Marjan Gantar

Turboinstitut,
Rovsnikova 7,
1210 Ljubljana,
Slovenia
e-mail: [email protected]

Dusan Florjancic
Brane Sirok
University of Ljubljana,
Faculty for Mechanical Engineering,
Askerceva c. 6,
1000 Ljubljana,
Slovenia

Hydraulic Axial Thrust in

Multistage PumpsOrigins and
Solutions
In this paper, axial thrust problems of multistage pumps are presented. The entire investigation has been focused on the pump design concept having all impellers in series
(facing in one direction) and is valid for barrel casing type as well as for segmental type.
The major part of the investigation has been experimental on one stage testing arrangement, using also Laser Doppler Anemometry (LDA) for determination of fluid rotation in
the impeller side chamber. Numerical flow analysis (NFA) has been applied for some flow
variants in order to verify whether a numerical approach could simulate the test results.
The phenomenon of fluid rotation in side chambers and its effect on impeller hydraulic
axial thrust have been determined for different leakage flow regimes. The influence of
increased wear ring radial clearance on axial thrust has been analyzed together with the
solutions for pump hydraulic axial thrust reduction. DOI: 10.1115/1.1454110
Keywords: Multistage Pump, Axial Thrust, Impeller Side Chamber, Fluid Rotation, Static
Pressure Distribution

Introduction

The investigation has been focused on the following:

High head multistage pumps play an important role in systems

where the need for high reliability is vital. For the proper and
economical design of axial thrust bearings, together with the hydraulic balancing devices, extensive knowledge of the origin of
hydraulic axial thrust is essential. The problems become even
more demanding when increased pump power concentration and
frequent off-design operation are required.
A typical multistage pump design with barrel casing is shown in
Fig. 1. Hydraulic axial thrust on the impellers is primarily balanced by the balancing piston, with the rest of the thrust loaded on
the axial bearing. The other possibility for balancing axial thrust
when all impellers are in series is with a balancing disk or a
combination piston/disk.
Many authors 1 4 have given a lot of thought to this subject
in the past. It is already known that the flow in the gaps between
impeller and casing wall influences the pressure distribution on
the impeller front and back shroud to a great extent. Flow in the
impeller side chambers is turbulent with separate boundary layers.
Within the flow in the peripheral direction, the core layer is
formed between the impeller shroud boundary layer and casing
wall boundary layer. This fluid rotates in the same direction as the
impeller but with reduced angular speed. Within the flow in the
radial direction, the impeller shroud boundary layer flow is radially outwards, in the casing wall boundary layer it is radially
inwards, and in the core layer there is no radial flow.
For the type of multistage pump under consideration, the flow
regimes in side chambers of normal impeller stages are different
from that in the last impeller stage Fig. 1. At all of the impeller
front shrouds, the direction of flow in the side chamber is radially
inwards. At the impeller back shrouds in normal stages, the flow
direction is radially outwards, while in the last stage it is inwards,
up to the piston gap entry. The leakage quantity is also increased
compared with normal stages. The quantity of flow in the impeller
side chambers changes with the pump operation regime, and
grows even more over time due to increased radial clearances in
wear rings and piston labyrinths.

to clarify in greater detail the flow field in the impeller side

chambers,
to investigate the possibilities for hydraulic axial thrust reduction without essential changes in the pump design and
efficiency,
to reduce the sensitivity of hydraulic axial thrust to increased
wear ring radial clearance.

Contributed by the Fluids Engineering Division for publication in the JOURNAL

OF FLUIDS ENGINEERING. Manuscript received by the Fluids Engineering Division
April 12, 2001; revised manuscript received September 12, 2001. Associate Editor:
Y. Tsujimoto.

336 Vol. 124, JUNE 2002

The major part of the investigation has been experimental, carried out on a test pump arrangement. Parallel with testing, a numerical flow analysis for impeller side chambers has been worked
out as well. The main contribution of the paper is to show the
possibility that with simple measures the fluid rotation in the impeller side chambers can be affected and consequently the hydraulic axial thrust can be reduced.

Experimental Investigation and Test Results

Description of Testing Arrangement. A series of tests was
carried out on the testing arrangement, which consisted of one
separate stage of a multistage pump. The bearing bracket was
designed to enable hydraulic axial thrust measurements using calibrated strain gauges. The dimensions of the impeller side chamber
can be altered and the flow in the side chamber can be changed in
direction and magnitude as well Fig. 2. The leakage fluid injection was performed by an auxiliary system, which consisted of a
circulating pump, valve combination, and flowmeter.
Tests were performed in the pump test rig in a closed system.
Pump flow rate was measured by a calibrated orifice plate, and all
pressures by absolute or differential pressure transducers. The casing wall of the impeller side chamber was built of a transparent
material in order to enable measurements of velocity distribution
by the LDA method Laser Doppler Anemometry. The device has
an argon-ion laser with an exit power of 4W and Braggs cells
with carrier frequency of 40 MHz. A separate lens system connected by a fiber optic cable was used in order to ensure simple
and accurate positioning of laser beam cross-sections in the axial
and radial directions. The measuring volume was 0.1 mm in diameter and 1.0 mm in length.
The basic geometrical parameters of the tested pump can be

Copyright 2002 by ASME

Transactions of the ASME

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 1 Multistage pump with barrel casing

Fig. 4 Dimensionless pump characteristics

seen in Fig. 3. Radial clearances A and B between impeller and

diffuser and overlapping U e have been designed according to
EPRI recommendations 5.
The rotational speed of the tested pump was on a level which
assures turbulent flow with separate boundary layers in the impel-

Fig. 2 Testing arrangement, pump specific speed nq 22 nsUSA

1135

Fig. 3 Geometry of tested pump

Journal of Fluids Engineering

ler side chamber (Re3.8.106 ). The definition of Re number is,

due to possible comparison, the same as that used by Daily and
Nece 1 and Senoo and Hayami 6.
The specific speed of the tested pump is nq 22 nsUSA 1135,
and the dimensionless characteristics are shown in Fig. 4. Measuring inaccuracies of experimental work were the following:
pump flow rate 0.5%, leakage flow rate 0.6%, pump head and
static pressures 0.3%, pump efficiency 0.7%, velocities in side
chamber 2%, hydraulic axial thrust 4%.
Test ResultsStatic Pressure Distributions. Static pressures were measured on the casing wall at 6 different radial positions. This gives a sufficient amount of data for the correct determination of the pressure distribution in the entire impeller side
chamber. The most important parameter, which affects the static
pressure distribution, is the flow in the impeller side chamber. In
Fig. 5 the pressure distributions are shown for the pump flow rate
Q opt and impeller central position relative to the diffuser. It can be
seen that the pressure distribution is greatly influenced by the
quantity and direction of flow in the side chamber through the
complete impeller periphery. The influence of q L is the lowest at
the outer impeller side chamber perimeter and increases in the
radial direction toward the impeller wear ring. Measured pressure
distributions, especially at leakage flow regimes q L , differ
greatly from the pressure distribution, presented by Stepanoff 7,
which has been taken in the past as the basis for hydraulic axial
thrust determination.

Fig. 5 Static pressure distributions for different leakage flow

conditions

JUNE 2002, Vol. 124 337

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 6 Static pressure distributions for different pump flow

conditions

As the static pressure at the impeller exit is changed, according

to the pump operating regime, pressure levels and internal flows
in the impeller side chambers Fig. 6 are also influenced. An
impact from eventual flow recirculation in the area of the impeller
exit/diffuser inlet was not observed in the impeller side chamber.
This can be explained by the fact that the main impeller passages
and side chambers are sufficiently separated by throttling via
small radial clearance A and overlapping U e . This also accounts
for the fact that at part flows, no significant instability can be
observed in the measured hydraulic axial thrust. Similar conclusions can also be found in Makay and Barrett 8. Additional test
results about the influence of gap A and overlapping U e on pump
characteristics are presented in Gantar et al. 9.
Test ResultsVelocity Distributions. The velocity distribution in the impeller side chamber has been measured by a twocomponent LDA measuring system. The method used is contact
free and does not disturb the flow field in the measured area.
Peripheral and radial velocity components were measured in the
impeller side chamber at six radial positions. At each radial position, 1015 measuring points have been taken through the side
chamber width. The main goal of this measurement is to determine the peripheral velocity component of the core layer in the
impeller side chamber and its distribution in the radial direction.
In Fig. 7 the peripheral velocity component is shown for different
relative leakage flows q L . When the leakage flow direction is
radially outwards, the rotation of the core layer is lower and opposite to that of the inward leakage flow direction. This result

Fig. 7 Measured peripheral velocity component in the middle

of impeller side chamber width

338 Vol. 124, JUNE 2002

Fig. 8 Calculated velocity distributions by NFA at diameter ratio DD2 0,63

matches logically with static pressure test results Fig. 5. Although at the outer impeller side chamber perimeter the peripheral
velocity component is dependent on the direction and quantity of
relative leakage flow q L , the difference increases radially toward
the impeller wear ring. We can summarize that the fluid rotation in
the core layer is dominant for the pressure distribution on the
impeller front and back shrouds, and consequently, for pump hydraulic axial thrust.

Numerical Flow Analysis

Varied software for numerical flow analysis is available and its
application in the field of hydraulic machinery is almost a standard tool. In this investigation a three-dimensional k- turbulent
code named Tasc-flow has been used, which is second-order accurate in space. The main goal of the NFA was to determine
whether the numerical approach could satisfactorily simulate the
test results. Due to simplification the flow analysis has been done
for impeller side chamber flow separately, uncoupled from the
main impeller flow.
Several flow conditions were analyzed and the results give a
clear and detailed picture of the velocity and pressure distributions
in the impeller side chamber. In Fig. 8 the velocity profiles are
shown in the peripheral and radial directions. Two separate
boundary layers with a central core layer can be observed. The
calculated velocity distribution of the core layer matches well
with LDA test results. An example is shown in Fig. 8 for the
regime q L .
The static pressure distribution in the side chamber is an important factor when judging the usefulness of NFA for hydraulic axial
thrust prediction. In Fig. 9 the comparison of measured and calculated static pressure distributions are shown. Matching is satisfactory and confirms the usefulness of NFA for pressure distribution calculation of different geometries of impeller side chambers
in order to optimize the pump elements in conjunction with axial
thrust balancing system.
It has to be stated that the important factor for realistic NFA
results is the introduction of proper boundary conditions at the
outer impeller periphery. This means application of the circumferTransactions of the ASME

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 9 Comparison of calculated and measured static pressure distributions

Fig. 11 Execution of FRI, FRD and MRBwr on one stage test
arrangement

ential velocity component as well as the basic static pressure in

gap A for both directions of Q L . The following boundary conditions were applied at gap A:
velocity direction: ratio between axial and peripheral velocity
component-1:20, radial velocity component-zero,
flow rate: calculated from relative leakage flow rate for three
conditions (q L 2,2%,0%,2,2%), and
average static pressure: measured values from the tapping in
the vicinity of gap A.
Impeller shroud and casing wall roughness have to be taken
into account when defining wall function in the NFA logarithmic
wall function, roughness: equivalent sand grain 2.105 . The numerical model used was axisymmetric with approximately 25,000
nodes. Computational time for one leakage flow variant on a PC
computer with a 450 MHz processor and 512 MB RAM was
about 60 minutes. Criteria for maximum residual level were
ERTIME103 .

Fluid Rotation in Impeller Side Chamber

Velocity profiles measured via LDA in the core layer of the
impeller side chamber shown in Fig. 7 were recalculated into the
form of relative fluid rotation k. Relative fluid rotation in the core
layer is the ratio between the angular velocity of fluid and the
angular velocity of the impeller: k / . In Fig. 10, relative fluid
rotation is presented, based on measurements in the core layer for
different leakage flow quantities and directions. It can be noticed
clearly that at constant relative leakage flow q L , the relative fluid

rotation k changes with the radial position in the side chamber. At

the outer impeller side chamber perimeter the influence of leakage
flow on fluid rotation is smaller, and increases radially inwards.
Inward leakage flow increases fluid rotation; the opposite is valid
for outward leakage flow.
When the distribution in the radial direction of relative fluid
rotation in the impeller side chamber is compared to static pressure distribution Fig. 5, it is then obvious that the pressure
changes at different leakage flows resulted from differences of
fluid rotation in the core layer. Another important conclusion can
be derived from the above stated results as well. In order to alter
the pressure distribution and accordingly the hydraulic axial thrust
on the impeller, the fluid rotation of the core layers in the side
chambers gaps E and F in Fig. 3 has to be changed artificially.

Measures for Hydraulic Axial Thrust Reduction

The main part of the total hydraulic axial thrust in a multistage
pump is generated in the normal stages, where the leakages in the
impeller front and back shroud side chambers flow in opposite
directions Fig. 1. The idea of axial thrust reduction is to increase
fluid rotation on the back shroud side chamber and to decrease it
on the front shroud side chamber. Efforts to alter fluid rotation
should be achieved by means that do not have a significant negative effect on pump characteristics, especially on the pump
efficiency.
Series of tests were carried out on the testing arrangement, Fig.
2. The Fluid Rotation Increaser FRI and the Fluid Rotation Decreaser FRD were built into the impeller side chamber, Fig. 11.
The FRI is effective for outward leakage flows and the FRD for
inward flows. Results are shown in Figs. 12 and 13. The FRI
consists of numerous radial ribs, located at the outer side of the
impeller back shroud at inner side chamber perimeter and rotates
together with the impeller. As expected, the FRI produces the
highest additional fluid rotation at the wear ring area, and the
effect decreases toward outer impeller side chamber perimeter
see effect on static pressure in Fig. 12. No effect on pump efficiency has been observed due to the introduction of the FRI. This
can be explained by the following facts:
for the leakage flow direction radial outwards the FRI represents a simplified small impeller, acting in the same direction
as the main impeller, and
disk friction losses are smaller at higher fluid rotation, according Senoo and Hayami 6.

Fig. 10 Relative fluid rotation in the impeller side chamber

Journal of Fluids Engineering

An effective FRD is type 2, which causes a considerable pressure increase in the impeller side chamber at the inner area and a
JUNE 2002, Vol. 124 339

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Fig. 12 Effect of FRI on static pressure distribution

corresponding impeller axial thrust reduction see effect on static

pressure in Fig. 13. The FRD2 consists of numerous stationary
radial ribs, located at the outer corner of the impeller side chamber. The negative aspect of the FRD2 is an efficiency reduction of
about 1% at Q opt due to additional mixing losses in the FRD area.
The FRI and FRD have to be installed on all stages of a multistage pump. They can be cast together with impeller FRI or
inter-stage casing FRD. In order to obtain the maximal effectiveness on fluid rotation in the impeller side chambers, the design
and execution of FRI and FRD should be studied in greater detail.

Effect of Increased Wear Ring Radial Clearance

During pump operation the radial clearances on impeller wear
rings can increase. How rapidly the clearances increase depends
on modes of operation, fluid properties, solid contents, material of
rings, speed of rotation, etc. According to API 610/8 the pump
elements balancing device with thrust bearing should be sized
for continuous operation, when all loads should be defined at design internal clearances and at two times as much design internal
clearances.
With wear of the radial wear rings, the volumetric losses and
consequently leakage flows Q L in both impeller side chambers
increase. The pressure distribution on both impeller sides is
changed, as schematically shown in Fig. 14 for a normal stage
impeller. The increased radial clearance causes hydraulic axial
thrust on the front shroud to decrease, while on the back shroud it
increases; consequently, axial thrust on the impeller increases.
We were looking for a solution to this problem of wear rings,
which would result in lower changes of leakage flow with in-

Fig. 13 Effect of FRD on static pressure distribution

340 Vol. 124, JUNE 2002

Fig. 14 Schematic pressure distributions on back and front

shroud-normal stage impeller

creased wear. This would have a positive effect on axial thrust as

well as on pump efficiency. The main efforts have been focused
on a design variation, which has a high pressure loss factor also at
increased radial clearances. A solution applied is called Multi Radial Bore wear ring MRBwr, shown in Fig. 11. The outer wear
ring dimensions are the same as a classical plain ring, but additional radial openings cavities were machined. The principle is
similar to honeycomb wear rings; however, the number and shape
of radial cavities is different. On the basis of the results from
Childs et al. 10 it can be concluded that an additional positive
effect of the MRBwr is improved pump rotordynamic behavior
due to changed stiffness and damping coefficient.

Modified Multistage Pump Design

Based on the results of the above described investigation a
modified pump design was worked out and tested. This design
includes the installation of FRI2 on all impeller front side chambers and the installation of MRBwr on all impeller front and back
wear rings. The positive effect of FRI is reduced hydraulic axial
thrust, and of MRBwr lower sensitivity of axial thrust and efficiency on increased wear ring radial clearance. The effectiveness
of FRI and MRBwr was verified on a three-stage pump constructed with the same impeller/diffuser elements as shown in
Figs. 2 and 3. Results are presented in Fig. 15 and are summarized
in Table 1 as a comparison between the modified design and the
classic multistage pump design.
There are several advantages of the modified design:

Fig. 15 Effect of modified pump design on hydraulic axial

thrust of three-stage pump

Transactions of the ASME

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Comparison between original and modified design of

three-stage pump of specific speed nq22 nsUSA 1135. Pump
flow condition: QQopt

P
Q
QL
R
SR
Ue
b
n
nq
ns USA
p
p opt
qL
u
v
z

lower hydraulic axial thrust,

smaller increase of axial thrust at twice the normal wear ring
clearances,
smaller efficiency drop at twice the normal wear ring clearance,
improved rotordynamic behavior due to changed stiffness and
damping effects of MRBwr.
A disadvantage of the modified design are the additional machining costs for MRBwr. The pump construction containing FRI and
FRD is the subject of patent application.

Conclusions
From the presented investigation the following main conclusions can be made:
1. The pressure distribution on the impeller front and back
shrouds can be influenced by introduction of Fluid Rotation Increasers FRI and Fluid Rotation Decreasers FRD, which results in axial thrust reduction. The final consequence is a smaller
piston diameter and a smaller axial bearing.
2. Sensitivity of hydraulic axial thrust to radial labyrinth wear is
considerably reduced by the introduction of Multi Radial Bore
wear rings MRBwr.
3. With proper execution of radial clearance A and overlapping
U e , flow in the side chamber can be effectively separated from
the influence of main flow at impeller exit. This is also important
to avoid axial thrust instability at part load operation of the pump.
4. Numerical flow analysis is an effective tool for axial thrust
prediction at the design stage.

Nomenclature
A
B
B 2 , B 3
D
Re
E
F
Fa
H

radial clearance
radial clearance vane passage
impeller, diffuser channel width
diameter
R 22 / -Reynolds number
axial width of impeller front shroud side chamber
axial width of impeller back shroud side chamber
axial thrust
head

Journal of Fluids Engineering

power
flow rate
leakage flow
radius
radial clearance
impeller/diffuser wall overlapping
impeller shroud thickness
speed of rotation
n(min 1 )Q(m 3 /s) 0.5/H(m) 0.75-specific speed
n(rpm)Q(G PM ) 0.5/H( f t) 0.75-specific speed
pressure
gH opt-pump total pressure at Q opt
Q L /Q opt-relative leakage flow
peripheral velocity
velocity
axial coordinate, impeller side chamber width
fluid angular velocity
efficiency
kinematic viscosity
fluid density
impeller angular velocity

back shroud
front shroud
optimal, best efficiency point
original
piston
radial
static
peripheral
wear ring
impeller outlet
diffuser inlet

Indices
bs
fs
opt
orig
p
r
st
u
wr
2
3

References
1 Daily, J. W., and Nece, R. E., 1960, Chamber Dimension Effects on Induced
Flow and Frictional Resistance of Enclosed Rotating Disks, J. Basic Eng., 83,
pp. 217232.
2 Kurokawa, J., and Toyokura, T., 1972, Study on Axial Thrust of Radial Flow
Turbomachinery, Proceedings, The Second International JSME SymposiumFluid Machinery and Fluidics, Tokyo, Japan, pp. 31 40.
3 Iino, T., Sato, H., and Miyashiro, H., 1980, Hydraulic Axial Thrust in Multistage Centrifugal Pumps, ASME J. Fluids Eng., 102, pp. 64 69.
4 Thomae, H., and Stucki, R., 1970, Axial Thrust Occurring in Multistage
Radial Pumps, Sulzer Technical Review, .
5 EPRI Report TR-102102, 1993, Feedpump Operation and Design Guidelines, Electric Power Research Institute, Palo Alto, CA, pp. 1/351/46.
6 Senoo, Y., and Hayami, H., 1976, An Analysis on the Flow in a Casing
Induced by a Rotating Disk Using a Four-Layer Flow Model, ASME J. Fluids
Eng., 99, pp. 192198.
7 Stepanoff, A. J., 1966, Centrifugal and Axial Flow Pumps, Wiley, New York.
8 Makay, E., and Barrett, J. A., 1984, Changes in Hydraulic Component Geometries Greatly Increased Power Plant Availability and Reduced Maintenance
Cost: Case Histories, Proceedings, The First International Pump Symposium,
Texas A&M University, College Station, Texas, pp. 8597.
9 Gantar, M., Florjancic, D., and Sirok, B., 2001, Changes in Hydraulic Component Geometries and its Influence on Axial Thrust in Multistage Pumps in
German, IndustriepumpenKompressoren, Vulkan Verlag, Essen, Germany.
10 Childs, D. W., Elrod, D., and Hale, K., 1988, Annular Honeycomb Seal Test
Results for Leakage on Rotordynamic Coefficients; Comparison to Labyrinth
and Smooth Configurations, ASME Paper 88-Trib-35, ASME/STLE 1988,
Tribology Conference.

JUNE 2002, Vol. 124 341

Downloaded 12 Nov 2007 to 128.104.189.231. Redistribution subject to ASME license or copyright, see http://www.asme.org/terms/Terms_Use.cfm