Numerical Study of Pressure Fluctuations Caused by Impeller-Diffuser Interaction in A Diffuser Pump Stage
Numerical Study of Pressure Fluctuations Caused by Impeller-Diffuser Interaction in A Diffuser Pump Stage
Shi
Research Lecturer,
Satellite Venture Business Lab.
H. Tsukamoto
Professor,
Department of Biological Functions
and Engineering,
Graduate School of Life Science
and Systems Engineering
Kyushu Institute of Technology
Kitakyushu-shi, 804-8550, Japan
Introduction
In a diffuser pump, the centrifugal impeller interferes with its
successive diffuser vanes and produces pressure fluctuations
downstream of the impeller. In the case of a small radial gap
between the impeller and diffuser vanes, the magnitude of these
pressure fluctuations may become as large as the total pressure
rise across the pump Arndt et al., 1,2; and Tsukamoto et al.
3. These fluctuations not only generate noise and vibration that
cause unacceptable levels of stress and reduce component life due
to fatigue, but also introduce unfavorable characteristics of pump
performance even at or near the design point. Therefore, there is a
need to understand the sources of unsteadiness i.e., potential and
wake interactions to control the pressure fluctuations and to improve the overall pump performance and reliability.
Dring et al. 4 indicated two distinct mechanisms of rotorstator interaction: i wake interactions and ii potential interactions. Potential interactions are presently fully understood as potential flow effects induced by inviscid interaction due to the
relative motion between rotor blades and stator vanes. Wake interaction is far more complicated than potential interaction. It
originates from the impingement and convection of wakes shed
from the impeller passages and moving through the successive
diffuser passages. Rotor-stator interactions have been studied extensively in axial-type gas turbomachines Gallus et al., 5;
Fleeter et al., 6; Rai, 7; Giles et al., 8; Lewis et al., 9; Rao
et al., 10; Sharma et al., 11; Ho et al., 12; Valkov et al., 13;
Arnone et al., 14; Chung et al., 15. However, only a few studies have focused on impeller-diffuser interaction in hydraulic machines, especially in centrifugal pumps. Some experimental contributions to wake interactions in pumps may be attributed to the
extensive PIV measurements by Dong et al. 16, and Akin et al.
17, and hot wire measurement by Ubaldi et al. 18. However, it
is difficult to understand the flow phenomena due to impellerdiffuser interaction from only experimental studies because of the
complicated flow structures in centrifugal pumps. A 2-D unsteady
flow calculation was presented by Fortes-Patella et al. 19, who
Contributed by the Fluids Engineering Division for publication in the JOURNAL
OF FLUIDS ENGINEERING. Manuscript received by the Bioengineering Division October 21, 1998; revised manuscript received April 12, 2001. Associate Editor:
B. Schiavello.
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
vent the decrease of natural frequency in the pressure measurement systems. Figures 1(b) and 1(c) illustrate the unsteady
pressure measurement stations on the shroud casing side of the
diffuser in the test pump. Because of the limited space in the
measuring sections the pressure taps for tangential traverse were
located only at one radial location in each passage of the diffuser.
Diffuser
Volute casing
Number of vanes Z d 8
Volute width b v 76.9 mm
Inlet diameter D 3 258 mm
Base circle diameter478.8 mm
Outlet diameter D 4 325 mm
Inlet and outlet width
b 3 b 4 45.8 mm
Blade inlet angle 3 16.97 deg
Blade outlet angle 4 15.44 deg
Flow rate Q 0 6.21 m3/min. ( 0 Q 0 / D 2 b 2 U 2 0.117)
Total head rise H29.2 m ( 0 gH/U 22 0.393)
Rotational speed N2066 rpm
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 2 Computational grid. a 2-D computational grid 41,800 cells; b 3-D computational grid 338,460 csells
and volute casing. This allowed portions of mesh in the rotary part
of the impeller to slide in small incremental moments relative to
those of the stationary part during transient simulations.
Boundary Conditions. Since the computational domain incorporated the entire stage of the diffuser pump, the boundary
conditions were easily specified at the inlet and outlet of the
pump. Corresponding to the specified flow rate, a steady uniform
radial velocity for 2-D calculations or axial inflow velocity distribution for 3-D calculations was given at the inlet boundary and
the inlet boundary was located as far from the impeller blades
leading edge, as possible. The outlet boundary was imposed at the
end port of the volute casing, and simple extrapolation boundary
conditions were employed. Further changes in locations of inlet
and outlet boundaries did not show differences in predicted static
pressure fluctuations. Nonslip wall boundary conditions were
specified for the impeller blades surface, diffuser vanes surface
and volute casing wall. The so-called wall functions were employed within the sub-layer adjacent to all the surfaces. Also, the
boundary values of turbulence energy and turbulence dissipation
rate were carefully set to make the solution stable and accurate.
Zero gradients of k and are specified at the outlet, while the
following relationships are selected for k and at inlet,
k 1 0.005 U 21 V 21 W 21 and 1 k 1.5
1 /0.01D 2
where U 1 , V 1 , and W 1 are inlet mean velocity components, and
D 2 the outlet diameter of the pump impeller. Variations in these
parameters showed little effect on the pressure fluctuation in the
diffuser. The sliding interface between the impeller and diffuser
was treated as an additional fluid zonal boundary in transient calculations with sliding meshes, and was updated implicitly after the
interior of computational domain has been updated.
Convergence Judgment and Numerical Error Control. All
the computations were performed on a computer VT-Alpha
500AXP 500 MHz Alpha processor, 512 MB RAM. The time
step, which is related to rotational speed and angular displacement
of the impeller between two successive computations, was set to
468 Vol. 123, SEPTEMBER 2001
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 3 Convergence history of time accurate computation during five rotations. a velocity fluctuation; b pressure fluctuation
present 2-D calculations, the numerical deviations from the experimental data can be attributed to physical modeling errors: 1
2-D and 3-D difference; 2 difference between CFD pump modeling and real pump modeling; 3 pump geometry deviation; and
4 turbulence models. While for the present 3-D simulations, the
discretization error due to inadequate grid resolution is a major
concern.
Unsteady Pressure in Diffuser Vane Passage. Figure 6 indicates the time histories of the unsteady part of the instantaneous
pressure coefficient on a representative pressure tap (r 1 ,c 1 ) D positioned close to the diffuser vane leading edge on the suction side
see Fig. 1(b). The relative position between impeller I-1 and
diffuser vane D-1 at t * 0.0 is shown in Fig. 1(c). The wave
form of unsteady pressure predicted by CFD shows better agreement with the experimental one than a wave form calculated by
the singularity method which is comprised of only low frequency
components. The poor waveform predicted by the singularity
method was due to exclusion of the blade thickness and viscous
wake-diffuser interaction from the calculation. As shown in Fig.
7, the incorporation of the wake vortices from the impeller passages into the singularity method with some assumptions can improve, to some degree, the unsteady static pressure in terms of the
fluctuation wave form Qin 26. The valley of instantaneous
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 7 Unsteady pressure at r 1 , c 1 predicted by Singularity method accounting for the viscous effects
pressure modeled by CFD is lower than observed in the experimental data or predicted by the singularity method but follows the
general pattern.
Figure 8 shows the blade-to-blade distributions of the peak-topeak values of unsteady pressure coefficient in the test diffuser
shown in Fig. 1(c). As compared to the results from the singularity method and the experimental data, the present CFD approach
is capable of predicting global trends in the pressure variations,
which are found to be larger on the suction side of diffuser vane
than on the pressure side at a given radius, and that unsteady
pressure decreases with increasing radius. The absolute magnitude
of the unsteady pressure coefficient from CFD simulation with the
two-equation model is larger than the experimental value. At the
current stage, nevertheless, the general features of the pressure
variations can be captured with the Navier-Stokes Solver. The
results obtained are in agreement with the available experimental
data in spite of the deviation in the magnitude of pressure fluctuations. Figure 9 shows both the measured and the CFD predicted
unsteady pressures in frequency domain at the location of
(r 1 ,c 1 ) D in the diffuser Fig. 1(b)). The frequencies are extracted
from the time domain with an external FFT program. The present
simulations demonstrate that the pressure in the diffuser passage
fluctuates with the impeller blade passing frequency Z i N and its
high harmonics, which cannot be accounted for in the singularity
method without viscous wake interactions. In conclusion, the
present whole stage calculations can predict frequency components of unsteady static pressures well, and provide fair approximations of pressure amplitudes.
Contour of Instantaneous Static Pressure 2-D CFD Results. Figure 10(a) and (b) shows the contour of instantaneous
static pressure in the stage of the test pump at two instants corresponding to P1 and P3 in Fig. 11 that indicates four pressure
peaks occurring at the pressure tap (r 1 ,c 3 ) D within one pitch
angular displacement of the impeller. At t * 0, when impeller
blade I-1 approaches closest to the leading edge of the diffuser
vanes D-1, static pressure at (r 1 ,c 3 ) D reaches the highest peak P1.
This pressure peak is due to potential interactions between impeller blades and diffuser vanes as indicated by the potential flow
analysis Qin and Tsukamoto, 22. When the trailing edge of the
impeller blade I-1 is passing by the pressure tap (r 1 ,c 3 ) D , the
pressure at (r 1 ,c 3 ) D is responding to another peak P3. This is
attributed by the wake interference from impeller blade I-1 as
described later.
Contour of Turbulent Kinetic Energy 2-D CFD Results.
Figure 12(a) and (b) shows the CFD calculated contour of turbulent kinetic energy in the stage of the studied pump at two
instants. The impeller wake with high level turbulent kinetic en470 Vol. 123, SEPTEMBER 2001
Fig. 8 Contour map of magnitude of p - p of pressure fluctuation in vaned diffuser passage at rated condition; experimental uncertainty in p - p 7.1 percent. a Measured; b calculated by singularity method; c CFD predicted
ergy moves out of the impeller discharge e.g., I-1 in Fig. 12(a),
and is chopped by diffuser vane when it approaches the leading
edge of the diffuser vanes. Then the wakes decay gradually in the
diffuser passage becoming insignificant after passing through the
diffuser passage, as can be seen in Fig. 12. The pressure peak P2
in Fig. 11 happens at the moment when the viscous wake shed
from blade I-1 passes by the location (r 1 ,c 3 ) D , as shown in Fig.
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 10 2-D CFD predicted static pressure at two instants. a t * 0.0; b t * 0.41378
12(b). In other words, the viscous wake shed from impeller passage can be interpreted as a kind of phantom blade causing
additional impeller-diffuser interactions.
Vorticity Contours 2-D CFD Results. Figures 13(a) and
(b) show the present CFD predicted contour of vorticity in the
stage of the test pump at two different instants. The wakes shed
from the impeller blade trailing edge traverse with the main flow
Fig. 12 2-D CFD predicted turbulent kinetic energy at two instants. a t * 0.0; b t * 0.13788
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Fig. 13 2-D CFD predicted contour of vorticity at two instants. a t * 0.41378; b t * 0.72418
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
1 Sources contributing to the pressure fluctuations in the diffuser passage of the present diffuser pump stage have been captured with a 2-D RANS solver: potential interactions cause the
highest peak of pressure in the diffuser; viscous wakes shed from
the impeller interfere with the successive diffuser vanes and result
in the presence of additional high pressure peaks.
2 The frequency components of the pressure fluctuations in the
diffuser passage are comprised mainly of the impeller blade passing frequency Z i N and its higher harmonics of 2Z i N and 3Z i N.
This indicates that the impeller-diffuser interaction is caused
chiefly by potential interaction and wake impingement with the
diffuser vanes.
3 The jet-wake flow structure at impeller discharge affects
the wake-diffuser interaction, but it is relatively small compared
with stronger viscous wake interactions in the present pump.
Further work is being focused on the numerical solutions on
real pump models including the leakage flow, which is believed to
be important factors for impeller-diffuser simulations in pumps.
Acknowledgments
Fig. 16 Static pressure contour at the impeller discharge
based on unsteady 3-D calculation. a t * 0.0; b t * 0.7758
Conclusions
A 2-D and 3-D unsteady incompressible Reynolds-averaged
Navier-Stokes equations with two equation turbulence models
was computed for the whole stage of a diffuser pump without
approximating the count ratios of rotor to stator. The calculated
results provided insights toward a better understanding of complicated unsteady rotor-stator interaction phenomena in a diffuser
pump. As a result of the comparison of the CFD predictions with
singularity calculations and experimental data, the following conclusions are derived:
Journal of Fluids Engineering
Nomenclature
c
D
k
f
N
PS
Ps
p
p
p
p*
R,r
time average p
unsteady component of p
relative pressurepPs
radius
SEPTEMBER 2001, Vol. 123 473
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
SS
T
Ti
t
t*
U,V,W
U2
Z
p- p
suction side
period of one revolution of impeller
time required to traverse one pitch of impeller blade
time
non-dimensional timet/T i
x-, y- and z- velocities in Cartesian coordinates
peripheral speed of impeller
number of blades
non-dimensional unsteady pressurep
* / U 22 /2
peak-to-peak magnitude of
dissipation rate of turbulence energy
density
instantaneous pressure coefficient(pPs )/( U 22 /2)
vorticity v / x u/ y
Subscripts
A,D,G
d
i
1, 2
3, 4
References
1 Arndt, N., Acosta, A J., Brennen, C. E., and Caughey, T. K., 1989, Rotor/
Stator Interaction in a Diffuser Pump, ASME J. Turbomach., 111, pp. 213
221.
2 Arndt, N., Acosta, A J., Brennen, C. E., and Caughey, T. K., 1990, Experimental Investigation of Rotor/Stator Interaction in a Centrifugal Pump with
Several Vaned Diffusers, ASME J. Turbomach., 111, pp. 213221.
3 Tsukamoto, H., Uno, M., Hamafuku, H., and Okamura. T., 1995, Pressure
Fluctuation Downstream of a Diffuser Pump Impeller, The 2nd Joint ASME/
JSME Fluids Engineering Conference, Forum of Unsteady Flow, FED-Vol.
216, pp. 133138.
4 Dring, R. P., Joslyn, H. D., Hardwin, L. W., and Wagner, J. H., 1982 Turbine
Rotor-Stator Interaction, ASME J. Eng. Power, 104, pp. 729742.
5 Gallus, H. E., 1979, Unsteady Flows in Turbines, VKI Lecture Series LS1979-3, Vol. 2, Von Karman Institute.
6 Fleeter, S., Jay, R. L., and Bennett, W. A., 1981, Wake Induced TimeVariant Fluids Aerodynamics Including Rotor-Stator Axial Spacing Effects,
ASME J. Fluids Eng., 103, pp. 5966.
7 Rai, M. M., 1987, Unsteady Three-Dimensional Navier-Stokes Simulations
of Turbine Rotor Stator Interactions, AIAA Paper No. 872058.
8 Giles, M. B., 1988, Calculation of Unsteady Wake/Rotor Interaction, J.
Propul. Power, 4, No. 4, pp. 356362.
9 Lewis, J. P., Delaney, R. A., and Hall, E. J., 1989, Numerical Prediction of
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
Downloaded 10 Jul 2010 to 129.173.72.87. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm