and measured results show a good agreement for a wide range of operating conditions. A number of mixed-ow impellers has been developed using the same design strategy in order to investigate the inuence of the number of blades and the leading-edge position (forward/backward swept) on cavitation inception. It is found that a larger number of impeller blades results in a atter cavitation inception curve for ow rates below the best cavitation point. Furthermore, the cavitation inception performance of the forward-swept impellers is superior to the backward-swept impeller.

n! r z x s g P0 pv v f

dimensionless specic speed,

radial coordinate [m] axial coordinate [m] meridional coordinate [m] dimensionless meridional coordinate [,] acceleration of gravity [m=s2 ] total inlet pressure [N=m2 ] vapor pressure of the uid [N=m2 ] absolute velocity [m=s] factor [,]

4 gH  3

pQ

[,]

Greek symbols angular velocity of impeller [rad=s] local blade angle [rad] angle [rad] angular coordinate [rad] density of the uid [kg=m3 ]

NOMENCLATURE NPSH net positive suction head [m] X meridional coordinate [m] Q ow rate [m3 =s] H head [m] C constant [,] 1

Subscripts i construction line index l leading-edge t trailing-edge m meridional opt best efciency point bcp best cavitation point Copyright c 1997 by ASME

INTRODUCTION In order to minimize size, weight and cost centrifugal and mixed-ow pump impellers are required to operate at the highest possible rotational speed. The rotational speed is limited by the phenomenon of cavitation. Cavitation occurs around an impeller blade when the local static pressure falls below the vapor pressure of the liquid being pumped. The parameter which characterizes the cavitation performance is the Net Positive Suction Head (NP SH ), dened as

where P0 is the total inlet pressure at the impeller inlet corresponding to cavitation, pv the vapor pressure, g the acceleration of gravity and the density of the uid. When the total pressure P0 is decreased from a cavitation-free ow condition, an inlet pressure will be reached where cavitation rst occurs. This corresponds to inception NPSH . A further reduction of the inlet pressure can cause a reduction of the delivered head. A head reduction of 3% is referred to as NP SH3 . Many authors (Avva et al (1995), de Lange (1996) and Drei et al (1996)) have proposed methods for dealing with the characteristic two-face phenomenon of cavitation. However, cavitation erosion damage is most likely to occur between inception NPSH and NPSH3 (McNulty, Pearsall (1982)). It also appears that for high energy pumps cavitation inception predominantly determines the safe operating conditions (Stoffel (1995)). Thus, prediction of cavitation inception is an important issue of the impeller design process. Previously used impeller design methods for optimization of inception NPSH are based on simple one-dimensional ow calculations along mean streamlines. The analytical aspects of these impeller inlet design methods are described in Pearsall (1973), Bunjes (1974), Noskievi (1976) and Adrizzon, Pavesi (1995). c In these methods empirical coefcients are used to describe the effects of incident angle, blade thickness and non-uniformity of the inlet velocity prole. Adrizzon and Pavesi (1995) have calculated some of these coefcients by applying a quasi threedimensional ow analysis. Cooper et al (1991) introduced a new method, known as biased-wedge blading, for designing impeller blades with a minimal drop in static pressure on the blade surface. This pressure drop leads to the formation of cavitation. This method uses a combination of experimental observations and computational ow analysis. However, these quasi three-dimensional approaches fail for fully three-dimensional ow patterns occurring in the impeller inlet. Therefore a fully three-dimensional potential ow analysis is performed here to predict cavitation inception. Such an approach allows pump designers to predict the cavitation inception behavior of the impeller with more condence. The impeller geometry is described using the geometric impeller model presented by van Os et al (1996). With this geo2

NPSH = P0 , pv g

(1)

metric impeller model the optimization process of the impeller design is greatly simplied. Modications to the geometry can be described in a simple and exible manner and input les for the potential ow analysis package are prepared automatically. This approach of a geometric model in combination with a three-dimensional potential ow analysis package is validated by comparing computed inception NPSH -values with experimental results. In order to investigate the effect of the number of blades and the position of the leading-edge on the cavitation inception performance a number of mixed-ow impellers has been developed using the same design strategy. 2 IMPELLER GEOMETRY DESCRIPTION The geometric impeller model is based on the Kaplan drawing method (Stepanoff (1957)) and cubic-spline functions. The description of the geometry of an impeller blade consists of two parts. The rst part represents the denition of the camber surface, including leading- and trailing-edge, by means of a number of camber lines that are located on different positions between hub and shroud. By means of a cubic-spline surface through these lines the camber surface is constructed. In the second part the pressure and suction surfaces are obtained from the camber surface by means of a thickness distribution perpendicular to this surface. By using different thickness distributions for the pressure and suction side an impeller blade with an asymmetric prole can be modeled. Further details are described in van Os et al. (1996).
y
camber surface

Xm r
rd  dx

z
impeller axis

x
camber line

Figure 1: A camber line on the camber surface of an impeller blade. Figure 1 shows three camber lines on the impeller blade camber surface. The projection of a single camber line on the meridional plane, the plane  = constant, is called the construction line. The position along the construction line is described by the meridional coordinate Xm , see gure 2. Copyright c 1997 by ASME

leading-edge

construction lines

shroud

Xm
hub

trailing-edge

Figure 4: Mixed-ow impeller (n!

= 1:6).

Figure 2: Construction line in the meridional plane. The curvature of the camber line is described by the local blade angle , dened as (Stepanoff (1957)) the assumption of incompressible, irrotational and inviscid ow (Re , Ma2 ) within the impeller channel. This is a reasonable assumption for well-designed impellers in which no ow separation occurs. The governing Laplace equation is solved numerically using the nite element method (FEM). This method is well suited to modeling very complex blade geometries. A detailed description of the numerical method is given in Kruyt et al. (1996). The ow is simulated within one of the impeller blade passages. Figure 5 shows a FEM-mesh of a mixed-ow impeller channel. The domain consists of an inlet region upstream of the impeller inlet, a single blade passage, and an outlet region extending downstream. The mesh contains nodes. A detailed description of the boundary conditions is given by van Esch et al (1995).

Starting from a construction line determined by the r; z coordinates in the meridional plane and a blade angle distribution xm , the angular coordinate  is computed from:

tan  = dxm : rd

(2)

 

xm = Xm  = xm = 0 +

shown in gure 3 and 4. The blades have a constant thickness :  D, with D the impeller diameter, and the blade noses are rounded elliptically. Figure 3 shows the meridional plane with the intersection lines  constant and the plane view of the mixed-ow impeller blade.

rxm  tan xm  : (3) An example geometry of a mixed-ow impeller (n! = 1:6) is
0

Xm

dxm

39432

0 017

inlet region

leading-edge

trailing-edge
X

Figure 3: Left the meridional plane with the intersec0) and right the plane tion lines  constant (  view of the mixed-ow impeller (n! : ).

= 10

=16

outlet region

Figure 4 gives a three-dimensional view of this mixed-ow impeller. For visual reasons the shroud surface is not shown. THREE-DIMENSIONAL COMPUTATION OF CAVITATION INCEPTION The inception NPSH is computed by using a fully threedimensional potential ow model. This model is based on 3 3

Figure 5: FEM-mesh of the blade passage of a mixedow impeller. The three-dimensional pressure eld is computed from the Bernoulli equation in the rotating frame of reference. Observation of iso-pressure surfaces indicates the occurrence and the position of cavitation inception. The NPSH inception is computed from the minimum pressure. Copyright c 1997 by ASME

A complete cavitation inception analysis takes less than one hour of CPU time on a SGI Power Challenge computer, using one processor, or less than three hours on a Pentium MHz PC. The maximum internal memory use is Mb.

Impeller design strategy The design strategy is based on a one-dimensional ow model. Hub, shroud, trailing-edge, blade angles at the trailing-edge and blade prole are identical to those of the mixed-ow impeller shown in gure 4. The inlet blade angles are computed from:

60

90

VALIDATION OF THE COMPUTED CAVITATION INCEPTION Cavitation inception measurements have been performed both visually by means of stroboscopic illumination and acoustically by means of hydrophones. Figure 6 shows computed and measured cavitation inception values for the mixed-ow impeller shown in gure 4.
50.00

tan l;i  = f  vm;i  cos i; ri =

p

(4)

40.00

30.00

20.00

10.00

0.00 0.60 0.80 1.00 1.20 1.40 1.60 1.80

Q/Qopt [-]

2 2 with l the blade angle at the leading-edge, vm vr vz the computed meridional velocity, the angular velocity, r the radius on the leading-edge, the angle between the meridional velocity and the construction line, f a factor, and i the index for the construction line number. The meridional velocity is computed from a meridional ow analysis that takes into account the number and the thickness of the impeller blades. Neumann (1991) proposed a factor f of : , : in order to improve the performance of a pump for ow rates larger than the best efciency point. Here a factor f : is used. The variation of the local blade angles between leading- and trailing-edge has been chosen similarly for all impellers. At d and from leading- to the leading- and trailing-edge dXm trailing-edge the local blade angle is obtained from:

NPSH [m]

1 05 1 25 = 1 05

=0

Figure 6: Computed and measured cavitation inception of the mixed-ow impeller; with , , , , M and N different cavitation inception measurements. Comparison of the computed and measured results shows a good agreement for : Q=Qopt : , where Qopt is the best efciency point. Furthermore, the predicted positions on the blade surface where cavitation inception occurs are in agreement with the visual observations of the cavitation inception measurements.

i s =

l;i , t;i

2 2

, ,

t;i, l;i

2 2

s cos , C  
; 2

sC
 

l;i , t;i

t;i, l;i

09

17

PARAMETRIC CAVITATION INCEPTION ANALYSIS A previous parametric cavitation inception study by van Os et al. (1996) showed that for identical camber surfaces a larger number of blades or an increase of the blade thickness cause a reduction of the best cavitation point (Qbcp). Furthermore, thickening the nose prole at the suction side (asymmetric blade prole) reduces the inception NP SH drastically for ow rates below Qbcp, whereas for higher ow rates the NP SH -value is the same. In the current parametric study a slightly different approach is followed in order to investigate the effect of the number of blades and the leading-edge position. All analyzed impellers are designed according to the same design method. Hence they are designed for the same ow rate and they are better comparable. 4

(5) with t the blade angle at the trailing-edge, s the dimensionless meridional coordinate and C a constant dening the point of maximum gradient of the blade angle variation. For all developed impellers C lies between : , : . A side effect of this design strategy is that the camber surfaces differ slightly at the leading-edge. Due to the unchanged blade angles at the trailing-edge the blade curvature at the exit is more or less similar for all developed impellers. Using this design strategy a number of impellers has been developed from a single reference impeller. The impeller with four blades and the same leading-edge position as the mixed-ow impeller is called the reference impeller. Figure 7 shows the meridional plane and the plane view of this reference impeller. The reference impeller differs slightly from the mixed-ow impeller of gure 4 at the leading-edge in the plane view, since the inlet blade angles differ. The other developed impellers differ from the reference impeller in the number of blades and the position of the leadingedge.

cos 1 + s,C 1,C

 2

; s C

0 35 0 65

Copyright c 1997 by ASME

Z
Y

Figure 7: Meridional plane and plane view of the reference impeller. Effect of the number of blades Figure 8 shows the computed cavitation inception for the developed mixed-ow impellers with three, four, ve and six blades respectively.
50.00

Effect of the position of the leading-edge Three impellers with different leading-edge positions have been designed. Figure 9 (top) shows the meridional plane of these impellers. The position of the leading-edge on the hub has been kept the same, whereas this position on the shroud differs. For impeller A this position lies more upstream, while for impeller C it lies more downstream than the reference impeller referred to as impeller B. Figure 9 (bottom) shows the plane view of the three impellers. Notice that impeller A and B are of the forward-swept type and impeller C of the backward-swept type.

Z
Z Z

40.00
Y

R
Y

R
Y

NPSH [m]

30.00

20.00
X
X

10.00

A
0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.00

Q/Qopt [-] 3 Blades 4 Blades 5 Blades 6 Blades

Figure 9: Meridional plane (top) and plane view (bottom) of the three developed impellers with different leading-edges; impeller A on the left side, impeller B in the center and impeller C on the right side. Figure 10 shows the computed cavitation inception for these three impellers. The forward-swept impellers (A and B) have better cavitation inception performance than the backward-swept impeller (C). For ow rates below Qbcp cavitation inception occurs at half-span of the leading-edge for the forward-swept impellers. For the backward-swept impeller it occurs close to the shroud. This is in agreement with the results obtained by Bunjes (1974) and Brown (1993). They described a positive effect of the forward-swept design on cavitation inception. The forwardswept design reduced the blade loading at the shroud side. The difference between the cavitation inception performance of the two forward-swept impellers is small, with impeller A showing a slightly better cavitation inception performance.

Figure 8: Cavitation inception performance for different number of blades. The three-bladed impeller has the steepest NP SH -curve for ow rates below Qbcp . This is caused by a reduced guidance of the ow around the impeller nose in comparison with the four-, ve- and six-bladed impellers. This result is similar to that presented by van Os et al. (1996) for different number of blades with identical camber surfaces. From gure 8 it can be observed that the six-bladed impeller has the lowest Qbcp, followed by the ve-, four- and three-bladed impellers. Only the ve- and six-bladed impellers have a Qbcp which approaches the design ow rate. The difference between Qbcp and the design ow for the fourand three-bladed impellers is attributed to blockage and a phenomenon called potential-ow vortex, see Visser et al. (1994). The effect of this vortex on the inlet velocity near the leadingedge is reduced when the number of blades is increased. Due to this vortex, the condition of shockless ow is reached at a higher ow rate for impellers with fewer blades. 5

DISCUSSION Three-dimensional potential ow analysis has been used for predicting cavitation inception of mixed-ow pump impellers. Computed and measured cavitation inception for the mixed-ow impeller are in good agreement for a wide range of operating Copyright c 1997 by ASME

50.00

40.00

30.00

20.00

10.00

0.00 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Q/Qopt [-] Impeller A Impeller B Impeller C

Figure 10: Effect of position leading-edge on cavitation inception performance.

conditions. Thus the potential ow solution, which can be obtained very quickly, can be used for a parametric design of impeller inlets. Observation of the cavitation inception behavior of the three-, four-, ve- and six-bladed impellers shows that the NPSH curve is less steep for the four-, ve- and six-bladed impellers compared to the three-bladed one. The lower Qbcp for a larger number of impeller blades is attributed to blockage and the potential-ow vortex. The cavitation inception performance of the forward-swept impellers is superior to the backward-swept impeller. This is due to a reduced blade loading at the shroud for the forward-swept impeller design. Therefore cavitation inception for the forwardswept impellers occurs at half-span of the leading-edge, while it occurs close to the shroud for the backward-swept impeller. The results obtained so far suggest that the cavitation inception performance of the mixed-ow impeller can be improved by using a forward-swept impeller design, a larger number of blades and an asymmetric nose prole. Future efforts will concentrate on further potential ow analysis for various inlet congurations in order to improve cavitation inception performance. Design parameters which will be investigated are impeller inlet diameter, inlet blade angle and the d development of the local blade angle ( dXm 6 ) close to the leading-edge.

=0

ACKNOWLEDGMENT J.A.G. Sierink (University of Twente, The Netherlands) is thanked for performing some computations. 6

REFERENCES Adrizzon G., and Pavesi G., (1995), Theoretical evaluation on the effects of the impeller entrance geometry and the incident angle on cavitation inception in centrifugal pumps, Proc. Instn. Mech. Engrs. Vol 209, pp. 29-38. Avva R.K., Singhal A.K., and Gibson D.H., (1995), An enthalpy based model of cavitation, FED-Vol. 226, pp. 63-70, Cavitation and Gas-Liquid Flow in Fluid Machinery Devices, ASME. Brown N.A., (1993), Impeller vane-edge design for improved cavitation performance, FED-Vol. 154, Pumping Machinery, pp. 21-29, ASME. Bunjes J.H., (1974), Design principles and test data of pump impellers with improved cavitation behavior, VMF/FDO Engineering Consultants. Cooper P., Sloteman D.P., and Graf E., (1991),Design of High-Energy Pump Impellers to Avoid Cavitation Instabilities and Damage, Proceedings of EPRI Symposium on Power Plant Pumps, Tampa, Florida. Drei A., and Sch tte V., (1996), Numerische und experiu mentelle Untersuchung von Kavitationszust nden im einer raa dialen Kreiselpumpe, Proceedings Pumpentagung Karlsruhe 1996, C5-1, Karlsruhe, Germany. Esch B.P.M. van, Kruyt N.P., and Jonker J.B., (1995), An efcient method for computing three-dimensional potential ows in hydraulic turbomachines, Proceedings Ninth International Conference on Finite Elements in Fluids - New Trends and Applications, pp.633-644, Venice, Italy. Kruyt N.P, Esch B.P.M. van, and Jonker J.B., (1996), A tool for the analysis of unsteady potential ows in centrifugal and mixed-ow pumps, Proceedings Pumpentagung Karlsruhe 1996, C8-2, Karlsruhe, Germany. Lange D.F. de, (1996), Observation and modelling of cloud formation behind a sheet cavity, Ph.D. Thesis, Department of Mechanical Engineering, University of Twente, Enschede, The Netherlands, ISBN 90-9009054-1. McNulty P.J., and Pearsall I.S, (1982), Cavitation inception in pumps, Journal of Fluids Engineering, ASME, Vol. 104, pp. 99-104. Neumann B., (1991), The interaction between geometry and performance of a centrifugal pump, Mechanical Engineering Publications Limited, Londen, p. 63, ISBN 0-85298-755-2. Noskievi J., (1976), The suction ability of centrifugal c pumps, International Conference on Pump and Turbine Design and Development, NEL. Os M.J. van, Jonker J.B., and Op de woerd J.G.H., (1996), A geometric model for the parametric design of pump impellers, Proceedings Pumpentagung Karlsruhe 1996, C5-3, Karlsruhe, Germany. Pearsall I.S., (1973), Design of pump impellers for optimum cavitation performance, Proc. Instn. Mech. Engrs, Vol 187, pp. 667-678. Copyright c 1997 by ASME

NPSH [m]

Stepanoff A.J., (1957), Centrifugal and axial ow pumps, Theory, Design and Application, JOHN WILEY & SONS, INC, New York. Stoffel B., (1995), Die Bedeutung der Kavitation f r den u sicheren Betrieb von Kreiselpumpen, Pumpen + Kompressoren, November 1995. Visser F.C., Brouwers J.J.H., and Badie, R., (1994), Theoretical analysis of inertially irrotational and solenoidal ow in two-dimensional radial-ow pump and turbine impellers with equiangular blades, Journal of Fluid Mechanics, Vol. 296, pp. 107-141.

Copyright c 1997 by ASME