1st International Conference “From Scientific Computing to Computational Engineering”

1st IC-SCCE
Athens, 8-10 September, 2004
© IC-SCCE

A NUMERICAL SIMULATION OF THE IMPELLER-VOLUTE INTERACTION IN

A CENTRIFUGAL PUMP
* † * *
Michalis D. Mentzos , Andronicos E. Filios , Dionisios P. Margaris and Dimitrios G. Papanikas
*
Fluid Mechanics Laboratory, Mechanical Engineering and Aeronautics Department
University of Patras
GR-26500 Patras, Greece
e-mail: margaris@mech.upatras.gr
†
Fluid Mechanics & Turbomachines Laboratory, Mechanical Engineering Department,
School of Pedagogical and Technological Education,
GR-141 21 Athens, Greece
e-mail: afilios@aspete.gr

Keywords: Centrifugal pump, Computational fluid dynamics, Navier-Stokes code, Pump performance.

Abstract. Pump designers are continually being challenged to provide machines that operate more efficiently,
quietly, and reliably at lower cost. With the aid of the CFD approach, the complex internal flows in water pump
impellers, which are not fully understood yet, can be well predicted, to speed up the pump design procedure.
The present work describes a numerical simulation of the internal flow in a centrifugal pump. The object of the
study is an industrial centrifugal water pump with backward curved blades, which is built within a vaneless
single tongue volute. For the numerical simulation, the viscous Navier-Stokes equations are handled with a
steady calculation utilizing the k-ε turbulence model and the multi reference frame technique is applied to take
into account the impeller-volute interaction. With them, it is possible to simulate correctly the blade passage in
front of the tongue and the flow and pressure distributions. The data obtained allow the analysis of the main
phenomena existent in the pump, such as: pressure changes in the volute for different flow rates; and the
incidence at the leading edge of the blade with different flow conditions.

1 INTRODUCTION
The design of hydraulic turbomachines has reached the stage were improvements can only be achieved
through a detailed understanding of the internal flow. The prediction of the flow in such equipment is very
complicated due to the rotation and the curved three-dimensional shape of the impellers. Furthermore, the flow
in turbomachines shows unsteady behaviour, especially at off-design conditions, as a result of interaction
between impeller and pump casing. Considering these complexities, computer simulations will become
increasingly important.
Computational fluid dynamics (CFD) analysis is being increasingly applied in the design of centrifugal
pumps. With the aid of the CFD approach, the complex internal flows in water pump impellers, which are not
fully understood yet, can be well predicted, to speed up the pump design procedure. Thus, CFD is an important
tool for pump designers. The use of CFD tools in turbomachinery industry is quite common today. Many tasks
can numerically be solved much faster and cheaper than by means of experiments. Nevertheless the highly
unsteady flow in turbomachinery raises the question of the most appropriate method for modeling the rotation of
the impeller.
Many CFD studies concerning the complex flow in all types of centrifugal pumps have been reported.
Goto[1] presented a comparison between the measured and computed exit-flow fields of a mixed flow impeller
with various tip clearances, including the shrouded and unshrouded impellers, and confirmed the applicability of
the incompressible version of the three-dimensional Navier-Stokes code developed by Dawes[2] for a mixed-
flow centrifugal pump. Van Esh et al.[3] developed a numerical method based on a fully 3D finite element
method for solving the inviscid, incompressible flow field in entire pump. The pump’s efficiency is determined
applying Thwaites’[4] method along with Green’s “lag entrainment method”[5]. The capabilities of the proposed
method are demonstrated by the analyses of the flow in two pumps. Zhou and Ng[6] and Ng et al.[7] also
developed a three-dimensional time-marching, incompressible Navier-Stokes solver using the pseudo-
compressibility technique to study the flow field through a mixed-flow water-pump impeller. The applicability
of the original code was validated by comparing it with many published experimental and computational results.
Oh and Ro[8] used a compressible time marching method, a traditional SIMPLE method, and a commercial
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.
program of CFX-Tascflow to simulate flow pattern through a water pump and compared the differences among
these methods in predicting the pump’s performance.
Recently, Kaupert et al.[9], Potts and Newton[10], Sun and Tsukamoto[11], and Byskov et al.[12] studied pump
design and off-design performance using the commercial software CFX- Tascflow, Fluent, StarCD and
Fine/Turbo, respectively. Although these researchers predicted reverse flow in the impeller shroud region at
small flow rates numerically, some contradictions still existed. For example, Kaupert’s[9] experiments showed
the simultaneous appearance of shroud-side reverse flow at the impeller inlet and outlet, but his CFD results
failed to predict the numerical outlet reverse flow. Sun and Tsukamoto[11] validated the predicted results of the
head-flow curves, diffuser inlet pressure distribution, and impeller radial forces by revealing the experimental
data over the entire flow range, and they predicted back flow at small flow rates, but they did not show an exact
back-flow pattern along the impeller outlet. Byskov et al.[12] investigated the flow field in a shrouded six bladed
centrifugal pump impeller using LES and RANS simulations and their predictions are compared with
experimental data obtained from particle image velocimetry. However, it is found that the two RANS models do
not reproduce the stall phenomenon observed at quarter load and are incapable of detecting the difference
between the two passages.
From such literature, it was found that most previous research, especially research based on numerical
approaches, had focused on the design or near-design state of pumps. Few efforts were made to study the off-
design performance of pumps. Centrifugal pumps are widely used in many applications, so the pump system
may be required to operate over a wide flow range in some special applications. Thus, knowledge about off-
design pump performance is a necessity. On the other hand, it was found that few researchers had compared
flow and pressure fields among different types of pumps. Therefore, there is still a lot of work to be done in
these fields.
In this work, the commercial CFD code Fluent, was used to study three-dimensional turbulent flow through
a commercial water-pump during design and off-design conditions. Fluent is a software package that can
predict laminar flow, turbulent flow, and heat transfer. It has been widely used in the field of turbomachinery,
and the simulation results have been proven by many researchers to be reliable (Sun and Tsukamoto[11],
Gonzalez et al.[13]). Fluent overcomes the meshing difficulties that arise in complex geometry by using a
powerful CAD-based preprocessor, Gambit, which generates a surface mesh of triangles. This surface mesh is
then converted into a volume mesh of tetrahedral elements. The selected commercial pump is the Norma 80-250
water-pump for which drawings were provided by the manufacturer[14]. In order to well suit the pump for CFD
analysis, modifications were done. These concern the fitting of a front shroud and the adjustment of the mean
camber line of the blades in a way that the relative w-velocity varies linearly from the inlet section of the
impeller to the outlet section. The resulted CAD-model of the above pump is shown in Figure 1 and its main
geometry and flow parameters are presented in Table 1.

Denomination Value
Suction pipe diameter Ds = 142,4mm
Impeller diameters D1 = 157mm, D2 = 274mm
Impeller widths b1= 31,0mm, b2= 20,8mm
Impeller angles β1 = 10°, β2 =30°
Number of blades z=6
Flow rate in best Q = 80m3/h,
efficiency point, rpm n = 1450rpm
Specific speed
nq = 22,86
(nq = n ·Q1/2/H3/4)
Reynolds number at
impeller exit Re = 2,84 ·106
(Re=U2·D2/ν)
Figure 1. CAD-model of the commercial centrifugal Table 1: Geometric details and flow parameters of the
pump under study. centrifugal pump under study.
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.
2 MODELS DESCRIPTION AND COMPUTATIONAL METHOD

2.1 Mathematical Models

The code solves the fully 3D incompressible Navier-Stokes equations, including the centrifugal force source
in the impeller and the unsteady terms. Turbulence is simulated with the standard k-ε model. Although grid size
is not adequate to investigate local boundary layer variables, global ones are well captured. For such
calculations, wall functions, based on the logarithmic law, have been used.
Basic Equations
For three-dimensional incompressible, unsteady flow, the continuity and momentum equations can be written
in the rotating coordinate system as follows:
∂ρ
+ ∇ ⋅ (ρU ) = 0 (1)
∂t
and
∂ρU
∂t
( (
+ ∇ ⋅ (ρU ⊗ U ) = ∇ ⋅ − pδ + µ eff ∇U + (∇U )T + S M )) (2)

where vector notation has been used, ⊗ is a vector cross product; U is the velocity; p is the pressure; ρ is the
density; δ is the identity matrix; and SM is the source term.
For flows in a rotating frame of reference that are rotating at the constant rotation speed Ω, the effects of the
Coriolis are modeled in the code. In this case, with r the location vector,
S M = − ρ [2 Ω ⊗ U + Ω ⊗ (Ω ⊗ r )] (3)

k-ε Turbulence Model

In Equation (2), µeff is the effective viscosity coefficient, which equals the molecular viscosity coefficient, µ,
plus the turbulent eddy viscosity coefficient, µt :
µ eff = µ + µ t (4)

The turbulent viscosity, µt, is modeled as the product of a turbulent velocity scale, Vt, and a turbulent length
scale, lt, as proposed by Kolmogorov[15]. Introducing a proportionality constant gives
µ t = ρc µ ltVt (5)

Both equation models take the velocity scale, Vt, to be the square root of the turbulent kinetic energy:

Vt = k (6)

The turbulent kinetic energy, k, is determined from the solution of a semiempirical transport equation.
In the standard k-ε two-equation model it is assumed that the length scale is a dissipation length scale, and
when the turbulent dissipation scales are isotropic, Kolmogorov determined that

k3/ 2 (7)
ε=
lt
where ε is the turbulent dissipation rate.
Therefore, the turbulence viscosity, µt, can be derived from Equations (5), (6), and (7) to link to the
turbulence kinetic energy and dissipation via the relation

k2
µt = Cµ ρ (8)
ε
where Cµ is a constant. Its value is 0,09.
The values of k, ε come directly from the differential transport equations for the turbulence kinetic energy
and turbulence dissipation rate:
∂ρk
+ ∇ ⋅ (ρUk ) − ∇ ⋅ (Γ k ∇k ) = p k − ρε (9)
∂t
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.
and
∂ρε ε
+ ∇ ⋅ (ρUε ) − ∇ ⋅ (Γε ∇ε ) = (Cε 1 p k − Cε 2 ρε ) (10)
∂t k
where the diffusion coefficients are given by
µt (11a)
Γk = µ +
σk
and
µt (11b)
Γε = µ +
σε
and Cε1=1,44; Cε2 =1,92; σk=1,0; and σε=1,3 are constants.
The pk in Equations (9) and (10) is the turbulent kinetic energy production term, which for incompressible
flow is

( 2
)
p k = µ t ∇U ⋅ ∇U + ∇U T − ∇ ⋅ U (µ t ∇ ⋅ U + ρk )
3
(12)

Equations (1), (2), (9), and (10) form a closed set of nonlinear partial differential equations governing the
fluid motion.
Log-Law Wall Functions
There are large gradients in the dependent variables near the wall. It is costly to fully resolve the solution in
this near-wall region as the required number of nodes would be quite large. Thus a common approach known as
“wall functions” is applied to model this region.
In the wall-function approach (Launder and Spalding[16]), the near-wall tangential velocity is related to the
wall shear stress by means of a logarithmic relation, which can be written as follows:

u+ =
κ
1
( )
ln y + + C (13)

where
ut
u+ = (13a)
uτ

ρ∆yut
y+ = (13b)
µ

τw (13c)
uτ =
ρ
τw is the wall shear stress, ut is the known velocity tangent to the wall at a distance of ∆y from the wall, κ is the
von Karman constant for smooth walls, and κ and C are constants, depending on wall roughness.
However, this form of the wall-function equations has the problem that it becomes singular at separation
points where the near-wall velocity, ut, approaches zero. In the logarithmic region, the alternative velocity scale,
u*, can be used instead of u+:

u* = c µ1 / 4 k (14)

This scale has the useful property of not going to zero if ut goes to zero (and in turbulent flow, k is never
completely zero). Based on this definition, the following explicit equation for the wall shear stress is obtained:

y* (15)
τ w = τ visc
u+
where
τvisc= µut/∆y; y*= ρu*∆y/µ; and u+ as given from equation (13).
The recommended practice is to locate near-wall nodes such that y* is in the range of 20 to 50 for smooth
walls. In the near-wall region, an estimate of the dissipation consistent with the log-law can be presented as
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.

c µ3 / 4 k 3 / 2
ε= (16)
u∆ y
The dissipation at the first interior node is set equal to this value. The boundary nodal value for k is
estimated via an extrapolation boundary condition. The near-wall production of turbulent kinetic energy is
derived to be
2
τ visc
pk = p k∗ (17)
µ
where
2
 y ∗  du +
p k∗ = +  (18)
 u  dy ∗
 

2.2 Modeling details

The computational grid was generated using the Fluent preprocessor Gambit. In a pre-step the walls of the
impeller limiting the fluid flow space were separated with a CAD-tool and some additional inner faces were
introduced. This was helpful since the complete unstructured grid was partially refined, as shown in fig. 2. The
spiral casing with its circular symmetric cross section and the draft tube part directly upstream of the impeller
inlet were quite difficult to mesh and this explains the adoption of the unstructured meshing, too. The resulting
grid for the entire pump under study is shown in fig. 3. For the use with the fluent solver, two blocks with
suitable boundary conditions form the complete computational domain. All parts of the pump rotating in the
inertial frame are made part of the impeller fluid zone that is calculated in a rotating reference frame when using
the FR-approach. The second fluid zone is steady, i.e. the spiral casing, the discharge nozzle that is extended
almost three outlet diameters for a less reacting outlet boundary condition, as well as the two impeller side gaps.
Grid interfaces are needed to couple fluid zones which are in relative motion to each other. The rotating fluid
zone is constructed with 114.908 tetrahedral elements, while the stationary fluid zone comprises 204.326
tetrahedral elements.

Z
Z Y
Y X
X

Figure 3. Sketch of the pump unstructured meshJul 08, 2004

Figure 2. Mesh structure in theFLUENT

impeller. Jul 08, 2004
6.0 (3d, segregated, ske) (outlet pipe portion is added).
FLUENT 6.0 (3d, segregated, ske)

Turbulence was modeled using the most popular and standard model for investigations not aiming at
turbulence model validation, the standard k-ε model with standard wall functions for smooth walls. As inlet
boundary condition the velocity according to the operating point flow rate was prescribed. Also, non-slip
boundary conditions have been imposed over the impeller blades and walls, the volute casing and the inlet pipe
wall. Turbulence intensity was not measured and estimated to 5%, the length scale to (Ds/2) mm. At the exit of
the discharge nozzle, assuming a fully developed turbulent flow, a practical zero velocity gradients was set.
Second order upwind schemes were chosen for space discretisation while the pressure correction equation was
solved with the SIMPLE algorithm. Convergence is checked by residuals decreasing about two orders. For
setting up the problem, a low rotational speed in the inputs for boundary conditions and for the angular velocity
of the reference frame was used. The rotational speed in the first attempt was selected as 10% of the actual
operating condition and gradually increased to 100%. A converged multi reference frame result needed about
2000 iterations, in a 3GHz Pentium IV PC, for the actual operating condition. The results are certainly may be
used as an initial flow field for a transient calculation.
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.
3 RESULTS FOR IMPELLER-VOLUTE INTERACTION
One rotational speed, 1450 rpm, was used in the computations. At this rotational speed, several different
flow rates were specified at the inlet boundary so as to study design and off design flow patterns. The
convergence criteria for each run were set to be 10-5 for root mean-square (RMS) residuals of mass/momentum
equations and 10-4 for RMS residuals of k-ε equations. It was clearly evident that after several hundred time
steps in each run, the above criteria could be satisfied, and the convergence was reached gradually.
Figure 4 shows the profiles of static pressure in the rotational zone of the domain at the design point. As for
pressure distribution, it can be seen clearly that the pressure increases gradually in a stream wise direction, and
normally it has higher pressure on the pressure surface than on the suction surface on each blade. Figure 5 shows
the w-relative velocity vectors through the impeller’s blade passages at the same operating conditions. From the
last figure it is evident that the flow is smooth and no recirculation is in presence.
Figures 6 and 7 show the static pressure profiles in the impeller-volute domain at the design point. In both
figures, the maximum outlet pressure is reached at the entrance of the attached pipe outlet portion.
Various volume flow rates were specified to study in constant rotational speed the off-design conditions. It
was found that with the change of inflow rate even as much us 50% of the design flow rate, the flow patterns
look similar to each other.
1.84e+01
1.77e+01
2.09e+00 1.69e+01
1.98e+00 1.62e+01
1.87e+00 1.55e+01
1.76e+00 1.47e+01
1.65e+00 1.40e+01
1.54e+00 1.33e+01
1.43e+00 1.25e+01
1.32e+00 1.18e+01
1.21e+00 1.11e+01
1.10e+00 1.03e+01
9.89e-01 9.60e+00
8.80e-01 8.86e+00
7.70e-01 8.13e+00
6.60e-01 7.39e+00
5.50e-01 6.66e+00
4.41e-01 5.92e+00
3.31e-01 5.19e+00
2.21e-01 4.45e+00
1.12e-01 3.72e+00
1.83e-03 2.98e+00
-1.08e-01 2.25e+00
-2.18e-01 1.51e+00 Z
Y
-3.27e-01 7.80e-01
X
-4.37e-01 YZ 4.47e-02
-5.47e-01
-6.56e-01
X

Figure 5. Relative velocity vector (m/s) through

relative-velocity Colored By Velocity Magnitude (m/s) Jul 08, 2004
FLUENT 6.0 (3d, segregated, ske)

Figure 4. Profiles of static pressure (atm).

Profiles of Static Pressure (atm) Jul 08, 2004
FLUENT 6.0 (3d, segregated, ske) the impeller’s blade passages at the design point.
2.09e+00
1.93e+00
1.78e+00
2.09e+00
1.93e+00
1.63e+00

1.78e+00 1.48e+00
1.63e+00 1.32e+00
1.48e+00 1.17e+00
1.32e+00 1.02e+00
1.17e+00 8.67e-01
1.02e+00 7.15e-01
8.67e-01
5.63e-01
7.15e-01
4.10e-01
5.63e-01
2.58e-01
4.10e-01
1.05e-01
2.58e-01
1.05e-01 -4.69e-02
-4.69e-02 -1.99e-01
-1.99e-01 -3.52e-01 Y
-3.52e-01 -5.04e-01
-5.04e-01 YZ Z X
X -6.56e-01
-6.56e-01

Figure 7. Contours of static pressure (atm) at the

Contours of Static Pressure (atm)
Figure 6.Pressure
Contours of Static Contours
(atm) of static pressure (atm) at the FLUENT 6.0 (3d
FLUENT 6.0 (3d, seg design point.
design point.

4 CONCLUSIONS
The commercially available 3D Navier-Stokes code called Fluent, which has a standard k-ε two-equation
turbulence model, was chosen to simulate the internal flow of a commercial centrifugal pump. The complete
centrifugal pump including impeller side gaps and discharge nozzle has been modeled and meshed with an
unstructured grid. Quasi-steady calculations have been performed for the design point as well as for a range of
volume rates with a deviation of 50% from the design value but with similar flow patterns. The predicted results
are presented in terms of pressure profiles and velocity vectors. Having a closer look at the flow rates in a
rotating impeller, the multiframe reference frame approach fails completely due to its fixed coupling
formulation. However, the present approach is useful for basically understanding the flow states in various
operating points. Certainly the scheduled transient calculations may serve as a real tool for understanding the
interaction between impeller and spiral casing. Further work will focus on a detailed statement on the influence
 Michalis D. Mentzos*, Andronicos E. Filios†, Dionisios P. Margaris* and Dimitrios G. Papanikas.
and a general validation of different turbulence models for use with pump calculation.

REFERENCES
[1] Goto, A. (1992), “Study of internal flows in a mixed-flow pump impeller at various tip clearances using
three-dimensional viscous flow computations”, ASME Journal of Turbomachinery, Vol.114, pp. 373–
382.
[2] Dawes, W. N. (1986), “A numerical method for the analysis of three dimensional viscous compressible
flow in a turbine cascade: Application to secondary flow development in a cascade with and without
dihedral”, ASME Paper 86-GT-145. New York: American Society of Mechanical Engineers.
[3] Van Esch, B.P.M., Kruyt, N.P., and Jonker J.B. (1997), “An inviscid-viscous coupling method for
computing unsteady flows in entire pump configurations”, The ASME Fluids Engineering Division
Summer Meeting, FEDSM’97, June 22-26, pp. 1-8.
[4] Thwaites, B.(1949), “Approximate calculation of the laminar boundary layer”, Aeronaut. Q., Vol. 1, pp.
245-280.
[5] Green, J.E., Weeks, D.J., and Brooman, J.W.F. (1972), “Prediction of turbulent boundary layers and
wakes in compressible flow by a lag-entrainment method”, RAE Technical Report 72231, pp. 49-53.
[6] Zhou, W. D., and Ng, E. Y. K. (1998), “3-D viscous flow simulation of mixed-flow water pump impeller
with tip-clearance effects”, Proceedings of the 4th International Conference and Exhibition on Pumps
and Systems, pp. 189–198, Singapore, HQ Link Pte Ltd.
[7] Ng, E. Y. K., Zhou, W. D., and Chan, W. K. (1998), “Non-Newtonian effects on mixed-flow water pump
using CFD approach”, Proceedings of the 19th International Association of Hydraulic Research
Symposium, Section on Hydraulic Machinery and Cavitation, pp.735–749, Singapore, International
Association of Hydraulic Research.
[8] Oh, J. S., and Ro, S. H. (2000), “Application of time marching method to incompressible centrifugal
pump flow”, Proceedings of the 2nd International Symposium on Fluid Machinery and Fluid
Engineering, pp. 219–225, Beijing: Tsinghua University Press.
[9] Kaupert, K. A., Holbein, P., and Staubli, T. (1996), “A first analysis of flow field hysteresis in a pump
impeller”, ASME Journal of Fluids Engineering, Vol.118, pp.685–691.
[10] Potts, I., and Newton, T. M. (1998), “Use of a commercial CFD package to predict shut-off behavior of a
model centrifugal pump: an appraisal”, IMechE Seminar Publication: CFD in Fluid Machinery Design,
London: Institute of Mechanical Engineers.
[11] Sun, J., and Tsukamoto, H. (2001), “Off-design performance prediction for diffuser pumps. Journal of
Power and Energy”, Proceedings of I. Mech. E., Part A, Vol. 215, pp. 191–201.
[12] Byskov, R.K., Jacobsen, C.B., and Pedersen N. (2003), “Flow in a centrifugal pump impeller at design
and off-design conditions – Part II: Large eddy simulations”, ASME Journal of Fluids Engineering, Vol.
125, pp. 73-83.
[13] Gonzalez, J., Fernandez, J., Blanco, E., and Santolaria C. (2002), “Numerical simulation of the dynamic
effects due to impeller-volute interaction in a centrifugal pump”, ASME Journal of Fluids Engineering,
Vol. 124, pp. 348-355.
[14] Drakos-Polemis S.A., Pump Catalogue.
[15] Kolmogorov, A. N. (1941), “Local structure of turbulence in incompressible viscous fluid for very large
Reynolds number”, Doklady Akademyia Nauk SSSR, Vol. 30, pp. 9-13.
[16] Launder, B.E., and Spalding, D.B. (1974), “The numerical computation of turbulent flows”, Complete
Methods of Applied Mechanical Engineering, Vol. 3, pp. 269-289.