60° Congresso Nazionale ATI, 13 –15 settembre, 2005, Roma

ART. 12/01
MULTISTAGE CENTRIFUGAL-PUMPS: ASSESSMENT OF
A MIXING PLANE METHOD FOR CFD ANALYSIS
P.Adamia - S.Della Gattaa - F.Martellia
L.Bertolazzib – D. Maestrib - G.Marencob - A.Pivab
a
Dipartimento di Energetica “S.Stecco”
Via S.Marta, 3 50139 Firenze, Italia
adami@ing.unifi.it, s.dellagatta@ing.unifi.it, martelli@ing.unifi.it
b
Weir Gabbioneta Srl
V.le Casiraghi, 68 20099 Sesto S.Giovanni-MI, Italia
laura.bertolazzi@weirgabbioneta.com , daniele.maestri@weirgabbioneta.com ,
giacomo.marenco@weirgabbioneta.com , alberto.piva@weirgabbioneta.com

Abstract

An accurate CFD investigation allows a thorough knowledge of the fluid dynamic field
inside a multistage centrifugal pump but only a correct interpretation of its results grants a
well-founded prediction of the machine performance curves. The paper introduces the
assessment of an improved methodology to predict head-flow and efficiency characteristics
of industrial pumps in real operating conditions. The described approach is based on
standard CFD analysis and experimental tests.
The numerical procedure is built-up of three different steps:
· CFD investigation of the single impeller and diffuser domains;
· impeller-diffuser interaction analysis;
· evaluation of the whole multistage pump performance curves by pressure field
interpretation and taking into account real machine secondary phenomena.
The stage computational model is made up of two 3D fluid domains with overlapping: one
of them is aimed to simulate the diffuser and is related to a stationary reference frame
while the other one models the impeller and is referred to a rotating system.
Moreover the impeller–diffuser interaction is studied by the employment of a mixing plane
technique. Numerical results are then compared to the real pump performance test at the
actual working conditions in order to validate the proposed methodology.
The paper will discuss all approach details and the main CFD features involved.

Symbols
w relative velocity
BEP best efficiency point u blade velocity
Q flow rate v flow velocity in seals
Qp parasitic flow rate N rotational speed
H head a,b flow angles
P power p pressure
h efficiency r density
c absolute velocity g gravity acceleration

1
R radius Subscript
A passage area RAD radial
L passage length AX axial
D diameter TG tangential
d seal diametral clearance min minimum
Hm meridional passage height max maximum
Jper periodicity angle exit exit
e roughness inlet inlet
a,b,c,k ,Cd,l leakage correlation
coefficients
zf ventilation coefficient

1. Introduction
A reliable and efficient methodology for accurate prediction of multi-stage centrifugal
pump performances has to handle stationary as well as rotating parts including secondary
effects and geometrical features such as leakages, disk friction and balancing ducts.
Experience gained from the collaboration between the Energetic Dept. and Pompe
Gabbioneta demonstrated how a mixing-plane stage analysis with domain overlapping
leads to an accurate estimate of the head-flow characteristic. Conversely, the efficiency
prediction and the correct best efficiency point flow rate evaluation (BEP) require a deeper
investigation including, as much as possible, all the loss sources experienced by the
pumped fluid. In a real centrifugal pump there are so many effects contributing to the
production of losses that an engineering feasible CFD prediction of the real working
conditions imposes the assessment of different strategies and models.
For instance balancing holes, frequently used to reduce axial load on bearings, are
responsible of a leakage flow through the impeller which does not contribute to pump head
and produces a loss of efficiency. Moreover, leakages flows and ventilation losses play
also an important role in the prediction of efficiency as well as the surface wall roughness.
According to these points, the paper describes a CFD multi-stage investigation
methodology aiming to include all the effects that have demonstrated not to be negligible
for the performance prediction. The assessed and tuned method is applied to two industrial
pump stage configurations and the numerical results are compared with experimental data
acquired by Pompe Gabbioneta thus demonstrating its effectiveness and reliability.

2. Pump geometry and working conditions

Two different machines have been selected among Pompe Gabbioneta VBN line which
includes vertical multistage pumps fully designed according to latest edition of API 610
Std [1]. VBN are applied in heavy duty process services and are characterized by closed
type impellers equipped with front and rear wear rings as well as balancing holes to
minimise the axial thrust. VBN line covers flow rate working conditions up to 1500 m3/h
with heads up to 2000 m.
A 4 stages VBN 200/50 B for low flow rates (BEP at 25 m3/h) and a 6 stages VBN 250/80
B for greater flow rates (BEP at 100 m3/h) have been chosen as reference pumps for the
present study. Both are equipped with 7 blades impeller with different diffuser
configurations related to the different range of working conditions (Tab. 1).

2
 Fig. 1 VBN pump configuration

BEP
STAGE BEP ROTATING
IMPELLER DIFFUSER FLOW
NUMBER HEAD SPEED
RATE
VBN Vaneless with
200/50 7 blades 12 return 4 25 m3/h 140 m 2960 rpm
B splitter
VBN
9 continuous
250/80 7 blades 6 100 m3/h 420 m 2980 rpm
channels
B
Tab. 1 Main characteristics of the investigated configurations

3. Experimental set-up
Experimental data to validate computational results come from functional tests which
have been carried out on a Pompe Gabbioneta automatic test loop which is fully in
compliance with ASME PTC 8.2-1990. VBN 200/50 B and VBN 250/80 B involved in the
present study have been equipped with mechanical seals and tested in accordance to latest
edition of API 610 Std [1].
All the operating parameters have been continuously measured by electronic
instrumentation connected to a control room PC. In detail:
· the capacity has been measured by electromagnetic flowmeters and its value has been
transmitted to the acquisition system by an analogical electric signal and read directly
on relevant monitors and in a PC testing room;
· speed has been measured by a transducer which can send a train of electric pulses to an
amplifier with frequency proportional to pump shaft speed;
· pump head has been determined by algebraic difference between total height of liquid
at discharge and suction sections: the total height value being given as the algebraic
addition of geodetic height, static height and kinetic height. Static heights are measured
by means of different pressure transducers. For vertical pumps in barrel, such as the
above mentioned VBN, the low pressure connection of differential pressure gauge
takes the air pressure inside the vertical vessel, the measure of air pressure is not
affected by geodetic variations and the kinetic head is only composed by the term
referred to discharge;
· pump shaft torque has been measured by an estensimetric torque-meter connected to a
signal amplifier. the pump absorbed power is evaluated as the product of torque and
speed;

3
· pump efficiency has been calculated as the ratio of the hydraulic power supplied by the
pump to the liquid and the mechanical power absorbed by the pump.

4. Numerical Investigation Features

4.1 The CFD solver

3D steady simulations solving RANS (Reynolds Averaged Navier Stokes) equations in the
fluid domain are carried out. A finite-volume pressure correction procedure for
incompressible flow is employed. A standard two-equation k-e turbulence model is applied
together with standard wall function which are able to model the boundary layer.

4.2 Stage coupling approach

To deal with the whole stage simulation, including stationary and rotating part, an interface
is usually defined between them. Unsteady simulation, frozen rotor and mixing plane are
the most typical approaches to model rotor – stator interaction [8].
The frozen rotor method assumes a steady relative flow in both rotor and stator and
requires the simulation of the entire geometry of diffuser and impeller in a fixed (“frozen”)
relative position. The interaction between rotor and stator is calculated by a steady
coupling, where the interface variables are transformed according to the relative motion
due to the rotating speed. According to this assumption, the approach is able to consider
partly for the non-uniformities and distortions between the impeller and diffuser because
the aerodynamic interaction and the local flows are effectively computed across the
interface. However, the unsteadiness remains unresolved while the physical relevance of
the stator-rotor interaction is limited to the fixed relative position assumed for the diffuser
and impeller. The main advantage of the frozen rotor approach is represented by the
reduction of computing time with respect to a full unsteady simulation.
Differently from the above technique, the mixing plane method circumferentially averages
the fluxes at interface before changing the reference frame. The axial symmetric flow
condition at the impeller-diffuser interface constitutes the main assumption of the method.
Therefore the mixing plane approach needs only one impeller and one diffuser passage to
be modelled with periodic boundary conditions. This results in a consistent reduction of
memory and CPU costs and removes the assumption of an unphysical fixed position
between the rotor and the diffuser.
However, despite the simplifications inherently assumed, the solution is often a reasonable
approximation of the time-averaged flow field.
As a drawback, the mixing-plane does not account for unsteady interaction effects and the
correct boundary conditions at interface have to be found by an iterative process. In case of
small interstage gaps, the placement of the mixing surface represents a further critical issue
because, at the mixing plane, the boundary flow condition is transferred using pitch
averaged distributions (in some applications even single surface averages) thus reducing
the physical 3D connotation of the numerical solution on the adjacent blade or walls that
are facing close to this boundary.
Both the frozen rotor and the mixing plane model assume steady conditions in the impeller
as well as in the diffuser thus neglecting the transient interaction effects. The assumption is
justified in all the cases where the rotor-stator interaction is weak or an approximate
solution for the system is desired. Differently, a fully coupled unsteady simulation may be
used in order to account for the actual geometry and the aerodynamic impeller-diffuser
interaction.

4
4.3 Computational domains and boundary conditions
As above mentioned, considering the axial symmetry of the problem, a mixing plane
method has to be preferred in the present study. The stage is divided into two sub-domains.
Only one of all the diffuser passages (twelve for the first pump, nine for the second one)
has to be meshed and analysed. Accordingly, only one of the seven impeller passages is
modelled, assuming an artificial domain overlapping. Fig. 2 shows the VBN 200/50 B
impeller and diffuser computational grids.

Fig. 2 Impeller (a) and diffuser (b) spatial grids for VBN200/50 B configuration

At the inlet and exit of a stage (i.e. at impeller inlet and diffuser outlet) the flow absolute
velocity and flow angle are assumed to be identical. This hypothesis allows to consider the
operating conditions of a single stage of a multistage centrifugal-pump. The inlet mass
flow, the components of the inlet vector velocity and the pressure outlet are imposed as
boundary conditions to each component.

5. Computational steps and mixing plane approach

The interstage pressure and velocity direction are defined by means of an iterative
procedure as reported in Fig. 3.

-Outlet direction velocity

IMPELLER DIFFUSER

-Interstage pressure
-Outlet direction velocity

Fig. 3 Iterative procedure of simulation

5
The steps of the procedure can be summarised as follows:
· CFD simulation of the impeller related to a rotating reference frame.
· Data analysis of the impeller flow field – Circumferential averages are computed at the
mixing plane interface to be imposed as boundary condition to the diffuser
· CFD simulation of the diffuser in the static reference frame
· Data analysis of the diffuser flow field – Circumferential averages are computed at the
mixing plane interface to be passed as boundary condition to the impeller
· Iteration from 1. to 4. until convergence is achieved.

5.1 CFD simulation of impeller

The CFD simulation of impeller is carried out in the rotating reference frame, imposing as
boundary conditions the inlet mass flow, the direction vector and the outlet static pressure.
The pitch-wise averaged distributions of the direction vector and of the outlet pressure are
retrieved from the diffuser field.

5.2 Data analysis of the impeller flow field

Data which have to be imposed as boundary condition in the subsequent simulation of the
diffuser are extrapolated from the impeller flow field. Velocity direction is taken out at the
impeller mixing surface and then applied to the diffuser inlet surface (i.e. the above
mentioned mixing plane approach).
A flow weighted tangential average is calculated to define the pitch and yaw angles:
Q Hm J PER Q Hm J PER

ò tga dq ò ò tga w RAD R dh dJ

~ ò tgb dq ò ò tgb w RAD R dh dJ
(1)
tga~ = 0
Q
= 0 0
tgb = 0
Q
= 0 0
Q Q
ò dq
0
ò dq
where wRAD , w AX and wTG are respectively 0 the radial, axial and tangential
velocity in the rotating frame.
w w
tga = AX and tgb = TG
wRAD wRAD

where Q is the flow rate, R is the constant radius of impeller mixing surface (Fig. 2), Hm is
the height of meridional passage at impeller exit, and finally θPER is the periodicity angle
of the impeller.
The outlet velocity from the analysis of the impeller is imposed at the diffuser inlet after
being transferred from the rotating reference frame to the stationary one.

5.3 CFD simulation of diffuser

The CFD simulation of diffuser is carried out in an absolute reference frame. The inlet
mass flow, the direction vector and the outlet static pressure obtained from the impeller
field are imposed as boundary conditions.

5.4 Data analysis of the diffuser flow field

Results from the diffuser simulation are employed to adjust the inlet condition for the next
analysis of the impeller domain. The actual interstage velocity direction, flow and static
pressure are iteratively computed reporting the averaged distributions of angle and static
pressure at the diffuser outlet mixing plane as new inlet condition to the impeller itself. The
iterative procedure of coupling the impeller and the diffuser implies a physical assumption
of kinematic repeatability of the stages in the real machine.

6
The diffuser outlet mixing plane is indicated in Fig.2. Here, as for the previous step, the
flow weighted averaged angles are computed:
Q Rmax JPER Q Rmax JPER

ò tga dq ò ò tga c RAD r dJ dr ~

ò tgb dq ò ò tgb c RAD r dJ dr
tga~ =
Rmin 0 Rmin 0
0
Q
= tgb = 0
Q
= (2)
Q Q
ò dq
0
ò dq
0

where Rmin e Rmax are respectively the minimum and maximum radius of the diffuser
channel height at the intersection with the mixing plane (Fig. 2); c RAD , c AX and cTG are
respectively the radial, axial and tangential velocities in the stationary frame defining the
c c
pitch and yaw angles: tga = RAD and tgb = TG .
c AX c AX
The impeller inlet static pressure is:
Q Hm JPER

~
ò p dq ò ò p c RAD R dh dJ
p= 0
Q
= 0 0
(3)
Q
ò dq
0
where R is the radius of the diffuser inlet surface and Hm the height of the passage.

5.5 Convergence criteria

Steps from 1. to 4. are repeated until the convergence is achieved. Three convergence
criteria are defined to monitor the variations of:
· impeller inlet velocity module
· diffuser inlet velocity module
· interstage (i.e. impeller outlet and diffuser inlet) static pressure
The procedure is arrested when the integral error computed over the above mentioned
planes for each variable is lower than a prescribed threshold.

6. Results and discussion

In a first example, the approach is applied to characterise the VBN200/50 pump with
vaneless diffusers. The hydraulic efficiency is calculated according to the following
definition
gH
h= (4)
(ucTG )exit - (ucTG )inlet

where g is the gravity acceleration, H is the stage head, u is the blade speed and cTG the
tangential velocity in the stationary frame. As for mixing plane, the global values used in
this expression are flow weighted at each location (here the impeller exit and inlet plane).
Head-rate and efficiency-rate curves obtained from the simulation loop for both
configurations are reported in Fig. 4 in comparison with experimental data obtained from
Pompe Gabbioneta. The numerical simulation correctly predicts both the trends and the
BEP flow rate, but fails to capture the head and efficiency in the whole operating range.
A critical analysis of this outcome leads to conclude that additional losses and a more
realistic model of the stage details are needed to achieve a satisfactory prediction.
Therefore the several secondary effects existing in the real centrifugal pump have not to be
neglected.

7
 2.00 2,00

1.80 1,80

exp CFD 1,60

head/head(BEP)
1.60
exp CFD
1,40
1.40

h/h BEP
1,20
1.20
1,00
1.00
0,80
0.80
0,60
0.60
0,40
0.40 0,50 0,60 0,70 0,80 0,90 1,00 1,10
0.50 0.60 0.70 0.80 0.90 1.00 1.10
Q/Q(BEP) Q/QBEP

Fig. 4 Non-dimensional head and non-dimensional efficiency vs non-dimensional flow rate

(experimental and numerical data). First simulation of the VBN 200/50 B pump.

First of all, friction losses are influenced by the non-smooth aspect of the metallic surfaces
of the pump due to the impeller and diffuser casting process. Considering the large extent
of the wet surface in these components, the influence of wall roughness in the boundary
layer model has been considered as reported in the literature [3].
As far as the turbulence model is concerned, being an a priori estimate of the coefficients
generally not available, a basic numerical tests campaign has to be performed to asses and
tune the values coherently with the observed surfaces roughness (e=1.6mm).
Even the leakage flow through the balancing holes is responsible for losses and it
contributes to a decrease in performance of the whole centrifugal-pump. Firstly the
leakage, the flow and the ducts itself disturb the average stream inside and at the discharge
of the impeller vanes. Further, even if pumped, the leakage flows does not provide any
useful pressure increase and it results in a loss of energy.

Fixed Rotating
surface surface

Fig. 5 Balancing hole, wear rings layout and corresponding spatial discretisation

Another source of loss is represented by the viscous action of the fluid filling the gap
between the impeller shrouds and the diffuser wall where the frictional losses between the
moving surfaces dissipate energy through ventilation. The evaluation of the balancing flow
rate and of the ventilation losses is here addressed to understand their actual importance in
predicting the efficiency and head of the whole stage. In Fig. 5, the balancing hole, the
wear rings and the leakage flow path are reported.
A simple and efficient way to include those effects trying to avoid any further CFD step
in the iterative loop previously explained is presented. To this aim, it is observed that

8
leakage flow mainly depends on the seals geometry, impeller rotational speed and head. In
case of fixed geometrical features and constant rotating speed, balancing flows and losses
therefore depend on the impeller head only.
As a consequence the characteristic curve ( Dp-Q) of the seals can be obtained by separate
CFD calculation at different mass flow inlet conditions. In the present study, this
preliminary computations are performed reproducing both the rotating and stationary
surfaces, while the computational domain extends to cover the whole leakage gap existing
between the impeller and the diffuser wear rings. Thanks to the details included, the
ventilation losses due to the presence of the rotating impeller are also accounted by the
same simulations. Both curves obtained for the balancing flow rate and ventilation losses
are reported in Fig.6.
0.14 1.050

vent loss/ vent loss (BEP)

balancing flow rate/BEP flow

0.12 1.040

0.10 1.030

0.08
rate

1.020

0.06
1.010
0.04
1.000
0.02

0.990
0.00
0.00 0.50 1.00 1.50 2.00 2.50 0.11 0.12 0.13 0.14 0.15 0.16 0.17

impeller head/impeller head (at BEP) balancing flow rate/flow rate (BEP)

Fig. 6 Calculated characteristic curve of the balancing holes

The results from this analysis have been employed also to assess the correlations used by
Pompe Gabbioneta. Concerning the leakage correlation, it utilizes the geometrical
characteristics, rotational speed and impeller head as follows [13][14]:
Q p = k 2 gH C d A (5)
0.5
æ ö
ç 1 ÷
Cd = ç + a÷ (6)
ç L ÷
çl ÷
è d ø
c
b é æuö ù
2

l = 0.25 ê1 + 0.5 ç ÷ ú (7)

Re êë è v ø úû

Here Qp is the leakage flow rate, H the head, A the passage area, L the passage length, d the
seal diametral clearance, u the blade velocity, v the mean flow speed inside the seal.
Coefficients (a,b,c) depend on geometrical characteristics.
The coefficient k is equal to 1. considering H as the head between inlet and outlet of
balancing hole. As proved by the tests, a value of k equal to 0.49 allows to use directly the
impeller head value.
As far as ventilation losses are concerned, in literature [11][12] expressions to estimate
ventilation losses are reported as follows:

Pvent losses = rz f Q p v 2 ( 8)
2
æ gH ö æ Hm ö
z f = f çç ÷
2 ÷ç ÷ (9)
è ( ND ) øè D ø
9
where both Qp and zp depend on H. For a fixed geometry and a constant rotating speed, it
is possible to graph zp (H).

Fig. 7 Ventilation losses coefficient zp vs. non-dimensional head

It is observed that the leakage flow due to balancing holes can reach the 30 % of the
main flow at low operating rates with an appreciable effect on performances. Therefore the
characteristic curve computed separately for the leakage flow has been introduced in the
iterative loop of the stage that has been modified as follows in Fig. 8.

Boundary Condition Mixing plane Tangential average

Inlet flow rate
Inlet velocity direction Outlet velocity direction
Outlet static pressure Inlet static pressure
Balancing flow rate

IMPELLER DIFFUSER

Mixing plane average Boundary Condition

Outlet velocity direction Inlet velocity direction
Impeller head Inlet flow rate
Balancing flow rate

Balancing flow rate= f(Impeller head)

Balancing flow rate BALANCING HOLE Ventilation losses

DH

Fig. 8 Simulation iterative loop including balancing holes

A further inlet surface corresponding to balancing hole is introduced in the impeller

domain, while in the diffuser domain outlet surfaces are introduced to consider the
presence of flow passage trough wear rings (see Fig. 2 and Fig. 5). The leakage flow rate
has been imposed following the relations above presented.
The efficiency definition becomes:
gQH
h* = ( 10 )
(Q + Q p )[(ucTG )exit - (ucTG )inlet ] + Pvent

10
 with Qp the parasitic flow and Pvent the ventilation losses. Here exit and inlet are
respectively referred to the impeller outlet and the inlet control surfaces. The new
efficiency and characteristic curves are reported in Fig. 9.
2.00 2,00

1.80 1,80
exp CFD
1.60 1,60
head/head(BEP)

exp CFD
1.40 1,40

h/h BEP
1,20
1.20
1,00
1.00
0,80
0.80
0,60
0.60
0,40
0.40 0,50 0,60 0,70 0,80 0,90 1,00 1,10
0.50 0.60 0.70 0.80 0.90 1.00 1.10
Q/Q(BEP) Q/QBEP

Fig. 9 Non-dimensional head and non-dimensional efficiency vs non-dimensional flow rate.

Second calculation of the vaneless diffuser test case including balancing holes.

As shown, owing to the secondary effects, the agreement between computational results
and experimental data greatly improves. The head-flow rate curve is fairly well predicted
while only a small difference is observed for the efficiency where the CFD analysis still
slightly over predicts the real pump behaviour. However, the above mentioned
discrepancies can be brought back to the experimental campaign which has been
performed using multistage pumps instead of the modelled single stage. Likely,
computational analysis ignores the effects of pump suction and discharge nozzle and,
finally, the discrepancies grow at low flow rate where further mechanical losses of real
machine may acquire relevance.

Fig. 10 Diffuser velocity in blade to blade surface (a) and pressure field in meridional plane(b),
impeller static pressure distribution in impeller flow field (c) for VBN 200/50 B. (a) and (b) show a
separated flow in the suction side due to the high deviation from inlet to exit. The stall zone extends
in the whole passage area.

The whole approach has been further assessed computing the performances of the VBN
250/80 B pump that is characterised by a nine continuous channels diffuser. The
corresponding computational grids (Fig. 11), the performance curves (Fig. 12) and the flow
field features (Fig. 13) are given herein.

11
 Fig. 11 Spatial discretisation for VBN 250/80 B

1,4 1.4

head/head(BEP)
CFD EXP CFD EXP
1,2 1.2
h/hBEP

1 1

0,8 0.8

0,6 0.6

0,4 0.4
0,6 0,7 0,8 0,9 1 1,1 1,2 1,3 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
flow rate/flow rate (BEP) flow rate/flow rate (BEP)

Fig. 12 Non-dimensional head and non-dimensional efficiency vs non-dimensional flow rate for
VBN 250/80 B including the effect of balancing holes.

The agreement between experimental and computational data confirms the capability of
the numerical tool.

Fig. 13 Flow field feature for VBN 250/80 B diffuser (a) and impeller (b)

12
Conclusion

An improved CFD approach is presented for the analysis of complex pump stages
performances. The method consists in steady simulations which use the mixing plane
technique to couple impeller and diffuser. Many secondary effects such as the presence of
leakage flows, wall friction and ventilation losses are not to be neglected in order to reach
an high level of accuracy.
The computational procedure gives results that fit very well the experimental data thus
confirming its effectiveness. Furthermore, it is successfully applied to two centrifugal
pumps which are equipped with different diffuser configurations and are designed for
different operating conditions.
Finally a quick and reliable prediction of head-rate and efficiency-rate curves in a wide
operating condition range is obtainable with the present approach. However the additional
parasitic effects has to be more deeply investigated at low flow rates where greater is their
influence.

REFERENCES
1) API Standard 610 9th Edition, 2003, American Petroleum Institute
2) Schlichting, H., 1968, Boundary Layer Theory, 6th Edition, McGraw-Hill, New York.
3) Launder, B.E., Spalding, D.B., 1974, The numerical computation of turbulent flows,
Comp. Meth. in Appl. Mech. and Eng., 3, pp. 269-289.
4) van Driest, E.R., 1956, On turbulent flow near a wall, J. Aeronaut. Sciences, 23,
pp.1007–1011,1036.
5) Japikse, D., Marscher, W.D., Furst, R.B., 1997, Centrifugal pump design and
performance, Concepts ETI, Wilder, Vermont.
6) Ubaldi, M., Zunino, P., Barigozzi, G., Gattanei, A., 1996, An Experimental
Investigation of Stator Induced Unsteadiness on Centrifugal Impeller Outflow, ASME
J. Turbomach., 118, pp. 41–54.
7) Wuibaut, G., Bois, G., Dupont, P., Caignaert, G., Stanislas M., 2002, PIV
Measurements in the Impeller and the Vaneless Diffuser of a Radial Flow Pump in
Design and Off-Design Operating Conditions, Journal of Fluids Engineering, 124,
pp.791-797.
8) Tamm, A., Gugau, M., Stoffel, B., 2002 , Experimental and 3-D Numerical Analysias
of the Flow Field in Turbomachines Part I, QUANSE 2002, International Congress on
Quality Assessment of Numerical Simulations in Engineering, Concepción – Chile.A
9) Hagelstein, D., Hillewaert, K., Van den Braembussche, R. A., Engeda A., Keiper , R.,
Rautenberg, M., 2000, Experimental and Numerical Investigation of the Flow in a
Centrifugal Compressor Volute, ASME J. Turbomach., 122, pp. 23-31
10) Brennen, C.E., 1994, Hydrodynamics of pumps, Concepts ETI, Wilder, Vermont.
11) Traupel, W., 1958, Thermische Turbomaschinen, Sprinter-Verlag, Berlin
12) Utz, C., 1972, Experimentelle Untersuchung der Strmungveluste in einer
mehrstufigen Axialturbine, Mitt.Ins.Therm.Turbomasch. ETH Zurich, Nr.19
13) Denny, D.F., 1954, Leakage flow through centrifugal pump wear rings, TN460,Beds:
BHRA, Cranfield
14) Worster, R.C., Thorne, E.W., 1959, Measurement of lakage flow tough the wearing
rings of a centrifugal pump and its effect onoverall performance, Beds: RR 619,
Cranfield

13