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

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

net/publication/253945015

HYDRODYNAMIC DESIGN OF RADIAL FLOW PUMP IMPELLER BY SURFACE


PARAMETERIZATION

Article · July 2005

CITATIONS READS

12 2,027

3 authors, including:

Dimitris E. Papantonis
National Technical University of Athens
71 PUBLICATIONS   1,373 CITATIONS   

SEE PROFILE

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

Assessment of two-stage ORC system under variable thermal load. View project

DEsign optimisation of Pelton and Turgo type water turbine's injectors View project

All content following this page was uploaded by John Anagnostopoulos on 23 August 2015.

The user has requested enhancement of the downloaded file.


1st International Conference on Experiments/Process/System Modelling/Simulation/Optimization
1st IC-EpsMsO
Athens, 6-9 July, 2005
© IC-EpsMsO

HYDRODYNAMIC DESIGN OF RADIAL FLOW PUMP IMPELLER


BY SURFACE PARAMETERIZATION
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
Laboratory of Hydraulic Turbomachines
School of Mechanical Engineering / Fluids Section
National Technical University of Athens, Greece
Heroon Polytechniou 9, Zografou, 15780 Athens, Greece
e-mail: vgrapsas@fluid.mech.ntua.gr

Keywords: Hydraulic turbine blades design, numerical simulation, Cartesian grid, shape parameterization,
optimization.

Abstract. In the present work a new approach for hydraulic turbomachinery blade design, which ties Bezier
representation of a radial flow pump impeller to a shape optimization methodology of two-dimensional,
hydraulic blades is presented. With the aim of constructing a comprehensive design optimization procedure of
hydraulic turbine blades, the first stage of an improved inverse optimization procedure is proposed. For the
numerical simulation, the viscous Navier-Stokes equations are handled with the control volume approach and
the k-ε turbulence model. An unstructured Cartesian grid is applied to treat the irregular boundaries of the
blade while adaptive grid refinement techniques are incorporated to enhance the accuracy of the geometry
representation. The numerical optimization is performed by a stochastic evolutionary algorithm, whereas Bezier
polynomials are employed for the parameterization of the leading edge of the turbine blades. The design
parameters are the Bezier control points, the number and coordinates of which are properly selected and
ranged to allow for a smooth and progressive profile variation of the blade, while satisfying the design
constraints. The geometry and especially the leading edge is optimized by means of a thorough computational
study, providing knowledge on the influence of certain flow characteristics, like the efficiency, the NPSH
behaviour or the H-Q curve.

1 INTRODUCTION

With the aid of computational fluid dynamics and the rapid increase in computing power, internal flows in
the complex domain of a hydraulic turbomachinery can be well predicted, thus facilitating the design of pumps.
However, the inverse design numerical optimization is still a difficult task, because it needs a large number of
flow field solutions. The progress of the CFD code performance enables to reduce the design time by coupling
CFD codes with optimization tools. Although optimization techniques are becoming more and more standard for
the design evaluation,[1] preliminary stages as well as most optimization algorithms are using inviscid simplified
fluid assumption. Surface parameterization of impeller blades for optimum design has been recently
developed,[2] where empirical coefficients are used to describe the effects of entrance geometry, incident angle,
blade thickness and non-uniformity of the inlet velocity profile.[3]
However the optimization procedure still employs difficult aspects, like cavitation. Although the numerical
modelling of cavitation has received a great deal of attention, it is still not possible to predict such complex
unstationary flow. Cavitation is one of the most important aspects that need to be considered while designing
centrifugal pumps, since it is a major contributor to failure and inefficiency. Many impellers appear to suffer
from premature wear due to cavitation attack on the blade leading edges. For this reason few impeller design
methods for optimization of cavitation inception in terms of Net Positive Suction Head (NPSH) have been
developed and are based on simple one-dimensional flow calculations along mean streamlines.[4,5] Recently a
new method for designing impeller blades with a minimal drop in static pressure on the blade surface which lead
to the formation of cavitation was introduced.[6] However, these approaches fail to relate the leading edge
geometry with the pressure change.
The discontinuities of the leading and trailing edge can be easily defined using curve approximation
techniques. Cubic B-splines and Bezier curves are often employed,[7] since they can provide accuracy and
flexibility in complex and smooth shapes with a reduced number of design parameters, avoiding any curvature
discontinuities. The application of 3-dimensional turbulent flow analysis tools for the impeller design increased
in the last years.[8-11] Unstructured, Cartesian grids to accelerate the grid generation are applied combined with
techniques for cells cut by the irregular boundaries of the blade.[12,13]
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
The present study focalizes on the correlation between the leading edge geometry and the minimum pressure
developed near the blade leading edge, which represents the cavitating behavior of the impeller: the less the
pressure the more the extend of the cavitation region when local pressure drops below the vaporization pressure.
Different types of leading edges are considered in this simulation. The calculated results for the minimum
pressures are presented over the entire flow range and the influence of the leading edge geometry in ηu-Q, Hu-Q
curves is discussed. The design optimization algorithm has been recently tested for laminar flows in simple
geometries [14] with quite encouraging results.

2 MODEL FORMULATION
2.1 Flow solver

For two-dimensional incompressible, turbulent flow, the continuity and momentum equations can be written
in the rotating coordinate system as follows:
r
Continuity: ∇⋅w = 0 (1)
r r r r r r r ∇p r
Momentum: w ⋅ ∇w = −2ω × w − ω × (ω × w) − +ν ∇2w (2)
ρ
r
where p is the fluid pressure; ρ is the density; ν is the kinematic viscosity; w is the fluid velocity in the rotating
system and ω is the rotation speed of the impeller. Τhe system of the above equations is numerically solved with
the finite volume approach and turbulence was modelled with the standard k-ε model. A pseudo-3D
incompressible term in dimensional form is added to the continuity equation to simulate the actual velocity and
pressure field in the 3D blade-to-blade domain.
An unstructured Cartesian grid is applied and adaptive grid refinement techniques are incorporated.[14,15] The
Cartesian grid is adaptively refined near the leading edge and along the pressure and suction sides, using two
refinement levels. Local refinement of the computational grid in these regions of steep gradients, can
significantly improve the accuracy of the results and of the blade geometry representation, without increasing
considerable the computer storage and the CPU-time requirements. The main drawback of the Cartesian grid
which is the simulation of irregular boundaries came over with a “partially blocked cell” technique [14,16] which
makes possible the solution of the grid cells that are crossed by the blade surfaces and are solved with an almost
second order accuracy. During the optimization process the evaluation algorithm is capable to automatically
generate the grid and solve the flow equations for any blade geometry.
When looking at the optimization process of a pump impeller, one needs to consider a number of different
stages. The main steps in the process that this study proposes are shown in the following flow chart. Therefore
the blade’s leading edge development process involves: geometry parameterization, mesh generation, CFD
computation and flow analysis.
Geometry

Mesh

Optimizer CFD

Analysis

Objective

Figure 1. Overview of the optimization process.

2.2 Geometry Representation and Leading Edge Parameterization


The geometry of the leading edge plays an important role in the blade design process. Modifications of the
leading edge geometry can be described in a simple and flexible manner using Bezier parameterization. The
shape of the leading edge can be controlled with a few control points. This allows the development of a
correlation between the control point locations representing the leading edge geometry and the corresponding
value of the minimum pressure.
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
The computational domain is a periodically symmetric section of the centrifugal impeller (z = 9 blades). The
impeller dimensions are regulated according to a commercial centrifugal pump operating in the Laboratory. The
impeller has 9 two-dimensional (non-twisted) blades with inlet and exit diameter D1=70 mm, D2=190 mm, exit
width b2=9 mm, and inlet and exit angle β1=26 deg, β2=49 deg. The nominal speed of rotation is fixed at
3000rpm. The blade thickness (b=5mm) as well as the trailing edge shape are kept unchanged.
The leading edge is described by a Bezier curve (actually, two different curves). A minimum number of four
(4) control points, shown in Figure 2, with one control point at the vertical and horizontal axe respectively, was
found to be adequate to obtain a variety of geometries of the leading edge. The design parameters are the
coordinates of the Bezier control points. By moving one or more of these points the curve’s shape deforms in a
smooth and flexible, ‘plastic’ way.

Figure 2. Blade leading edge parameterization. Figure 3. Blade angle distribution.

For the representation of the pressure (PS) and suction (SS) surfaces of the blade the triangle error method[17]
was used, where the local blade angle β in the i random triangle (Fig. 3) is described by the wrap angle dθ, in
polar coordinates as
dm dm (3)
tgβ i = =
r ⋅ dθ du
where β(r) is the blade angle distribution, r the radius corresponding to each blade angle β and dm the
meridional intersections for each radius.

2.3 Impeller head and power calculations


The energy gained by the fluid through the impeller, H is computed from the total energy of the fluid at the
impeller inlet and exit periphery (Fig. 4):

1 ⎛ p 2 − p 1 c 22 − c12 ⎞
Qu ∫
Η = H 2 − H1 = ⋅ ⎜ ⎟ dq (4)
⎜ ρ g + 2g ⎟
⎝ ⎠
where c is the absolute velocity of the fluid, Qu the flow rate through the impeller and g the gravity acceleration,
while the subscripts 1 and 2 denote impeller inlet and exit conditions, respectively (Fig. 4). The head given by
the impeller, Hu, can be calculated from the torque Mu developed on the blades:
r2
N u = ρ ⋅ g ⋅ Qu ⋅ H u = ω ⋅ M u = ω ⋅
∫ r1
[(rr × nr ) ⋅ p + (rr × τrw ) ⋅ cot β ]⋅ b ⋅ dr (5)

c2 ωr2
β2
e
sid

r2 wr w2
re

w
ssu
pre

wu
su c t

ωr1 n
c1
io n s

w1 τw
id e

r1 β
β1
ω r
1 2

Figure 4. Sketch of a centrifugal pump impeller.


Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
Hence the impeller head Hu can be computed as :
ω ⋅Mu (6)
Hu =
ρ ⋅ g ⋅ Qu
and the hydraulic efficiency of the impeller can be finally defined as:
H
ηu = (7)
Hu

2.4 Numerical optimization


The optimal design of the leading edge geometry of hydraulic turbomachinery blades is based on the
maximization of the minimum pressure cost function, which is realized through an optimization algorithm based
on evolutionary strategies. The Evolutionary Algorithms SYstem (EASY) optimization software [1] used in this
study is a recently developed and brought to market by the Lab. of Thermal Turbomachinery, NTUA. It makes
use of stochastic optimization methods and has been successfully applied in various engineering applications for
inverse design.

3 RESULTS
3.1 General

The design of the leading edge of centrifugal pumps is strongly related to cavitation flow phenomena, which
may occur in either the rotating impeller or the stationary parts of the machine. Furthermore, avoiding cavitation
should be one of the goals of runner blades design in hydraulic machines. The possibility of correlating the
development of cavitation with the geometry of the leading edge allows the study of the influence of the
geometry. In a preliminary validation of the effect of the geometry of the leading edge on the minimum
developed pressure and thus on the cavitation characteristics, five (5) different cases-geometries are selected.
The leading edge as already mentioned, is described by two Bezier curves. By varying the location of the control
points various leading edges can be formed, from a slim to a rounded edge, like the examples of Figure 5. The
leading edge of blade A is rounded following a cyclic arc (“normal” case). Blade’s B leading edge is blunt, C
slim, and D and E show the other two blunt-slim combinations.

E
A B C D

Figure 5. Geometries of the leading edge.

The computational domain shown in Figure 6a presents the impeller with the “normal” blade, and a more
detailed view of the grid close to the leading edge is shown in Figure 6b.

Figure 6. a) Computational grid for a 2D pump impeller, and b) detailed view at the blade’s leading edge.

The characteristic curves H–Q, Hu-Q, Nu-Q and ηu–Q for the normal leading edge are shown in Figure 7. The
predicted results for the head-flow curves in these cases are presented over the entire flow range. The impeller
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
head Hu is, as expected, quite higher than the head H that the water acquires in the impeller, due to the head
losses. In the ηu-Q curve the Best Efficiency Point (BEP) is obtained for Q ≈ 62,5m3/h. The flow field in the
impeller is illustrated in Figure 8a along with a detailed view in the vicinity of the leading edge region in Figure
8b. The region of the minimum pressure is located at the blade’s suction side where the flow accelerates.

Figure 7. Computed characteristic curves of the pump with the normal leading edge.

Figure 8a. Pressure contours and flow Figure 8b. Details of the flow
streamlines in the computational domain. field around the normal edge.
It is remarkable that the computed Hu-Q, Nu-Q and ηu–Q curves for all the different leading edge geometries
of Figure 5 do not present substantial changes over the entire flow range, hence no particular influence of the
leading edge to the characteristic curves was observed.

3.2 Preliminary Study

Performing a systematic variation of the leading edge geometry, an optimum in terms of minimum pressure
pmin can be revealed. However, in order to predict the cavitation performance of the impeller within a flow rate
range ±30% around BEP, where it might operate, the construction of the pmin-Q characteristic curve of each case
is needed. Five such curves corresponding to the blades of Figure 5 are plotted in Figure 9. In all curves the
pressure reduces as the flow rate becomes lower, because the inlet flow angle is reduced and the streamline
curvature in the blade suction side increases. The minimum pressure obtained around the BEP was found lower
for Slim and Upper Edge, whereas the three other edges exhibit a smooth behavior with higher minimum
developed pressures. The intermediate normal case shows the best cavitation behavior over the other extreme
cases.
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
-40000
30 40 50 60 70 80 90

-60000 Normal Case


Lower Edge
Slim Edge
-80000
Blunt Edge
Upper Edge
-100000
Pmin (Pa)

-120000

-140000

-160000

-180000
3
Q (m /h)

Figure 9. Pmin-Q characteristic curve of each pump.

In Figure 10a to d the flow field for the BEP in the vicinity of the blade leading edge region is illustrated.
The cases of the Slim leading edge (b) and partly of the Upper leading edge (c) exhibit lower minimum pressure
at the inlet part of the suction side. The location of the minimum pressure is for these two cases just behind the
sharp edge, as well as at the second discontinuity point downstream (Fig. 10b, c). Although the minimum
pressure is very low at these points the area of the low pressure region is smaller than the other two cases (Fig.
10a, b). In the latter cases, as well as in the Normal case (Fig. 8b), the minimum pressure is observed at the
suction side where the streamlines curvature maximizes. The minimum pressure area is mow more extended,
although the minimum values are higher (Fig. 9).

Figure 10. Flow Streamlines and Pressure contours for a) Blunt; b) Slim; c) Upper and d) Lower leading edges
Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis

3.3 Optimization Procedure

The objective of the optimization is to maximize the minimum pressure at the blade suction side, by
changing the shape of the leading edge. An additional parameter is the desired operation range of the impeller,
and hence two different studies were performed: in the first (OPT1) the target is to maximize the minimum
pressure only at the BEP, whereas in the second (OPT2) for flow rate values up to -15% of the BEP (average
pressure). The convergence history of the optimization algorithm is shown in Figure 11a. The convergence rate
is quite fast and the final shape is practically obtained after about 70 evaluations for both cases. The resulting
optimal geometries for the above two studies are sketched in Figure 11b. The shape of the two blades is
remarkably different, and the impeller operation range seems to be an important parameter in order to determine
the optimal design. However, as can be seen in Figure 12, the performance of the OPT2 blade edge below the
BEP is much better than that of the OPT1, whereas the two blades exhibit similar behavior above BEP.
Therefore the maximization of the average minimum pressure is a more reliable target and leads to an optimal
design with considerably improved behavior for a wide operation range of the impeller, against the initially
considered as normal design (Fig. 12).

105000

100000
OPT2
95000

90000 OPT1 OPT1 OPT2


Objective

85000

80000

75000

70000

65000

60000
0 50 100 150
Evaluations

Figure 11a. Convergence history of the Figure 11b. Optimum geometries of the
optimization procedure. blade leading edge.

-40000
30 40 50 60 70 80 90

-60000

-80000

-100000
Pmin (Pa)

-120000
OPT1

-140000 OPT2

Normal Case
-160000

-180000
3
Q (m /h)

Figure 12. Pmin-Q characteristic curve of optimum leading edges.


Vasilios A. Grapsas, John S. Anagnostopoulos and Dimitrios E. Papantonis
4 CONCLUSIONS
A numerical hydrodynamic design optimization methodology is developed and regulated, combining a fast
and automated flow evaluation algorithm, a shape parameterization technique and an optimization software. The
method is applied here to improve the cavitation characteristics of a centrifugal pump impeller by omptimizing
the blade leading edge geometry, but can be applied to other parts of a hydrodynamic machine, or using different
objectives (e.g. increase of efficiency). The optimum blade shape is obtained quite fast and exhibits a
considerably improved performance over a wide operation range. A more elaborate design would require the
detailed parameterization and solution of the entire 3D inlet section of the impeller, which is currently
underway.

ACKNOWLEDGEMENTS

The project is co-funded by the European Social Fund (75%) and National Resources (25%) - Operational
Program for Educational and Vocational Training II and particularly the Program Pythagoras.

REFERENCES
[1] Giannakoglou, K.C. (2002), “Design of optimal aerodynamic shapes using stochastic optimization methods
and computational intelligence”. Progress in Aerospace Sciences, Vol 38, pp. 43-76.
[2] Ferrando, L., Kueny, J. and Avellan F., (2004), “Surface Parameterization of a Francis Runner Turbine for
Optimum Design”, Proceedings of the 22nd IAHR Symposium on Hydraulic Machinery and Systems,
Stockholm, Sweden, Jun 29-July2.
[3] 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. Engirs.
Vol. 209, pp. 29-38.
[4] Noskievic, J., (1976), “The suction ability of centrifugal pumps”, International Conference on Pump and
Turbine Design and Development, NEL.
[5] 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. Engirs.
Vol. 209, pp. 29-38.
[6] Cooper, P., Sloteman, D., 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.
[7] Ferrando, L., Kueny, J.L., Avellan, F., Pedretti, C., and Tomas, L., (2004), “Surface Parameterization of a
Francis Runner Turbine for Optimum Design”, Proceedings 22nd IAHR Symposium on Hydraulic
Machinery and Systems, Stockholm, Sweden, June 29 – July 2, A02-2.
[8] Benra, F.K. (2001), “Economic development of efficient centrifugal pump impellers by numerical
methods”, World Pumps, May 2001, pp. 48-53.
[9] Yulin, W., Jianming, Y. and Shuliang, C (1999), “Advanced design for large hydraulic turbine runner”,
Proceedings ASME/JSME FEDSM’ 99, San Francisco, California, July 18-23, S-287.
[10] Garrison, L.A., Richard, K.F., Sale, M.J. and Cada, G. (2002), “Application of biological design criteria
and computational fluid dynamics to investigate fish survival in Kaplan turbines”, Proceedings HydroVision
2002, Portland, Oregon, July 29 – August 2.
[11] Swiderski, J., Martin, J.N., Norrena, R. (2001), “Automated runner blade design optimization process based
on CFD verification”, Proceedings Waterpower XII, Salt lake City.
[12] Ye, T., Mittal, R., Udaykumar, H.S. and Shyy, W. (1999), “An accurate Cartesian grid method for viscous
incompressible flows with complex immersed boundaries”, J. of Computational Physics, 156, pp. 209-240.
[13] Fadlun, E.A., Verzicco, R., Orlandi, P. and Mohd-Yusof, J. (2000), “Combined immersed-boundary finite-
difference methods for three-dimensional complex flow simulations”, J. Computational Physics, 161, pp.
35-60.
[14] Grapsas, V. and Anagnostopoulos, J. (2004), “Numerical optimization of the hydrodynamic shape of fluid
flow systems”, Proceedings 1st IC-SCCE, Athens, Greece, 8-10 September.
[15] Anagnostopoulos, J. (2003), “Discretization of transport equations on 2D Cartesian unstructured grids
using data from remote cells for the convection terms”, Intl. J. for Numerical Methods in Fluids, 42, pp.
297-321.
[16] Anagnostopoulos, J. (2005), “A numerical Simulation Methodology for Hydraulic Turbomachines”, 5th
GRACM International Congress on Computational Mechanics, Limassol, Cyprus, 29 June-1 July.
[17] Stepanoff, A.J. (1993), “Centrifugal and Axial Flow Pumps, Theory, Design and Application”, Krieger
Publishing Company Malabar, Florida.

View publication stats

You might also like