Numerical Analysis of a
Waterjet Propulsion System
Norbert Willem Herman Bulten
Acknowledgement:
The research described in this thesis was supported by
Wärtsilä Propulsion Netherlands B.V.
Cover: Michelle Tjelpa
Photo: Bram Kruyt
Printing: Printservice Technische Universiteit Eindhoven
Copyright © 2006 by N.W.H. Bulten, The Netherlands
All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system or transmitted in any form or by any means, electronically, meachanically, by pho
tocopying, recording, or otherwise, without the written permission of the author.
A catalogue record is available from the Library Eindhoven University of Technology
ISBN10: 9038629885
ISBN13: 9789038629889
Numerical Analysis of a
Waterjet Propulsion System
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de
Technische Universiteit Eindhoven, op gezag van de
Rector Magnificus, prof.dr.ir. C.J. van Duijn, voor een
commissie aangewezen door het College voor
Promoties in het openbaar te verdedigen
op woendag 15 november 2006 om 16.00 uur
door
Norbert Willem Herman Bulten
geboren te Winterswijk
Dit proefschrift is goedgekeurd door de promotoren:
prof.dr.ir. J.J.H. Brouwers
en
prof.dr.ir. H.W.M. Hoeijmakers
Copromotor:
dr. B.P.M. van Esch
Numerical analysis of a waterjet propulsion system 1
Table of contents
Chapter 1 Introduction ................................................................ 5
1.1 Waterjet layout................................................................................ 6
1.2 Relation of waterjet propulsion system to other turbo machinery... 7
1.3 Aim of the analysis........................................................................ 10
1.4 Outline of this thesis ..................................................................... 11
1.5 Nomenclature ............................................................................... 12
1.6 References ................................................................................... 12
Chapter 2 Waterjet propulsion theory ..................................... 15
2.1 Characteristic velocities in a waterjet system............................... 16
2.1.1 Wake fraction .................................................................... 17
2.1.2 Inlet velocity ratio ............................................................... 20
2.1.3 Jet velocity ratio ................................................................. 21
2.1.4 Summary ........................................................................... 22
2.2 General pump theory.................................................................... 22
2.2.1 Dimensionless performance parameters ........................... 22
2.2.2 Pump geometry parameters .............................................. 24
2.2.3 Cavitation parameters ....................................................... 25
2.2.4 Correlation with propeller performance parameters .......... 26
2.3 Thrust ........................................................................................... 27
2.3.1 General thrust equation ..................................................... 27
2.3.2 Open propeller thrust ......................................................... 28
2.3.3 Waterjet thrust ................................................................... 30
2.3.4 Concluding remarks ........................................................... 34
2.4 Pump head ................................................................................... 34
2.5 Overall propulsive efficiency......................................................... 37
2.5.1 Cavitation margins ............................................................. 39
2.5.2 Limitations in specific speed .............................................. 40
2.5.3 Limitations in jet velocity ratio ............................................ 41
2.5.4 Limitation of power density ................................................ 42
2.6 Waterjet selection ......................................................................... 43
2.7 Closing remark.............................................................................. 44
2.8 Nomenclature ............................................................................... 44
2.9 References ................................................................................... 46
Chapter 3 Nonuniform distribution of pump entrance velocity
field .......................................................................... 49
3.1 Representation of nonuniform velocity distribution...................... 49
2
.
3.1.1 Experimental setup ........................................................... 50
3.1.2 Nondimensional representation ........................................ 52
3.1.3 Twodimensional representation ........................................ 52
3.2 Local flow rate fluctuations............................................................ 53
3.3 Impeller velocity triangles.............................................................. 55
3.4 Origin of the nonuniform velocity distribution............................... 57
3.4.1 Boundary layer ingestion ................................................... 57
3.4.2 Deceleration of the flow ..................................................... 58
3.4.3 Obstruction of the flow due to the shaft ............................. 59
3.4.4 Bend in the inlet duct ......................................................... 60
3.4.5 Closing remark ................................................................... 60
3.5 Nonuniform inflow velocity distributions in other turbo machinery60
3.6 Nomenclature................................................................................ 61
3.7 References.................................................................................... 62
Chapter 4 Mathematical treatment ........................................... 63
4.1 Requirements of mathematical method ........................................ 63
4.1.1 Incompressibility ................................................................ 64
4.1.2 High Reynolds number ...................................................... 64
4.1.3 Time dependency .............................................................. 65
4.1.4 Nonuniformity of impeller inflow ........................................ 65
4.1.5 Tip clearance flow .............................................................. 66
4.1.6 Final remarks ..................................................................... 67
4.2 Conservation laws......................................................................... 67
4.3 Reynolds Averaged NavierStokes (RANS) flow.......................... 68
4.3.1 Reynolds averaging ........................................................... 68
4.3.2 Eddy viscosity turbulence models ...................................... 70
4.4 Twodimensional test cases.......................................................... 75
4.4.1 Isolated NACA 0012 profile ............................................... 75
4.4.2 Cascades with NACA 65410 profiles ................................ 80
4.4.3 Sensitivity of errors in drag on thrust and torque ............... 82
4.5 Nomenclature................................................................................ 85
4.6 References.................................................................................... 86
Chapter 5 Numerical analysis of waterjet inlet flow ............... 89
5.1 Review of CFD analyses on waterjet inlets................................... 89
5.2 Geometry and mesh generation ................................................... 91
5.3 Numerical approach...................................................................... 94
5.3.1 Boundary conditions .......................................................... 94
5.3.2 Fluid properties .................................................................. 95
5.3.3 Discretisation and solution algorithm ................................. 95
5.4 Validation with experimental data ................................................. 96
Numerical analysis of a waterjet propulsion system 3
5.4.1 Comparison of static pressure along the ramp centre line 97
5.4.2 Comparison of cavitation inception pressure at cutwater 101
5.4.3 Comparison of total pressure at impeller plane ............... 105
5.4.4 Comparison of velocity field at impeller plane ................. 108
5.4.5 Results obtained with kω turbulence model ................... 113
5.4.6 Mesh convergence study ................................................. 115
5.4.7 Closing remarks ............................................................... 117
5.5 Analysis of the suction streamtube............................................. 117
5.5.1 Visualisation of suction streamtube ................................. 118
5.5.2 Determination of suction streamtube shape .................... 118
5.6 Evaluation of wall shear stress ................................................... 123
5.7 Nomenclature ............................................................................. 125
5.8 References ................................................................................. 125
Chapter 6 Numerical analysis of waterjet pump flow ........... 127
6.1 Geometry and mesh generation ................................................. 127
6.2 Numerical approach.................................................................... 131
6.2.1 Boundary conditions ........................................................ 131
6.2.2 Fluid properties ................................................................ 132
6.2.3 Impeller rotation ............................................................... 132
6.2.4 Calculation of global pump performance ......................... 132
6.3 Validation with experimental data............................................... 134
6.3.1 Quasisteady flow calculations with the MFR method ..... 135
6.3.2 Transient flow calculations with moving mesh ................. 138
6.3.3 Rotorstator interaction forces ......................................... 144
6.4 Influence of nonuniform axial inflow .......................................... 147
6.4.1 Pump performance for nonuniform inflow ...................... 147
6.4.2 Background of radial forces acting on the impeller .......... 148
6.4.3 Flow rate fluctuations in the impeller channel .................. 149
6.4.4 Radial forces for nonuniform inflow ................................ 150
6.4.5 Concluding remark .......................................................... 155
6.5 Nomenclature ............................................................................. 155
6.6 References ................................................................................. 156
Chapter 7 Analysis of a complete waterjet installation ........ 159
7.1 Generation of the numerical model............................................. 159
7.2 Evaluation of volume flow rate.................................................... 160
7.3 Evaluation of waterjet thrust ....................................................... 161
7.3.1 Integration of solid wall forces ......................................... 161
7.3.2 Momentum balance ......................................................... 163
7.3.3 Results ............................................................................ 163
7.4 Evaluation of required power ...................................................... 164
4
.
7.5 Analysis of vertical force on waterjet structure............................ 165
7.6 Pressure distribution on streamtube surface .............................. 168
7.6.1 Evaluation of momentum balance in vertical direction ..... 168
7.6.2 Calculation of vertical force on streamtube ...................... 169
7.6.3 Concluding remark ........................................................... 171
7.7 Nomenclature.............................................................................. 171
7.8 References.................................................................................. 172
Chapter 8 Concluding remarks ...............................................173
8.1 Conclusions ................................................................................ 173
8.1.1 Theory of thrust prediction for waterjet systems .............. 173
8.1.2 Numerical aspects ........................................................... 174
8.1.3 Waterjet inlet flow characteristics .................................... 174
8.1.4 Waterjet mixedflow pump analyses ................................ 175
8.2 Recommendations...................................................................... 175
8.2.1 Research topics for marine propulsion systems .............. 175
8.2.2 Application of RANS methods ......................................... 176
Appendix A Stability of nonuniform ..velocity distribution 177
A.1 Test case with nonuniform pipe flow.......................................... 178
A.2 References.................................................................................. 180
Appendix B Fourier analyses of transient flow calculations 181
Summary ................................................................................... 189
Samenvatting ........................................................................... 193
Dankwoord ............................................................................... 197
Curriculum Vitae ...................................................................... 199
Numerical analysis of a waterjet propulsion system 5
Chapter 1 Introduction
The desire to travel faster and further is probably as old as mankind itself.
There has been an enormous development in the way people use to travel
from one place to another. At first it was only over land, and later also over
sea. And since about a century is it possible to travel through air as well.
Achievements in automotive and aerospace technology are widely
recognized. But probably, most readers do not realize the substantial
development in high speed ship transportation. At the end of the 20th century,
fast ferry catamarans sailing at 50 knots (equivalent to about 90 km/h) were in
commercial service all over the world. However, this type of vessel had
entered the market less than two decades before.
The considerable development in the high speed craft can be partly
contributed to the application of waterjet propulsion systems. Currently used
stern mounted waterjets are based on principles as applied by Riva Calzoni
in 1932 [1]. However, the first type of waterjet propulsion was invented
already 300 years earlier, by David Ramseye [2]. He stated in 1630 in English
Patent No. 50 that he was able ‘to make Boates, Shippes and Barges to goe
against Stronge Winde and Tyde’. It is supposed that he had a waterjet in
mind for the propulsion, since at that time there was a great interest in using
steam to raise water and to operate fountains. In 1661, English patent no.
132 was granted to Thomas Toogood and James Hayes for their invention of
’Forceing Water by Bellowes [...] together with a particular way of Forceing
water through the Bottome or Sides of Shipps belowe the Surface or Toppe of
the Water, which may be of siguler Use and Ease in Navagacon’. This
concept was based on a waterjet without a doubt. However, they did not
6
Chapter 1. Introduction
manage to develop a working prototype. This invention and the subsequent
development of the waterjet until 1980 is described in much more detail by
Roy [3]. From 1980 onwards the use of waterjets in commercial applications
really started to grow [4].
At the start of the 21th century the sizes of installed waterjets have increased
to diameters of about 3 meter. This has led to installed powers of 25 MW per
installation. Luxury high speed motor yachts have achieved ship speeds well
above 65 knots, which is about 120 km/h [5].
1.1 Waterjet layout
A sternmounted waterjet installation as used in commercial applications, can
be divided into four components: the inlet, the pump, the nozzle and the
steering device. Figure 1.1 shows a drawing of a typical waterjet installation,
with the main components labelled.
The main component is the pump, which delivers the head to produce the jet
at the nozzle exit. In general the stator bowl and the nozzle are integrated in
one part. In the remainder of the thesis, the combination of the pump unit and
the nozzle is regarded as the waterjet pump.
The ducting system upstream of the pump is called the inlet. The waterjet in
figure 1.1 shows a flush mounted inlet duct. This is used, for example, in fast
ferries and high speed motor yachts. Kruppa et al. [6] have given an overview
of the basic concepts of waterjet inlet ducting systems. Besides the flush
mounted inlet, ram and scoop type inlets are mentioned. The latter two have
an opening that is situated more or less perpendicular to the flow direction,
whereas the flush mounted inlet opening is parallel to the flow. The ram and
scoop intake will not be considered in this thesis.
Downstream of the nozzle there is a steering device, which can deflect the jet
in order to create steering and reversing forces. There are also installations
for the deflection of the jet possible, with only the reversing option. This can
be useful for quick crashstop manoeuvres. If the waterjet has no steering
device at all, it is called a booster waterjet.
1
2
3
4
1. inlet
2. pump
3. nozzle
4. steering device
Figure 1.1 Threedimensional view of a waterjet installation
1.2 Relation of waterjet propulsion system to other turbo machinery
Numerical analysis of a waterjet propulsion system 7
1.2 Relation of waterjet propulsion system to other turbo
machinery
If the very early 17th century developments are neglected, waterjet
propulsion is relatively new. For further development of the installation it may
be useful to look at related engineering applications. Figure 1.2 shows a box
with eight different types of apparatus. The three faces which are connected
to the waterjet share a common property.
The front face is formed by four installations which are designed to produce
thrust. This group contains, besides the waterjet, the ship propeller and the
two main aeroplane propulsion systems. Any thrust production by the
installations at the back face (mixedflow pump, compressor, ventilator and
mixer) is an undesirable side effect.
If history is reviewed an interesting parallel can be recognised. In aerospace
the propeller has been replaced by the jet engine, which was necessary to
reach higher speeds. Application of waterjets in marine industry shows a
similar trend where the waterjet propelled vessels reach higher speeds.
Many relations which describe the principles of waterjet propulsion are
directly derived from propeller theory, with the same nomenclature. This can
lead to misunderstandings, if the same waterjet is described as a mixedflow
pump, with the accompanying pump nomenclature. For example, often Q is
used for torque in propeller theory and for flow rate in pump theory.
The two side planes of the box show the difference in type of flow. The left
side is formed by external flow machines and the right side by internal flow
machines. Transmission of the forces in an external flow machine can only be
done through the shaft. Internal flow machines can also transfer forces
through the surrounding structure.
The top plane of the box shows four installations which operate in water,
whereas the applications on the bottom plane operate in air. So here the fluid
is the distinguishing factor. Cavitation is a typical problem for installations
operating in water. Another important fluid property of water is its very low
compressibility. Both phenomena can be important in the selection of
numerical solution methods. Numerical methods used for the analysis of
compressors and other flow machinery often require a certain amount of
compressibility, what makes these methods less suitable for the analysis of a
waterjet propulsion system.
The box model will be used to relate the occurring phenomena in a waterjet
installation to known ones in other machines, like the ship propeller, the
aeroplane jet engine and the mixedflow pump.
A ship propeller seems to be the most logical connection to a waterjet for a
description of the propulsion system. Typical parameters used in propeller
8
Chapter 1. Introduction
theory are the thrust loading coefficient and the cavitation number [7]. These
parameters can be employed to describe the performance of a waterjet as
well. Moreover the concept of wake fraction, which represents the difference
between the free stream advance speed and actual inflow velocity, can be
used to account for the effect of the hull boundary layer ingestion.
It is wellknown that the inflow velocity distribution to the waterjet impeller is
strongly nonuniform. This is similar to the wake field of a ship propeller. Due
to this wake field the loading of a propeller blade fluctuates during a
revolution. This results in fluctuations of the pressure distribution on the
blades and in a radial force on the shaft. These phenomena will also be
present in a waterjet. Therefore the choice of a propeller as a starting point
for the analysis of a waterjet installation seems to be logical. However, there
is a very important difference between a propeller and a waterjet installation.
A propeller is an external flow machine whereas the waterjet installation is
mainly an internal flow system. The thrust of a propeller will always be guided
through the shaft into the ship. In a waterjet installation the forces can be
transferred to the vessel via the shaft and via the ship structure. In fact it is
possible to have a higher thrust acting on the shaft than the net thrust of the
installation [8]. In that case a negative force will work on the transom stern
and the inlet ducting.
Figure 1.2 Box model of connections of waterjet to other types of
turbomachinery
1.2 Relation of waterjet propulsion system to other turbo machinery
Numerical analysis of a waterjet propulsion system 9
Because the ship propeller operates as an external flow machine, the ship
speed can be used as a governing parameter for the operating point. In non
dimensional notation it is called advance ratio (see for example [10], [10]):
(1.1)
where v
ship
is the ship (or advance) speed, n the shaft speed and D the
diameter of the propeller.
The working point of the waterjet installation is based on the volume flow rate
Q through the system. In this system the pump head curve matches the
system resistance curve, which is based on the required head to produce the
jet velocity and the head to overcome the hydraulic losses. The influence of
ship speed on the operating condition is small.
As a consequence, the available set of propeller equations cannot be used
for a good description of the waterjet propulsion system.
The theory of aeroplane jet engines may provide the missing equations to
describe the performance of a waterjet system. A turbojet engine is a thrust
producing internal turbomachine, just like the waterjet. The turbojet engine
can be divided into five major components: intake, compressor, combustion
chamber, turbine and nozzle (see for example [11]). These components
include the power generating part of the jet engine, i.e. the compressor is
driven by the turbine. In a waterjet a separate diesel engine or gas turbine is
needed to supply the required power to the shaft.
Net thrust of a turbojet engine is based on the change of momentum:
(1.2)
where is the mass flow through the system, v
out
the jet velocity leaving the
engine and v
in
the velocity of the air entering the intake, which is equal and
opposite to the forward speed of the aircraft. Strictly spoken the mass flow in
the system increases due to the addition of fuel, but this increase in mass
flow is negligible. According to equation (1.2), the thrust of a waterjet system
is directly related to the volume flow rate, since the flow is incompressible:
.
The definition of the propulsive efficiency of a turbojet engine can be found in
literature [11]:
(1.3)
which is often denoted as Froude efficiency. The ratio between intake and
nozzle velocity is called nozzle velocity ratio (NVR = v
out
/v
in
). At zero speed
J
v
shi p
nD
 =
F m
·
v
out
v
i n
– ( ) =
m
·
m
·
ρQ =
η
p
F v
i n
⋅
P
shaft

2
1 v
out
v
i n
⁄ ( ) +
 = =
10
Chapter 1. Introduction
the NVR becomes infinite, therefore the reciprocal value is used in literature
for waterjets; this is known as jet velocity ratio µ [12].
Although the working principle of the aeroplane jet engine and the waterjet
seem to be similar, it should be kept into mind that cavitation and nonuniform
inflow, two important issues in waterjet propulsion, are not dealt with in
jetengine research.
The third type of turbomachinery which may provide part of the basic theory
to describe system performance is a mixedflow pump. At first sight this is a
bit strange, because normally the axial thrust in pump operation is not
exploited. Nevertheless, the head curve of the pump and the system
resistance curve provide sufficient information to determine the volume flow
rate Q through the system. To get a first estimation of the thrust of the
system, only the average velocity of the ingested flow and the dimension of
the nozzle diameter have to be known.
1.3 Aim of the analysis
In this thesis a detailed analysis of a waterjet propulsion system is made.
Results of Computational Fluid Dynamics (CFD) calculations are used to get
an impression of the flow phenomena occurring in such systems and to
quantify system parameters, such as flow rate, torque and thrust. With the
application of a numerical method some flow features are easier to determine
than in a model scale test. Typical complicating factors in the analysis of
waterjets are the boundary layer ingestion and the nonuniform velocity
distribution just upstream of the pump. Unfortunately, both the boundary layer
ingestion as well as the nonuniformity of the velocity distribution are
inevitable in commercial waterjet propulsion systems with a flush type of inlet.
The major problem of the impeller inlet velocity distribution is the large
variation of the velocity in circumferential direction. This will give rise to a
blade loading, which varies strongly with time. This may lead to a decrease in
system performance, like a reduced efficiency, a deterioration in cavitation
behaviour and an increase of forces acting on the impeller. These
phenomena will increase the noise and vibrations in the installation.
The aim of the analysis presented in this thesis is (i) to quantify the effects of
the nonuniform inflow to the mixedflow pump and the resulting non
stationary flow in the pump on the system performance and (ii) to quantify the
forces on the complete waterjet installation in both axial and vertical direction.
The currently used theory to determine system performance includes some
assumptions about the influence of the pressure distribution on the
streamtube of the ingested water. These assumptions will be reviewed to
check their validity.
1.4 Outline of this thesis
Numerical analysis of a waterjet propulsion system 11
1.4 Outline of this thesis
In chapter 2 the conventional theory of waterjet propulsion systems will be
discussed in detail. This will give insight in the governing parameters of the
total propulsion system. Some connections will be made with standard
propeller theory to show the similarities and the differences. Some of the
underlying assumptions made will be discussed to enable assessment of
these assumptions later on. The analysis also reveals the basic principles of
waterjet selection which is suitable for most of the current applications.
Values for pump parameters in literature are based on uniform inflow.
However, it is wellknown that a waterjet impeller has to operate in a non
uniform inflow velocity field. The nature of the velocity distribution will be
discussed in chapter 3. Results of measurements will be shown to give an
impression of the level of nonuniformity. It is concluded that the typical non
uniform velocity distributions are inevitable in waterjet installations with flush
mounted inlets, based on an analysis of the development of the non
uniformity in the duct upstream of the impeller.
Chapter 4 deals with the choice of a mathematical method to analyse the flow
through the system. An evaluation of several methods, such as potential flow,
Euler and RANS, will be presented. An important requirement is the capability
to capture the effects of the nonuniform inflow to the pump.
The chosen method for the calculation of the flow through a waterjet inlet will
be validated with available experimental data in chapter 5. In these
calculations, the mass flow rate is prescribed as a boundary condition, since
the pump is not included in the model. The results of the numerical analysis
of the inlet will also be used to evaluate the shape of the streamtube
upstream of the inlet duct. Determination of this streamtube enables a more
detailed analysis of the momentum distribution of the ingested water.
Chapter 6 will deal with the numerical analysis of the nonstationary flow
through the mixedflow pump. Results of calculations are compared with
model scale measurements of the pump performance. Transient calculations
with both uniform and nonuniform velocity distributions will show the
presence of fluctuating radial forces. These forces are strongly related to the
level of nonuniformity in the flow.
Chapter 7 will show the results of the analysis of a complete full scale
waterjet propulsion system. Overall performance indicators, like volume flow,
thrust and power, will be analysed. Comparisons are made with performance
prediction software of Wärtsilä Propulsion Netherlands (WPNL). A more
detailed analysis of the streamtube will reveal some new insights into the
forces acting on the installation in vertical direction.
Finally, the conclusions of the present research will be presented in chapter
8.
12
Chapter 1. Introduction
1.5 Nomenclature
D propeller diameter m
F thrust N
J advance ratio of propeller (= v
ship
/nD) 
mass flow rate kg/s
n rotational speed 1/s
NVR nozzle velocity ratio (NVR = v
out
/v
in
) 
P
shaft
shaft power W
Q volume flow rate m
3
/s
v
ship
advance velocity of propeller m/s
v
in
advance velocity of jetengine m/s
v
out
jet velocity of jetengine m/s
Greek symbols
η
p
propulsive efficiency 
µ jet velocity ratio (= 1 / NVR) 
ρ
fluid density kg/m
3
1.6 References
[1] Voulon, S., ‘Waterjets and Propellers, Propulsors for the future’, Pro
ceedings SATEC’96 conference, Genoa, Italy, 1995
[2] Ramseye, David, ‘Manufacture of Saltpetre, Raising Water, Propelling
Vessels, &c.’, English patent no. 50, 1630
[3] Roy, S.M., ‘The evolution of the modern waterjet marine propulsion
unit’, Proceedings RINA Waterjet Propulsion conference, London,
1994
[4] Warren, N.F., & Sims, N.,’Waterjet propulsion, a shipbuilder’s view’,
Proceedings RINA Waterjet Propulsion conference, London, 1994
[5] Bulten, N. & Verbeek, R.,’Design of optimal inlet duct geometry based
on operational profile’, Proceedings FAST2003 conference Vol I, ses
sion A2, pp 3540, Ischia, Italy, 2003
[6] Kruppa, C., Brandt, H., Östergaard, C., ‘Wasserstrahlantriebe für
Hochgeschwindigkeitsfahrzeuge’, Jahrbuch der STG 62, Band 1968,
Nov., pp. 228258, 1968
[7] Terwisga, T.J.C. van,’Waterjet hull interaction’, PhD thesis, Delft, 1996
m
·
1.6 References
Numerical analysis of a waterjet propulsion system 13
[8] Verbeek, R.,’Waterjet forces and transom flange design’, RINA waterjet
propulsion conference, London, 1994
[9] Newman, J.N., ‘Marine hydrodynamics’, MIT press, 1977
[10] Lewis, E.V., ‘Principles of naval architecture’, Volume II, Society of
Naval Architects and Marine Engineers, Jersey City, 1998
[11] Cohen, H., Rogers, G.F.C., Saravanamuttoo, H.I.H., ‘Gas turbine the
ory’, Longman Group, London, 1972
[12] Verbeek, R.,’Application of waterjets in highspeed craft’, in Hydrody
namics: Computations, Model Tests and Reality, H.J.J. van den Boom
(Editor) Elsevier Science Publication, 1992
14
Chapter 1. Introduction
Numerical analysis of a waterjet propulsion system 15
Chapter 2 Waterjet propulsion theory
In this chapter the basic principles of waterjet propulsion will be discussed.
The equations of the waterjet theory will be based on standard nomenclature
used in the description of pump performance. Where possible, equivalent
nomenclature of commonly used propeller theory will be mentioned as a
reference.
In the first section some specific velocities, as used in waterjet theory, will be
defined. These definitions form the basis for the remainder of the chapter. In
the second section, the generally applied standard parameters are defined,
which are used to describe the overall pump performance.
In the commonly used waterjet propulsion theory, equations for the derivation
of thrust of a waterjet propulsion system are based on open propeller theory.
The transition from open propellers to waterjets will be reviewed in detail, in
order to reveal possible deficiencies in the waterjet theory.
The equations for the waterjet thrust can be coupled to the required pump
head and flow rate. This will be discussed in section 2.4. It will be shown that
a certain thrust can be achieved with different combinations of flow rate and
pump head. Determination of the optimal combination of flow rate and pump
head is obtained with the aid of the overall propulsive efficiency. This will
result in the design operating point of the pump in the waterjet installation.
In some conditions, the optimal pump operating point can not be reached due
to severe cavitation in the pump. This limitation in the optimization process
will be discussed in section 2.5.1.
16
Chapter 2. Waterjet propulsion theory
In the selection of a waterjet installation for ship propulsion the weight of the
installation is an important issue. To minimize the weight of the system, the
size of the waterjets is selected as small as possible. The shaft speed of the
pump is then maximised. It will be shown that for a given available power the
minimum required pump size depends on the ship speed. The available
power is governed by the installed diesel engine or gasturbine. This dictates
the selection procedure to a large extent.
2.1 Characteristic velocities in a waterjet system
In the equations for pump performance and thrust, use is made of some
specific velocities. Four main velocities are distinguished and will be used
throughout this thesis:
1. ship speed (v
ship
)
2. mass averaged ingested velocity at duct inlet (v
in
)
3. averaged axial inflow velocity at the pump entrance (v
pump
)
4. averaged outlet velocity at the nozzle (v
out
)
Figure 2.1 shows a sketch of a waterjet installation with the four velocities
indicated. A nonuniform velocity distribution is sketched to indicate the
development of the boundary layer along the hull surface, upstream of the
inlet. This figure is also used to give an impression of the dividing streamline.
By definition there will be no mass flow across this line. In three dimensions,
this line is extended to a dividing streamtube. The curved part of the inlet,
where the streamline ends, is denoted as inlet lip or cutwater.
The inlet velocity is determined at a crossflow plane just upstream of the
waterjet inlet, where the influence of the waterjet is not yet noticeable. The
ingested velocity distribution is massaveraged over the crosssectional
shape of the streamtube to find the actual inlet velocity v
in
:
(2.1)
v
out
v
ship
v
in
v
pump
Figure 2.1 Characteristic velocities in waterjet propulsion system
z
x
l
v
in
1
Q
 v z ( )v
n
A d
A
∫
=
2.1 Characteristic velocities in a waterjet system
Numerical analysis of a waterjet propulsion system 17
where v(z) is the velocity distribution in the boundary layer.
The four velocities are related by three parameters; wake fraction, inlet
velocity ratio and jet velocity ratio. These three parameters are discussed in
detail in this section.
2.1.1 Wake fraction
The water that is ingested into the waterjet inlet channel partly originates from
the hull’s boundary layer. The mass averaged velocity of the ingested water
(v
in
) is lower than the ship speed due to this boundary layer. The velocity
deficit is expressed as the momentum wake fraction (w), which is defined as:
(2.2)
Calculation of the wake fraction is rather complex, since the crosssectional
shape of the streamtube is not known a priori. Experiments have revealed
that the crosssection of the streamtube has a semielliptical shape under the
hull [1]. This is often simplified by a rectangular box with a width of 1.3 times
the pump diameter. Some comparisons have been made with experimental
results [2], [3] and it is concluded that the resulting value for the wake fraction
can be determined within acceptable limits, if the rectangular box
approximation is used.
For a given volume flow rate through the waterjet the height of the box can be
calculated once the velocity distribution in the boundary layer is known.
Standard theory for a flat plate boundary layer, as described in several
textbooks ([4], [5]) can be used to get a first indication of the velocity
distribution. It is convenient to use a power law velocity profile for the
boundary layer velocity distribution:
(2.3)
where v denotes the local velocity in the boundary layer at a distance z
normal to the wall, the undisturbed velocity, δ the local boundary layer
thickness and n the power law index.
Besides the thickness of the boundary layer δ there are also derived
quantities like the displacement thickness δ
1
and momentum thickness δ
2
of
the boundary layer.
w 1
v
i n
v
shi p
 – =
v
U
∞

z
δ

\ .
 
1
n

=
U
∞
18
Chapter 2. Waterjet propulsion theory
The momentum thickness can be related to the wall friction coefficient c
f
(l) for
a flat plate:
(2.4)
where l is the wetted length. This relation gives the frictional drag of the flat
plate in terms of the development of the boundary layer.
Substitution of the power law velocity distribution in the definitions of the
boundary displacement thickness δ
1
, the momentum thickness δ
2
and the
energy thickness δ
3
results in a set of the following relations:
(2.5)
(2.6)
(2.7)
Combination of equations (2.4) and (2.6) gives an expression for the friction
coefficient c
f
(l) as function of the boundary layer thickness δ(l) and the power
law exponent n. For turbulent flow a value of n = 7 is often used. With aid of
the analysis of developed turbulent pipe flow, an expression for the flat plate
boundary layer thickness is derived:
(2.8)
where Re
l
is the Reynolds number based on the wetted length. The wall
friction coefficient for n=7 becomes:
(2.9)
Comparison with experimental data shows good agreement for Reynolds
numbers between 5x10
5
and 10
7
(see [4]).
In general, full scale waterjet installations operate at Reynolds numbers of
about 10
9
, which is 2 orders of magnitude larger. The wall friction coefficient
for a flat plate cannot be based on equation (2.9) at these high Reynolds
numbers. Several logaritmic equations for the flat plate wall friction coefficient
are defined for high Reynolds numbers. A typical example is the ITTC’57
c
f
l ( )
τ
w
1
2
ρU
∞
2
 ≡ 2
l d
dδ
2
=
δ
1
δ
n 1 +
 =
δ
2
δn
n 1 + ( ) n 2 + ( )
 =
δ
3
2δn
n 1 + ( ) n 3 + ( )
 =
δ
n 7 =
0.370 l Re
l
1 5 ⁄ –
⋅ ⋅ =
c
f
l ( ) 0.0576 Re
l
1 5 ⁄ –
⋅ =
2.1 Characteristic velocities in a waterjet system
Numerical analysis of a waterjet propulsion system 19
friction line, which is commonly used to extrapolate the viscous resistance
component of a model scale ship to full scale dimensions.
The logaritmic friction line gives the wall friction coefficient as function of
Reynolds number. Based on equations (2.4) and (2.6), there is a relation
between friction coefficient c
f
(l), boundary layer thickness δ(l) and power law
exponent n. The actual power law exponent n is determined from velocity
profile measurements by Wieghart. Results of measurements at different
Reynolds numbers are presented in Schlichting [4]. For a certain Reynolds
number, the corresponding boundary layer thickness can be calculated once
the wall friction coefficient c
f
(l) and the power law exponent n are established.
Full scale measurements of the hull boundary layer velocity distribution are
presented by Svensson [7]. Velocity profiles are measured on two different
vessels and at different ship speeds. This results in a large variation of
Reynolds numbers. A reasonable fit of a power law profile with n = 9 and the
measured values is found. The equation for the boundary layer thickness, as
given in equation (2.8), is modified for n = 9:
(2.10)
It can be noticed that both the constant as well as the power of the Reynolds
number have to be changed when the value of n is changed. This is in
accordance with measurements of Wieghardt (see [4]). Adjustment of these
values should result in the right boundary layer thickness and in an accurate
prediction of the velocity profile and the wall friction. For Reynolds numbers of
order 10
9
the power law exponent becomes 10 to 11.
With known boundary layer thickness and volume flow through the pump, the
average incoming velocity and thus the wake fraction can be calculated. A
typical value for the wake fraction w is 0.10 to 0.14 for a fast ferry.
The accuracy of the rectangular box approximation will be reviewed in
chapter 5 when CFD calculations of the flow through the waterjet inlet duct
are discussed. With the numerical method it is possible to visualize the actual
shape of the streamtube and determine the massaveraged velocity by
numerical integration. This numerical method is based on the computed
shape of the streamtube, whereas the determination of the wake fraction in
experiments is based on an approximated shape of the streamtube.
Consequently, the wake fraction obtained from the numerical results is more
accurate than the one obtained from experimental data.
δ
n 9 =
0.270 l Re
l
1 – 6 ⁄
⋅ ⋅ =
20
Chapter 2. Waterjet propulsion theory
2.1.2 Inlet velocity ratio
The averaged axial inflow velocity of the pump is denoted by v
pump
. This
velocity can be written as:
(2.11)
where Q is the volume flow through the pump and D
inlet
the diameter at the
suction side of the pump. This velocity is an important parameter to describe
the flow phenomena in the inlet, where the speed is changed from the ship
speed to the pump velocity. The pump velocity is related to the ship speed
through the Inlet Velocity Ratio (IVR):
(2.12)
At normal operating condition, IVR will be around 1.3 to 1.8. The reciprocal of
equation (2.12) is used in literature as well ([2], [8]) and used by the ITTC; this
results in values of this quantity, at operating conditions below 1 and a value
of infinite for zero ship speed. Use of the definition in equation (2.12) is
preferred since the operating range is bounded between 0 and about 2.5.
IVR is used to denote the flow conditions in the waterjet inlet duct. At
relatively low ship speed, e.g. during manoeuvring in harbour, IVR will be
smaller than 1. This means that the flow is accelerated upon entering the inlet
duct. In this condition the stagnation point of the dividing streamline is located
at the hull side of the inlet lip (or cutwater). This might lead to cavitation and/
or separation in the inlet at the upper side of the lip. Figure 2.2 shows a
sketch of the flow phenomena at low IVR condition.
If the vessel sails at design speed, the inlet flow phenomena are quite
different. As mentioned, the design IVR will be around 1.3 to 1.8. IVR values
v
pump
Q
π
4
D
i nl et
2
 =
IVR
v
shi p
v
pump
 =
Increased risk for
Figure 2.2 Flow phenomena at low IVR
cavitation and/or separation
2.1 Characteristic velocities in a waterjet system
Numerical analysis of a waterjet propulsion system 21
of more than 2.0 are known for high speed motor yachts (>60 knots). This
implies a significant deceleration of the flow in the inlet. In this condition the
stagnation point is located at the inlet side of the cutwater. The critical
location for cavitation is located at the hull side of the lip for this condition.
The deceleration of the flow in the inlet duct leads to an adverse pressure
gradient in the inlet. If this pressure gradient becomes too large, flow
separation is likely to occur at the top side of the inlet. The possible flow
phenomena at high IVR are sketched in figure 2.3.
Whether or not cavitation or separation really occurs in a practical situation,
strongly depends on the actual geometry of the inlet duct. With a good inlet
design cavitation and separation free operation is possible up to about 44
knots [9], which is a commonly used design speed for fast ferries.
It should be kept in mind, that an inlet has to be designed to cope with the low
IVR and the design IVR condition, because each vessel has to start from zero
ship speed.
2.1.3 Jet velocity ratio
The velocity v
out
at the outlet of the waterjet nozzle, is related to the volume
flow through the pump and the diameter of the nozzle as:
(2.13)
The outlet velocity is related to the incoming velocity by the jet velocity ratio µ:
according to [10]:
(2.14)
Increased risk of cavitation
Increased risk of separation
Figure 2.3 Flow phenomena at high IVR
v
out
Q
π
4
D
nozzl e
2
 =
µ
v
i n
v
out
 =
22
Chapter 2. Waterjet propulsion theory
The importance of the parameter µ will be shown in section 2.5, where the
overall propulsive efficiency of the waterjet system is derived. It will be shown
that typical values are in the range of 0.5 to 0.7.
2.1.4 Summary
In this section four velocities are introduced; v
ship
, v
in
, v
pump
and v
out
. The
relations between these velocities are defined by three ratios: wake fraction
w, inlet velocity ratio IVR and nozzle velocity ratio µ. The theory of waterjet
propulsion will be based on these velocities and ratios.
2.2 General pump theory
In this section a short overview of the standard pump theory is given in order
to introduce a set of parameters to describe the pump performance. This
theory can be found in many textbooks about centrifugal pumps, see for
example [11], [12].
2.2.1 Dimensionless performance parameters
Performance of a pump can be expressed in terms of a set of non
dimensional parameters. The performance is expressed in terms of flow rate,
head and cavitation behaviour. In dimensionless form, the flow rate through
the pump is given as the flow coefficient ϕ:
(2.15)
where Q is the flow rate in m
3
/s, Ω the speed of the impeller in rad/s and D
the impeller diameter in m. The head coefficient ψ of a pump is defined as:
(2.16)
where H is the head in m. It can be shown that geometrically similar pumps
have equal values for flow and head coefficient. This forms the basis of the
socalled similarity method. If the performance of a pump for a certain size
and shaft speed is known, equations (2.15) and (2.16) can be used to predict
the performance for different sizes and shaft speeds. Elimination of the
diameter D from equations (2.15) and (2.16) results in:
(2.17)
ϕ
Q
ΩD
3
 ≡
ψ
gH
ΩD ( )
2
 ≡
1
ΩQ
1 2 ⁄
gH ( )
3 4 ⁄

ψ
3 4 ⁄
ϕ
1 2 ⁄
 ⋅ =
2.2 General pump theory
Numerical analysis of a waterjet propulsion system 23
which leads to the definition of the specific speed of:
(2.18)
where Ω is the speed of the impeller in rad/s, Q the flow rate in m
3
/s and H
the head in m. It is also found that the similarity method implies that
geometrically similar pumps have equal values of specific speed:
(2.19)
The value of the specific speed of a specific pump gives a good indication of
its type: typical axial flow pumps have a specific speed above 2.4, whereas
radial flow pumps have low values of the specific speed (typically below 1.0).
Mixedflow pumps have intermediate values for the specific speed.
Pump efficiency η
pump
is defined as the ratio between the hydraulic power
P
hydr
, which is the product of flow rate and pressure rise, and the required
shaft power P
shaft
.
(2.20)
where T
q
is the shaft torque. The required shaft power can be expressed in a
nondimensional specific power P
*
:
(2.21)
The specific power is related to the flow coefficient, head coefficient and
pump efficiency. Combination of equations (2.15), (2.16) and (2.20) yields:
(2.22)
Strictly speaking, similarity of performance is only valid in cases of both
geometrically and dynamically similar internal flows. In this analysis viscosity
is not taken into account. Since hydraulic losses do scale differently,
additional empirical relations are used to predict the effect of these losses on
pump efficiency and specific power.
n
ω
Ω Q
g H ⋅ ( )
3 4 ⁄
 ≡
n
ω
ϕ
1 2 ⁄
ψ
3 4 ⁄
 =
η
pump
P
hydr
P
shaf t

ρgHQ
ΩT
q
 = =
P
*
P
shaf t
ρΩ
3
D
5
 =
P
* ϕψ
η
pump
 =
24
Chapter 2. Waterjet propulsion theory
2.2.2 Pump geometry parameters
It is shown in the preceding section, that the specific speed is found from the
expressions for flow coefficient and the head coefficient, when the diameter is
eliminated. In a similar way, the specific diameter δ is found, if the rotational
speed Ω is eliminated:
(2.23)
so that:
(2.24)
and:
(2.25)
The specific speed and specific diameter are based on the same two
parameters, namely ϕ and ψ. The relation between the two is represented in
the so called Cordierdiagram [14], which is based on experience from actual
pumps. Waterjet pump designs may deviate from this empirical rule for
conventional pumps due to the difference in functionality as outlined in
chapter 1.
The basic geometry of the impeller of conventional centrifugal pumps is
strongly related to the specific speed of a pump, however. The large
similarities in pump geometry lead to comparable efficiencies for different
pumps with the same specific speed. The statistically attainable optimal pump
efficiency can be derived from several published prediction formulas, based
on measured performances. An example of such empirical formula is given in
[15]:
(2.26)
where Q
ref
is set equal to 1 m
3
/s in order to maintain the nondimensional
representation. Figure 2.4 shows the expected maximum pump efficiency for
three flow rates. The highest efficiency is found at a specific speed of 1.0.
Decrease in efficiency is rather slow when the specific speed is increased to
values above 1.0. In general, waterjet pumps have a specific speed around
2.03.0.
1
gHD
4
Q
2

ϕ
2
ψ
 ⋅ =
δ D
gH ( )
1 4 ⁄
Q
 ⋅ ≡
δ
ψ
1 4 ⁄
ϕ
1 2 ⁄
 =
η
pump
0.95
0.05
Q Q
ref
⁄
3
  – 0.125 n
ω
( ) log [ ]
2
– =
2.2 General pump theory
Numerical analysis of a waterjet propulsion system 25
Achievable pump efficiencies around 90% for large pumps seem to be a
reasonable estimate. This value will be used in the remainder of this chapter
for estimates of the overall waterjet efficiency.
2.2.3 Cavitation parameters
For cavitation free operation the pump requires a certain available pressure
at the inlet, or suction side. This is denoted with the inception net positive
suction head (NPSH
i
), which is a pressure expressed in meters water
column. In general, pump operation is still possible beyond the cavitation
inception level, i.e. for lower NPSH levels. Therefore the criterion for the inlet
suction head is based on a certain loss of pump performance (for example 1
or 3% head loss or a certain percentage of pump efficiency decrease, see
[16]). Based on the choice for the admissible head loss, a required NPSH is
defined.
The required net positive suction head (NPSH
R
) can be made non
dimensional in a similar way as the head to form the suction coefficient κ:
(2.27)
Another, wellknown method to present the NPSH in dimensionless
representation is the Thoma number, defined as:
(2.28)
50%
55%
60%
65%
70%
75%
80%
85%
90%
95%
100%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Q= 0.1 m3/s
Q= 1.0 m3/s
Q= 10 m3/s
Figure 2.4 Maximum pump efficiency as function of specific speed, based
on equation (2.26).
η
p
u
m
p
[
−
]
Pump specific speed Nω [−]
κ
gNPSH
R
ΩD ( )
2
 =
σ
NPSH
R
H
 =
26
Chapter 2. Waterjet propulsion theory
The nondimensional parameters are related by the head coefficient.
The required NPSH can also be related to the flow rate and the rotational
speed of the impeller, similar to the pump specific speed. This gives the
suction specific speed of the pump n
ωs
, defined as:
(2.29)
The suction specific speed of a pump is more or less constant for all pump
types. Values of about 4.0 are common in commercial pumps [1417]. In
order to create some extra margin to accommodate cavitation, a design value
of 3.5 for a waterjet impeller is adopted.
2.2.4 Correlation with propeller performance parameters
The flow coefficient ϕ of a pump can be related to the propeller advance ratio
J (as given in eqn. (1.1)) with substitution of equations (2.11) and (2.12):
(2.30)
This relation shows the fundamental difference between an open propeller
and a waterjet installation, where in the waterjet IVR is introduced as an
additional parameter. This parameter is needed because of the principle of
internal flow of the pump compared to the external flow of the propeller.
In a similar way, the nondimensional head can be related to the thrust
coefficient of an open propeller. For a propeller the thrust coefficient is
defined as [18]:
(2.31)
The head H of a pump is related to the total pressure increase generated by
the impeller according to:
(2.32)
In actuator disk theory the production of thrust of an open propeller equals the
product of the pressure rise and the crosssectional area of the propeller:
(2.33)
It is assumed that the static pressure rise is equal to the total pressure rise,
due to the infinitesimal thickness of the actuator disk. This results in a relation
n
ωs
Ω Q
g NPSH
R
⋅ ( )
3 4 ⁄
 =
ϕ
πv
pump
4ΩD
 
J
8IVR
 = =
K
T
T
ρn
2
D
4
 =
∆p
t ot
ρgH =
T ∆p A
prop
⋅ =
2.3 Thrust
Numerical analysis of a waterjet propulsion system 27
between the head coefficient ψ of a pump and the thrust coefficient K
T
of a
propeller:
(2.34)
It is concluded that the QH curves of a pump are equivalent to the JK
T
curves of an open propeller. The main difference is caused by the used inflow
velocity.
2.3 Thrust
2.3.1 General thrust equation
The purpose of a propulsion installation is to produce thrust to propel a
vessel. Water is accelerated in the installation, which results in a reaction
force on the ship structure. The thrust can be derived from the momentum
balance for an incompressible fluid [5]:
(2.35)
The momentum balance states that the sum of all surface forces F
s
and all
body forces F
b
acting on the spatially fixed control volume V equals the rate
of change of momentum in the control volume with surface A. The surface
force is defined as:
(2.36)
where p is the static pressure, I the unit tensor and σ the viscous stress
tensor.
In the remainder of this section the steady flow situation will be analysed. As
a consequence, the first term on the right hand side of equation (2.35)
vanishes. Moreover, the body forces, like gravity, acting on the fluid will be
neglected.
In the following subsections the momentum balance will be derived for both
an open propeller and a waterjet.
ψ
∆p ( ) ρ ⁄
4π
2
n
2
D
2

K
T
π
3
 = =
F F
s
F
b
+
t ∂
∂
vρ V v
A
∫
+ d
V
∫
ρv A d ⋅ = =
F
s
p – I σ + ( ) A d ⋅
A
∫
=
28
Chapter 2. Waterjet propulsion theory
2.3.2 Open propeller thrust
An expression for the thrust of an open propeller is determined with equation
(2.35) [6]. The propeller is treated as an actuator disk, which is a singularity
modelled by a body force acting over an infinitesimal thin disk. The control
volume consists of the streamtube of fluid which passes through the propeller
plane area. Figure 2.5 shows a sketch of the control volume of an open
propeller with the nomenclature of the velocities.
Evaluation of the momentum balance is split in two parts; the contribution of
the momentum fluxes and the contribution of the surface forces. The
contributions of the momentum fluxes in xdirection result in a net momentum
flux component in xdirection of:
(2.37)
This can be rewritten, with aid of the continuity condition, as:
(2.38)
The contributions of the surface forces in xdirection are defined as:
(2.39)
It is assumed that the pressure at the inlet (far upstream) and at the outlet (far
downstream) is equal to the ambient pressure . Moreover, the contribution
of the viscous forces is neglected on the inlet and outlet area as well as on
v
in
v
out
v
prop
p
tube
Figure 2.5 Control volume for the momentum balance applied to an
propeller within a streamtube
A
in
A
out
A
tube
z
x
p p
∞
=
p p
∞
=
φ
mx
ρv
out
2
A
out
ρv
i n
2
A
i n
– =
φ
mx
ρv
prop
A
prop
v
out
v
i n
– ( ) =
F
x
T
prop
– p p
∞
– ( ) A d
A
i n
∫
p p
∞
– ( ) A d
A
out
∫
+ – p p
∞
– ( )x A d ⋅
A
t ube
∫
+ =
p
∞
2.3 Thrust
Numerical analysis of a waterjet propulsion system 29
the streamtube surface. Combination of equations (2.38) and (2.39) gives the
final thrust equation for an open propeller, based on the momentum balance:
(2.40)
where A
prop
is the crosssectional area of the propeller plane, x the unit
vector in xdirection and A
tube
the streamtube surface. The contribution of the
pressure acting on the streamtube to the thrust vanishes, based on the
paradox of d’Alembert, if the streamlines are aligned in xdirection far
upstream and downstream.
If Bernoulli’s theorem is applied along the streamlines in the part of the
control volume upstream and downstream of the propeller, a second relation
for the propeller thrust is found:
(2.41)
Combination of the momentum balance and Bernoulli’s law, leads to a simple
relation between the inlet and outlet velocity and the volume flow through the
propeller disk (see [18]):
(2.42)
It can be seen that the velocity through the disk is the average of the
upstream and downstream velocities. The difference between the velocity
through the disk and the incoming velocity is called the induced velocity v
ind
.
Thrust loading coefficient
Loading of an open propeller is often expressed by the propeller loading
coefficient, defined as [18]:
(2.43)
where A
prop
is the crosssectional area of the propeller disk, based on the
propeller diameter. The propeller loading coefficient can be expressed in
terms of the ratios as defined in section 2.1. Substitution of equation (2.41),
with the inflow velocity equal to the ship speed, i.e. v
in
=v
ship
, yields:
T
prop
ρA
prop
v
prop
v
out
v
i n
– ( ) p p
∞
– ( )x A d ⋅
A
t ube
∫
– =
T
prop
∆p A
prop
⋅ ρA
prop
1
2
 v
out
2
v
i n
2
– ( ) ⋅ = =
v
prop
1
2
 v
i n
v
out
+ ( ) v
i n
v
i nd
+ = =
C
Tprop
T
prop
1
2
ρv
ship
2
A
prop
 =
30
Chapter 2. Waterjet propulsion theory
(2.44)
With µ<1 the propeller loading coefficient is thus directly related to the jet
velocity ratio.
The jet velocity ratio can be related to the IVR, if equation (2.42) is substituted
into equation (2.12):
(2.45a)
It can be seen that open propellers always operate at IVR values below 1.
After rearranging this equation, it is shown that the IVR is equal to Froude
efficiency as given in equation (1.3):
(2.45b)
Although the term IVR is not used in the theory for open propellers, it is
already present as the Froude efficiency.
2.3.3 Waterjet thrust
For the determination of the thrust of a waterjet installation in general the
same approach as for the open propeller is used. The control volume will be
bounded by the streamtube surface on one side and the solid wall on the
other side. It is assumed that the inlet and exit planes are perpendicular to the
xdirection and the hull is parallel to the xaxis. Figure 2.6 shows the control
volume and the contributing terms to the momentum balance. The forces
acting on the waterjet structure, which are included in this control volume, are
denoted as T
wj,tube
.
It is noted that the control volume based on the streamtube of the ingested
water does not take into account the part of the waterjet inlet structure at the
hull side near the cutwater lip, which is excluded from the streamtube control
volume. The thrust or drag on that part of the waterjet structure will be
denoted will T
wj,hull
. At high IVR conditions a significant part of the cutwater
geometry belongs to the excluded cutwater region. The subdivision of the
complete waterjet inlet structure into the part, which is included in the
streamtube approach, and the part which is excluded is shown in figure 2.7.
C
Tprop
1
2
ρ v
out
2
v
in
2
– ( )A
prop
1
2
ρv
i n
2
A
prop

v
out
v
i n

\ .
 
2
1 –
1 µ
2
–
µ
2
 = = =
IVR
v
i n
v
prop

2v
i n
v
i n
v
out
+

2µ
µ 1 +
 = = =
IVR
2
1
1
µ
 +

2
1 v
out
v
in
⁄ ( ) +
 η
p
= = =
2.3 Thrust
Numerical analysis of a waterjet propulsion system 31
The total thrust T
wj,all
of a waterjet is therefore:
(2.46)
Application of the momentum balance for a waterjet learns that there are two
momentum flux terms that contribute to the force in xdirection; these are the
fluxes at the nozzle exit surface A
out
and at the plane A
in
upstream of the
inlet:
v
out
v
in
p
tube
Figure 2.6 Control volume for a momentum balance on the streamtube of
the ingested water of a waterjet installation
z
x
A
in
A
tube
A
out
A
wj
A
hull
p p
∞
=
p p
∞
=
all solid wall cells
streamtube solid wall cells
remaining hull cells
Figure 2.7 Subdivision of all solid wall cells of the waterjet installation into
group belonging to streamtube control volume (left) and group
of remaining cells on hull (right)
T
wj al l ,
T
wj t ube ,
T
wj hul l ,
+ =
32
Chapter 2. Waterjet propulsion theory
(2.47)
where v
in
is the mass averaged inflow velocity. With aid of the continuity
condition, this becomes:
(2.48)
where Q is the flow rate through the waterjet installation. The contributions of
the surface forces in xdirection are defined as:
(2.49)
Similar to the open propeller, it is assumed that volumetric forces and viscous
forces can be neglected, while the pressure levels at the inlet (far upstream)
and at the outlet (far downstream) are equal to the ambient pressure .
Effect of the viscous forces is neglected also on these two planes, though
there is a nonuniform velocity distribution present at the inlet plane A
in
.
Contribution of this shear stress force is assumed to be negligible. With
equations (2.48) and (2.49) can be combined to get the expression for the
waterjet thrust in xdirection based on the streamtube momentum balance:
(2.50)
The contribution of the streamtube pressure can not be quantified analytically,
since the shape of the streamtube and the pressure distribution are unknown.
Even with numerical methods it is a very complex task to determine this
value, due to the threedimensional shape of the streamtube surface and the
dependency of the shape on IVR. In chapter 7 the contribution of the
streamtube pressure term will be reviewed in more detail.
The thrust of the complete waterjet installation is found, when equation (2.50)
is substituted in equation (2.46), which yields:
(2.51)
The last two terms on the righthandside are assumed to be small compared
to the first term, and often neglected in waterjet propulsion literature. The
φ
mx
ρv
out
2
A
out
ρv
i n
2
A
i n
– =
φ
mx
ρQ v
out
v
in
– ( ) =
F
x
T
wj t ube ,
– p p
∞
– ( ) A d
A
i n
∫
p p
∞
– ( ) A d
A
out
∫
+ – p p
∞
– ( )x A d ⋅
A
t ube
∫
+ =
p
∞
T
wj t ube ,
ρQ v
out
v
i n
– ( ) p p
∞
– ( )x A d ⋅
A
t ube
∫
– =
T
wj al l ,
ρQ v
out
v
i n
– ( ) px A d ⋅
A
t ube
∫
– T
wj hul l ,
+ =
2.3 Thrust
Numerical analysis of a waterjet propulsion system 33
influence of this simplification will be addressed in more detail in chapter 7.
The resulting simplified thrust equation for a waterjet becomes [10]:
(2.52)
Despite neglecting the streamtube and hull surface forces, this simplified
equation can be used to explain the main theory on waterjet propulsion. This
equation shows the three main parameters of a waterjet propulsion system:
the volume flow rate Q through the system, the nozzle exit area A
nozzle
and
the jet velocity ratio µ.
Thrust loading coefficients
The thrust loading coefficient of a waterjet installation can be based on the
nozzle outlet area or the pump inlet area. The thrust loading coefficient based
on nozzle exit area is discussed in [13]. With the nozzle area as reference
area, the relation between jet velocity ratio and the thrust loading coefficient
becomes:
(2.53)
where w is the wake fraction according to equation (2.2). The wake fraction
becomes zero, when the inflow velocity is equal to the ship speed, i.e.
v
in
=v
ship
. This is equivalent with an open water test of a propeller with uniform
inflow. The resulting loading coefficient for a waterjet with undisturbed inflow
yields:
(2.54)
Comparison with the open propeller thrust loading coefficient (equation
(2.44)) reveals a difference between the waterjet and the open propeller. This
is due to the fact that a waterjet is an internal flow machine. For a waterjet the
ratio between the inlet and nozzle area is fixed, whereas it is related to the
thrust for an open propeller.
The waterjet thrust loading coefficient can also be based on the pump inlet
diameter. In this way the dimensions of the complete installation are
recognised more clearly. This approach is more in agreement with the open
propeller thrust loading coefficient, where the propeller diameter is used.
T
wj
ρQ v
out
v
i n
– ( )
ρQ
2
A
nozzl e
 1 µ – ( ) = =
C
Tnozzl e
T
1
2
ρv
shi p
2
A
nozzl e

2 1 µ – ( ) 1 w – ( )
2
µ
2
 = =
C
Tnozzl e
w 0 =
2 1 µ – ( )
µ
2
 =
34
Chapter 2. Waterjet propulsion theory
(2.55)
The thrust loading coefficient based on the pump inlet diameter shows that
the IVR is introduced to describe the system performance. This gives the
designer of waterjets another optimization option, compared to open
propellers.
2.3.4 Concluding remarks
In a waterjet there is no direct relation between the IVR and µ like there is for
an open propeller. Since it is an internal flow machine, part of the thrust can
be transferred to the hull structure via the transom stern and the inlet ducting.
On the other hand, it can also appear that the thrust acting on the shaft will
exceed the total thrust of the installation [19]. In such condition a negative
thrust acts on the transom stern or the inlet ducting. For conventional pumps
the axial thrust is to be kept as low as possible. Thrust production is not
regarded as an important performance indicator, like efficiency and head as
function of the mass flow.
In case of a waterjet, the thrust can be calculated, if the values for the
velocities v
in
, v
pump
and v
out
are known. These can be related to the mass
flow for a given geometry of the waterjet installation. This mass flow through
the system is related to the pump head. In this way the standard pump
performance characteristics, like head curve, efficiency and cavitation
behaviour, can be used to evaluate the performance of a waterjet installation.
2.4 Pump head
The required head of a waterjet installation will be discussed in this section.
The head H of a pump represents the increase of total pressure in a pump
measured in meters liquid water column as given in equation (2.32).
The volume flow rate through the system follows from the intersection of the
required system head curve and the pump head curve. The pump head curve
depends on the type of pump used in the waterjet system. In general, mixed
flow pumps have a headcurve with a negative slope in the design point to
ensure a stable operating point. For lower volume flow rates the slope may
become zero or even negative. For the sake of simplicity, the pump head
curve, as used in the examples in this section, is assumed to be a linear
function of flow rate.
C
Tpump
T
1
2
ρv
shi p
2
A
pump

2 1 µ – ( ) 1 w – ( )
2
IVR µ ⋅
 = =
2.4 Pump head
Numerical analysis of a waterjet propulsion system 35
The required system head curve can be regarded as a pipe resistance curve
of the waterjet installation. The acceleration of the fluid in the nozzle requires
a certain pressure difference. Additional head is required to overcome the
hydraulic losses in the inlet and the nozzle. However, the energy of the
ingested fluid can be used partly, which is beneficial for the head
requirement. Finally, the waterjet nozzle may be positioned above the
waterline, which will require some more pump head. All contributions together
give the equation for the required system head H
R
:
(2.56)
where φ is the nozzle loss coefficient, ε the inlet loss coefficient and h
j
the
nozzle elevation above the waterline. The elevation of the nozzle is limited by
the selfpriming requirement of the waterjet installation. In general, the
elevation h
j
can be neglected relative to the other contributions in equation
(2.56).
Equation (2.56) shows a positive contribution from the incoming velocity,
therefore the system performance is coupled to the ship speed. Strictly
speaking, the average ingested velocity v
in
should be based on a mass
averaged dynamic pressure term:
(2.57)
whereas v
in
in equation (2.56) is based on the mass averaged velocity as
given in equation (2.1). The difference between the two methods can be
expressed in the powerlaw exponent, assumed that the water is ingested
completely out of the boundary layer:
(2.58)
The difference between the two methods of averaging is less than 1% for a
powerlaw exponent of n=9. The error will be even smaller if the water is
ingested from the undisturbed fluid. In general, the introduced deviation is
compensated for in the determination of the loss coefficient.
At constant ship speed, the required system head H
R
can be approximated
as a quadratic function of the flow rate Q. The slope of this quadratic curve
depends on the nozzle diameter. Figure 2.8 shows an example of a pump
head diagram with a pump head and efficiency curve and two system lines for
a constant ship speed and different nozzle sizes. The assumption of constant
H
R
v
out
2
2g
 1 φ + ( )
v
i n
2
2g
 1 ε – ( ) – h
j
+ =
v
i n
˜
1
Q
 v z ( )
2
v
n
A d
A
∫
1 2 ⁄
=
v
in
˜
2
v
i n
2

n 2 + ( )
2
n 3 + ( ) n 1 + ( )
 =
36
Chapter 2. Waterjet propulsion theory
ship speed is a hypothetical condition, since in actual situations, the ship
speed will depend on the delivered thrust, which is in turn related to the flow
rate.
The system lines are based on different nozzle diameters. An increase of the
nozzle diameter, results in a lower nozzle velocity for constant volume flow.
This leads to a lower required head. The work point of the pump can be
controlled by the size of the nozzle exit area.
The effect of the ship speed on the volume flow through the installation is
shown in figure 2.9.In this pump head diagram, the system lines of ship
speeds of 20, 30 and 40 knots are plotted. The nozzle size is kept constant in
this figure.
It can be observed from this diagram, that the increase in volume flow
between 20 and 40 knots is only 6%. This increase in velocity through the
impeller results in a small change of the pump operating point. It is concluded
that a waterjet installation can operate in a relatively small range of flow rates.
2
1
0
1
2
3
4
5
6
7
8
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Pump_head
System_head_nozzle1
System_head_nozzle2
Work point
Pump efficiency
INCREASING
NOZZLE DIAMETER
Figure 2.8 Pump head diagram for different nozzle sizes and constant
ship speed
ψ
[
−
]
η
p
u
m
p
[
−
]
ϕ [−]
2.5 Overall propulsive efficiency
Numerical analysis of a waterjet propulsion system 37
2.5 Overall propulsive efficiency
This section deals with the influence of the parameter µ on the overall
propulsive efficiency. If the propulsion system is regarded as a black box,
then engine power P
shaft
is input and thrust T at a certain ship speed is
output. The overall propulsive efficiency η
d
of this black box is then based on
the bare hull resistance R
bh
of a vessel [18]:
(2.59)
where R
bh
is the bare hull ship resistance and P
shaft
the power at the waterjet
shaft.
In conventional naval architecture theory, the resistance of a ship with an
active propeller is found to be different from the bare hull resistance. Due to
the action of the propeller, a low pressure region at the rear of the vessel is
created, which results in an increased drag of the vessel. The difference
between the bare hull resistance R
bh
and the required thrust T at a certain
ship speed is expressed in terms of the thrust deduction factor t according to:
(2.60)
2
1
0
1
2
3
4
5
6
7
8
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Pump_head
system_head_20 knots
system_head_30 knots
system_head_40 knots
work point
Pump efficiency
INCREASING
SHIP SPEED
Figure 2.9 Pump head diagram for different ship speeds
ψ
[
−
]
η
p
u
m
p
[
−
]
ϕ [−]
η
d
R
bh
v
shi p
⋅
P
shaft
 =
R
bh
1 t – ( ) T ⋅ =
38
Chapter 2. Waterjet propulsion theory
For a propeller the thrust deduction factor is always positive, which leads to a
higher ship resistance due to the action of the propeller and therefore a
higher required thrust.
In waterjet propulsion theory, the thrust deduction factor t can be used to
account for the effects of (i) the neglected surface forces such as the force on
the streamtube and the force on the region aft of the waterjet inlet and (ii) a
change in the pressure distribution along the hull. This approach is used by
Van Terwisga [2], where a jet thrust deduction factor t
j
and a resistance
increment factor 1+r are introduced.
Substitution of equations (2.2), (2.20) and (2.60) in equation (2.59) gives:
(2.61)
In the next step, equations (2.14), (2.52) and (2.56) are substituted into
equation (2.61). After rearranging of all variables, the equation for overall
propulsive efficiency becomes:
(2.62)
where the first term is denoted as hull efficiency:
(2.63)
Eqn (2.62) shows that the overall propulsive efficiency is mainly a function of
the jet velocity ratio µ, since the hull efficiency η
hull
and the pump efficiency
η
pump
as well as the inlet and nozzle loss coefficients may be regarded as
constant values in a first approximation.
Figure 2.10 shows the overall propulsive efficiency for three inlet loss
coefficients. Thrust deduction is set to t=0.02, wake fraction to w=0.12, pump
efficiency is 90% (η
pump
= 0.90) and outlet loss coefficient is φ=0.02. Also
plotted is the ideal efficiency, where all losses are neglected. This efficiency is
defined already as Froude efficiency in equation (2.45) for an open propeller.
It is obvious that the optimum propulsive efficiency can be obtained, if the jet
velocity ratio is in the range of 0.65 to 0.75 depending on the inlet loss
coefficient ε. In general the design point is chosen at a jet velocity ratio, which
is slightly below the best efficiency point. This part of the curve is relatively
flat, which results in a stable working point, when the inflow conditions show
some variation.
η
d
1 t – ( )
1 w – ( )
η
pump
T v
i n
⋅
ρgHQ
 =
η
d
1 t – ( )
1 w – ( )
 η
pump
2µ 1 µ – ( )
1 φ + ( ) µ
2
1 ε – ( ) –
 ⋅ ⋅ =
η
hul l
1 t – ( )
1 w – ( )
 =
2.5 Overall propulsive efficiency
Numerical analysis of a waterjet propulsion system 39
2.5.1 Cavitation margins
The waterjet pump needs a certain level of the pressure at the suction side of
the pump in order to prevent cavitation. This required pressure is expressed
in the required net positive suction head (NPSH
R
), which is introduced in
section 2.2.3. Pump operation is allowed as long as the available NPSH
exceeds the required NPSH. The available suction head is the total head at
the inlet of the pump minus the vapour pressure of the liquid. For a waterjet
installation the available suction head is determined by the waterjet operating
point. The NPSH
A
can be expressed as function of the ship speed:
(2.64)
where h
j
represents the elevation of the pump above the waterline. Due to
selfpriming constrains of the pump, this elevation is negligible in most cases.
The pump will perform well as long as the required inlet suction head
(NPSH
R
) is below the available inlet suction head (NPSH
A
):
(2.65)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
.
.
.
Figure 2.10 Overall propulsive efficiency η
d
as function of jet velocity ratio
µ. Thrust deduction factor t=0.02, wake fraction w=0.12, pump
efficiency η
pump
=90%, nozzle loss coefficient φ=0.02
η
d
[
−
]
Jet velocity ratio µ [−]
increasing
inlet loss
coefficient ε
ideal case
ε = 0.10
ε = 0.20
ε = 0.30
NPSH
A
p
∞
p
v
–
ρg

v
i n
2
2g
 1 ε – ( ) + =
p
∞
p
v
–
ρg

v
shi p
2
2g
 1 ε – ( ) 1 w – ( )
2
h
j
– + =
NPSH
R
NPSH
A
≤
40
Chapter 2. Waterjet propulsion theory
Combination of equations (2.18) and (2.29), with the requirement of equation
(2.65) yields:
(2.66)
2.5.2 Limitations in specific speed
The expressions for pump head (2.56) and available suction head (2.64), with
negligible pump elevation, can be substituted into equation (2.66). Given
values for the suction specific speed n
ωs
, the wake fraction w and the inlet
and outlet loss coefficients ε and φ, a function of the ship speed and the jet
velocity ratio is found for the maximum allowable specific speed:
(2.67)
This equation shows that the allowable specific speed of the pump will be
limited for constant jet velocity ratio µ, when the ship speed increases. This
phenomenon is illustrated in figure 2.11 for different values of the jet velocity
ratio.
n
ω
n
ωs
NPSH
R
H

\ .
 
3 4 ⁄
n
ωs
NPSH
A
H

\ .
 
3 4 ⁄
≤ =
n
ω
n
ωs
2 p
∞
p
v
– ( )
ρv
shi p
2
1 w – ( )
2
 1 ε – ( ) +
1 φ +
µ
2

\ .
 
1 ε – ( ) –

3 4 ⁄
⋅ ≤
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
20 25 30 35 40 45 50 55 60 65 70 75
.
.
.
.
Figure 2.11 Maximum allowable pump specific speed as a function of ship
speed for various jet velocity ratios. Inlet loss ε = 0.20, outlet
loss φ = 0.02, wake fraction w = 0.12 and suction specific
speed n
ωs
= 3.5.
P
u
m
p
s
p
e
c
i
f
i
c
s
p
e
e
d
n
ω
[
−
]
Ship speed v
ship
[knots]
increasing jet
velocity ratio µ
allowable specific speed n
ω
µ=0.7
0
µ=0.6
5
2.5 Overall propulsive efficiency
Numerical analysis of a waterjet propulsion system 41
For a given design speed and a chosen jet velocity ratio the maximum
allowable pump specific speed can be determined. On the other hand, for a
certain available pump type, with a known specific speed, the range of
possible jet velocity ratios can be determined for a given design ship speed.
2.5.3 Limitations in jet velocity ratio
In practice, to cover the complete speed range waterjet manufacturers use a
set of standard pumps with different specific speeds. The allowable jet
velocity ratio for given specific speed can be determined after rearranging
equation (2.67). Figure 2.12 shows the minimum allowable jet velocity ratio µ
for a number of specific pump speeds n
ω
. This diagram shows that the
optimal jet velocity ratio µ can be selected over a large range of ship speeds
with a limited number of different pumps.
A jet velocity ratio of 0.7 is possible up to 35 knots for a pump with a specific
speed of 4.0. A pump with a specific speed of 3.0 can be used at 65 knots for
this jet velocity ratio, however. It is also shown that the range of allowable jet
velocity ratios at very high ship speeds (>60 knots) increases significantly for
a radialflow type pump (n
ω
=2.0) compared to an axial flow type pump. A
pump with a specific speed of 4.0 requires a minimum jet velocity ratio of 0.78
at 65 knots. This condition may be at the right side of the optimum propulsive
efficiency curve, as shown in figure 2.10. Since this is undesirable, a pump
with lower specific speed is to be used.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
20 25 30 35 40 45 50 55 60 65 70 75
.
.
.
nω=4.
0
nω=3.
Figure 2.12 Minimum allowable jet velocity ratio as a function of ship
speed for various pump specific speeds. Inlet loss ε = 0.20,
outlet loss φ = 0.02, wake fraction w = 0.12 and suction
specific speed n
ωs
= 3.5.
J
e
t
v
e
l
o
c
i
t
y
r
a
t
i
o
µ
[
−
]
Ship speed v
ship
[knots]
allowable jet velocity ratio µ
increasing pump
specific speed n
ω
42
Chapter 2. Waterjet propulsion theory
2.5.4 Limitation of power density
Weight reduction of a waterjet installation is an important issue. This can be
achieved with highly loaded, small installations. This criterion can be
expressed in terms of the power density P/D
2
given in terms of the specific
power P* (according to equation (2.22)):
(2.68)
with (ΩD) twice the tip speed of the impeller:
(2.69)
Cavitation behaviour of different pump sizes with the same specific speed
can be compared with the net positive suction head coefficient κ, as defined
in equation (2.27). This gives a relation between the required NPSH and the
tip speed. Since the available NPSH is a function of the ship speed, it is found
that the allowable tip speed increases with ship speed. Consequently, the
allowable power density P/D
2
increases also with increasing ship speed.
Effect of the ship speed on the allowable power density is shown in figure
2.13. It should be obvious that an increase of the impeller diameter reduces
the power density for given engine power. Consequently, the minimum
allowable impeller diameter as function of the ship speed can be determined
P
shaf t
D
2
 ρP
*
ΩD ( )
3
=
v
t ip
πnD
1
2
ΩD = =
Figure 2.13 Maximum allowable power density as function of ship speed.
Inlet loss ε = 0.20, wake fraction w = 0.12, nondimensional
flow coefficient ϕ = 0.2, specific speed n
s
= 3.0 and suction
specific speed n
ωs
= 3.5.
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
10000
20 25 30 35 40 45 50 55 60 65 70 75
Maximum power density
OPERATION
ALLOWED
Ship speed v
ship
[knots]
P
o
w
e
r
d
e
n
s
i
t
y
P
/
D
2
[
k
W
/
m
2
]
2.6 Waterjet selection
Numerical analysis of a waterjet propulsion system 43
from the power density. It is found that the minimum allowable size depends
on the cavitation behaviour of the pump, the ship speed and the power
applied to the pump.
2.6 Waterjet selection
Up to this point, the entire analysis has been based on optimal performance
at one single design operating condition. In many cases waterjet installations
have multiple operating points. If the waterjets are applied for example in
large fast ferries or planing hulls, the resistance line is quite different from the
one of a regular displacement vessel. There is an additional resistance at a
speed range between about 20 and 30 knots. The speed, at which the local
maximum resistance occurs, is denoted as hump speed. The waterjet
installation has to provide sufficient thrust to exceed the resistance at the
hump speed. This requirement may lead to a larger waterjet than necessary
for the design operating point.
In all shown examples, values for the wake fraction µ, pump efficiency η
pump
,
inlet losses, etc., have been considered as constants. In actual installations
all these parameters depend on the ship speed and/or the flow rate Q through
the installation. Implementation of all of these dependencies will result in a
complex waterjet performance prediction program.
For a realistic comparison of various installations with different sizes over the
complete range of ship speeds, such program should be used in order to take
the actual values for the loss coefficients and the efficiencies into account.
Figure 2.14 shows the output of a waterjet performance prediction program
as used at the authors’ company for three different sized waterjets. The thrust
is kept constant and the size is changed to show the effect of power density
on the cavitation margins. The hatched area represents the noncontinuous
operational region. In this region severe cavitation in the waterjet will be
present. The upper line represents the thrust breakdown line and the lower
line of this area is denoted as the 1% cavitation line. Thrust breakdown
occurs when the mass flow through the system collapses due to extreme
cavitation.
In the same figure an indication of a typical resistance curve of a fast ferry is
plotted. The increased resistance at the hump speed of 25 knots can be
noticed. The effect of waterjet size is obvious from this figure. The resistance
at the hump speed is about equal to the maximum thrust of the small waterjet.
This means that the vessel has no extra thrust available to accelerate. The
maximum speed will not exceed 25 knots in this case, which is only 55% of
the design speed. The larger waterjets, with lower power densities, have
sufficient margin up to the design speed. To get a good balance between
cavitation performance and weight of the installation, the midsize jet will be
selected for this application.
44
Chapter 2. Waterjet propulsion theory
2.7 Closing remark
All equations and empirical values in this chapter are based on a uniform
inflow to the pump. However, the inflow velocity field of a waterjet shows a
strong nonuniform distribution, with variations in radial and tangential
direction. This inflow might have an influence on the empirical values or the
pump performance itself. Effects of the nonuniformity on pump performance
will be reviewed in more detail in chapter 7, where the complete waterjet
installation is analysed numerically.
2.8 Nomenclature
A area m
2
c
f
(l) local wall friction coefficient 
D diameter m
F force N
g gravitational acceleration m/s
2
H pump head m
h
j
nozzle elevation m
IVR inlet velocity ratio (v
ship
/v
pump
) 
l wetted length m
NPSH Nett positive suction head m
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
0 5 10 15 20 25 30 35 40 45 50
P/D^2 = 7000 kW/m^2
P/D^2 = 5000 kW/m^2
P/D^2 = 3000 kW/m^2
Resitance
P/D
2
=7000
kW/m
2
P/D
2
=5000
Figure 2.14 Maximum thrust curves for three different waterjet sizes and
resistance curves for fast ferry.
Ship speed v
ship
[knots]
T
h
r
u
s
t
[
%
]
7000 kW/m
2
5000 kW/m
2
3000 kW/m
2
2.8 Nomenclature
Numerical analysis of a waterjet propulsion system 45
n
ω
pump specific speed 
n
ωs
suction specific speed 
n propeller/pump shaft speed 1/s
n boundary layer power law exponent 
P power W
P* specific power 
p pressure N/m
2
Q volume flow rate m
3
/s
R resistance N
Re
l
Reynolds number based on length 
T thrust N
t thrust deduction factor 
U
∞
undisturbed free stream velocity m/s
v velocity m/s
w momentum wake fraction 
z distance normal to the wall m
Greek symbols
δ
specific pump diameter 
δ
boundary layer thickness m
δ
1
boundary layer displacement thickness m
δ
2
boundary layer momentum thickness m
δ
3
boundary layer energy thickness m
ε
inlet loss coefficient 
φ
momentum flux N
φ
nozzle loss coefficient 
η
efficiency 
ϕ
flow coefficient 
κ
net positive suction head coefficient 
µ
jet velocity ratio 
ρ
fluid density kg/m
3
σ
Thoma number 
τ
w
wall shear stress N/m
2
Ω
angular velocity rad/s
ψ
head coefficient 
Subscripts
1 inflow plane
46
Chapter 2. Waterjet propulsion theory
A available
in inflow
ind induced
out outlet / nozzle
prop propeller
R required
tube streamtube
wj,all complete waterjet structure
wj,hull excluded part of waterjet streamtube control volume
wj,tube included part of waterjet streamtube control volume
2.9 References
[1] Roberts, J.L. & Walker, G.J.,’Boundary layer ingestion in flush waterjet
intakes’, Proceedings RINA Waterjet Propulsion II conference, Amster
dam, 1998
[2] Terwisga, T.J.C. van,’Waterjet hull interaction’, PhD thesis, Delft Uni
versity,1996
[3] Wilson, M.B., Chesnakas, C., Gowing, S., Becnel, A.J., Purnell, J.G.,
Stricker, J.G., ‘Analysis of hull boundary layer velocity distributions with
and without active waterjet inlets’, RINA Waterjet IV conference, Lon
don, 2004
[4] Schlichting, H., ‘Boundary layer theory’, McGrawHill, New York, 1968
[5] Fox, R.W., & McDonald, A.T.,’Introduction to fluid mechanics’, Third
Edition, John Wiley & Sons, New York, 1985
[6] Newman, J.N., ‘Marine hydrodynamics’, MIT press, Cambridge, 1977
[7] Svensson, R. & Grossi, L.,’Trial result including wake measurements
from the world’s largest waterjet installation’, Proceedings RINA Water
jet Propulsion II conference, Amsterdam, 1998
[8] Seil, G.J., ‘Development of waterjet inlets for 100 knots’, Proceedings
FAST’99 conference, pp. 853868, Seattle, 1999
[9] Bulten, N. & Verbeek, R.,’Design of optimal inlet duct geometry based
on operational profile’, Proceedings FAST2003 conference Vol I, ses
sion A2, pp 3540, Ischia, Italy, 2003
[10] Verbeek, R.,’Application of waterjets in highspeed craft’, in Hydrody
namics: Computations, Model Tests and Reality, H.J.J. van den Boom
(Editor) Elsevier Science Publication, 1992
2.9 References
Numerical analysis of a waterjet propulsion system 47
[11] Stepanoff, A.J., ‘Centrifugal and axial pumps; theory, design and appli
cation’, John Wiley & Sons, New York, 1957
[12] Wislicenus, G.F., ‘Fluid mechanics of turbomachinery’, Dover, New
York, 1965
[13] MacPherson, D.M., ‘A universal parametric model for waterjet perform
ance’, Proceedings FAST’99 conference, pp. 879882, Seattle, 1999
[14] Bohl, W.,’Strömungsmachinen (Aufbau und Wirkungsweise)’, Vogel
verlag, Würzburg, 1977
[15] Arnold, J., Nijhuis, G.J., ‘Selection, design and operation of rotody
namic pumps’, Published by Nijhuis Pompen, first edition, 2005
[16] Gülich, J.F., ‘Kreiselpumpen’, Springer Verlag, Berlin, 1999
[17] Os, M.J. van,’On the flow and cavitation inception of mixedflow impel
lers’, PhD thesis, Twente University, 1997
[18] Lewis, E.V., ‘Principles of naval architecture’, Volume II, Society of
Naval Architects and Marine Engineers, Jersey City, 1998
[19] Verbeek, R.,’Waterjet forces and transom flange design’, RINA water
jet propulsion conference, London, 1994
48
Chapter 2. Waterjet propulsion theory
Numerical analysis of a waterjet propulsion system 49
Chapter 3 Nonuniform distribution of
pump entrance velocity field
The theory as presented in chapter 2 is based on experience with pumps with
uniform inflow. It is known that in normal operating conditions the inflow
velocity in waterjet pumps is far from uniform. The first developers of
waterjets have accepted this phenomenon; most probably since it has clear
similarities with a shippropeller wake field.
In the following subsection some methods will be presented to represent the
nonuniform pump inflow velocity distribution. With such representation it is
possible to derive an estimation of the flow rate fluctuations through an
impeller channel and the variations of the inflow angle at the leading edge of
the blade.
Obviously, the nonuniformity should be kept minimal from a hydrodynamic
point of view. Therefore, a thorough analysis of the contributing factors to the
nonuniformity is made. In section 3.4 it will be shown that the major
contributions are unavoidable in waterjet applications with flush type inlets.
3.1 Representation of nonuniform velocity distribution
In order to get an impression of the type of nonuniform velocity distributions
discussed here, typical distributions will be shown in this section. The
distributions are derived from experimental results. The distributions can be
expressed as nondimensional parameters or with a twodimensional
representation. Both methods will be discussed in this section as well.
50
Chapter 3. Nonuniform distribution of pump entrance velocity field
3.1.1 Experimental setup
Measurements have been carried out on a model scale inlet, which is
mounted on the Tom Fink cavitation tunnel [1]. A sketch of the test setup is
shown in figure 3.1. The model scale inlet has an inlet diameter of 150 mm. In
the model scale testrig the actual waterjet pump is not included. Though, the
impeller drive shaft, with a diameter of 22 mm, is included in the test setup.
The tunnel has a square crosssection of 600 x 600 mm.
At a crosssectional plane just upstream of the pump, which will be denoted
as impeller plane, the velocity distribution is measured with a 3 hole Pitot
tube.
Experiments have been carried out with a constant tunnel speed v
tunnel
of 8
m/s. Upstream of the inlet an serrated edge is applied to thicken the natural
tunnel wall boundary layer. The shape of the edge is selected after an
extensive test procedure. During the tests the growth of the boundary layer
thickness and the smoothness of the profiles was evaluated.
The mass flow through the inlet is adjusted to get the desired IVR values, as
defined in equation (2.12). The tunnel v
tunnel
is used to represent the ship
speed v
ship
. The measured conditions are listed in the table below. Reynolds
number for these conditions is based on the diameter of the inlet D and the
averaged pump velocity v
pump
.
Figure 3.2 shows the axial velocity distributions for two inlet velocity ratios,
derived from the measurements. The condition with an IVR of 1.68
represents a normal cruising speed of a fast ferry. The figure on the right
shows the distribution at very high IVR. This condition can occur in high
speed motor yacht applications (> 60 knots). It should be noted that the shaft
Figure 3.1 Sketch of test setup with inlet mounted on top of cavitation
tunnel
pitottube
v
tunnel
v
pump
Z
x
y
z
p
ref
3.1 Representation of nonuniform velocity distribution
Numerical analysis of a waterjet propulsion system 51
diameter (d=22 mm) is about half the diameter of the non measured (blank)
region. Moreover, it should be taken into account that the shaft was not
rotating during experiments, which results in a symmetric velocity profile.
Similar velocity distributions are found for another inlet geometry in this test
rig. The measurements in the cavitation tunnel indicate that the level of non
uniformity is almost independent of the actual inlet geometry. Measurements
in a windtunnel with systematic changes of the geometry confirm this
behaviour [2]. It is concluded that the nonuniformity is strongly related to the
inlet velocity ratio and only weakly related to the precise shape of the inlet.
Based on a survey of the available (confidential) experimental and numerical
data at the authors’ company, it is concluded that within the design space for
commercial applications all possible inlet geometries show more or less the
same type of velocity distribution.
Table 3.1 Parameters of conditions of the measured velocity
distributions
IVR
[]
V
tunnel
[m/s]
V
pump
[m/s]
Re
inl
[]
1.68 8.00 4.76
7.14 x 10
5
1.87 8.00 4.28
6.42 x 10
5
2.03 8.00 3.94
5.91 x 10
5
2.19 8.00 3.65
5.48 x 10
5
Figure 3.2 Experimentally determined nonuniform velocity distributions
for 2 IVR conditions. Left: medium IVR of 1.68, right high IVR
of 2.19.
y
z
52
Chapter 3. Nonuniform distribution of pump entrance velocity field
3.1.2 Nondimensional representation
For easy comparison of different inlet geometries and operating conditions,
the level of nonuniformity is expressed as a single value ζ [3]:
(3.1)
where v is the local axial velocity and v
pump
the average axial velocity. The
relation between nonuniformity and IVR can be assessed with this
parameter.
3.1.3 Twodimensional representation
Hu&Zangeneh [4] use a circumferentially averaged velocity distribution to
investigate the effects of the nonuniform velocity distribution on the waterjet
pump performance. However, for a detailed analysis of the timedependent
effects of nonuniformity on pump performance, this quantification is not
sufficient. At least a two dimensional description, based on the radius r and
the angle θ, of the inflow field in front of the impeller is required to capture the
time varying phenomena.
Such a description of the axial velocity distribution can be obtained if the
velocity field is approximated as a Fourier series:
(3.2)
where the coefficients a
n
(r) are taken as quadratic functions of the radius and
m is the number of harmonics. For the present Fourier approximation 4
harmonics have been used. The quadratic functions a
n
(r) are based on
coefficients at five different radii. Results of the two dimensional velocity field
description are shown in figure 3.3 for the condition IVR=2.19 together with
the measured values.
Agreement between measurements and the numerical approximation is
satisfactory for all radii along the complete circumference. A similar
agreement is found at lower IVR.
ζ
1
Q
 v v
pump
– ( )
2
A d
∫
≡
v
x
r θ , ( ) a
0
r ( ) a
n
r ( ) nθ ( ) cos ( )
n 1 =
m
∑
+ =
3.2 Local flow rate fluctuations
Numerical analysis of a waterjet propulsion system 53
3.2 Local flow rate fluctuations
It is expected that the nonuniform inflow velocity distribution will result in a
varying flow rate through the channels between the impeller blades during a
revolution. The amplitude of the fluctuation is not only dependent on the level
of nonuniformity of the velocity distribution but also on the number of impeller
blades. An estimation of the channel flow rate as function of the rotor position
can be calculated with:
(3.3)
where the axial velocity v
x
is integrated over the impeller passage inlet area,
with N the number of impeller blades, Q
bb
the volume flow rate in an impeller
channel, r
1
the hub radius and r
2
the tip radius at the impeller inlet.
Figure 3.3 Comparison of axial velocity derived from measurements with
the velocities based on Fourier series for velocity distribution at
IVR = 2.19
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 45 90 135 180 225 270 315 360
Measurement r=20 mm Fourier r=20 mm
Measurement r=40 mm Fourier r=40 mm
Measurement r=55 mm Fourier r=55 mm
Measurement r=65 mm Fourier r=65 mm
Measurement r=70 mm Fourier r=70 mm
Angle θ [degrees]
v
x
(
r
,
θ
)
/
v
p
u
m
p
[

]
Q
bb
v
x
r θ , ( ) r ⋅ r θ d d
r
1
r
2
∫
θ
π
N
 –
θ
π
N
 +
∫
=
54
Chapter 3. Nonuniform distribution of pump entrance velocity field
In figures 3.4 and 3.5 the normalised local flow rate estimate is presented for
medium and high IVR and for impellers with different number of blades. In
these figures the channel flow rate Q
bb
is normalised with the averaged flow
rate per channel (Q/N).
As expected, the amplitude increases with increasing IVR. The six bladed
impeller shows a flow rate deficit of 30% for the medium IVR in figure 3.4,
which increases to 46% for the high IVR condition, as shown in figure 3.5.
For impellers with only three blades the amplitude of the variation is smaller.
The flow rate deficit is 20% at medium IVR and 35% at high IVR.
The actual local flow rate through an impeller channel for different IVR
conditions will be determined in chapter 6.
50%
60%
70%
80%
90%
100%
110%
120%
130%
140%
150%
0 45 90 135 180 225 270 315 360
2 bladed impeller
3 bladed impeller
4 bladed impeller
6 bladed impeller
y
z
Figure 3.4 Local flow rate estimate as a function of impeller channel
position for medium IVR of 1.68, using data of figure 3.3.
Angle θ [degrees]
Q
b
b
/
(
Q
/
N
)
[
%
]
θ
2 blades
6 blades
3.3 Impeller velocity triangles
Numerical analysis of a waterjet propulsion system 55
3.3 Impeller velocity triangles
The axial velocity distribution can be used to derive the fluctuations in inlet
flow angles. This angle determines for a great deal the loading of the impeller
blade. A lower inlet flow angle will lead to higher blade loading in general.
Moreover the cavitation behaviour of the impeller will depend on the
fluctuations of the inflow angle.
The design inlet blade angle is based on the inlet velocity triangle with
uniform flow and without prerotation, as shown in figure 3.6:
(3.4)
50%
60%
70%
80%
90%
100%
110%
120%
130%
140%
150%
0 45 90 135 180 225 270 315 360
2 bladed impeller
3 bladed impeller
4 bladed impeller
6 bladed impeller
y
z
Figure 3.5 Local flow rate estimate as a function of impeller channel
position for high IVR of 2.19, using data of figure 3.3.
Angle θ [degrees]
Q
b
b
/
(
Q
/
N
)
[
%
]
θ
2 blades
6 blades
Figure 3.6 Velocity triangle of inlet flow angle.
v
x
Ωr
β
design
β
design
v
x
Ωr

\ .
 
atan =
56
Chapter 3. Nonuniform distribution of pump entrance velocity field
The actual inflow angle will vary due to the axial velocity variations:
(3.5)
where v
x
(r,θ) is the local axial velocity. The incidence angle is the difference
between the design angle and the actual inflow angle. Note that crossflow
plane velocity components are neglected in this approach. This simplification
is allowed, since the tangential velocities are small compared to the tangential
velocity component Ωr of the impeller. Typical crossflow plane velocities are
about 5% of the tip speed of the impeller.
Figure 3.7 shows the estimation of the incidence angle for high IVR. The
deviations vary from +10 degrees to 10 degrees at the outer radii. It is to be
expected that such amplitude of inflow angles will result in significant
variation of the impeller blade loading during a revolution of the blade.
β
actual
v
x
r θ , ( )
Ωr

\ .
 
atan =
Figure 3.7 Estimation of the incidence angle as function of impeller
channel position for the velocity distribution at IVR = 2.19
20
15
10
5
0
5
10
15
0 45 90 135 180 225 270 315 360
radius = 20 mm
radius = 40 mm
radius = 55 mm
radius = 65 mm
radius = 70 mm
Angle θ [degrees]
I
n
c
i
d
e
n
c
e
a
n
g
l
e
[
d
e
g
r
e
e
s
]
3.4 Origin of the nonuniform velocity distribution
Numerical analysis of a waterjet propulsion system 57
3.4 Origin of the nonuniform velocity distribution
The development of the nonuniform velocity distribution can be explained
with basic fluid dynamics theory. There are four phenomena which contribute
to the nonuniformity of the velocity distribution:
1. boundary layer ingestion
2. deceleration of the flow
3. obstruction of the flow due to the shaft
4. bend in the inlet duct
A sketch of the phenomena is presented in figure 3.8. The above items are
discussed in more detail in the remainder of this section.
3.4.1 Boundary layer ingestion
Waterjets with flush mounted inlet ducts ingest water from the boundary layer
below the hull. A typical ship speed for a waterjet propelled vessel is 40 knots
(about 20 m/s). With a wetted length of 80 m, the Reynolds number becomes
1.6 x 10
9
. At these high Reynolds numbers the boundary layer thickness can
be approximated with a power law velocity distribution, as shown in equation
(2.3), with a power n of about 10.
The contribution of the boundary layer ingestion to the nonuniformity
depends on the amount of water that is ingested from the boundary layer.
The ingestion of boundary layer water can be expressed with the ratio
between the suction depth h and the boundary layer thickness δ. If this ratio
(h/δ) is smaller than 1, then all the water is taken from the hull boundary layer.
4
2
1
3
Figure 3.8 Phenomena which contribute to development of a nonuniform
inflow velocity distribution to the pump. 1. boundary layer
ingestion. 2. deceleration of the flow. 3. obstruction of the flow
by shaft. 4. bend in duct.
h
58
Chapter 3. Nonuniform distribution of pump entrance velocity field
The suction depth h depends on the total ingested volume flow rate and the
assumed shape of the streamtube. As mentioned before, waterjet
manufacturers and ship model basins use a simplified method, in which the
shape of the streamtube is approximated as a rectangular box with a width of
1.3 times the inlet diameter.
When the suction depth h is smaller than the boundary layer thickness, the
suction depth h is calculated with:
(3.6)
where Q is the volume flow, v
ship
the undisturbed velocity and λD the
assumed width of the rectangular box with λ equal to 1.3.
If the suction depth h exceeds the boundary layer thickness, then the
displacement thickness can be used to calculate the suction depth h:
(3.7)
The relations for the boundary layer thickness (for example equation (2.10)
for n=9) show that the length of the vessel has a major influence on the
development of the boundary layer. This effect is shown more clearly in figure
3.9, where the boundary layer ingestion ratio h/δ is shown for three different
vessel lengths. The flow rate Q is adjusted for each ship speed, to take the
effect of increasing flow rate with increasing ship speed into account.
The ratio h/δ decreases with ship speed for all ship lengths. With higher
incoming velocity and constant suction box width, the suction height h will
reduce.
It can be seen that the boundary layer ingestion ratio h/δ is about 0.90 at 45
knots for the short vessel. The ratio is reduced to 0.50 when the length is
doubled. Since the velocity profile has the largest gradients near the wall, this
results in increased nonuniformity, for longer vessels. Decrease of the
boundary layer ingestion ratio h/δ is coupled to an increase of IVR. As a
consequence, the nonuniformity will increase with increasing IVR as well.
3.4.2 Deceleration of the flow
Waterjets operate in IVR conditions of 1.3 to 1.8 in general. The averaged
velocity of the water just in front of the pump is thus smaller than the ship
speed. Even if the velocity in front of the pump is compared with the mass
averaged incoming velocity, which is smaller due to boundary layer
development, there is still a significant retardation of the flow in the inlet duct.
The retardation of the flow can be regarded as a (subsonic) diffuser flow
h
Q
v
ship
λD

n 1 +
n
 δ
1 n ⁄
⋅ ⋅
\ .
 
n n 1 + ( ) ⁄
h δ < ( ) =
h
Q
v
ship
λD
 δ
1
+
Q
v
ship
λD

δ
n 1 +
 h δ > ( ) + = =
3.4 Origin of the nonuniform velocity distribution
Numerical analysis of a waterjet propulsion system 59
phenomenon. According to the theory of standard diffusors, the velocity
profile in a conventional diffusor depends on the diffusor angle. Nikuradse
measured the velocity profiles in convergent and divergent channels with
diffusor angles between 4°and +8°(shown in Schlichting [5]). The velocity
profiles in convergent channels, which have negative diffusor angles, become
more uniform. In contrast, the velocity profile becomes less uniform in
divergent channels (or in other words, it corresponds to a lower n value in a
power law profile). Therefore, retardation of the flow increases the level of
nonuniformity in the velocity distribution. Consequently, the increase of non
uniformity is related to the value of IVR.
The standard diffusor theory assumes a uniform inflow at the entrance of the
diffusor. However in a waterjet the boundary layer creates a nonuniform
distribution at the beginning of the diffusor. When a velocity profile is non
uniform at the entrance a diffusor, the level of nonuniformity will even
increase. This phenomenon is explained by Betz [6] with a simple example.
3.4.3 Obstruction of the flow due to the shaft
The impeller shaft forms an obstruction of the flow in the inlet, comparable to
a propeller shaft in a ship. A narrow wake region with low velocity is found
near the shaft. If the pump shaft is equipped with a stationary sleeve, then it
can lead to unstable vortex shedding. Hu & Zangeneh [7] and Seil [8]
0.00
0.25
0.50
0.75
1.00
1.25
1.50
25 30 35 40 45 50
0.5
0.75
1
1.25
1.5
1.75
2
Wetted length 50 m
Wetted length 75 m
Wetted length 100 m
IVR
Figure 3.9 Boundary layer ingestion ratio h/δ for three different vessel
lengths l as function of ship speed v
ship
with corresponding
IVR.
Ship speed [knots]
I
V
R
[

]
h
/
δ
[

]
60
Chapter 3. Nonuniform distribution of pump entrance velocity field
investigated the effect of the rotation of the shaft. They show that the flow
stabilizes due to rotation of the shaft. In general a waterjet installation does
not have a stationary sleeve around the shaft, so a stable flow is found in the
vicinity of the shaft.
3.4.4 Bend in the inlet duct
The fourth contributing factor to the nonuniform velocity distribution is the
bend in the inlet duct. Inside the bend a variation in velocity is found between
the inner and outer part of the bend. This nonuniformity will restore to a
certain extent downstream of the bend. In a waterjet however, the length of
the cylindrical pipe between the bend and the pump is in general smaller than
the inlet diameter. As a result part of the created nonuniformity due to the
bend remains in the velocity distribution.
The convex shape of the first part of the roof of the inlet may be regarded as
a bend as well, which might have a positive influence on the velocity
distribution. The radius of curvature is about six times larger than the radius
of curvature of the bend, so the effect of this curvature can be neglected in
the analysis and is therefore not marked in figure 3.8.
3.4.5 Closing remark
The major contributions to the nonuniformity are caused by the IVR related
phenomena of boundary layer ingestion and retardation of the flow. As long
as waterjets are operated at IVR values above 1.5, a substantial nonuniform
flow will exist.
3.5 Nonuniform inflow velocity distributions in other
turbo machinery
In the literature relatively little attention has been paid so far to the effects of
nonuniform inflow to the waterjet pump. This may be attributed to its close
relation to the conventional ship propeller, which also operates in a non
uniform wake field in general. This wake field is determined by the shape of
the vessel, so the propulsion system designer has to cope with it. A lot of
research has been done on the effects of nonuniform inflow to a ship
propeller already. For example the work of Van Manen [9] in the 1950s can be
mentioned. Variations in effective inflow angle for the tested propeller were
between 1° and +4°. This range is significantly smaller than the expected
variations for the waterjet pump, as shown in figure 3.7.
Nonuniform inflow to mixedflow or centrifugal pumps may occur if the pump
is mounted close to an upstream disturbance, like a pipe bend. For an
acceptable inflow to the pump, most manufacturers prescribe a minimum
3.6 Nomenclature
Numerical analysis of a waterjet propulsion system 61
required length of straight pipe on the suction side of the pump, typically
several pipe diameters long. Examples can be found in a pump handbook of
Sulzer [10]. The use of an acceleration nozzle just upstream of the pump is
also recommended to enhance the uniformity of the inflow for certain types of
pumps.
The requirement of long straight suction pipes can not always be met, like the
flow into the first stage of a boiler feed pump. Cooper [11] shows the
calculated flow pattern at the exit of a radial inlet passageway for such a
pump. A maximum velocity of 1.75 times the average inflow velocity is found
for this example. This level of nonuniformity seems to be more in line with
that found in waterjet applications. Similar flow phenomena will occur in
double suction pumps and inline pumps.
Another typical inflow is found for pumps with a sump, as used in cooling
water pump applications in power stations. Here a suitable design of the
inflow and inlet chamber is to obtain an acceptable inflow pattern. Additional
acceleration can be applied for further improvement.
3.6 Nomenclature
A area m
2
D diameter m
h suction depth m
IVR inlet velocity ratio (v
ship
/v
pump
) 
N number of impeller blades 
n boundary layer power law value 
Q flow rate m
3
/s
Q
bb
channel flow rate 
r radius m
v velocity m/s
Greek symbols
β blade angle rad
δ boundary layer thickness m
δ
1
boundary layer displacement thickness m
θ angle rad
λ suction tube width factor 
ζ
nonuniformity 
Ω
angular velocity rad/s
62
Chapter 3. Nonuniform distribution of pump entrance velocity field
Subscripts
x axial direction
3.7 References
[1] Brandner, P. & Walker, G.J., ‘A waterjet test loop for the Tom Fink cavi
tation tunnel’, Proceedings Waterjet Propulsion III conference, Gothen
burg, 2001
[2] Bulten, N.W.H.,‘Influence of boundary layer ingestion on waterjet per
formance parameters at high ship speeds’, Proceedings of the 5th
international conference on Fast Sea Transportation, pp 883892,
Seattle, 1999
[3] Verbeek, R. & Bulten, N.W.H.,’Recent developments in waterjet
design’, Proceedings RINA Waterjet Propulsion II conference, Amster
dam, 1998
[4] Hu, P. & Zangeneh, M.,’CFD calculation of the flow through a waterjet
pump’, Proceedings Waterjet Propulsion III conference, Gothenburg,
2001
[5] Schlichting, H.,’Boundary layer theory’, Mc GrawHill, New York, 1968
[6] Betz, A, ‘Introduction to the theory of flow machines’, Pergamon Press,
Oxford, 1966
[7] Hu, P., Zangeneh, M., ‘Investigations of 3D turbulent flow inside and
around a waterjet intake duct under different operating conditions’,
ASME Journal of Fluids Engineering, Vol 121, pp. 396404, 1999
[8] Seil, G.J.,’The effect of the shaft, shaft rotation and scale on the flow in
waterjet inlets’, Proceedings Waterjet Propulsion III conference,
Gothenburg, 2001
[9] Manen, J.D. van, ’Invloed van de ongelijkmatigheid van het snelheids
veld op het ontwerp van scheepsschroeven’, PhD thesis, Wageningen,
1951
[10] Sulzer Brothers Ltd, ‘Sulzer centrifugal pump handbook’, Elsevier
Applied Science, London, 1989
[11] Cooper, P.,’Perspective: the new face of R&D  A case study of the
pump industry’, ASME Journal of Fluids Engineering, Vol 118, pp. 654
664, 1996
Numerical analysis of a waterjet propulsion system 63
Chapter 4 Mathematical treatment
To get more insight in the behaviour of the flow through a waterjet installation,
experiments can be done and calculations can be made. With an
experimental setup of a waterjet inlet velocity, pressure and forces can be
measured. In this thesis, however these quantities will be determined by
numerical simulations.
A numerical method is based on a mathematical model of the physics of the
flow. It constitutes a set of governing equations derived from first principles,
often complemented with empirical relations. The numerical method is
obtained by discretisation of the governing equations. The present type of
flow can be described by the NavierStokes equations. Before these equation
can be solved numerically for the application of interest, it is necessary to
make some simplifications. A review of the important flow features is given in
section 4.1 to get an impression of the feasible simplifications without major
loss of accuracy. This analysis will result in a selection of the best suitable
mathematical method. The chosen method will be described in more detail in
section 4.3. Some known weaknesses of the mathematical method will be
discussed in more detail to get an indication of the obtainable accuracy.
4.1 Requirements of mathematical method
In this section the requirements for the mathematical method will be
formulated. The goal of the mathematical analysis is the prediction of the flow
through a waterjet installation. Based on the different requirements for the
prediction three different mathematical methods will be evaluated: the
64
Chapter 4. Mathematical treatment
potential flow model, the model based on the Euler flow and the model based
on the Reynolds averaged NavierStokes (RANS) equations. RANS methods
are also denoted as viscous methods, whereas the other two methods do not
include terms due to viscosity and heat conduction. They require, for example
effects of viscosity, such as flow separation, to be modelled. In potential flow
methods, the flow is irrotational, i.e. the vorticity equals zero everywhere,
except in infinitesimal regions such as vortex sheets and vortex filaments.
4.1.1 Incompressibility
The first requirement for the mathematical method is the capability of
handling incompressible flow. The maximum velocity in a full scale waterjet
installation is about 50 m/s, which is much smaller than the speed of sound in
water, which is about 1450 m/s. The resulting Mach number is thus
sufficiently small to treat the flow as incompressible. It is noted that the speed
of sound is based on a noncavitating flow. In a cavitating flow, the speed of
sound reduces significantly, which results in a compressible flow behaviour.
All three mentioned mathematical methods can satisfy this requirement.
Nevertheless it can be mentioned that this requirement represents one of the
major differences between pumps and compressors, as discussed in section
1.2. mathematical methods for compressor analysis cannot be used directly
for incompressible flow. For compressible flow the Euler equations are
hyperbolic in time for any Mach number. This gives the opportunity to use a
single numerical technique for subsonic, supersonic or mixed subsonic
supersonic (transonic) flow problems [1]. These methods are commonly
called time marching methods. In these methods compressibility is employed,
which results in amongst others travelling pressure waves through the
domain. In an incompressible flow, pressure fluctuations are instantaneous in
the whole domain. Reconditioning, i.e. some form of artificial compressibility
has to be introduced in the compressible flow solver in order to analyse an
incompressible flow.
4.1.2 High Reynolds number
For full scale waterjet installations the typical Reynolds numbers for the inlet
and the impeller are both very large. For the impeller the Reynolds number is
defined as:
(4.1)
where ρ is the density, µ the dynamic viscosity, v
tip
is the tip speed of the
impeller blade and D
inl
the diameter of the inlet at the pump suction side. For
Re
imp
ρv
ti p
D
inl
µ
 =
4.1 Requirements of mathematical method
Numerical analysis of a waterjet propulsion system 65
typical waterjet applications the tip speed can be about 50 m/s. For an inlet
diameter of 1.2 m this gives a Reynolds number Re
imp
of 6x10
7
.
Calculation of the Reynolds number that characterises the flow in the inlet
requires another typical velocity. This can be either the ship speed or the
pump speed (as defined in equation (2.11)). The actual choice is not so
critical, since both values do not differ by more than a factor of two. In order to
represent the flow phenomena inside the inlet best, the Reynolds number for
the inlet is defined as:
(4.2)
For the same typical configuration as above, the Reynolds number Re
inl
will
be about 10
7
, which is 6 times smaller than the impeller Reynolds number.
Model scale experiments are performed at lower Reynolds numbers, as
shown in table 3.1 on page 51. These values still exceed 5x10
5
, which can be
regarded as high Reynolds number flow. It is to be expected that, at least at
design conditions, viscous effects play a limited role.
4.1.3 Time dependency
There are two different reasons for the flow to be timedependent in a
waterjet installation: (i) the nonuniformity of the flow at the impeller entrance
results in a time varying onset flow of the rotating blades and (ii) the
interaction between moving rotor and stationary stator blades at the impeller
outlet is unsteady. In the rotating frame of reference, phenomena associated
with the first reason have a frequency related to the shaft frequency. The
frequency of the interaction phenomena between the rotor and stator blades
depends on the number of stator blades and the shaft frequency.
In the stationary frame of reference the frequency of the fluctuations due to
rotorstator interaction is related to the number of impeller blades. The effect
of the nonuniform inflow will give a steady component with superposed
fluctuations with the blade passing frequencies.
4.1.4 Nonuniformity of impeller inflow
In chapter 3 the velocity distribution upstream of the impeller is discussed. It
is shown that the velocity field is strongly nonuniform even at normal waterjet
operating conditions. This velocity distribution has to be reproduced by the
mathematical method in order to obtain a correct prediction of the inlet flow
phenomena. For an analysis of the isolated pump, the nonuniform inflow to
the impeller has to be implemented as an inflow boundary condition.
Re
inl
ρv
pump
D
inl
µ
 =
66
Chapter 4. Mathematical treatment
The development of the velocity distribution in the inlet is attributed to the
ingestion of the boundary layer and to the deceleration of the flow.
Generation of vorticity plays an important role in these flow phenomena.
Application of a mathematical method based on the potential flow assumption
or an inviscid flow model is therefore not suitable for the analysis of inlet flow
phenomena.
Potential flow analyses of mixedflow pumps have been made by Van Esch
[2] and Van Os [3]. These calculations are based on a uniform inflow velocity
distribution. Implementation of a nonuniform velocity distribution introduces
velocity gradients for the axial velocity component. The constraint of
irrotational flow results in additional velocity gradients in the directions
perpendicular to the axial inflow direction. The cross components of the inflow
velocity distribution have such a dominating influence on the overall
development of the nonuniformity, that the velocity distribution is almost
uniform within about one diameter pipe length. The analysis of the stability of
the nonuniform velocity distribution is discussed in more detail in appendix A.
It is concluded that for the analysis of both a waterjet inlet as well as a pump
with nonuniform inflow a suitable mathematical method should take the
presence of vorticity into account.
4.1.5 Tip clearance flow
In general, waterjets are equipped with unshrouded mixedflow or axial
impellers. Unshrouded impellers have a small clearance between the blade
tips and the stationary housing. This housing is called the seatring. The
pressure difference between the pressure and suction side of the blade
causes some leakage flow through the clearance. This leakage flow should
be kept as low as possible to maintain a high efficiency of the pump. The
distance between the impeller tip and the seatring is therefore very small,
about 12% of the diameter. On the other hand, the velocity difference
between the rotating blade tip and the stationary seatring can be about 50 m/
s, which leads to large velocity gradients in the clearance.
As a result of the available pressure difference between pressure and suction
side of the blade and the occurring viscous losses in the clearance a certain
flow will establish. Effects of viscosity can not be neglected, if this flow is to be
determined. Second effect of the viscosity is the development of boundary
layers on the blade tip and the seatring. This is confirmed by Kunz et al. [4] in
the investigation of tip clearance phenomena in an axial compressor cascade
with an Euler and a NavierStokes method. It is found that the NavierStokes
method shows better agreement with experimental data than the Euler
method. Prediction of the tip clearance mass flow rate is presumed to be
more accurate with the NavierStokes method.
4.2 Conservation laws
Numerical analysis of a waterjet propulsion system 67
The flow in a mixedflow pump with various tip clearances has been analysed
by Goto [5]. For this analysis he uses the unsteady Reynoldsaveraged
NavierStokes equations in a rotational frame of reference. Results of the
calculations show reasonably good agreement with the experimental data.
Moreover a variety of jetwake flow patterns at the exit of the impeller are
predicted well.
4.1.6 Final remarks
Based on the listed requirements and the capabilities of the different models,
the choice for a viscous flow model is justified. In the next two chapters,
results of a detailed validation study of numerical simulations for the waterjet
inlet and the mixedflow pump are presented. It is acknowledged that a
viscous (i.e. RANS) flow method requires significantly more computational
resources than an Euler method and certainly much more than a potential
flow method.
Cavitation will not be taken into account in the analyses. Since cavitation
models are presently developed for most commercial RANS codes, is it to be
expected that application of methods with some form of cavitation model will
be feasible in the near future.
The calculations presented in this study are carried out with the commercial
CFD method StarCD. This method is based on a finite volume numerical
method. Both tetrahedral and hexahedral cell types can be used for the
mesh. In this study all meshes are generated with hexahedral cells only.
4.2 Conservation laws
To describe the flow phenomena in a waterjet installation, two conservation
laws are used. These are the laws for conservation of mass and that of
momentum. Conservation of mass is also denoted as the continuity equation
(see for example [6]):
(4.3)
where ρ is the density of the fluid and the velocity.
In many cases the flow can be assumed to be incompressible. This is
allowed, whenever variations in density are small. These variations in density
are caused by variations in the pressure. Since in the present applications the
velocities are much smaller than the speed of sound in the water, which is
about 1450 m/s, the variation in density will be negligible. For an
incompressible flow the continuity equation (4.3) reduces to:
(4.4)
t ∂
∂ρ
∇ + ρv ⋅ 0 =
v
∇ v ⋅ 0 =
68
Chapter 4. Mathematical treatment
Conservation of momentum is given by (see also [6]):
(4.5)
where p is the pressure, µ is the dynamic viscosity, which is assumed to be
constant in the derivation of equation (4.5) and is the acceleration of
gravity. This set of equations is known as the NavierStokes equation. For an
incompressible flow, equation (4.4) can be substituted into equation (4.5).
This results in the NavierStokes equation for incompressible flow:
(4.6)
The NavierStokes equation can be solved numerically directly without further
assumptions. This requires direct numerical simulations (DNS) to obtain the
timeaccurate solution of equations (4.4) and (4.6) on a grid that is sufficiently
fine to resolve all flow details. Such a method is not suitable for practical
engineering analyses, however. In the following section an approach will be
discussed, which enables numerical solution of the flow field, within a
practical context.
4.3 Reynolds Averaged NavierStokes (RANS) flow
In order to convert the NavierStokes equations into a set of equations, that
can be solved numerically for general engineering applications, the concept
of splitting the flow variables in a mean and a fluctuating part is employed.
Substitution of this decomposition in the NavierStokes equation and time
averaging the continuity and the NavierStokes equations, results in a set of
equations for the meanflow field variables. The concept of time averaging
was introduced by Reynolds in 1895. The resulting equation is therefore
called the Reynoldsaveraged NavierStokes (RANS) equation.
4.3.1 Reynolds averaging
The principle of Reynolds averaging is based on a decomposition of the
variables in a time averaged value and a fluctuating part:
(4.7)
ρ
t ∂
∂
v ρ v ∇ ⋅ ( )v + ∇p – µ∇
2
v
1
3
µ∇ ∇ v ⋅ ( ) ρg + + + =
g
ρ
t ∂
∂
v ρ v ∇ ⋅ ( )v + p µ v ρg + ∇
2
+ ∇ – =
v
i
v
i
v
i
' + =
4.3 Reynolds Averaged NavierStokes (RANS) flow
Numerical analysis of a waterjet propulsion system 69
The time averaged variable is defined as:
(4.8)
where t
1
has to be larger than the time scale of the smallest fluctuations.
Therefore,
(4.9)
The average of the fluctuating part is zero by definition. Nevertheless, if the
product of two variables is considered, not all fluctuating terms vanish when
they are correlated, thus:
(4.10)
Substitution of the decomposed terms in the NavierStokes equation for the
velocity and pressure and application of the method of Reynolds averaging
gives:
(4.11)
The term is called the Reynolds stress term, where is a diadic
product defined by . This tensor contains the correlations of the
fluctuating terms of the velocity components. The term with the Reynolds
stress can be treated in several ways. Boussinesq proposed a closure
hypothesis for the Reynolds stress term. In index notation the closure is
defined as:
(4.12)
which is referred to as an eddy viscosity model with µ
T
the turbulent or eddy
viscosity and k the turbulent kinetic energy defined as:
(4.13)
Turbulence models are used to obtain a value for the eddy viscosity. An
alternative approach is to derive from the original timedependent Navier
v
i
1
T

T t
1
→
lim v
i
t d
t
t T +
∫
=
v
i
'
1
T
 v
i
v
i
– [ ]
t
t T +
∫
T t
1
→
lim dt 0 = =
v
i
v
j
v
i
v
j
v
i
' v
j
' + =
ρ
t ∂
∂
v ρ v ∇ ⋅ ( ) + v p µ v ρg ∇ ρ v' v' ( ) ⋅ – + ∇
2
+ ∇ – =
ρ v' v' ( ) v' v' ( )
v' v' ( )
i j
v
i
' v
j
' =
ρv
i
' v
j
' – µ
T
x
j
∂
∂v
i
x
i
∂
∂v
j
+
\ .

 
2
3
ρkδ
i j
– =
k
1
2
 v
i
' v
i
' ( )
1
2
 v
x
' v
x
' v
y
' v
y
' v
z
' v
z
' + + ( ) = =
70
Chapter 4. Mathematical treatment
Stokes equations a transport equation for each of the Reynolds stress
components, however. These equations require another group of closure
relations before the equations can be solved numerically. Such a model is
called a Reynolds stress turbulence model.
Substitution of equation (4.12) into the Reynolds averaged NavierStokes
equation (4.11) gives:
(4.14)
The next step is the combination of the pressure term, the turbulent kinetic
energy term and the gravitational term. The modified pressure is defined as:
(4.15)
where it is assumed that gravity is directed in the zdirection. The resulting
NavierStokes equation with the Reynolds stress terms included, becomes:
(4.16)
where the molecular viscosity µ and turbulent viscosity µ
T
have been
combined into the effective viscosity µ
eff
.
4.3.2 Eddy viscosity turbulence models
Eddy viscosity turbulence models are used to determine a value for the eddy
viscosity µ
T
. Once this eddy viscosity is known, the Reynolds stresses are
known and consequently the Reynoldsaveraged NavierStokes equations
can be solved. The available eddy viscosity turbulence models can be divided
into 3 groups:
1. mixing length or algebraic models or zeroequation models
2. oneequation models
3. twoequation models
In a zeroequation model, the eddy viscosity µ
T
is based on the mixing length
concept. This mixing length is based on algebraic relations. The eddy
viscosity for an algebraic model is defined as:
(4.17)
where l
mix
is Prandtl’s mixing length.
ρ
t ∂
∂
v ρ v ∇ ⋅ ( )v + p ∇ µ µ
T
+ ( ) v v ∇ ( )
T
+ ∇ ( ) ρg ∇
2
3
 ρk – + ⋅ + ∇ – =
p
∗
p
2
3
ρk ρgz – + =
ρ
t ∂
∂
v ρ v ∇ ⋅ ( )v + p
∗
∇ µ ⋅
eff
∇v ∇v ( )
T
+ ( ) + ∇ – =
µ
T
ρl
mix
2
dv
dy

=
4.3 Reynolds Averaged NavierStokes (RANS) flow
Numerical analysis of a waterjet propulsion system 71
The oneequation model uses a transport equation for the turbulent kinetic
energy k and an algebraic relation for the mixing length scale. This model is
thus an extension of the mixing length model. The eddy viscosity for a one
equation model is defined as:
(4.18)
which uses a transport equation for k and an algebraic relation for the mixing
length l
mix
.
Twoequation models use a transport equation for both the turbulent energy k
as well as the turbulent length scale or an equivalent. Two wellknown
examples are the dissipation ε for the kε model and the dissipation per unit
turbulence kinetic energy ω for the kω model.
kε turbulence model
The eddy viscosity in the kε turbulence model is defined as:
(4.19)
with C
µ
a dimensionless constant. The standard kε turbulence model will be
discussed in more detail. This model was first presented by Jones and
Launder in 1972. The exact transport equation for the turbulent kinetic energy
can be derived from the NavierStokes equation. The transport equation is
given for example by Wilcox [7]:
(4.20)
where the terms on the right hand side represent production, diffusion and
dissipation respectively. The diffusion term has three components; diffusion
by viscosity, turbulent velocity fluctuations and pressure fluctuations. The
latter two require a closure term in order to enable the transport equation to
be solved. The turbulent transport and pressure diffusion terms are modelled
by:
(4.21)
where σ
k
is a dimensionless closure coefficient.
µ
T
ρk
1 2 ⁄
l
mix
=
µ
T
ρC
µ
k
2
ε
 =
ρ
t ∂
∂k
ρv
j
x
j
∂
∂k
+ ρv
i
' v
j
'
x
j
∂
∂v
i
–
x
j
∂
∂
µ
x
j
∂
∂k 1
2
ρv
i
' v
i
' v
j
' p' v
j
' – – ρε – + =
1
2
ρv
i
' v
i
' v
j
' p' v
j
' +
µ
T
σ
k
 –
x
j
∂
∂k
=
72
Chapter 4. Mathematical treatment
The final transport equation for k becomes:
(4.22)
where the production term of the turbulent kinetic energy P
k
is defined as:
(4.23)
The exact equation for the dissipation ε can be derived from the Navier
Stokes equation, but it requires a considerable amount of algebra to arrive at
the final equation as given by Wilcox [7]. The exact equation for ε has a
number of unknown double and triple products, so a number of closure terms
are still required. The entire equation for ε can also be regarded as a model in
a similar form as the transport equation for k (see [8]). The final equation
becomes:
(4.24)
with dimensionless closure coefficients C
ε1
, C
ε2
and σ
ε
. Similar to equation
(4.22) a production, diffusion and dissipation term can be recognized on the
righthandside.
The kε turbulence model contains five closure coefficients. The values for the
coefficients of the kε turbulence model, as used in the CFD method
employed in the present study, are listed in the table below. These values are
found in the manual [9].
A value of 1.3 for the constant σ
ε
can be found in literature as well ([7], [10]).
kω turbulence model
The kω turbulence model has a similar set of equations as the kε turbulence
model. The kω model uses the dissipation per unit turbulence kinetic energy
ω, whereas the dissipation ε is used in the latter. The basis of the kω model
was postulated by Kolmogorov. Further development of this model has led to
Table 4.1 Values for kε turbulence model closure coefficients
C
µ
C
ε1
C
ε2
σ
k
σ
ε
0.09 1.44 1.92 1.0 1.22
ρ
t ∂
∂k
ρv
j
x
j
∂
∂k
+ P
k
x
j
∂
∂
µ
µ
T
σ
k
 +
\ .
 
x
j
∂
∂k
ρε – + =
P
k
ρv
i
' v
j
'
x
j
∂
∂v
i
– =
ρ
t ∂
∂ε
ρv
j
x
j
∂
∂ε
+ C
ε1
P
k
ε
k

x
j
∂
∂
µ
µ
T
σ
ε
 +
\ .
 
x
j
∂
∂ε
C
ε2
ρ
ε
2
k
 – + =
4.3 Reynolds Averaged NavierStokes (RANS) flow
Numerical analysis of a waterjet propulsion system 73
the following set of equations. The eddy viscosity hypothesis for the kω
turbulence model is defined as:
(4.25)
The equation for the turbulent kinetic energy is similar to equation (4.22)
except for the dissipation term and the closure coefficients:
(4.26)
with the production term P
k
according to equation (4.23) and the closure
coefficients β* and σ*. The transport equation for the dissipation per unit
turbulence kinetic energy ω is also adapted in a similar way, as shown by
Wilcox [7]:
(4.27)
where α, β and σ are three more closure coefficients. The values for all five
closure coefficients of the kω turbulence model are shown in table 4.2
Wall functions
Use of one of the many available high Reynolds turbulence models implies
the application of wall functions to replace the noslip boundary condition.
The wall functions employ special algebraic formulas for the representation of
the distribution of the velocity and turbulence within the part of the boundary
layer closest to the wall. This is necessary to relax the requirement of grid
resolution in the boundary layer. The dimensionless distance from the wall to
the cell centre of the first cell is generally denoted as y
+
and is, according to
[10], defined as:
(4.28)
where and ∆y
p
is the distance from the wall to the near wall cell
centre. Note that the dimensionless distance y
+
is a kind of Reynolds number,
Table 4.2 Values for kω turbulence model closure coefficients
α β β* σ σ∗
5/9 3/40 9/100 1/2 1/2
µ
T
ρ
k
ω
 =
ρ
t ∂
∂k
ρv
j
x
j
∂
∂k
+ P
k
x
j
∂
∂
µ σ
∗
µ
T
+ ( )
x
j
∂
∂k
β
∗
ρkω – + =
ρ
t ∂
∂ω
ρv
j
x
j
∂
∂ω
+ αP
k
ω
k

x
j
∂
∂
µ σµ
T
+ ( )
x
j
∂
∂ω
βρω
2
– + =
y
+
∆y
p
ν

τ
w
ρ
 ⋅
∆y
p
u
τ
ν
 = =
u
τ
τ
w
ρ ⁄ =
74
Chapter 4. Mathematical treatment
based on the distance from the wall and the velocity near to the wall to the
power 1/2. For accurate application of the wall function the y
+
value should be
in the range of 30 to 100, according to the manual [9]. For y
+
values smaller
than the dimensionless velocity u
+
is calculated from:
(4.29)
For y
+
values larger than , the u
+
is calculated from:
(4.30)
where satisfies the equation:
(4.31)
with κ the von Kármán constant and an empirical constant E.
The dimensionless velocity u
+
is defined as:
(4.32)
where u is the tangential fluid velocity and u
w
the velocity of the wall. The
empirical constants of equation (4.30) can be coupled to the universal loga
rithmic velocity distribution, known also as loglaw [11]:
(4.33)
with κ the von Kármán constant equal to 0.4187 and C a constant with a
value between 4.9 and 5.5 ([12]). The default value for the empirical constant
E in the CFD method used is 9.0. This results in a value for C of 5.25.
The definition of y
+
according to equation (4.28) will lead to numerical prob
lems at points where the flow is about to separate. At such locations the wall
shear stress τ
w
becomes zero, which results in an y
+
of zero. This is solved
with the modified definition for y
+
based on the turbulent kinetic energy k:
(4.34)
y
m
+
u
+
y
+
=
y
m
+
u
+ 1
κ
ln Ey
+
( ) =
y
m
+
y
m
+ 1
κ
 Ey
m
+
( ) ln – 0 =
u
+
u u
w
– ( )
u
τ

u u
w
– ( )
τ
w
ρ

 = =
u
+ 1
κ
ln y
+
( ) C + =
y
+
ρ∆y
p
µ
 C
µ
1 4 ⁄
k ⋅ =
4.4 Twodimensional test cases
Numerical analysis of a waterjet propulsion system 75
It is assumed that there will always be turbulent fluctuations at a flow separa
tion point, which eliminates the problem.
4.4 Twodimensional test cases
In order to get an indication of the obtainable accuracy of both the
mathematical and the numerical methods, the flow around a number of well
known NACA profiles has been analysed. The flow along isolated profiles as
well as profiles in a cascade have been calculated. For all configurations
experimental data for the lift and drag is available.
4.4.1 Isolated NACA 0012 profile
The mesh for the numerical analysis of the flow along the NACA 0012 profile
is shown in figure 4.1. The chord length of the profile is 600 mm. The domain
is divided in three subregions. The first region is meshed with an Ogrid
around the profile. This gives good control of the quality of the boundary layer
cells along the surface. A rectangular box is placed around the first region to
make a transition from the Ogrid to an Hgrid. In this second region also an
Ogrid type of mesh is applied.
The third domain extends the numerical domain to either the tunnel walls in
the experiments or to a distance to impose the farfield boundary conditions.
At the interface between the second and third domain an arbitrary coupling
method is employed, which allows nonmatching cells at both sides of the
interface.
The Ogrid region closest to the profile surface can not be recognised well in
the mesh plot due to the large number of mesh lines. The other two regions
can be distinguished more clearly. In this approach the mesh around the
profiles is identical for all calculated conditions. Regions 1 and 2 can be
rotated to obtain the desired angle of attack for the profile.
Mesh dependency studies have been carried out to evaluate the variation in
lift and drag prediction for an angle of attack of 4 degrees. The number of
cells around the profile and in the direction perpendicular to the profile have
been varied. The number of cells in the Ogrid around the profile has been
varied from 150 to 330 cells. In the normal direction the number of cells is
increased from 32 to 48. The number of cell in normal direction in the first
region is kept constant to keep a constant y
+
value of about 110. This is in
accordance with the requirements for the use of the wall functions.
At the upper and lower boundary of the domain two types of boundary
conditions can be applied: (i) wall boundary conditions or (ii) constant
pressure boundary conditions. The first type can be used if the experimental
data is obtained from wind tunnel tests. The second type is suitable for an
unbounded region. For sufficiently large numerical domains both types will
76
Chapter 4. Mathematical treatment
give comparable results. At the inlet boundary a uniform velocity distribution
of 10 m/s is prescribed and the constant density of water is used. The
Reynolds number for the calculations becomes . Turbulence intensity
is set to 0.01% and the length scale is set to a small fraction of the tunnel
height. All calculations have been carried out with the standard kε turbulence
model and employing wall functions. Solution is based on a second order
MARS (= Monotone Advection and Reconstruction Scheme) discretisation
scheme for the momentum equations. This second order method is least
sensitive to the mesh structure and skewness [9]. The kε model turbulence
equations are discretised with a first order upwind differencing scheme.
Convergence behaviour of one of the calculated conditions is shown in figure
4.2. The convergence criterion for all calculations is set to 10
4
for the
momentum, mass and turbulent kinetic energy equations.
6 10
6
⋅
Figure 4.1 Plot of a part of the mesh as used in calculations of isolated
NACA 0012 profile.
4.4 Twodimensional test cases
Numerical analysis of a waterjet propulsion system 77
Figure 4.3 shows a comparison of the calculated and measured lift and drag
for the isolated NACA 0012 profiles. The experimental data is taken from
Abbott & von Doenhoff [13], where the results of the measurements with
smooth profiles at a Reynolds number of are used. Compressibility
has been negligible for the tested conditions.
The numerical values for lift and drag are based on the integrated pressure
and shear forces acting on the profile surface. The dimensionless lift and
drag coefficients are defined as:
(4.35)
(4.36)
where ρ is the density, v the freestream velocity and A the surface area of
the wing, i.e. here the chord length times the width in span wise direction. L is
the lift force and D the drag force.
1.0E06
1.0E05
1.0E04
1.0E03
1.0E02
1.0E01
1.0E+00
1.0E+01
0 50 100 150 200 250
Uvelocity residu
Vvelocity residu
Mass residu
Turbulent energy residu
Epsilon residu
Figure 4.2 Plot of convergence of a calculation of NACA 0012 profile at
angle of attack of 4 degrees for mesh with 222 cells around
profile and 40 cells in normal direction.
Iteration number
mass
R
e
s
i
d
u
a
l
u velocity
v velocity
dissipation
turbulent energy
6 10
6
⋅
c
l
L
1
2
ρv
2
A
 =
c
d
D
1
2
ρv
2
A
 =
78
Chapter 4. Mathematical treatment
Agreement is good for the lift up to an angle of attack of about 6 degrees. The
comparison of calculated and measured drag shows a clear overprediction.
The relative error increases from 33% at 0 degrees angle of attack to 55% at
an angle of attack of 4 degrees. The deviation between the measurements
and the calculations continuously increases with larger angles of attack.
The results of the mesh sensitivity study are shown in table 4.3 for the lift
prediction and in table 4.4 for the drag prediction at an angle of attack of 4
degrees. The relative difference in lift coefficient between the minimum and
maximum lift is about 3%. For the drag coefficient a variation of about 9% is
found. The results of the finer meshes do not show a reduction of the
deviation with the experimental data. The error in prediction of drag might be
related to an error in the production term of the turbulence model at the
stagnation point, as described by Moore & Moore [14].
0
0.2
0.4
0.6
0.8
1
0 1 2 3 4 5 6 7 8 9
0
0.01
0.02
0.03
0.04
0.05
cl  EXP
cl  CFD
cd  EXP
cd  CFD
Figure 4.3 Comparison of measured and calculated lift an drag for NACA
0012 profile. Reynolds number of experiments and
calculations is 6.0x10
6
. Mesh is based on 222 cells around
profile and 32 in normal direction
Section angle of attack [degrees]
c
d
[

]
c
l
[

]
4.4 Twodimensional test cases
Numerical analysis of a waterjet propulsion system 79
The sensitivity of the turbulence model is evaluated by a variation of the
turbulence intensity of the free stream flow at the inlet boundary condition.
Calculations are carried out with a mesh with 222 cells around the profile and
32 in normal direction. The level of the turbulence intensity at the inlet
boundary condition is increased from 0.01% to 1.0%. The results for the
prediction of lift and drag coefficient are presented in table 4.5. Variation in lift
coefficient is 4%, whereas the change in drag is about 42%.
The test calculations with the NACA 0012 profile show that the error in
prediction of profile drag remains after some mesh refinement steps.
Moreover the deviation between experimental data and calculations
increases significantly when the level of turbulence intensity at the inlet
boundary increases. It should be noted that the low turbulence levels as used
in the experiments are not representative for the inflow to the waterjet pump.
Table 4.3 Lift coefficient for mesh convergence study. Columns
show number of cells in normal direction and rows show
cells in Ogrid around the profile
c
l
[] 32 cells 40 cells 48 cells
150 cells 0.4418 0.4386 0.4407
222 cells 0.4384 0.4375 0.4366
330 cells 0.4291 0.4289 0.4302
Table 4.4 Drag coefficient for mesh convergence study. Columns
show number of cells in normal direction and rows show
cells in Ogrid around the profile
c
d
[] 32 cells 40 cells 48 cells
150 cells 0.010754 0.011324 0.011493
222 cells 0.010522 0.010988 0.011332
330 cells 0.010784 0.011243 0.011495
Table 4.5 Lift and drag coefficients for calculations with variation of
input values for turbulence intensity
Turbulence intensity [%] c
l
[] c
d
[]
0.01 % 0.4384 0.010522
0.10 % 0.4287 0.012713
0.50 % 0.4242 0.014147
1.00 % 0.4206 0.014980
80
Chapter 4. Mathematical treatment
It is concluded that with the currently used cell sizes, which is comparable
with the sizes to be used in the three dimensional pump mesh, a significant
deviation between calculated and measured profile drag will remain.
4.4.2 Cascades with NACA 65410 profiles
Apart from calculations on isolated profiles, also the flow around profiles in a
cascade have been analysed. Extensive experimental data is available on
tests with NACA 65 compressor blade profiles [15]. Data were reported for
NACA 65 profiles with various camber lines and a maximum thickness of
10% of the chord. Cascades of profiles can be described with two additional
parameters. These are the solidity and the blade angle. The solidity is a
measure for the distance between two profiles in relation to the chord length.
The blade angle is defined as the angle between the profile base line and the
line connecting all leading edges. An example of the cascade is shown in
figure 4.4.Calculations have been carried out for a NACA 65 410 profile with a
blade angle of 20 degrees and a solidity of 1.0 and 1.5. The designation of
the profile is based on a design lift coefficient of 0.4 and a maximum
thickness of 10%. The basic camber line used is the a =1.0 mean line (see for
example [13]).
The mesh is created in a similar way as for the isolated NACA 0012 profile.
Periodic boundaries are applied to simulate a cascade with an infinite number
of profiles. Turbulence model, discretisation scheme and convergence criteria
are identical to the calculations for the NACA 0012 profile.
Figure 4.4 Geometry of NACA 65410 cascade. Cascade solidity σ = 1.0,
blade angle β = 20 degrees.
β
σ
α
4.4 Twodimensional test cases
Numerical analysis of a waterjet propulsion system 81
Figure 4.5 shows the nondimensional pressure distribution C
p
for the
cascade with a solidity of 1.0 and a blade angle of 20 degree. The
dimensionless pressure C
p
coefficient is defined as:
(4.37)
where p is the static pressure, the reference pressure, ρ the density and
the freestream velocity.
The angle of attack varies from 0.5 to 12.5 degrees. Agreement between
calculations and measurements is acceptable for most pressure taps for the
different conditions.
Lift and drag are derived from the CFD results in the same manner as for the
NACA 0012 profile. These results are compared with the experimental data in
figure 4.6. The lift is predicted quite well for most conditions. The deviations
C
p
p p
∞
–
1
2
ρv
∞
2
 =
p
∞
v
∞
1.5
1
0.5
0
0.5
1
0 20 40 60 80 100
Experimental values
CFD results
1.5
1
0.5
0
0.5
1
0 20 40 60 80 100
Experimental values
CFD results
1.5
1
0.5
0
0.5
1
0 20 40 60 80 100
Experimental values
CFD results
1.5
1
0.5
0
0.5
1
0 20 40 60 80 100
Experimental values
CFD results
Figure 4.5 Pressure coefficient distribution along surface of NACA
65_410 profile for different angles of attack. Cascade solidity =
1.0, blade angle = 20 degrees.
α = 0.5 degrees α = 4.5 degrees
α = 12.5 degrees
α = 8.5 degrees
Chord length [%] Chord length [%]
Chord length [%] Chord length [%]
C
p
[

]
C
p
[

]
C
p
[

]
C
p
[

]
82
Chapter 4. Mathematical treatment
between the measured and calculated drag vary between about 10% and
50%. This is in agreement with the results of the isolated NACA 0012 profile.
It can be concluded that the trend of overprediction of drag occurs for both
isolated profiles as well as for profiles in a cascade. In both cases the lift is
predicted much better. In the next subsection the consequences of an error in
lift or drag on the prediction of thrust and torque are discussed.
4.4.3 Sensitivity of errors in drag on thrust and torque
Torque and thrust of a waterjet impeller are related to the tangential and axial
force experienced by the blade sections. These forces can be derived from
the lift and drag of these profiles. Consequently, if the blade profile drag is
overpredicted, there will be an effect on the prediction of the torque and the
thrust. Lift and drag are transformed into axial and tangential forces with:
(4.38)
where φ is the inflow angle, L the lift force and D the drag force. The inflow
angle φ is related to the blade angle β and the angle of attack α of the flow
with respect to the chord line:
(4.39)
Figure 4.6 Comparison of measured and calculated lift and drag for
NACA 65_410 profiles in cascade with solidity of 1.0 and blade
angle of 20 degrees.
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
4 0 4 8 12 16 20
0.02
0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
cl  exp
cl  CFD
cd  exp
cd  CFD
Angle of attack [degrees]
c
d
[

]
c
l
[

]
F
ax
φ L ⋅ cos φ D ⋅ sin – =
F
tan
φ L ⋅ sin φ D ⋅ cos + =
φ β α – =
4.4 Twodimensional test cases
Numerical analysis of a waterjet propulsion system 83
The drag prediction based on CFD calculations can be expressed as:
(4.40)
where ε represents the relative overprediction of the profile drag. It is
assumed that:
(4.41)
This can be up to about 50%. Figure 4.7 shows a sketch of the forces acting
on a profile. From this sketch the effect of a higher drag can already be
recognised.
Substitution of equation (4.40) in equation (4.38) gives an expression for the
axial and tangential force prediction based on CFD results.These forces can
be related to the exact solutions to determine the resulting relative error in
axial and tangential direction. The relative error for the axial force can be
calculated with:
(4.42)
where φ is the inflow angle, L/D is the lift over drag ratio. The relative error in
tangential direction yields:
(4.43)
D
CFD
D
exact
1 ε + ( ) =
L
CFD
L
exact
=
Figure 4.7 Sketch of forces acting on profile
∆torque
∆
d
r
a
g
∆thrust
φ
α β
F
ax_CFD
F
ax_exact
 1
ε
1
φ tan

L
exact
D
exact
 ⋅ 1 –
 – =
F
tan_CFD
F
tan_exact
 1
ε
φ tan
L
exact
D
exact
 ⋅ 1 +
 + =
84
Chapter 4. Mathematical treatment
To get an indication of the influence of the error in drag prediction on the axial
and tangential force, the data for the NACA 65410 cascade will be used. The
lift over drag ratio as function of the angle of attack is derived from figure 4.6.
The relative errors in axial and tangential force are plotted as function of the
angle of attack in figure 4.8. The drag overprediction is set to 20%, 40% and
60%.
From this diagram is becomes clear that the effect on axial force is very small
even for large overprediction of drag. The estimated error on thrust will be
less than 1% for most cases. On the other hand the error in tangential force
remains significant for realistic values of the angle of attack α and drag error
factor ε. The error in tangential force can be about 2.0 to 8.0%. For cascades
with smaller blade angles β, the error in tangential force will increase even
more.
It is to be expected that this effect will be noticeable in both impeller and
propeller torque calculations. Calculation of thrust seems to be insensitive to
an error in drag. It is expected that the calculation of pump head shows the
same behaviour, based on the similarity between propeller thrust and pump
head as shown in equation (2.34).
2%
1%
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
11%
12%
0 2 4 6 8 10 12 14 16 18


Figure 4.8 Relative errors in tangential and axial forces due to over
prediction of profile drag. Data is based on NACA 65410
cascade with solidity of 1.0 and blade angle of 20 degrees.
Angle of attack [degrees]
F
C
F
D
/
F
e
x
a
c
t
[

]
Tangential force error  ε=0.20
Tangential force error  ε=0.40
Tangential force error  ε=0.60
Axial force error  ε=0.20
Axial force error  ε=0.40
Axial force error  ε=0.60
4.5 Nomenclature
Numerical analysis of a waterjet propulsion system 85
4.5 Nomenclature
A area m
2
c
d
drag coefficient 
c
l
lift coefficient 
C
p
pressure coefficient 
D drag N
F force N
g gravitational acceleration m/s
2
k turbulent kinetic energy m
2
/s
2
L lift force N
l
mix
Prandtl’s mixing length m
P production term of turbulence m
3
/s
2
p pressure N/m
2
Re Reynolds number 
t time s
u
+
dimensionless velocity 
v velocity m/s
∆y
p
distance from wall m
y
+
dimensionless wall distance 
z coordinate in vertical direction m
Greek symbols
α angle of attack degrees
β
blade angle degrees
ε
dissipation m
2
/s
3
ε
relative error in drag prediction 
φ
inflow angle degrees
κ Von Kármán constant 
µ
dynamic viscosity kg/ms
ρ
fluid density kg/m
3
τ
w
wall shear stress N/m
2
ω
vorticity, dissipation rate 1/s
Subscripts
ax axial direction
86
Chapter 4. Mathematical treatment
CFD based on CFD results
eff effective (=laminar + turbulent)
exact based on exact formulations
i,j directions
imp impeller
inl inlet
T turbulent
tan tangential direction
tip impeller blade tip
w wall
x,y,z carthesian coordinate system directions
Superscripts
v’ fluctuating part
v time averaged value
4.6 References
[1] Aungier, R.H.,’Centrifugal compressors’, ASME Press, New York, 2000
[2] Van Esch, B.P.M.,’Simulation of threedimensional unsteady flow in
hydraulic pumps’, PhD thesis, University of Twente, 1997
[3] Van Os, M.J.,’On the flow and cavitation inception of mixedflow impel
lers’, PhD thesis, University of Twente, 1998
[4] Kunz, R.F., Lakshminarayana, B., Basson, A.H., ‘Investigation of tip
clearance phenomena in an axial compressor cascade using Euler and
NavierStokes procedures’, Journal of Turbomachinery, Vol 115, pp
453467, 1993
[5] Goto, A.,’Study of internal flows in a mixedflow pump impeller at vari
ous tip clearances using threedimensional viscous flow computations’,
Journal of Turbomachinery, Vol 114, pp 373382, 1992
[6] Fox, R.W., McDonald, A.T.,’Introduction to fluid mechanics’, Third Edi
tion, John Wiley & Sons, New York, 1985
[7] Wilcox, D.C.,’Turbulence modeling for CFD’, Griffin Printing, Glendale,
1993
[8] Ferziger, J.H., Peric, M., ‘Computational methods for fluid dynamics’,
Second edition, Springer, Berlin, 1999
4.6 References
Numerical analysis of a waterjet propulsion system 87
[9] Computational Dynamics Limited, ‘StarCD methodology, version
3.150’, 2001
[10] Versteeg, H.K., Malalasekera, W., ‘An introduction to Computational
Fluid Dynamics’, Longman Scientific & Technical, Essex, 1995
[11] Schlichting, H., ‘Boundary layer theory’, McGrawHill, New York, 1968
[12] Cebeci, T., Smith, A.M.O., ‘Analysis of turbulent boundary layers’, Aca
demic press, New York, 1974
[13] Abbott, I.H., Von Doenhoff, A.E. ,’Theory of wing sections’, Dover Pub
lications, Inc., New York, 1958
[14] Moore, J.G., Moore, J., ‘Controlling overproduction of turbulence in
twoequation models by limiting the anisotropy of the Reynolds normal
stresses’, 1997 ASME Fluids Engineering Division Summer Meeting,
1997
[15] Emery, J.C., Herrig, L.J., Erwin, J.R., Felix, A.R., ‘Systematic two
dimensional cascade tests of NACA 65series compressor blades at
low speeds’, NACA report 1368, Langley Aeronautical Laboratory, Lan
gley Field, 1958
88
Chapter 4. Mathematical treatment
Numerical analysis of a waterjet propulsion system 89
Chapter 5 Numerical analysis of waterjet
inlet flow
In this chapter the flow through the waterjet inlet will be analysed in more
detail. First a critical review of several published analyses of the flow through
waterjet inlets will be given, which provides additional information about the
suitable numerical approach for the inlet flow CFD analysis. Discussion of the
application of the numerical method is divided into two parts: (i) the mesh
generation for the three dimensional inlet geometry based on the
experimental setup for the finite volume analysis and (ii) the selection of
boundary conditions. In order to validate the computational method the CFD
results are compared with experimental data for a model scale waterjet inlet
duct. During the validation process, the effects of the confinement of the flow
due to the cavitation tunnel walls are also addressed.
The CFD results can be used to visualise the flow behaviour in more detail.
Besides pressure and velocity distribution, it is also interesting to quantify
wall friction and to determine the shape of the dividing streamtube.
5.1 Review of CFD analyses on waterjet inlets
Calculations for a three dimensional waterjet inlet have been reported by
Førde et al. [1]. For the calculations an Euler method was employed, but
computed results were not validated with measurements. The presented
velocity distribution at the impeller plane is not in agreement with a typical
flush type waterjet inlet. This is caused by the neglect of the boundary layer
velocity profile at the inlet and absence of viscous effects in the method,
90
Chapter 5. Numerical analysis of waterjet inlet flow
necessary for describing the development of the boundary layer in the inlet
duct.
A potential flow method has been applied by Van Terwisga [2]. Comparison of
the calculated pressure distribution along the ramp and the cutwater (or inlet
lip) with experimental data shows large deviations. This is due to the neglect
of viscous losses and due to small deviations in the calculated velocity field
near the stagnation point at the cutwater. A small deviation of the angle of
attack at the cutwater can lead to relatively large differences in the prediction
of the static pressure distribution.
Pylkkänen ([3], [4]) presented results of CFD calculations obtained from a
RANS code. A twodimensional model of a waterjet inlet is used for these
analyses. Differences between calculations and measurements of pressure
are about 11 to 15%. The experimental data is based on measurements of a
threedimensional inlet on a windtunnel. It is known that an actual three
dimensional waterjet inlet ingests water from a region that is wider than the
inlet itself. In a twodimensional situation this phenomenon cannot be
reproduced, and therefore the pressure distribution at the lip section will be
different. In the second part of the inlet, the rectangular crosssection
transitions to a circular crosssection in general. This transition is not taken
into account in a twodimensional analysis either. Therefore it is expected that
the pressure distribution in this part of the inlet will show a deviation from the
threedimensional case as well.
The necessity of considering the threedimensional geometry is shown by
Van der Vorst et al. [5]. Calculations were made for a two and for a three
dimensional case. Results of both calculations were compared with
experimental data. The calculated pressure distribution along the inlet ramp
centre line shows agreement with the measurements for the three
dimensional analysis, whereas large deviations are found for the results of
the calculations of the twodimensional geometry. The pressure
measurements were performed for a model scale waterjet inlet, mounted on
top of a windtunnel.
Another example of the use of viscous methods for the calculation of three
dimensional inlet flows was reported by Seil et al. [6]. In these calculations,
the geometry of the pump is included. The effect of the impeller is modelled
with an actuator disk, however. Calculated results show that the location of
the stagnation point at the cutwater depends on the IVR value. It is also found
that the level of nonuniformity of the velocity profile at the bend increases
with increasing IVR. This nonuniformity vanishes towards the impeller plane,
where the actuator disk is located. This behaviour may be a result of the
implementation of the actuator disk. Yang et al. [7] have also calculated the
flow around a waterjet inlet with a viscous flow method. A complete hull is
included in the computational domain. Agreement between measurements
and CFD results is poor for the pressure distribution along the ramp and lip
centre lines.
5.2 Geometry and mesh generation
Numerical analysis of a waterjet propulsion system 91
Experimental data from windtunnel tests has been used for comparison with
calculations of the viscous flow through a threedimensional inlet geometry
[8]. Agreement between measurements and calculations is good for the static
pressure distribution along the ramp for a range of IVR values. Moreover, the
typical nonuniform velocity distribution in the impeller plane is reproduced
well for the conditions considered. A comparison between the calculated and
measured velocity distribution is shown for an IVR of 1.59.
Hu and Zangeneh [9], [10] have presented an optimization algorithm for
waterjet inlet geometries. This method optimizes the twodimensional
symmetryplane geometry of the inlet, which is then extended in the third
direction. Both the twodimensional as well as the threedimensional
geometries are analysed using a viscous flow method.
The development described above has resulted in a widespread use of three
dimensional viscous flow calculations for the analysis of the flow phenomena
in waterjet inlets. However, these calculations are not always validated with
measurements of static pressure and velocity.
Validation of the CFD method requires an accurate set of experimental data.
This data can be obtained from cavitation tunnel experiments, for example. In
a test setup with an inlet mounted on top of a cavitation tunnel, all governing
parameters of the operating condition can be measured accurately. If a wind
tunnel is used for measurements instead of a cavitation tunnel, the air density
has to be monitored simultaneously in order to be able to determine the flow
rate accurately. Relatively small variations in flow rate will result in a deviation
of the IVR. Another important aspect in the experimental setup is the
capability to create an incoming hull boundary layer of sufficient thickness, in
accordance with normal waterjet applications. Accurate measurements of the
static pressure at the ramp and the velocity distribution at the impeller plane
were obtained in the Tom Fink cavitation tunnel [11]. These measurements
are used for validation purposes in this chapter.
Measurements on actual waterjet installations can provide validation data as
well. Some typical problems of model scale testing, like Reynolds scaling
effects, artificial boundary layer thickening and confinement of the flow by
tunnel walls are eliminated in these measurements. On the other hand, it is
very difficult to determine the exact operating conditions of the waterjet
installation. This limits the use of measurement data obtained from full scale
waterjet installations.
5.2 Geometry and mesh generation
The analysis presented in this chapter is based on the geometry of the inlet
as used in the experimental setup, as described in subsection 3.1.1. The
calculations are carried out to reproduce the flow phenomena of the
measured conditions.
92
Chapter 5. Numerical analysis of waterjet inlet flow
The geometry of a waterjet inlet duct can be described by a number of
parameters, e.g. the pump inlet diameter, the inlet angle, the radius of
curvature of the bend and the shape of the cutwater. It is therefore convenient
to develop a fully parametric threedimensional geometry and mesh
generator, based on a list of geometric parameters. Figure 5.1 shows a
sketch of a twodimensional inlet geometry with the main parameters. Based
on the geometric parameters the threedimensional shape of the inlet is
calculated. This geometrical data is used in the CFD preprocessor. From this
input data the block definitions are created. The topology of the blocks is kept
identical for all waterjet inlets. Figure 5.2 shows a typical output of the used
block structure. A thin layer of cells is created at the walls of the inlet, in order
RAMP RADIUS
TRANSITION LENGTH
CYLINDRICAL
LENGTH
BENDING RADIUS
SHAFT
DIAMETER
INLET
ANGLE
TRANSITION
ANGLE
DIAMETER
INLET
SHAFT
DIAMETER
RAMP
POINT
TANGENCY
CUTWATER
Figure 5.1 Inlet geometry with main parameters and specific
nomenclature listed
Figure 5.2 Wire frame plot of block structure of the inlet mesh
5.2 Geometry and mesh generation
Numerical analysis of a waterjet propulsion system 93
to get high quality cells in the boundary layer. The averaged y
+
value of the
cells near the wall is about 60 for this mesh.
Near the socalled cutwater (or inlet lip), a local refinement of the mesh is
applied to capture the gradients of the flow field better. The final mesh is
shown in figure 5.3a and in figure 5.3b a detail of the cutwater with local
refinement and the region around the shaft is shown. In case of computations
Figure 5.3a Final mesh of waterjet inlet (half of the complete domain)
Figure 5.3b Detail of cutwater mesh with local refinement and regular mesh
coupling on the block boundaries (left) and mesh around the
(stationary) shaft
94
Chapter 5. Numerical analysis of waterjet inlet flow
with a stationary shaft or without a shaft, the computational domain is
restricted to half the geometry for reasons of symmetry with respect to the
vertical plane through the centreline of the pump.
5.3 Numerical approach
5.3.1 Boundary conditions
All cell faces at the boundary of the computational domain require some type
of boundary condition.
At the inlet of the computational domain an inlet type boundary condition is
applied. This type of boundary condition requires a prescription of the
velocities in all three directions and values for the turbulence intensity and the
length scale, if a turbulence model is used. The velocity profile, which
represents the hull boundary layer, is input through a userroutine. In this
routine a powerlaw velocity profile is calculated for all cells that are located in
the boundary layer. The presented calculations are based on a boundary
layer profile with a powerlaw value n=7 and a thickness δ of 0.3D, with D the
diameter of the inlet. An undisturbed uniform velocity is prescribed in the
remainder of the cells. The turbulence intensity is set to 2.0% and the length
scale to 0.05 m, which is equivalent to about 8% of the tunnel inlet hydraulic
diameter.
At the impeller plane the total mass flow leaving the domain through the inlet
duct is imposed. This is established with a fixed flow outlet boundary
condition. This type of condition allows for a nonuniform velocity distribution
over the surface. The pressure distribution in the outflow plane is part of the
solution as well.
For the outflow plane of the cavitation tunnel a constant pressure boundary
condition is used. It is assumed that the static pressure is uniform at large
enough distance from the waterjet inlet. The resulting velocity distribution will
be nonuniform, however.
The side plane and the bottom plane of the domain are placed at the location
of the cavitation tunnel walls. For these walls, the slip condition is applied as
the wall boundary condition. The mesh near these walls can be made
relatively coarse, since the boundary layer is not resolved. With this boundary
condition the normal velocity is set to zero, which simulates the effect of the
wall on the flow. The effects of the development of the natural tunnel wall
boundary layer, characterised by the displacement and momentum thickness,
are neglected. The effects of the actual boundary layer in the cavitation tunnel
are analysed during the tests and it is concluded that the effect of blockage
was less than 0.2%.
For conventional waterjet inlet CFD analyses the tunnel walls are not taken
into account. The side and bottom planes of the domain are modelled as
5.3 Numerical approach
Numerical analysis of a waterjet propulsion system 95
constant pressure planes. Using pressure boundary conditions, additional
inflow of water is allowed. The difference between wall boundary conditions
and constant pressure boundary conditions is investigated later on in this
thesis.
Calculations for a half model geometry require symmetry conditions at the
symmetry plane. If the complete model is analysed with shaft rotation, an
additional wall boundary condition is applied to the shaft surface in order to
model the rotation.
5.3.2 Fluid properties
Selection of the fluid properties for the waterjet inlet analysis is
straightforward. As discussed in the previous chapter, the flow can be
considered as incompressible, which results in a constant density. All model
scale calculations have been carried out with a density ρ of 1000 kg/m
3
. The
dynamic fluid viscosity µ is set to 0.001 kg/ms.
Turbulent flow behaviour is modelled with a turbulence model. Though the
CFD method provides several different turbulence models with different
levels of complexity, the well established highReynolds number flow kε
turbulence model is applied for all calculations. This also implies the use of
wall functions to impose the noslip boundary condition. It is acknowledged
that the standard kε turbulence model has a moderate performance for some
types of flow. These are (i) some external unconfined flows, (ii) flows with
large extra strains (e.g. the flow in curved boundary layers, swirling flows), (iii)
rotating flows and (iv) fully developed flows in noncircular ducts, see [12].
Results from the validation process will show whether the choice of this
turbulence model for the present flow is acceptable.
The effect of gravity is accounted for in the analyses, but it should be
mentioned that this is only an additional postprocessing feature, since the
density is constant.
5.3.3 Discretisation and solution algorithm
Solution of the partial differential equations requires a discretisation scheme.
As for the turbulence models, several methods are provided within the CFD
method used. Here all calculations have been performed employing the
second order MARS scheme (short for Monotone Advection and
Reconstruction Scheme) for the momentum equations. This second order
method is least sensitive to the mesh structure and skewness [13]. The kε
model turbulence equations are discretised with a first order upwind
differencing (UD) scheme.
The coupling between the velocity and pressure field is resolved with an
iterative solution strategy based on the SIMPLE pressurecorrection method.
96
Chapter 5. Numerical analysis of waterjet inlet flow
In this algorithm, originally put forward by Patankar and Spalding [14], the
convective fluxes are evaluated from an estimated velocity field. Furthermore,
an estimation for the pressure distribution is used to solve the momentum
equations. The continuity equation then yields a pressure correction
equation, which yields a pressure correction field. This pressure correction
term is used in turn to update the estimated velocity and pressure field. This
process is iterated until the velocity and pressure fields are converged.
The system of discretised partial differential equations is solved with an
Algebraic MultiGrid (AMG) algorithm.
Figure 5.4 shows the convergence behaviour of a calculation for an IVR of
1.87. In order to accelerate the convergence, first 125 iterations are carried
out with the first order UD scheme for the velocity components. The restart
with MARS discretisation causes the step in the residuals of the momentum
and the mass. The convergence criterion is set to 10
4
for the momentum,
mass and turbulent kinetic energy equations.
5.4 Validation with experimental data
The experimental data used in this chapter is measured at the Tom Fink
cavitation tunnel [11] as described in subsection 3.1.1. In this test program
measurements are made on two different inlet geometries. The experimental
program consists among others of static pressure measurements along the
1.0E06
1.0E05
1.0E04
1.0E03
1.0E02
1.0E01
1.0E+00
1.0E+01
0 50 100 150 200 250 300
Uvelocity
Vvelocity
Wvelocity
Mass
Turbulent energy
Dissipation
Figure 5.4 Plot of convergence of a calculation of waterjet inlet flow
phenomena for IVR = 1.87.
Iteration number
R
e
s
i
d
u
a
l
mass
u velocity
v velocity
dissipation
turbulent
w velocity
energy
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 97
ramp centre line, total pressure measurements at the plane just upstream of
the impeller, cavitation inception observations at the cutwater and
visualisation of streamlines. In the following sections, measured data will be
used for comparisons with CFD results.
5.4.1 Comparison of static pressure along the ramp centre line
The static pressure is measured at the ramp centre line at 12 different
locations. The locations are determined by the distance along the ramp
centre line from the impeller plane towards the entrance of the inlet. The
entrance of the inlet, sometimes denoted as ramp tangency point, is located
at a distance of 1000 mm for this model scale inlet.
The static pressure is made nondimensional using the density ρ and the
tunnel speed v
tunnel
:
(5.1)
The reference static pressure p
ref
and the tunnel speed are taken at a
location upstream of the inlet, since the velocity, and the pressure
downstream of the inlet vary with the value of IVR value (see equation
(2.12)).
The static pressure is measured for eight different IVR values. Table 5.1
shows the conditions, used for the measurements. The Reynolds number
Re
inl
is defined in equation (4.2).
Table 5.1 Parameters of measured conditions for the static pressure
distributions. D
inlet
= 150 mm.
IVR []
(=v
tunnel
/v
pump
)
v
tunnel
[m/s]
v
pump
[m/s]
Re
inl
[]
1.07 8.00 7.48
11.21 x 10
5
1.21 8.00 6.61
9.92 x 10
5
1.29 8.00 6.20
9.30 x 10
5
1.50 8.00 5.33
8.00 x 10
5
1.70 8.00 4.71
7.06 x 10
5
1.87 8.00 4.28
6.42 x 10
5
2.03 8.00 3.94
5.91 x 10
5
2.19 8.00 3.65
5.48 x 10
5
C
p
p p
ref
–
1
2
ρv
tunnel
2
 =
98
Chapter 5. Numerical analysis of waterjet inlet flow
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0 200 400 600 800 1000 1200
Experiments
CFD
IVR = 1.21
IVR = 1.50
IVR = 1.87
IVR = 2.19
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0 200 400 600 800 1000 1200
Experiments
CFD
IVR = 1.07
IVR = 1.29
IVR = 1.70
IVR = 2.03
Figure 5.5 Comparison of measured and calculated pressure coefficient
Cp along the ramp centre line. Values are given as a function
of the distance from the impeller plane. The data for different
values of IVR are divided in 2 groups for improved visibility.
Distance along ramp [mm]
C
p
[

]
Distance along ramp [mm]
C
p
[

]
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 99
The comparison with CFD results is shown in figure 5.5. Values of pressure
coefficient Cp along the ramp centre line are given as function of the distance
from the impeller plane. The two diagrams present the data for alternate
values of IVR.
Both figures show a good agreement between measurements and CFD
calculations along the entire ramp section. The discontinuity in calculated
pressure between 350 mm and 400 mm is due to the presence of the shaft.
This influence is strongest at low IVR since the velocity in the inlet duct is
then highest. The static pressure at the impeller plane increases with
increasing IVR (due to decreasing v
pump
) according to the expectations.
However, at high IVR values the static pressure is more or less constant. This
is due to increased nonuniformity and hydraulic losses.
Figure 5.6 shows the pressure distribution in the symmetry plane of the
configuration for an IVR of 1.07 (top) and 2.03 (bottom). It can be observed
that the effect of IVR on the pressure distribution is not restricted to the ramp.
Clear differences in pressure can be recognized at the cutwater and in the
bend of the inlet.
From the pressure distributions can be seen that the location of the
stagnation point at the cutwater changes with variation of IVR. As a
consequence, the location of minimum value of the pressure also changes.
This phenomenon is reported before by Seil [6]. It is shown in more detail in
figure 5.7, where the pressure distribution along the cutwater is presented for
all calculated IVR conditions. On the left part of the diagram the negative s
coordinates represent the lower part of the cutwater, whereas the positive s
coordinates represent the upper part of the cutwater. The influence of IVR on
the results can be recognised clearly. The stagnation point moves from a
negative s coordinate in the positive direction for increasing IVR. The
locations of the minimum pressure can be divided in two groups. For IVR
conditions up to 1.29 the minimum pressure is found at a location with
positive s coordinate. For higher IVR, the location of minimum pressure is
found at a negative s coordinate.
The movement of the stagnation point, and consequently the location of the
minimum pressure, is related to the change in the shape of the dividing
streamtube with varying IVR. The streamtube analysis will be given in section
5.5.
Analysis of the pressure in 5.7 learns that the static pressure at the tunnel
outlet plane is not equal to the reference pressure at the inlet plane and
moreover, this variation is dependent on IVR. This deviation in static pressure
might influence the experimental determination of the cavitation inception
pressure, when the location of the minimum pressure is at the lower side (or
hull side) of the inlet for high IVR conditions. The cavitation inception results
will be compared with CFD results in the next section.
100
Chapter 5. Numerical analysis of waterjet inlet flow
Figure 5.6 Pressure coefficient Cp at symmetry plane for IVR of 1.07 (top)
and IVR of 2.03 (bottom).
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 101
5.4.2 Comparison of cavitation inception pressure at cutwater
Cavitation below the cutwater occurs at high IVR conditions. For four IVR
values the inception pressure is determined by visual observation. During the
tests the reference pressure p
ref
in the tunnel is gradually reduced until a
small cavity is observed. With this procedure only cavities at the lower side of
the cutwater could be observed. These are typical for high IVR conditions.
The measuring point for the reference pressure p
ref
is located upstream of
the inlet at half the tunnel height, as shown in figure 3.1 on page 50.
The cavitation inception pressure is presented in nondimensional form,
according to:
(5.2)
where p
v
is the vapour pressure of the fluid, h the height correction between
the cutwater and the centre plane of the cavitation tunnel and v
pump
the
average axial inflow velocity of the pump. The comparison of the
experimental results and the CFD results is shown in figure 5.8.
2.5
2
1.5
1
0.5
0
0.5
1
1.5
100 80 60 40 20 0 20 40 60 80 100
IVR = 1.07
IVR = 1.21
IVR = 1.29
IVR = 1.50
IVR = 1.70
IVR = 1.87
IVR = 2.03
IVR = 2.19
Distance s along cutwater [mm]
C
p
[

]
IVR=1.07
IVR=1.07
IVR=2.19
IVR=2.19
s<0
s>0
s=0
Figure 5.7 Calculated pressure coefficient Cp along the cutwater for
different IVR values. Negative s coordinate represents lower
side of cutwater and positive coordinate represents upper side.
σ
v_pump
p
ref
p
v
– ρgh –
1
2
ρv
pump
2
 =
102
Chapter 5. Numerical analysis of waterjet inlet flow
It should be noted that the pump inflow velocity is used in the definition of
σ
v_pump
, whereas the ship speed is used for C
p
. With the current σ definition
the inception behaviour of different inlet geometries at high IVR can be
evaluated better.
Agreement between experiments and CFD calculations is good for all four
tested conditions. The calculated conditions at lower IVR values show the
expected behaviour. At a certain IVR the dividing streamline is optimally
aligned with the cutwater geometry, which results in a minimum value for
σ
v_pump
. Further decrease of IVR results in the point of minimum pressure,
i.e. the cavitation inception point, moving to the inner side of the inlet.
Effect of tunnel walls
At normal waterjet operating conditions cavitation inception occurs below the
cutwater in general. Figure 5.7 shows that the pressure in this region is not
constant and depends on IVR. This is due to the effect of tunnel walls and
conservation of mass in the complete system. The mass flow which enters
the tunnel is split into a part which leaves the domain through the waterjet
inlet and a part which leaves through the tunnel exit section. As long as the
pump exit is open the mass flow rate at the exit of the tunnel section is lower
than at the inlet. Because the tunnel crosssection is constant, this results in
0
1
2
3
4
5
6
7
1 1.2 1.4 1.6 1.8 2 2.2 2.4
Measurements
CFD calculations
Figure 5.8 Comparison of measured and calculated cavitation inception at
cutwater
IVR []
σ
v
_
p
u
m
p
[

]
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 103
lower velocities and consequently a higher static pressure. The pressure
increase as a function of IVR can be estimated as:
(5.3)
where A
pump
is the crosssectional area at the impeller plane and A
tunnel
the
crosssectional area of the tunnel. The effect of viscous losses is neglected in
this estimation.
The diameter of the crosssectional area at the impeller plane is 150 mm and
the cavitation tunnel has a square section of 600 x 600 mm. The velocity at
the outlet of the tunnel shows a drop of about 5% for an IVR = 1.0 to 2% for
IVR = 2.0. Figure 5.9 shows the pressure difference ∆Cp based on the CFD
results and the theoretical value. Agreement between the analytical value
and the numerical result is good over the complete IVR range.
Equation (5.3) neglects the effect of the displacement thickness of the
boundary layer on the tunnel walls in the calculation of the mass fluxes.
Inclusion of this displacement thickness will result in a slightly lower mass
flow entering the tunnel section, which is equivalent to a lower IVR. The
analytical pressure difference estimate will increase with less than one
percent in this case. This simplification is assumed to be justified. Moreover,
hydraulic losses in the tunnel are not taken into account in the estimation.
These hydraulic losses will reduce the actual pressure increase in the tunnel.
However, for a first indication of the effect of the confinement of the tunnel
∆C
p
2A
pump
IVR A
t unnel
⋅

A
pump
IVR A
tunnel
⋅

\ .
 
2
– =
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
1 1.2 1.4 1.6 1.8 2 2.2 2.4
CFD
Equation (5.3)
Figure 5.9 Dimensionless pressure increase in tunnel section based on
CFD results and analytical formula
IVR []
∆
C
p
[

]
104
Chapter 5. Numerical analysis of waterjet inlet flow
walls on the pressure distribution, equation (5.3) can be used as a first
estimate.
In actual conditions at open sea the equivalent cross sectional area goes to
infinity and consequently the pressure difference tends to zero. This is in
accordance with the expectations.
Effect of the tunnel walls on static pressure distribution along the ramp is
shown in figure 5.10 for two IVR values. These are the same conditions as
shown in the symmetry plane pressure distributions of figure 5.6. Clear
deviations between the calculation with and without the walls can be seen in
the first part of the inlet. In this region the pressure is influenced most by the
local tunnel pressure. Moreover, the deviations are larger for the low IVR
condition. This is in line with equation (5.3). Further downstream in the inlet
duct the pressure distribution is similar for both configurations.
It is to be expected that the largest influence of the tunnel pressure increase,
due to the wall confinement, is found in the tunnel downstream of the
cutwater. This is the location where the cavitation inception occurs at medium
and at high IVR. The effect of the tunnel walls on cavitation inception
pressure is shown in figure 5.11. The deviations between the two series of
calculations are large. For a cavitation free design, the cavitation inception
diagram of the inlet is matched with the available ambient pressure. The
cavitation inception diagram represents the required pressure to avoid
cavitation. As long as the required pressure is lower than the available
ambient pressure, cavitation free operation is possible. Figure 5.11 shows
that the required pressure, based on experiments or calculations with tunnel
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0 200 400 600 800 1000 1200
IVR = 1.07  wall
IVR = 1.07  no wall
IVR = 2.03  wall
IVR = 2.03  no wall
Inlet geometry
Figure 5.10 Comparison of calculations of static pressure distribution along
ramp for configuration with and without tunnel walls
Distance along ramp [mm]
C
p
[

]
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 105
walls is too optimistic for medium and high IVR. The actual required pressure
is higher, as can be seen from the calculations without tunnel walls. This
effect should always be taken into consideration when interpreting cavitation
inception results in a test setup with finite dimensions.
5.4.3 Comparison of total pressure at impeller plane
Good agreement between measurements and calculations is shown in the
previous subsections for the static pressure along the ramp and the cavitation
inception that occurs at the cutwater. In this subsection the total pressure
distribution in the impeller plane is evaluated. The impeller plane is defined as
the crosssectional area at the end of the inlet, just upstream of the impeller,
as shown in figure 3.1.
Total pressure measurements have been made with a pitotrake positioned at
different radii. In the tangential direction steps of 10 degrees are made
between the measurement locations.
Figures 5.12a to 5.12d show the comparison between the measured and the
calculated total pressure distribution for four different IVR conditions. The
lowest IVR is a normal operating condition of 1.68 and the highest IVR is a
very high speed condition of 2.19. The other two IVR conditions are in
between with values of 1.87 and 2.03. The results are made nondimensional
with the tunnel speed in a similar way as the static pressure, i.e.:
0
1
2
3
4
5
6
7
1 1.2 1.4 1.6 1.8 2 2.2 2.4
Measurements
CFD with tunnel walls
CFD without tunnel walls
Figure 5.11 Comparison of calculations of cavitation inception pressure for
configuration with and without tunnel walls
IVR []
σ
[

]
106
Chapter 5. Numerical analysis of waterjet inlet flow
(5.4)
The experimental data shows totalpressure losses up to 65% at the highest
pump speeds (low IVR) in the wall boundary layer. In the region around 12
o’clock, i.e. the region affected by the flow around the stationary shaft losses
are found up to 50%.
C
ptot
p
tot
p
ref
–
1
2
ρv
tunnel
2
 =
Figure 5.12a Comparison of measured (left) and calculated total pressure
distribution in impeller plane for IVR = 1.68
Figure 5.12b Comparison of measured (left) and calculated total pressure
distribution in impeller plane for IVR = 1.87
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 107
For all four conditions the qualitative agreement between the measurements
and the calculations is good. The typical distribution and the effect of the shaft
is reproduced well within the CFD model. It appears that the computed
boundary layer is thinner than in the experiment. However, measuring with a
total pressure tube close to the wall will be troublesome.
Comparison of the four IVR conditions shows that in the region affected by
the presence of the shaft, a decrease of the total pressure level with
increasing IVR takes place. This means an increase of hydraulic losses with
increasing IVR. This increase is in accordance with the expectations and this
is partly due to the increased retardation of the flow as discussed in chapter
3.
Figure 5.12c Comparison of measured (left) and calculated total pressure
distribution in impeller plane for IVR = 2.03
Figure 5.12d Comparison of measured (left) and calculated total pressure
distribution in impeller plane for IVR = 2.19
108
Chapter 5. Numerical analysis of waterjet inlet flow
5.4.4 Comparison of velocity field at impeller plane
Total pressure measurements of the previous subsection are used to derive
an axial velocity field. In this derivation a constant static pressure over the
crosssection is assumed. moreover, the influence of the inplane velocity
components is neglected. The axial velocity is derived from the experimental
data according to:
(5.5)
Integration of the axial velocity over the impeller plane showed that the flow
rate was predicted within 2.5% compared to the measured flow rate. The
comparison between the value of the axial velocity derived from the
measured total pressure and the calculated axial velocity component is
shown in figures 5.13a to 5.13d in a similar way as for the total pressure
distribution. The axial velocity is normalised with the averaged axial velocity.
v
axial
p
tot
p
stat
–
1
2
ρ
 =
Figure 5.13a Comparison of axial velocity derived from measured total
pressure (left) and calculated axial velocity distribution in
impeller plane for IVR = 1.68
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 109
Figure 5.13b Comparison of axial velocity derived from measured total
pressure (left) and calculated axial velocity distribution in
impeller plane for IVR = 1.87
Figure 5.13c Comparison of axial velocity derived from measured total
pressure (left) and calculated axial velocity distribution in
impeller plane for IVR = 2.03
110
Chapter 5. Numerical analysis of waterjet inlet flow
Qualitative agreement between experimental data and calculated velocity
field is good for all four IVR conditions. The location and magnitude of
minimum and maximum axial velocity is predicted well with the numerical
method.
The numerical results of the velocity distribution are compared to the
representation of the measured data by Fourierseries (see section 3.1). The
deviation between the two is used to quantify the relative error of the
calculated axial velocity. Figure 5.14 shows the relative difference for the both
low IVR of 1.68 and the high IVR of 2.19.
The relative difference is defined as:
(5.6)
with v
EXP
based on the two dimensional Fourier representation of the
measured data.
The deviations are below 15% for a significant part of the crosssectional
area. It is concluded that the CFD analysis of the inlet flow reproduces the
typical nonuniform velocity distribution well. Therefore, the CFD method
employed in the present investigation seems suitable for the investigation of
the effect of nonuniform inflow into the mixedflow waterjet pump.
Figure 5.13d Comparison of axial velocity derived from measured total
pressure (left) and calculated axial velocity distribution in
impeller plane for IVR = 2.19
∆
v
CFD
v
EXP
–
v
aver
 =
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 111
The velocity distributions show a clear increase of nonuniformity with
increasing IVR,i.e. decreasing v
pump
. The nonuniformity can be represented
in a single value, if calculated according to equation (3.1). Figure 5.15 shows
the nonuniformity parameter ζ as a function of IVR. Both the results of
calculations with wall and pressure boundary conditions have been used. The
relation between the IVR and the nonuniformity is shown clearly for the two
types of boundary conditions. The small deviations are negligible.
Figure 5.14 Relative difference between experimental data and calculated
axial velocity distribution at impeller plane for IVR = 1.68 (left)
and IVR = 2.19
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Wall boundaries
Pressure boundaries
Figure 5.15 Nonuniformity parameter ζ in impeller plane as function of IVR
IVR []
N
o
n

u
n
i
f
o
r
m
i
t
y
ζ
[

]
112
Chapter 5. Numerical analysis of waterjet inlet flow
Figure 5.13d shows a region with very low axial velocity in the upper part of
the crosssectional area. For such conditions there may even be a region of
flow recirculation upstream of the impeller plane.
The distribution of the magnitude of the velocity in the symmetry plane for two
different values of IVR, is shown in figure 5.16. At low IVR conditions the
effect of the bend on the velocity distribution can be recognised. At the inner
corner of the inlet the velocity reaches a maximum value which reduces
further downstream. This effect is less pronounced at high IVR conditions.
In the region in the bend above the shaft a very low velocity magnitude is
observed. Here flow separation is likely to occur at sufficiently high IVR, i.e.
low values of v
pump
. The method to determine boundary layer separation will
be discussed in section 5.6, where additional flow phenomena in a waterjet
inlet are reviewed.
Figure 5.16 Distribution of magnitude of velocity in symmetry plane for IVR
1.07 (top) and 2.03 (bottom)
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 113
5.4.5 Results obtained with kω turbulence model
The use of turbulence models in CFD calculations always provides reason for
discussion. Though an indepth study of effects of turbulence models on the
flow in waterjets does not fit within the scope of this thesis, results of
calculations with the kω turbulence model will be shown for comparison with
results of the kε turbulence model. The kω turbulence model is used widely
just like the kε turbulence model. Often, the level of accuracy or the absence
of accuracy of a CFD calculation is attributed to the turbulence model used.
Comparison of the experimental data with the results of calculations
employing the kω turbulence model can show whether the result is sensitive
to the choice of a particular turbulence model.
Comparison of static pressure along ramp centre line
Figure 5.17 shows the static pressure distribution along the ramp as
calculated with the kω turbulence model. Agreement between calculations
and experimental data is good for all four presented conditions. Similar
results have been found for intermediate IVR conditions. The accuracy of the
calculations is comparable to the calculations with the kε turbulence model,
as shown in figure 5.5. The main differences are in the region downstream of
the shaft, where the pressure predicted by the method employing the kω
model gives higher values than the ones predicted by the method using the k
ε model. This suggest that flow separation occurs later for the method
employing the kw model.
Figure 5.17 Comparison of measured and calculated static pressure along
ramp centre line obtained with kω turbulence model
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0 200 400 600 800 1000 1200
Experiments
CFD
IVR = 1.07
IVR = 1.29
IVR = 1.70
IVR = 2.03
Distance along ramp [mm]
C
p
[

]
114
Chapter 5. Numerical analysis of waterjet inlet flow
Comparison of velocity field at impeller plane
The assessment of the performance of the kω turbulence model is not only
based on the prediction of static pressure, but also on the prediction of the
velocity field. Figures 5.18a and b show the comparison of the velocity
distribution at the impeller plane for two IVR conditions. Agreement is good
for both conditions.
The relative difference between calculations and data derived from the
measurements is shown in figure 5.19 for both conditions. The majority of the
crosssectional area has a difference below +/ 10%. Quantitative agreement
seems to be slightly better with the kω turbulence model than with the kε
model, see figure 5.14.
Figure 5.18a Comparison of axial velocity derived from measurements (left)
and calculated axial velocity distribution with kω turbulence
model in impeller plane for IVR = 1.68
Figure 5.18b Comparison of axial velocity derived from measurements (left)
and calculated axial velocity distribution with kω turbulence
model in impeller plane for IVR= 2.19
5.4 Validation with experimental data
Numerical analysis of a waterjet propulsion system 115
5.4.6 Mesh convergence study
The mesh as used in this study has been made coarser and finer to evaluate
to mesh convergence of the applied mesh. For the coarse mesh, the block
divisions of the blocks near the symmetry plane of the original mesh are
reduced by a factor of two in the direction perpendicular to the symmetry
plane. In the direction along the cutwater the number of cells is reduced by a
factor of two as well for the blocks near the cutwater. The reduction of cells
from the original mesh to the coarse mesh is about 30.000, but it should be
noted that this reduction is achieved mainly in the region of the cutwater.
The fine mesh is derived from the original mesh by doubling all block
divisions in all three directions. The number of cells in the normal direction of
the extrusion layer has been kept constant to remain at the same y
+
values.
It should be noted, that the original medium mesh size has been developed
about 5 years ago. The number of cells of the original mesh, of about
200.000, were governed by the hardware constrains of that time. Currently,
the tools for generation of inlet meshes are used in the design procedure for
waterjet inlet geometries. In order to have calculation times, which are
acceptable during the design phase, the default number of cells has not been
increased. Nevertheless it is now possible to run the refined mesh, which has
about 1.5 million cells.
The pressure distribution along the cutwater, as shown in figure 5.7 will be
reviewed for the different meshes, because the largest gradient are present in
this region. Figure 5.20 shows the pressure coefficient for IVR = 1.87 along
the cutwater for the three different meshes.
Figure 5.19 Relative difference between experimental data and calculated
axial velocity distribution with kω turbulence model in impeller
plane for IVR = 1.68 (left) and IVR = 2.19
116
Chapter 5. Numerical analysis of waterjet inlet flow
1.2
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
10 9 8 7 6 5 4 3 2 1 0
Coarse mesh (175.596 cells)
Medium mesh (206.884 cells)
Fine mesh (1.564.448 cells)
1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1.2
100 80 60 40 20 0 20 40 60 80 100
Coarse mesh (175.596 cells)
Medium mesh (206.884 cells)
Fine mesh (1.564.448 cells)
Distance s along cutwater [mm]
C
p
[

]
s<0
s>0
s=0
Figure 5.20 Calculated pressure coefficient Cp along the cutwater for
different mesh sizes for IVR of 1.87. Top figure shows s range
of 100 to 100 and bottom figure shows region with location of
minimum pressure, located between s range of 10 and 0.
Negative s coordinate represents lower side of cutwater and
positive coordinate represents upper side.
Distance s along cutwater [mm]
C
p
[

]
s<0
s>0
s=0
5.5 Analysis of the suction streamtube
Numerical analysis of a waterjet propulsion system 117
The region with the location of the minimum pressure is shown in detail in the
second diagram. The markers in this figure represent the corner points of the
cells.
Increase of the number of cells with a factor of 8 results in a difference of the
prediction of minimum pressure ∆c
p
=0.06 for this condition. The difference is
found along a small part of the cutwater region of about 2 mm, which is 1.3%
of the inlet diameter.
The difference in pressure distribution along the cutwater between the
original mesh and the fine mesh is regarded to be representative for the
complete mesh. Since the deviations between the two meshes are limited to
a small region near the cutwater and the magnitude of the difference is small,
is it concluded that the cell sizes of the original mesh are suitable for the
presented study.
5.4.7 Closing remarks
Agreement between the experimental data and the results of the calculations
is quite good, despite the mentioned deficits of the kε turbulence model.
Performance of the kω turbulence is comparable. It is not an aim to
benchmark turbulence models in this study, and therefore the kε model is
used in the remainder of the analyses.
One aspect of the use of RANS methods has not been addressed yet. This is
the capability to calculate the flow for geometries according to full scale
dimensions with full scale boundary conditions. Calculations of full scale
waterjet inlets only require a refinement of the cells near the solid walls in
order to keep acceptable y
+
values. The thickness of the actual hull boundary
layer can be applied at the inflow boundary condition. In this way, the effect of
a thicker boundary layer, due to an increase in the length of a vessel, can be
taken into account for example.
5.5 Analysis of the suction streamtube
The preceding section covers the comparison of measured and calculated
results. Typical quantities which can be measured in experimental facilities
are total and static pressure. Velocities can be derived from this data
afterwards. The CFD results provide a wide range of additional post
processing capabilities to get more insight into the behaviour of the flow.
Examples include the calculation of shear stresses along the inlet surface
and the determination of the suction streamtube. The CFD results used in this
section are the results obtained for the waterjet inlet without tunnel walls, this
in order to avoid possible effects of the confinement on the flow of the
presence of the cavitation tunnel walls.
118
Chapter 5. Numerical analysis of waterjet inlet flow
5.5.1 Visualisation of suction streamtube
The CFD method provides a way to introduce a concentration scalar as a
passive traces in the flow field. This concentration scalar can be implemented
at inlet or pressure boundaries and it can be used as a weight function for
further analysis.
The shape of the streamtube is not known in advance. In fact, only at the
impeller plane the streamtube is known to coincide with the impeller plane
boundary. Setting the concentration factor to 100% in the impeller plane
allows for the determination of the complete streamtube with a upstream
tracing method. In this approach the flow field is reversed and frozen, which is
allowed in a steady flow problem. With the frozen velocity field, the solution of
the scalar only takes a few iterations for the complete numerical domain.
The shape of the streamtube is derived from an isosurface plot of the
concentration. In this way a clear representation of the streamtube interface
can be obtained. An example of the threedimensional streamtube
visualisation is shown in figure 5.21 from different view angles.
5.5.2 Determination of suction streamtube shape
The shape of the suction streamtube at the inlet boundary of the domain can
be used to calculate the average ingested velocity v
in
. This velocity is used in
the calculation of the wake fraction w and the thrust of the installation.
Determination of the crosssectional area of the streamtube at several
locations upstream of the impeller plane provides information about the
diffusor effect in the streamtube. The values of the crosssectional areas can
be used to calculate the equivalent diffusor angles. This will give an indication
of the risk of the onset of boundary layer separation in the inlet duct.
Figure 5.21 Visualisation of threedimensional suction streamtube for
IVR=1.29
5.5 Analysis of the suction streamtube
Numerical analysis of a waterjet propulsion system 119
Crosssectional shape of streamtube at inlet boundary
The shape of the suction streamtube at the inlet boundary of the numerical
domain can be approximated with a semielliptical shape. An example is
given in figure 5.22 for a range of IVR values. The boundary of the
streamtube can be approximated by the ellipse:
(5.7)
where 2w
0
and h
0
are the maximum width and maximum height resp. These
two parameters are determined by employing a leastsquare fit of the data
determined from the computed streamtube surface crosssection at the inlet
plane of the computational domain.
The elliptical curve fit can be used to determine the mass averaged inflow
velocity of the ingested fluid out of the boundary layer. In the calculations a
powerlaw exponent n=7 and a boundary layer thickness of 0.3D, with D the
inlet diameter, is used.
An efficient method to determine the inflow velocity is based on the
integration of the streamtube velocity. Once the concentration scalar is
y
w
0

\ .
 
2
z
h
0

\ .
 
2
+ 1 =
Figure 5.22 Computed semielliptical shape of suction streamtube at inlet
of numerical domain for various IVR conditions
0.6
0.5
0.4
0.3
0.2
0.1
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
IVR 1.07
IVR 1.29
IVR 1.70
IVR 2.03
IVR 1.07  elliptic fit
IVR 1.29  elliptic fit
IVR 1.70  elliptic fit
IVR 2.03  elliptic fit
y/D []
z
/
D
[

]
120
Chapter 5. Numerical analysis of waterjet inlet flow
available, the mass averaged inflow velocity can be determined by simple
integration:
(5.8)
In this equation the concentration scalar is denoted by c and x is the direction
normal to the inflow area A
inflow
.
Figure 5.23 shows the calculated wake fraction w (as defined in equation
(2.2)) based on the elliptical streamtube shape and on direct integration of the
CFD results. The rectangular box approach is also plotted for three different
widths of the box.
The method of direct integration and the method of the elliptical curve fit give
more or less comparable results for the wake prediction. The method, based
on a rectangular streamtube with a width of 1.3D, gives an underestimation of
the wake fraction of about 20%. The other two curves for the rectangular box,
based on 1.6D and 1.9D, show that a width depending on IVR is required to
obtain a good fit. It is acknowledged that the value of 1.3D is an empirical
factor, which has been derived in the past to obtain good correlations with the
actual sailing fleet.
v
i n
ρ c v
x
2
⋅ ( ) A d
A
i nfl ow
∫
ρ c v
x
⋅ ( ) A d
A
i nfl ow
∫
 =
Figure 5.23 Calculated wake fraction w based on elliptical fit, rectangular
box method and direct integration of CFD results.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Ellips curve fit
Direct integration
Rectangular box  1.3 D
Rectangular box  1.6 D
Rectangular box  1.9 D
IVR []
W
a
k
e
f
r
a
c
t
i
o
n
w
[

]
5.5 Analysis of the suction streamtube
Numerical analysis of a waterjet propulsion system 121
Crosssectional area of streamtube
In the preceding section the crosssectional shape of the streamtube at the
inflow boundary has been discussed. Here the development of the cross
sectional area of the streamtube will be analysed in more detail. The cross
sectional area has been determined at 9 different stations. The cross
sections are taken perpendicular to the ramp surface from the inlet boundary
to the cutwater, as shown in figure 5.24. The area is determined by
integration of the concentration factor over the whole plane:
(5.9)
where n is the integration plane index number, in the range 1 to 9.
Figure 5.25 shows the development of the crosssectional area of the
streamtube. The areas are normalised with the streamtube area at plane 9.
This area is denoted as intake throat area [2]. The distances are calculated
from the first plane to the throat area.
A
tube
c A d
A
n
∫
=
Figure 5.24 Location of streamtube crosssections. The parameter s is the
length along the waterjet inlet contour in the plane of
symmetry.
122
Chapter 5. Numerical analysis of waterjet inlet flow
Far upstream, near the inflow plane, a constant crosssectional area is found
for each IVR condition. The influence of the ramp curvature, at crosssection
numbers 4 and 5, can be recognised as a small region of reduced area,
which is due to an acceleration of the flow. The typical retardation of the flow
at higher IVR (i.e. lower v
pump
) conditions is reflected in a steep increase of
the streamtube area when approaching the inlet lip.
The development of the streamtube in streamwise direction can also be
expressed in an equivalent diffuser angle α
diff
. This diffusor angle is based on
the equivalent streamtube diameter and the distance dl between the different
crosssections:
(5.10)
The equivalent diameter is based on a circular section with area identical to
the area of the streamtube crosssection. The variation of the equivalent
diameter along the streamtube in streamwise direction has been fitted with a
fifth order polynomial. With this curve fit the diffuser angle α
diff
can be derived
in small steps of the length dl.
The development of the diffusor angle along the streamtube up to the
cutwater is shown in figure 5.26 for various IVR values.
Figure 5.25 Streamtube crosssectional area normalised by A
9
. The
numbers refer to the integration plane index numbers, as
shown in figure 5.24.
0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5 6 7 8
IVR = 1.07
IVR = 1.21
IVR = 1.29
IVR = 1.50
IVR = 1.70
IVR = 1.87
IVR = 2.03
IVR = 2.19
Dimensionless distance s/D []
N
o
r
m
a
l
i
s
e
d
a
r
e
a
A
/
A
9
[

]
2 3 4 5 6 7 8 9
α
diff
D
n 1 +
D
n
–
dl

\ .
 
atan =
5.6 Evaluation of wall shear stress
Numerical analysis of a waterjet propulsion system 123
At the ramp curvature (around s/D = 3.5), all conditions show a small
negative value of α
diff
, which represents the contraction of the streamtube. At
IVR conditions of 1.87 and higher, the diffuser angle exceeds 8 degrees. At
such high diffuser angles, flow separation is likely to occur in normal circular
diffusers [15]. The occurrence of flow separation will be discussed in more
detail in the following subsection.
5.6 Evaluation of wall shear stress
Evaluation of the wall shear stress in the inlet duct can provide information
about the probability of flow separation in the inlet. Flow separation will lead
to increased nonuniformity of the flow in the impeller plane and higher
hydraulic losses. This is a deterioration of the performance of the whole
propulsion system and should be avoided in the range of normal operating
conditions. Boundary layer separation will occur inside the inlet duct at the
ramp side, because of the strong adverse pressure gradient acting on the
flow. Occurrence of this large pressure gradient at high IVR (i.e. low v
pump
)
has already been demonstrated in figure 5.5 in subsection 5.4.1.
Determination of a possible region of separated flow is based on the
magnitude of the wall shear stress along the duct. It is assumed that the
onset of separation occurs at the location where the wall shear stress
Figure 5.26 Equivalent diffuser angle α
diff
of streamtube
4
2
0
2
4
6
8
10
12
0 1 2 3 4 5 6 7 8
IVR = 1.07
IVR = 1.21
IVR = 1.29
IVR = 1.50
IVR = 1.70
IVR = 1.87
IVR = 2.03
IVR = 2.19
Dimensionless distance s/D []
α
d
i
f
f
[
d
e
g
r
e
e
s
]
124
Chapter 5. Numerical analysis of waterjet inlet flow
reduces to a zero value [16]. The dimensionless wall shear stress is denoted
as the dimensionless friction velocity v
f
:
(5.11)
The nondimensional representation can be used to get a more
comprehensive comparison of calculations for varying pump velocity.
For the standard flushtype inlet geometries, possible flow separation is
located inside the duct near station 9 (as shown in figure 5.24). The wall
shear stress component in axial direction will change sign in case of
boundary layer separation in this region. For the detection of flow separation,
the friction velocity is multiplied with the sign of the axial wall shear stress.
Figure 5.27 shows the result of the evaluation of the minimum wall shear
stress at the waterjet inlet duct part. Results of calculations with both wall as
well as pressure boundaries are used in this evaluation. The difference
between the results of both series of calculations is negligible.
According to the wall shear stress criterion, flow separation will occur for IVR
values higher than 1.75. This is in accordance with the maximum allowable
diffuser angle criterion of 8 degrees (as shown in figure 5.26).
For a practical inlet design, flow separation should not occur at normal
operating conditions. With a dedicated inlet geometry design it is possible to
avoid flow separation in the inlet for all operating conditions.
v
f
τ
w
ρv
pump
2
  =
0.03
0.02
0.01
0
0.01
0.02
0.03
0.9 1.1 1.3 1.5 1.7 1.9 2.1 2.3
Wall boundaries
Pressure boundaries
Figure 5.27 Dimensionless wall shear stress as function of IVR
IVR []
F
r
i
c
t
i
o
n
v
e
l
o
c
i
t
y
v
f
[

]
5.7 Nomenclature
Numerical analysis of a waterjet propulsion system 125
5.7 Nomenclature
A area m
2
c concentration 
C
p
pressure coefficient 
g gravitational acceleration m/s
2
h height from tunnel centre line to cutwater m
h
0
maximum suction depth m
IVR inlet velocity ratio (v
ship
/v
pump
) 
p static pressure N/m
2
p
0
static ambient pressure in tunnel N/m
2
p
v
vapour pressure N/m
2
v velocity m/s
v
f
friction velocity 
w
0
maximum suction half width m
y,z coordinates m
Greek symbols
ρ
fluid density kg/m
3
σ
cavitation inception pressure 
τ
w
wall shear stress N/m
2
Subscripts
n normal direction
pump based on values just upstream of the impeller
tunnel based on tunnel values
x axial direction
5.8 References
[1] Førde, M., Ørbekk, E., Kubberud, N., ‘Computational Fluid Dynamics
applied to high speed craft with special attention to water intake for
water jets’, FAST’91 conference, pp. 6989, Trondheim, 1991
[2] Terwisga, T.J.C. van,’Waterjethull interaction’, PhD thesis, Technical
University of Delft, 1996
[3] Pylkkänen, J.V.,’Test cases of application of CFD code to predict
waterjet inlet flows’, Technical report VTT VAL B11, VTT Manufactur
ing Technology, Maritime Research, Espoo, 1994
126
Chapter 5. Numerical analysis of waterjet inlet flow
[4] Pylkkänen, J.V.,’CFD analysis and iterative design of waterjet inlet’,
International conference on ship and marine research, Rome, 1994
[5] Vorst, H.J.A. van der, Verbeek, R., Hendriks, A.J.A.M., ‘Design consid
erations of waterjet propulsion systems’, 13th Fast Ferry International
Conference, Singapore, 1997
[6] Seil, G.J., Fletcher, C.A.J., Doctors, L.J., ‘The application of computa
tional fluid dynamics to practical waterjet propulsion system design and
analysis’, Proceedings of the 3rd international conference on Fast Sea
Transportation, pp 13791390, LubeckTravemunde, 1995
[7] Yang, C.K., Lee, Y.B., Choi, H.S., ‘A numerical analysis of the flow
around the waterjet inlet’, Proceedings of the 3rd international confer
ence on Fast Sea Transportation, pp 13911401, LubeckTravemunde,
1995
[8] Bulten, N.W.H.,‘Influence of boundary layer ingestion on waterjet per
formance parameters at high ship speeds’, Proceedings of the 5th
international conference on Fast Sea Transportation, pp 883892,
Seattle, 1999
[9] Hu, P., Zangeneh, M., ‘Investigations of 3D turbulent flow inside and
around a waterjet intake duct under different operating conditions’,
ASME Journal of Fluids Engineering, Vol 121, pp. 396404, 1999
[10] Hu, P., Zangeneh, M., ‘A method for automatic optimisation of the
intake duct geometry of marine waterjets’, Proceedings of the 5th
international conference on Fast Sea Transportation, pp 843851,
Seattle, 1999
[11] Brandner P., Walker, G.J.,’A waterjet test loop for the Tom Fink cavita
tion tunnel’, Proceedings Waterjet Propulsion III conference, Gothen
burg, 2001
[12] Versteeg, H.K., Malalasekera, W., ‘An introduction to Computational
Fluid Dynamics’, Longman Scientific & Technical, Essex, 1995
[13] Computational Dynamics Limited, ‘StarCD methodology, version
3.150’, 2001
[14] Patankar, S.V., Spalding, D.B., ‘A calculation procedure for heat, mass
and momentum transfer in threedimensional parabolic flows’, Int. J.
Heat Mass Transfer, Vol 15, pp. 17871806
[15] Schlichting, H., ‘Boundary layer theory’, McGrawHill, New York, 1968.
[16] Fox, R.W., McDonald, A.T., ’Introduction to fluid mechanics’, Third edi
tion, John Wiley & Sons, New York, 1985
Numerical analysis of a waterjet propulsion system 127
Chapter 6 Numerical analysis of waterjet
pump flow
This chapter deals with the numerical analysis of the mixedflow waterjet
pump. The geometry and mesh generation procedure will be discussed in
section 6.1. The numerical domain encloses the impeller, the stator bowl and
a straight suction pipe. Subsequently, in section 6.2, the choice of boundary
conditions and the options for implementation of impeller rotation are
reviewed. In section 6.3, results of the CFD calculations with uniform inflow
are compared with available experimental data to get an indication of the
accuracy of the numerical method used. The experimental data is measured
during the development of the waterjet pump and is only available at the
authors’ company. This data is used for the prediction of the performance of
the full scale installations.
The influence of a nonuniform inflow on the performance is presented in
section 6.4. Its effect on the radial loading of the impeller is treated as well.
The axial velocity distributions as shown in the previous chapters, are used
as input velocity distributions.
6.1 Geometry and mesh generation
The numerical domain for the analysis of the flow through the mixedflow
pump includes the complete impeller and the stator bowl with guide vanes.
Preliminary calculations assuming periodic flow, i.e. reducing the calculation
to computing the flow through a single impeller channel showed poor
agreement with experimental data. This can be partly attributed to the applied
boundary condition of constant pressure at the outlet part of the numerical
128
Chapter 6. Numerical analysis of waterjet pump flow
domain. The choice to model the complete impeller and stator enables the
use of a constant pressure boundary condition at the outlet. Upstream of the
pump a straight suction pipe is incorporated in the numerical domain. A
sketch of the rotorstator configuration is shown in figure 6.1. The two dotted
lines indicate the locations of the sliding interfaces between the stationary
and the rotating domains in the mesh.
The geometry of the impeller is based on seven blade profile sections equally
spaced in radial direction and the blade root fillet geometry. The blade profiles
provide detailed information of the shape of the blade sections near the
leading and trailing edges. Figure 6.2 shows the blade geometry for two
blades, based on the streamline profile sections.
Figure 6.1 Sketch of rotorstator pump configuration
inlet
nozzle
sliding
interfaces
Figure 6.2 Blade geometry based on streamline profile sections
6.1 Geometry and mesh generation
Numerical analysis of a waterjet propulsion system 129
Around the crosssections of the blades, an Ogrid is used to ensure good
orthogonality of the boundary layer cells along the impeller surface. These
cells are also used in the region near the hub surface. The remainder of the
volume between the impeller blades is filled with additional hexagonal cells.
The stator bowl is meshed with another group of hexagonal cells, which
follow the guide vanes curvature. Thickness of the guide vanes is taken into
account in the model. The shaft is included in the model at the suction side of
the impeller, like in a real waterjet installation.
Figure 6.3 shows the topology of the mesh between two impeller blades in a
plane at halfspan and the block topology from hub to tip.
At the seatring, very fine cells are created from an extrusion layer. In this
region high gradients occur due to the seatring being stationary with respect
to the rotating unschrouded impeller. The layer of extruded cells fills the tip
region between the impeller blades and the seatring. In this way, water can
flow over the tip from pressure to the suction side of the blade. Treatment of
the tip clearance is in agreement with the findings of several studies of turbo
machinery addressing tip clearance flow phenomena [14].
Figure 6.3 Mesh topology for the impeller of the mixedflow pump in a
plane at halfspan (top) and in the direction from hub to tip
130
Chapter 6. Numerical analysis of waterjet pump flow
Figure 6.4 shows the surface plot of the impeller blades, the stator blades, the
hub and the shaft. The final mesh of the complete mixedflow pump is
presented in figure 6.5. The total number of cells is about 950.000 for the
complete mesh.
Figure 6.4 Surface plot of impeller blades, stator blades, hub and shaft
inlet nozzle
Figure 6.5 Complete mesh of the mixedflow pump with about 950.000
cells (top) and surface mesh of the impeller
inlet
rotor
stator
nozzle
6.2 Numerical approach
Numerical analysis of a waterjet propulsion system 131
This mesh shows the straight suction pipe, the impeller and the stator with the
nozzle. The surface mesh of the impeller is shown as well. The block
topology of the mesh along the fillet can be recognized, to a certain extent in
the detailed view of the impeller surface mesh.
6.2 Numerical approach
The description of the numerical approach is divided into three parts, first the
applied boundary conditions are discussed, followed by the description of the
fluid properties. Finally, the implementation of the impeller rotation is
discussed.
6.2.1 Boundary conditions
The numerical domain of the mixedflow pump is bounded by a number of
surfaces at which different types of boundary conditions are imposed. In the
inflow plane, an inlet type boundary condition is applied. This requires the
prescription of the three velocity components and the properties of the
turbulence model. The velocity distribution in the plane can be either constant
or nonuniform. The two measured nonuniform velocity distributions as
shown in figure 3.2 can be implemented using a Fourierseries
approximation.
It was already mentioned in section 1.2 that the waterjet mixedflow pump
belongs to the group of internal flow machines. At the nozzle outlet plane two
types of boundary conditions are available: normal outlet boundary condition
or a prescribed static pressure condition. The normal outlet condition only
prescribes the mass flow rate. The actual distribution of the axial velocity may
be nonuniform over the exit plane.
The pressure condition is used with a prescribed constant value of the static
pressure over the whole outlet plane in general. The actual behaviour of the
waterjet pump can be modelled with a value of the static pressure equal to
the ambient pressure at the nozzle exit. The computational method enforces
conservation of mass and the resulting velocity distribution can be non
uniform.
A wall boundary condition can be applied to the remaining boundaries. The
default wall boundary condition assumes zero velocity of the wall in normal
and tangential direction (noslip) in the computational frame of reference.
Special attention is required in turbo machinery calculations for two types of
wall boundary conditions: (i) stationary walls in a rotating frame of reference,
as discussed in section 6.2.3 and (ii) rotating walls in the stationary frame of
reference.
132
Chapter 6. Numerical analysis of waterjet pump flow
6.2.2 Fluid properties
The fluid properties are specified similar to the ones specified for the
calculations of the flow through the waterjet inlet as discussed in section
5.3.2. The density and viscosity are taken constant for the fluid. Turbulence is
modelled with the kε turbulence model.
6.2.3 Impeller rotation
The basic idea is to split the numerical domain into three different parts:
stationary inlet part, a rotating impeller part and a stationary stator bowl part.
Rotation of the impeller can be accounted for in two different ways: (i) a
quasisteady approach with a multipleframesofreference (MFR) method, or
(ii) a fully transient method with a mesh moving with the impeller and with
sliding interfaces with the stationary part of the mesh.
The MFR method is a relatively fast method, which can be used with a
steadyflow solution method. This method is identical to the method used for
the waterjet inlet flow calculations. In a rotating frame of reference rotation of
the impeller results in centrifugal and Coriolis forces on the fluid elements.
These are implemented by additional source terms in the momentum
equations.
In case of a moving mesh with a sliding interface, a fully transient flow
solution method is required. The CFD method provides a routine, which
automatically rotates the mesh after each timestep and connects the rotating
and stationary domains. Fully transient flow calculations use the PISO
(Pressure Implicit with Splitting of Operators) algorithm to couple the
momentum and pressure terms. This algorithm was originally developed for
the noniterative computation of unsteady compressible flow [5]. It involves
one predictor step and two or more corrector steps, which may be seen as an
extension of the SIMPLE algorithm with additional corrector steps.
All calculations presented in this chapter have been made for an impeller
rotational speed of 1920 RPM, (32 Hz). Reynolds numbers based on inlet
pump diameter and rotational speed (according to equation (4.1)) are in
excess of . Note that full scale waterjet installations normally operate at
Reynolds numbers which are 3 to 5 times larger.
6.2.4 Calculation of global pump performance
Pump head
By definition, pump head is based on the difference in total pressure
upstream and downstream of the pump (see (2.32)). These values can be
derived from the CFD results by integration of the total pressure over a cross
10
7
6.2 Numerical approach
Numerical analysis of a waterjet propulsion system 133
sectional area upstream and downstream of the impeller. It follows from the
energy balance, that mass averaged quantities are to be used to determine
the total pressure:
(6.1)
where A denoting the crosssectional plane. The actual mass averaged head
H
ma
is calculated as:
(6.2)
In an experimental setup it is common practice to divide the total pressure
into a static and a dynamic pressure component. Static pressure is measured
at the inlet and outlet pipe circumference through pressure taps. Estimation of
the dynamic pressure contribution is based on the inlet and outlet pipe
diameter and the volume flow through the pump. Within this approach it is
assumed that the velocity and pressure distributions are uniform over the
measurement planes, and that there are no significant velocity components in
the tangential and radial direction (i.e. swirl).
The calculation of the head from the CFD calculations can also be based on
the static pressure distribution and the flow rate. The areaaveraged static
pressure is calculated according to:
(6.3)
where A represents the crosssectional area. It is acknowledged, that in
actual experiments, the static pressure is only measured at the outer radius
and subsequently averaged. This difference is neglected in the analysis. The
areaaveraged head becomes:
(6.4)
There are therefore two expressions for the head of the pump: the correct
value H
ma
, based on a mass average, and an estimated value H
aa
, based on
area averages. The nonuniformity of the axial inflow velocity is taken into
account with the mass averaged method correctly, but not with the area
averaged method.
p
tot
1
ρQ
 ρp
tot
v
n
A d
A
∫
=
H
ma
p
t ot o , ut
p
t ot i n ,
–
ρg
 =
p
st at
1
A
 p
st at
A d
A
∫
=
H
aa
p
st at out ,
p
stat i n ,
–
ρg

Q
2
2g

1
A
out
2

1
A
i n
2
 –
\ .

 
+ =
134
Chapter 6. Numerical analysis of waterjet pump flow
Shaft torque
In an experimental setup, shaft torque can be measured with strain gauges
on the shaft. The measured torque can be corrected for additional friction
losses in seals and bearings. In the CFD calculations torque can be
determined in two ways. The first method is based on a direct summation of
wall forces acting on the impeller and shaft surface. The second method is
based on a momentum balance in circumferential direction. The equation for
the moment of momentum is given by (see for example [6]):
(6.5)
where CV represents the control volume and CS all surfaces of the control
volume, r is the distance from the axis of rotation, v is the velocity and τ
w
the
viscous stress tensor. In general turbo machinery applications, the axial
component of this vector equation has to be evaluated only. For steady flow
conditions the equation for moment of momentum reduces to:
(6.6)
where v
t
is the circumferential velocity component, v
n
the normal velocity at
control surfaces and τ
w,t
the tangential shear force at the surfaces of the
stationary housing. This force will be a result of the shear forces acting on the
surface of the socalled seatring. Evaluation of the first term on the right hand
side of equation (6.6) can be limited to the surfaces through which the flow
enters and leaves the control volume, since the normal velocity vanishes at
wall surfaces. The actual equation for the determination of the torque within
the numerical method becomes:
(6.7)
6.3 Validation with experimental data
The numerical method used to compute the flow in the waterjet mixedflow
pump is validated with the aid of experimental data of pump performance.
Extensive series of measurements have been made on a modelscale pump
at the authors’ company. The experimental data include the head and torque
curves as a function of flow rate. The pump performance data is used in the
waterjet performance prediction software of Wärtsilä Propulsion Netherlands
T
shaf t
t ∂
∂
r v ×
CV
∫
ρdV r vρv × A d ⋅
CS
∫
r τ
w
×
CS
∫
– + =
T
shaft
ρrv
t
v
n
A d
CS
∫
rτ
w t ,
CS
∫
– =
T
shaf t
ρrv
t
v
n
A d
A
out
∫
ρrv
t
v
n
A d
A
i n
∫
rτ
w t ,
A
seat ri ng
∫
– – =
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 135
BV. The performance prediction software has been used for the prediction of
all full scale installations, which are currently in service.
Flow rate and head are expressed as nondimensional quantities with
equations (2.15) and (2.16) respectively. Torque is presented as a non
dimensional quantity as well. This is similar to the torque coefficient Kq used
for propellers. The nondimensional torque is defined as:
(6.8)
Pump efficiency is derived from the values for the flow rate, head and torque,
based on equation (2.20).
6.3.1 Quasisteady flow calculations with the MFR method
Results of CFD calculations with the method using the quasisteady multiple
frame of reference approach are presented in this section.
Head curve
First the head curve based on the CFD calculations will be compared with the
experimental data. Figure 6.6 shows the dimensionless head curves from the
measurements and the calculations. The calculated head is based on
equations (6.2) and (6.4). The presented head curves are normalised with a
constant given flow rate and pump head of the design condition. Agreement
between the calculations and the experimental data is good over a large
range of flow coefficients.
Differences between the two methods to determine the head from the CFD
results are limited to the low flowrate conditions. Near the design point the
deviations between the two approaches are negligible. The differences
between the two methods at low flow rates can be attributed to a nonuniform
velocity distribution at the outlet.
The numerical results are used to express the differences between mass
averaged and area averaged determination of the head. These calculations
are based on a pump configuration with a nozzle, which has ambient
pressure at the exit area. The measurements are made in a closedloop
system however. The measured head will give the correct value,
corresponding with the mass averaged numerical prediction, when the static
measurement locations are located sufficiently far upstream and downstream
of the pump to ensure uniform velocity distributions at those locations.
T
* T
ρΩ
2
D
5
 =
136
Chapter 6. Numerical analysis of waterjet pump flow
Gülich et al [7] show the obtained accuracy of calculated pump head for a
large number of different pumps with a RANS method. Comparisons with
measured data show deviations up to about 4% for mixedflow type pumps.
Deviations of 5% in head prediction are also presented for a complete stage
of a boiler feed pump [8]. The currently found deviations between
measurements and calculations are of the same order or smaller. It can be
concluded that the prediction of head with the CFD method is sufficiently
accurate for further analysis of the complete waterjet installation in chapter 7.
Shaft torque
A comparison of the calculated and measured shaft torque is shown in figure
6.7. Calculation of torque is based on the integration of the wall forces and on
the moment of momentum balance (equation (6.7)). Differences between the
results of the two numerical methods are negligible for all conditions.
Agreement between the numerical results and the measured data is
acceptable over the whole range of analysed conditions.
The magnitude of the differences appears to be related to the flow rate,
where an underestimation is observed at low flow rates and an over
estimation at high flow rates. The differences are limited to a few percent at
most, however, in the range considered.
Figure 6.6 Comparison of measured and calculated head curves, based
on equations (6.2) and (6.4).
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
200%
50% 60% 70% 80% 90% 100% 110% 120%
CFD_H_massaveraged
CFD  H_areaaveraged
Measurements
Q/Q
bep
[%]
H
/
H
b
e
p
[
%
]
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 137
Efficiency
The efficiency of a pump has been defined in equation (2.20) as the ratio of
the hydraulic power and the shaft power. Hydraulic power is the product of
volume flow rate and the produced pressure head. Shaft power is the product
of torque and the impeller angular speed. Based on the graphs of head and
torque, it is to be expected that the calculated pump efficiency will show some
deviations compared to the experimental data.
Figure 6.8 shows the calculated and the measured normalised efficiency of
the pump. The calculated efficiency is based on the head according to
equation (6.2), and the torque based on integration of the wall forces. The
pump efficiency is normalised with the efficiency at the design point.
The deviation between the calculated and the measured efficiency is about
1% near the design flow rate.
Figure 6.7 Comparison of measured and calculated shaft torque based on
integration of wall forces and moment of momentum balance
50%
60%
70%
80%
90%
100%
110%
120%
130%
140%
150%
50% 60% 70% 80% 90% 100% 110% 120%
CFD_wall force integration
CFD_moment of momentum balance
Measurements
Q/Q
bep
[%]
T
*
/
T
*
b
e
p
[
%
]
138
Chapter 6. Numerical analysis of waterjet pump flow
6.3.2 Transient flow calculations with moving mesh
Apart from calculations with the MFR method for quasisteady flow,
calculations have been made with the method for fully transient flow,
including the moving mesh option. With this method the actual movement of
the rotor with respect to the stator blades is taken into account. This enables
a detailed analysis of the interaction forces between the rotor and the stator.
Evaluation of convergence behaviour
In a fully transient flow calculation it takes a number of impeller revolutions
before the solution becomes periodic. The level of periodicity of the flow field
is monitored at some monitoring points in the numerical domain. Three points
are located about half a diameter upstream of the impeller in the inlet pipe
and three points are located in between the rotor and stator blades in the
stationary part of the mesh. The locations of the three points are chosen
arbitrarily.
The flow field variables at the monitoring points between the impeller and the
stator blades should give a periodic solution with a frequency equal to the
impeller blade passing frequency. Figure 6.9 shows the axial velocity at the
monitoring points for the design flow condition during the fifth revolution of the
impeller. The axial velocity has been normalised with the mean axial velocity.
Figure 6.8 Comparison of measured and calculated pump efficiency.
50%
60%
70%
80%
90%
100%
110%
120%
50% 60% 70% 80% 90% 100% 110% 120%
Measurements
CFD
Q/Q
bep
[%]
η
/
η
b
e
p
[
%
]
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 139
For the monitoring points upstream of the impeller constant values can be
observed, due to the prescribed axial inflow velocity. Downstream of the
impeller a periodic solution is found for all three monitoring points. The
average value and the amplitude depend on the location of the monitoring
points. The signal shows that during the fifth revolution of the impeller the
signal is periodic with frequency equal to the blade passing frequency (BPF).
The fluctuating pressure coefficients at the monitoring points is shown in
figure 6.10. The pressure coefficient is defined as:
(6.9)
where p
0
is the reference pressure at the nozzle exit plane, ρ the density, Ω
the shaft speed and D the diameter of the inlet.
The monitoring points downstream of the impeller show the expected periodic
behaviour of the pressure. The monitoring points upstream of the impeller
show fluctuations also containing higher frequencies. Pressure fluctuations at
the inlet side are related to the choice of the boundary conditions at both the
inlet and outlet boundary. The constant pressure boundary condition implies
the pressure being steady at the nozzle exit surface. As a consequence, any
fluctuation in the overall pressure head is experienced at the inlet. On the
Figure 6.9 Normalised axial velocity at monitoring points during fifth
impeller revolution for design condition. Points 1,2 and 3 are
located upstream of the impeller, points 4,5 and 6 are located
in between the impeller and stator.
0.8
0.9
1
1.1
1.2
1.3
1.4
4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00
Point 1
Point 2
Point 3
Point 4
Point 5
Point 6
Number of revolutions []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

]
3
2
1
6
4
5
C
p
p
st at
p
0
–
1
2
ρΩ
2
D
2
 =
140
Chapter 6. Numerical analysis of waterjet pump flow
other hand, in an actual pump, small fluctuations in mass flow will be present,
which may limit the high frequency pressure fluctuations. In the calculations
the mass flow rate is a constant prescribed value due to the prescription of a
fixed velocity at the inlet side.
Evaluation of the periodic behaviour of the solution is based on a Fourier
transformation of the fluctuations. Figures 6.11 and 6.12 show the Fourier
transforms for both the axial velocity and the pressure at the design flow rate
condition for the monitoring points, which are located downstream of the
impeller, i.e. in between the impeller and the stator.
The peaks at the impellerblade passing frequency and its higher harmonics
can be recognised easily. The Fourier transformations show a clear harmonic
solution downstream of the impeller for both the axial velocity as well as the
static pressure. Results of the Fourier analyses for other flow rates are
presented in Appendix B.
Figure 6.10 Static pressure coefficient at monitoring points during fifth
impeller revolution for design condition. Points 1,2 and 3 are
located upstream of the impeller, points 4,5 and 6 are located
in between the impeller and stator.
0
0.05
0.1
0.15
0.2
0.25
4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00
Point 1
Point 2
Point 3
Point 4
Point 5
Point 6
Number of revolutions []
P
r
e
s
s
u
r
e
c
o
e
f
f
i
c
i
e
n
t
C
p
[

]
3
2
1
6
4
5
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 141
Figure 6.11 Fouriertransform of fluctuating normalised axial velocity at
monitoring points at location in between the impeller and
stator. Frequency is normalised by the shaft frequency
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

]
Figure 6.12 Fouriertransform of fluctuating static pressure coefficient at
monitoring points at location in between the impeller and the
stator. Frequency is normalised by the shaft frequency.
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Pressure point 4
Pressure point 5
Pressure point 6
x 10
3
Normalised frequency []
P
r
e
s
s
u
r
e
c
o
e
f
f
i
c
i
e
n
t
C
p
[

]
142
Chapter 6. Numerical analysis of waterjet pump flow
Performance diagrams
Transient flow calculations have been made for a number of flow rates. The
pump performance parameters, like head and torque are averaged over a
number of timesteps.
Figure 6.13 shows the mass averaged head curve based on equation (6.2)
for both the quasisteady and the transient flow calculations. The
experimental data is plotted as reference.The results for the head curves
based on the area averaged quantities, according to equation (6.4), is shown
in figure 6.15.
Both methods show the tendency of a slight increase of the head for the
transient flow calculations. This phenomenon is also observed in calculations
based on a potential flow method by Van Esch [9]. Agreement with the
experimental data is still acceptable for both the quasisteady and the
transient flow calculations.
The comparison for the impeller torque is shown in figure 6.14. Both
numerical methods as well as the experimental data are shown in this figure.
A similar trend is seen for the torque as for the head, where transient flow
results show slightly higher values. The efficiency of the pump is shown in
figure 6.16 for the quasisteady and transient flow calculations. This efficiency
is based on the mass averaged head. It is observed that the transient flow
calculations predict a small increase of efficiency compared to the quasi
steady flow results.
Figure 6.13 Comparison of results of quasisteady and transient flow
calculations of mass averaged head based on equation (6.2)
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
200%
50% 60% 70% 80% 90% 100% 110% 120%
Measurements
CFD_quasi_steady
CFD_transient
Q/Q
bep
[%]
H
/
H
b
e
p
[
%
]
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 143
Figure 6.14 Comparison of results of quasisteady and transient flow
calculations of area averaged head based on equation (6.4)
0%
20%
40%
60%
80%
100%
120%
140%
160%
180%
200%
50% 60% 70% 80% 90% 100% 110% 120%
Measurements
CFD_quasi_steady
CFD_transient
Q/Q
bep
[%]
H
/
H
b
e
p
[
%
]
Figure 6.15 Comparison of results of quasisteady and transient flow
calculations of torque.
50%
60%
70%
80%
90%
100%
110%
120%
130%
140%
150%
50% 60% 70% 80% 90% 100% 110% 120%
Measurements
CFD_quasi_steady
CFD_transient
Q/Q
bep
[%]
T
*
/
T
*
b
e
p
[
%
]
144
Chapter 6. Numerical analysis of waterjet pump flow
6.3.3 Rotorstator interaction forces
The fully transient flow calculations with the moving mesh can be used to
evaluate the radial force acting on the impeller due to rotorstator interaction.
Given the number of rotor and stator blades, an estimation of the behaviour of
the interaction force can be made. The currently analysed configuration has
six rotor and seven stator blades. According to Brennen [10] and Dubas [11]
this should give a radial force counter rotating at the blade passing frequency.
The counter rotating force is derived from a model which is based on the
assumption of a variation in the pressure field when a rotating impeller blade
passes a stationary guide vane in the stator bowl. These fluctuations in
pressure distribution result in a fluctuating force on the impeller. Disturbances
in the symmetry of the blade pressure distribution will result in a radial force.
For this pump the angles between all the possible combinations of rotor and
stator blade pairs are determined and inserted in a 6x7 matrix. Such a matrix
can be used to determine the sequence in which stator blades encounter
passing rotor blades. The resulting matrix is shown in table 6.1.
Realizing that angular distances decrease during rotation of the impeller, the
sequence of stator blades with an impeller blade passing will be: 7  6  5  4 
3  2  1. The resulting interaction force has thus a counterrotating direction
compared to the impeller rotation and can possibly lead to backward whirling.
Figure 6.16 Comparison of results of quasisteady and transient
calculations flow calculations of efficiency
50%
60%
70%
80%
90%
100%
110%
120%
50% 60% 70% 80% 90% 100% 110% 120%
Measurements
CFD_quasi_steady
CFD_transient
Q/Q
bep
[%]
η
/
η
b
e
p
[
%
]
6.3 Validation with experimental data
Numerical analysis of a waterjet propulsion system 145
Figure 6.17 shows the force map for different flowrates. The forces are made
nondimensional to obtain the force coefficients Cf
x
and Cf
y
defined as:
(6.10)
The two force components are combined to get the radial force component
Cf
r1
:
(6.11)
During the calculations, output is generated with a frequency of 960 Hz,
which is equivalent to five times the blade passing frequency. The amount of
scatter of the force signal at each of these five points in subsequent cycles
gives an indication of the higher order fluctuations in the solution. The time
averaged magnitude of the radial force Cf
r1
for each flow rate is plotted as a
circle in the force map.
The forces show a periodic behaviour for all flow rates. The centre of the
concentric circles, which represent the timeaveraged magnitude of the force,
is located at the impeller axis. This means that the timeaverage of both the
horizontal component Cf
x
as well as the vertical component Cf
y
of the
interaction force diminish. This is in accordance with the expectations, given
a situation with uniform axisymmetrical inflow and outflow.
The average of the blade interaction force has a minimum value at the design
condition, denoted as 100% flow rate. This quantity increases with decreas
ing flow rate. However, at 60% of the design flow a significant reduction of the
magnitude of the radial force is observed. This effect can be observed more
clearly in figure 6.18, where the time averaged magnitude Cf
r1
of the rotor
Table 6.1 Angular distance between rotor and stator blades for
pump with 6 rotor blades and 7 stator blades
Stator
1
Stator
2
Stator
3
Stator
4
Stator
5
Stator
6
Stator
7
Rotor1 0.0 51.4 102.9 154.3 205.7 257.1
308.6
Rotor2 300.0 351.4 42.9 94.3 145.7 197.1
248.6
Rotor3 240.0 291.4 342.9 34.3 85.7 137.1
188.6
Rotor4 180.0 231.4 282.9 334.3 25.7 77.1
128.6
Rotor5 120.0 171.4 222.9 274.3 325.7 17.1
68.6
Rotor6 60.0 111.4 162.9 214.3 265.7 317.1 8.6
Cf
x
F
x
ρΩ
2
D
4
 = Cf
y
F
y
ρΩ
2
D
4
 =
Cf
r1
Cf
x
2
Cf
y
2
+ =
146
Chapter 6. Numerical analysis of waterjet pump flow
stator interaction force and the corresponding standard deviation are plotted
as function of the flow rate. The standard deviation shows a minimum near
the design point, which is in agreement with the expectations.
Figure 6.17 Plot of rotorstator interaction force for different flow rates:
instantaneous values (markers) and timeaverage of
magnitude of force coefficient (dotted curves)
0.6
0.4
0.2
0
0.2
0.4
0.6
0.6 0.4 0.2 0 0.2 0.4 0.6
Radial force  60% flow rate
Radial force  70% flow rate
Radial force  80% flow rate
Radial force  90% flow rate
Radial force  100% flow rate
Radial force  110% flow rate
Time average  60% flow rate
Time average  70% flow rate
Time average  80% flow rate
Time average  90% flow rate
Time average  100% flow rate
Time average  110% flow rate
C
f
y
[

]
Cf
x
[]
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
50% 60% 70% 80% 90% 100% 110% 120%
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
Timeaveraged force
Standard deviation
Figure 6.18 Time averaged rotorstator interaction force coefficient Cf
r1
and standard deviation as function of the flow rate
T
i
m
e
a
v
e
r
a
g
e
d
r
a
d
i
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
C
f
r
1
[
Q/Q
bep
[%]
S
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n
[

]
6.4 Influence of nonuniform axial inflow
Numerical analysis of a waterjet propulsion system 147
Model scale measurements on the same waterjet pump have confirmed the
presence of multiple local minima and maxima of the radial force [12]. These
results have not been published yet.
6.4 Influence of nonuniform axial inflow
The results discussed in the previous section are obtained for uniform inflow
velocity distributions. Waterjets operate in a strongly nonuniform inflow
distribution as described in chapter 3. The effect of a nonuniform velocity
distribution on the pump performance and the impeller forces will be
investigated in more detail in this section. Four distributions are used in the
analysis, the measured distributions are shown in given in figures 5.13a to
5.13d. The nonuniform velocity distributions are imposed as boundary
condition in the numerical analysis with the aid of a Fourier series
approximation of the velocity distribution, according to equation (3.2).
All calculations performed with the nonuniform inflow distribution are carried
out with the transient moving mesh option of the CFD method. Results of the
transient flow calculations for uniform inflow show a radial force due to rotor
stator interaction. This force originates from the interaction between the
impeller trailing edge and the statorblade leading edge. With a nonuniform
inflow velocity distribution, an additional source for a radial force is
introduced, since the velocity distribution is not axisymmetric.
First the pump performance for the nonuniform inflow conditions will be
reviewed. In the remainder of this section the effects of the inflow on the
radial forces will be analysed.
6.4.1 Pump performance for nonuniform inflow
The calculations with uniform inflow velocity already showed an interesting
difference between the two methods for the determination of the pump head
based on either massaveraged or areaaveraged quantities. For the
calculations for nonuniform inflow velocity distributions, the differences
between the results of these two methods may increase even more. Tables
6.2 and 6.3 show the normalised pump performance for four different inflow
conditions. The flow rate for all calculations is equal to the design flow rate.
The values in the tables have been normalised with the result from the
corresponding calculation with a uniform inflow distribution and design flow
rate of 100%.
The results of the method based on the massaveraged total pressure are
shown in table 6.2. It can be observed that the variations in both head and
torque are limited to less than two percent for all inflow conditions. The
efficiency shows a small decrease for increasing nonuniformity.
148
Chapter 6. Numerical analysis of waterjet pump flow
On the other hand, the results as shown in table 6.3, are in disagreement with
the general expectation. A clear increase of the pump head and consequently
the efficiency is found with increasing nonuniformity. Although an efficiency
increase has been reported by Kooiker et al. [13], who claim an increase of
5% due to nonuniform inflow, this is (at least partly) a result of the way in
which the pump head is measured. From equation (6.4) it is clear that the
averaging of a nonuniform inflow velocity distribution will underestimate the
suction head, which results in a higher estimate of the produced head by the
pump. Therefore, it is concluded that the method of areaaveraged head
determination is invalid for nonuniform velocity distributions.
The actual pump performance shows only a small decay due to the non
uniform inflow velocity distribution. This deviation can be neglected for
practical nonuniform inflow distributions.
6.4.2 Background of radial forces acting on the impeller
Steady fluidinduced radial forces can be generated by a nonuniform
pressure distribution at the impeller periphery or by an imbalance in the blade
torque. The first phenomenon is known to occur in centrifugal pumps with a
spiral volute casing (see [14][16]). The tongue introduces a clear asymmetry
Table 6.2 Normalised pump performance based on mass averaged
head (equation (6.2)) for various inflow distributions,
related to IVR.
IVR
[]
H
ma
/H
ma
_uniform
[%]
T*/T*_uniform
[%]
η/η_uniform
[%]
1.68 101.0 101.7 99.6
1.87 101.0 101.6 99.7
2.03 100.1 101.5 99.0
2.19 100.0 101.4 98.8
Table 6.3 Normalised pump performance based on area averaged
head (equation (6.4)) for various inflow distributions,
related to IVR.
IVR
[]
H
aa
/H
aa
_uniform
[%]
T*/T*_uniform
[%]
η/η_uniform
[%]
1.68 103.0 101.7 101.2
1.87 103.5 101.6 101.9
2.03 104.3 101.5 102.7
2.19 105.3 101.4 103.8
6.4 Influence of nonuniform axial inflow
Numerical analysis of a waterjet propulsion system 149
in the geometry. Waterjet installations with a stator bowl do not suffer from
such type of asymmetry, however. This source for a radial force can
therefore be excluded. The nonuniform inflow velocity may create a non
symmetrical pressure distribution on the hub surface. This may result in a net
radial force.
Imbalance in blade torque may give a more significant contribution to the
radial forces acting on the impeller. The concept of this imbalance is shown in
figure 6.19.
Nonuniformity of the inflow velocity distribution will result during a revolution
in variations of both the flow angle at the leading edge and the flow rate
through an impeller channel. This will result in a torque contribution that is
different for each blade. The difference in torque contributions between two
opposite blades will result in a net radial force on the impeller axis.
6.4.3 Flow rate fluctuations in the impeller channel
The CFD results are used to determine the local flow rate Qbb through the
impeller channel between two consecutive blades. The volume flow rate is
determined at a crosssectional plane at midchord of the impeller blades.
Figure 6.20 shows the normalized local flow rate as a function of the rotor
position for the four different inflow conditions.
Even for the case of maximum level of nonuniformity, i.e. for IVR=2.19, the
maximum deviation is less than 7% of the average. The variations obtained
from the CFD results are much smaller than the predictions based on the
analytical approach used in section 3.2. It appears that the nonuniform inflow
velocity distribution is smoothed in the first part of the pump. This will
influence the local flow angles at the leading edge of the impeller blade. As a
consequence, the blade loading and the blade torque will be influenced. This
eliminates the possibility to derive a simple analytical model to estimate the
radial and tangential forces on the impeller.
Figure 6.19 Concept of radial force due to unbalanced blade torque
net radial force
150
Chapter 6. Numerical analysis of waterjet pump flow
6.4.4 Radial forces for nonuniform inflow
The nonuniform inflow velocity distribution has an effect on the radial forces
acting on the impeller. For a uniform inflow distribution a periodic rotorstator
interaction force is found, as shown in subsection 6.3.3. In this section the
forces are evaluated that are caused by the nonuniformity of the inflow.
Figure 6.22 shows a graph of the horizontal and vertical component of the
force acting on the impeller for different IVR, each with a typical nonuniform
inflow velocity distribution, at the design flow rate.
The forces acting on the impeller can be approximated by a mean component
Cf
r0
and harmonic components Cf
r1
and Cf
r2
:
(6.12)
This is shown in a sketch in figure 6.21 for the mean component Cf
r0
and the
first harmonic Cf
r1
.
90%
92%
94%
96%
98%
100%
102%
104%
106%
108%
110%
0 45 90 135 180 225 270 315 360
IVR = 1.68
IVR = 1.87
IVR = 2.03
IVR = 2.19
Figure 6.20 Normalised channel flow rate Qbb as a function of the rotor
position for various inflow distributions, which are related to the
IVR
Angle θ [degrees]
Q
b
b
/
(
Q
/
N
)
[
%
]
Cf
r
Cf
r0
Cf
r1
ωt ⋅ Cf
r2
2ωt ⋅ + + =
6.4 Influence of nonuniform axial inflow
Numerical analysis of a waterjet propulsion system 151
From the results as presented in figure 6.22 can be seen that the mean radial
component Cf
r0
shows a strong relation with the level of nonuniformity, which
is in accordance with expectations. The direction of the mean force seems to
be constant for all inflow conditions. This direction may be related to the flow
rate through the pump, which is equal for all presented conditions in figure
6.22.
The condition with the most severe nonuniform velocity distribution (IVR =
2.19) shows quite some scatter of the time dependent force. This seems to
be an indication of higher order harmonics (i.e. Cf
r2
and Cf
r3
) in the system at
these inflow conditions.
Figure 6.21 Sketch of radial forces acting on the impeller due to non
uniform inflow.
x
y
Cf
r0
Cf
r1
ω
1
0
1
2
3
4
5
1 0 1 2 3 4 5 6 7 8
Uniform inflow
IVR = 1.68
IVR = 1.87
IVR = 2.03
IVR = 2.19
x 10
3
x 10
3
Figure 6.22 Horizontal and vertical component of timedependent impeller
force coefficient for different inflow velocity distributions, which
are related to IVR, at design flow rate
Horizontal force coefficient Cf
x
[]
V
e
r
t
i
c
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
C
f
y
[

]
152
Chapter 6. Numerical analysis of waterjet pump flow
Figure 6.23 shows the plot of the forces for 80% of the design flow rate. The
occurrence of increasing mean force Cf
r0
with increasing nonuniformity is
confirmed. The magnitude of both the mean force Cf
r0
and the first harmonic
Cf
r1
are different compared to the results for 100% flow rate. Moreover, the
direction of the steady force has changed for this flow rate.
Analysis of mean radial force Cf
r0
The results of the calculations for different flow rates and levels of non
uniformities show that both parameters influence the mean radial force Cf
r0
.
A quantitative assessment is made in table 6.4, where the time averaged
magnitude and the direction of the nondimensional radial force are listed.
The direction angle α is calculated according to:
(6.13)
The direction of the radial force shows a small increase with increasing non
uniformity. Results at the design capacity show a variation of less than 1.5
degrees. On the other hand, the differences due to variation in flow rate are
significant. At an IVR of 2.03 a deviation of 4.6 (= 30.5  25.9) degrees is
found.
1
0
1
2
3
4
5
1 0 1 2 3 4 5 6 7 8
Uniform inflow
IVR = 1.68
IVR = 1.87
IVR = 2.03
IVR = 2.19
x 10
3
x 10
3
Figure 6.23 Horizontal and vertical component of timedependent impeller
force coefficient for different inflow velocity distributions, which
are related to IVR, at 80% of the design flow rate
Horizontal force coefficient Cf
x
[]
V
e
r
t
i
c
a
l
f
o
r
c
e
c
o
e
f
f
i
c
i
e
n
t
C
f
y
[

]
α
Cf
y
Cf
x

\ .

 
atan =
6.4 Influence of nonuniform axial inflow
Numerical analysis of a waterjet propulsion system 153
The variation in direction seems to be related to the flow rate. This may be
due to an increase of the residence time of the fluid in the impeller, in which
the impeller will rotate over a larger angle. The direction of the radial force will
rotate accordingly.
Analysis of unsteady forces
The unsteady radial force can be examined in more detail, if the mean force
components Cf
x0
and Cf
y0
are subtracted from the results. The resultant of
the vertical and horizontal component is plotted in figure 6.24 for the design
flow rate and in figure 6.24 for 80% of design flow rate.
The results of the calculations with uniform inflow indicate a rotating radial
force with a mean magnitude. The condition with the highest level of non
uniformity at design flow rate shows a chaotic behaviour of the unsteady
forces. This is a result of higher order fluctuations of the forces.
At 80% of design flow rate, the behaviour of the fluctuating forces is more
regular. Figure 6.24 shows that the shape of the locus of the force
components becomes more and more elliptical, when the level of non
uniformity is increased. This means that the higher order harmonics of the
radial force, i.e. Cf
r2
and Cf
r3
, become more important for these conditions.
This is confirmed by the Fourier transformation of the fluctuating component
of the radial force, as shown in figure 6.26. The conditions at 80% of the
design flow rate and at the design flow rate and moderate levels of non
uniformity show two clear peaks at one and two times the blade passing
Table 6.4 Timeaveraged magnitude Cf
r0
and direction α of radial
force coefficient
Flow rate IVR Cf
r0
*1000 α
100% 1.68 3.819 24.9
100% 1.87 4.569 25.3
100% 2.03 5.550 25.9
100% 2.19 6.478 26.2
80% 1.68 2.368 28.9
80% 1.87 2.826 29.7
80% 2.03 3.423 30.5
80% 2.19 3.977 31.3
154
Chapter 6. Numerical analysis of waterjet pump flow
frequency. The conditions at design flow rate and an IVR above 2 show
numerous additional peaks at other frequencies. This is represented in the
chaotic behaviour of the radial force.
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
Uniform inflow
Curvefit
IVR = 1.68
IVR = 1.87
IVR = 2.03
IVR = 2.19
x 10
3
x 10
3
Figure 6.24 Unsteady impeller force components for different inflow
velocity distributions at design flow rate.
Horizontal force Cf
x
Cf
x0
[]
V
e
r
t
i
c
a
l
f
o
r
c
e
C
f
y

C
f
y
0
[

]
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
Uniform inflow
Curvefit
IVR = 1.68
IVR = 1.87
IVR = 2.03
IVR = 2.19
x 10
3
x 10
3
Figure 6.25 Unsteady impeller force components for different inflow
velocity distributions at 80% of design flow rate.
Horizontal force Cf
x
Cf
x0
[]
V
e
r
t
i
c
a
l
f
o
r
c
e
C
f
y

C
f
y
0
[

]
6.5 Nomenclature
Numerical analysis of a waterjet propulsion system 155
6.4.5 Concluding remark
The radial forces, which are presented in this chapter, are determined for an
impeller with a centred axis. This type of radial forces are also denoted as
excitation forces. In practice, the impeller will move away from its centred
position due to the radial forces. The offcentred motion of the rotating
impeller is called whirling [10]. This whirling motion of the impeller has not
been taken into account in the currently presented calculations, however. The
influence of the whirling motion on the occurrence of socalled reaction forces
has to be investigated further.
6.5 Nomenclature
A area m
2
Cf force coefficient 
Cp pressure coefficient 
D diameter m
F force N
F
s
surface shear force N
g gravitational acceleration m/s
2
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 2 4 6 8 10 12 14
FFT100168
FFT100187
FFT100203
FFT100219
FFT080168
FFT080203
Figure 6.26 Fourier transformation of the unsteady radial impeller force
component (Cf
r
 Cf
r0
) for different inflow velocity distributions
and flow rates.
Frequency/shaft frequency []
M
a
g
n
i
t
u
d
e
[

]
156
Chapter 6. Numerical analysis of waterjet pump flow
H head m
p
stat
static pressure Pa
p
tot
total pressure Pa
Q flow rate m
3
/s
r radial distance m
T* torque coefficient 
T torque Nm
v
n
normal velocity m/s
Greek symbols
α
radial force direction deg
ρ
density kg/m
3
Ω shaft speed rad/s
Subscripts
in inlet plane
n normal direction
out outlet plane
r0 mean component of radial force coefficient
r1, r2 harmonics of rotorstator interaction force coefficient
t tangential direction
x horizontal direction
y vertical direction
6.6 References
[1] Goto, A., ‘Study of internal flows in a mixedflow pump impeller at vari
ous tip clearances using threedimensional viscous flow computations’,
Journal of Turbomachinery, Vol 114, pp. 373382, 1992
[2] Nilson, H., Davidson, L., ‘A numerical comparison of four operating
conditions in a Kaplan water turbine, focusing on tip clearance flow’,
Proceedings 20th IAHR symposium, Charlotte, 2000
[3] Aschenbrenner, T., Göhringer, M., Moser, W., ‘Numerical and experi
mental flow analysis in a Kaplan turbine’, Proceedings 20th IAHR sym
posium, Charlotte, 2000
[4] Kunz, R.F., Lakshminarayana, B., Basson, A.H., ‘Investigation of tip
clearance phenomena in an axial compressor cascade using Euler and
6.6 References
Numerical analysis of a waterjet propulsion system 157
NavierStokes procedures, Journal of Turbomachinery, Vol 115, pp.
453467, 1993
[5] Versteeg, H.K. & Malalasekera, W., ‘An introduction to Computational
Fluid Dynamics’, Longman Scientific & Technical, Essex, 1995
[6] Fox, R.W., & McDonald, A.T.,’Introduction to fluid mechanics’, Third
Edition, John Wiley & Sons, New York, 1985
[7] Gülich, J.F., Favre, J.N., Denus, K., ’An assessment of pump impeller
performance predictions by 3DNavier Stokes calculations’, ASME Flu
ids Engineering Division summer meeting, 1997
[8] Cugal, M., Baché, G., ’Performance prediction from shutoff to runout
flows for a complete stage of a boiler feed pump using computational
fluid dynamics’, ASME Fluids Engineering Division summer meeting,
1997
[9] Van Esch, B.P.M.,’Simulation of threedimensional unsteady flow in
hydraulic pumps’, PhD thesis, University of Twente, 1997
[10] Brennen, C.E., 'Hydrodynamics of pumps', Oxford University Press,
1994
[11] Dubas, M., ‘Über die Erregung infolge der Periodizität von Turboma
chinen’, IngenieurArchiv, vol. 54, pp. 413426., 1984
[12] Esch, B.P.M. van, personal communication.
[13] Kooiker, K., Van Terwisga, T., Verbeek R., Van Terwisga, P., 'Perform
ance and cavitation analysis of a waterjet system on a cavitation tun
nel', proceedings FAST 2003 conference, session A1 pp 5762, Ischia,
Italy, 2003
[14] Stepanoff, A.J., “Centrifugal and axial flow pumps  Theory, design and
application”, 2nd edition, John Wiley, 1964
[15] Badie, R., ‘Analysis of unsteady potential flows in centrifugal pumps’,
PhD thesis, University of Twente, 1993
[16] Jonker, J.B. & Van Essen, T.G., ‘A finite element pertubation method
for computing fluidinduced forces on a centrifugal impeller rotating in a
volute casing’, International journal for numerical methods in engineer
ing, Vol. 40, pp. 269294, 1997
158
Chapter 6. Numerical analysis of waterjet pump flow
Numerical analysis of a waterjet propulsion system 159
Chapter 7 Analysis of a complete waterjet
installation
In the preceding two chapters the waterjet inlet and the mixedflow pump
were analysed separately. In this chapter the complete waterjet installation is
considered. Results are presented of the numerical analysis of such a
propulsion system on full scale. An analysis is made of the overall
performance indicators, like volume flow rate, thrust and power. Comparisons
are made with performance prediction and selection software of Wärtsilä
Propulsion Netherlands (WPNLselect). This software is partially based on
semiempirical relations, which are tuned to the performance of the
propulsion systems in the currently sailing fleet.
A detailed analysis of the streamtube will reveal some new insights into the
forces acting on the installation in both vertical and axial direction.
7.1 Generation of the numerical model
The numerical domain of the complete waterjet installation is a combination
of the two separate grids of the waterjet inlet and the mixedflow pump. The
interface between the two grids is located at the crosssection downstream
of the inlet bend (see figure 5.1 for the location of these parts). At this
interface a coupling method is used to create a fully connected domain,
based on arbitrary matching of the cells. It should be noted that this interface
between the models of the inlet and the pump is different from the two sliding
interfaces, which are required for the rotation of the impeller.
Care should be taken that the dimension of the nozzle exit area is equal to
the exit area of the actual fullscale installation. The sensitivity of the flow rate
160
Chapter 7. Analysis of a complete waterjet installation
as function of the nozzle size has been illustrated in figure 2.8 on page 36.
Figure 7.1 shows a plot of the final mesh of the full scale waterjet installation.
The boundary conditions and parameters of the numerical method are
identical to the ones used for the calculations of the separate parts. At the
inlet side of the domain a prescribed velocity distribution is applied, with a
boundary layer velocity profile. At the other sides on the domain beneath the
hull and at the nozzle exit plane constant pressure boundary conditions are
applied. As a consequence, the actual flow rate through the waterjet
becomes a part of the solution, whereas it was prescribed by the boundary
conditions in the calculations for the isolated inlet and pump.
7.2 Evaluation of volume flow rate
The calculations are made with a constant rotational speed of the pump and a
varying ship speed. In this way a relatively large range of IVR conditions is
covered. The variation in volume flow rate will be much smaller, as indicated
by the results presented in figure 2.9 on page 37.
The calculated volume flow rate through the waterjet is compared with the
results of WPNLselect. To a large extent, this program is based on the theory
as discussed in chapter 2. The loss coefficients in this method are empirical
values taken from experimental data. Further finetuning is made with
additional empirical data. The accuracy of the method is continuously
increased by adding the measured performance of newly installed waterjet
installations. Since the currently sailing fleet performs according to the
Figure 7.1 Mesh of the complete waterjet installation, number of cells is
about 1.2 million.
inlet
side
nozzle
exit
pump
7.3 Evaluation of waterjet thrust
Numercal analysis of a waterjet propulsion system 161
predictions, it is assumed that the accuracy of the performance prediction
software is adequate.
It is shown in equation (2.52) that the thrust is related to the square of the
volume flow rate. Consequently, a relatively small error in the prediction of the
volume flow rate can lead to a large error in the thrust of the installation. It is
believed that the prediction of thrust in WPNLselect is quite accurate, and
consequently the prediction of the volume flow rate will be accurate as well.
Differences between the results of WPNLselect and the CFD analyses are
within 1%. According to the expectations, only a rather limited difference in
volume flow rate is found between the low and high ship speed condition.
This trend has been captured well by the numerical method.
Agreement is satisfactory and it appears that the numerical method is
suitable for further use to investigate the flow phenomena in the waterjet
installation.
7.3 Evaluation of waterjet thrust
The second quantity that can be compared with the results of WPNLselect is
the thrust of the complete installation. The thrust can be determined from the
CFD results by a summation of the forces acting on the solid walls or by the
simplified momentum balance method given by equation (2.52) on page 33.
7.3.1 Integration of solid wall forces
The procedure for the selection of the cells at the solid walls requires some
attention. It is possible to either select all solid wall cells of the numerical
domain, or to select only the solid wall cells, that are part of the streamtube.
The two options have been shown in figure 2.7 on page 31. With the selection
of all wall cells of the numerical domain, a large part of the hull structure is
also taken into account. The net thrust will be reduced due to the contribution
of the drag of the hull.
Table 7.1 Comparison of volume flow rate through waterjet based
on prediction software and CFD calculations
v
ship
[kn]
Q
select
[m
3
/s]
Q
CFD
[m
3
/s]
Deviation
[%]
31 13.0 12.93 < 1%
35 13.1 13.08 < 1%
39 13.3 13.19 < 1%
43 13.4 13.31 < 1%
47 13.6 13.49 < 1%
162
Chapter 7. Analysis of a complete waterjet installation
Selection of the solid wall cells of the streamtube only, can not be done with a
welldefined procedure, however. The wall cells near the streamtube surface
will be partly included in the streamtube. The results show a significant
influence of the choice whether these cells are included or excluded from the
selection. The process of wall cell selection is avoided, when all solid wall
cells of the numerical domain are taken into account in the force evaluation.
The additional drag of the hull structure can be compensated for, given the
dimensions of the numerical domain.
The method using drag compensation is also preferable from a ship builder’s
point of view. This is due to a subtle difference between a vessel with waterjet
propulsion and a vessel with conventional propeller propulsion. For such a
propeller ship, the hull and the propeller can be split into two subsystems. A
waterjet propelled vessel can not be split into a bare hull and a propulsion
unit, without leaving a hole in the hull structure. In a bare hull resistance test,
additional hull drag, due to the development of the boundary layer, will be
measured at the location of the inlet duct opening. The additional drag of the
boundary layer will change once the waterjet installation with the inlet duct is
installed.
It is concluded that the thrust of the installation has to be computed using all
solid wall cells and that frictional resistance of the original hull area has to be
compensated for. The total waterjet thrust is then:
(7.1)
where T
wj,all
is the integrated force of the pressure and wall friction on all solid
wall cells and D
hull
the drag force of the equivalent rectangular hull area.
Calculation of the drag of the hull boundary layer is based on flat plate
boundary layer theory for high Reynolds numbers. This formula is used,
because it is also used to extrapolate model scale bare hull resistance data to
full scale predictions. The friction coefficient is given by [1]:
(7.2)
where the friction coefficient C
f
is defined as:
(7.3)
with D
hull
the actual friction drag of the equivalent hull, v
ship
the undisturbed
velocity, and A
hull
the area of the equivalent hull, which is length times width
of the used mesh.
T
wj
T
wj al l ,
D
hul l
– =
C
f
0.455
Re
l
( )
2.58
log
 =
C
f
D
hul l
1
2
 ρv
shi p
2
A
hul l
 =
7.3 Evaluation of waterjet thrust
Numercal analysis of a waterjet propulsion system 163
The Reynolds number for the flat plate is defined as:
(7.4)
where L
wetted
is the wetted length of the hull upstream of the inlet. The actual
drag of the hull area of the numerical domain depends on the dimensions of
the area A and the wetted length of the hull upstream of the waterjet inlet.
7.3.2 Momentum balance
The thrust based on the momentum balance, can be calculated with the
simplified equation (2.52). In this equation the contribution of the pressure
distribution on the streamtube below the hull and the aft part of the hull is
neglected. The introduced error can be compensated with the socalled thrust
deduction factor t
j
[2]. The mass averaged inflow velocity is determined using
a concentration scalar as described in detail in section 5.5.
7.3.3 Results
Figure 7.2 shows the thrust of the fullscale installation based on WPNL
select and on the CFD calculations. In order to show the differences between
results of the two methods that use the results of the CFD calculations, as
described above, both results are presented in the graph. The results have
been normalised with the design thrust at a ship speed of 39 knots.
The method of integration of the solid wall forces shows a very good
agreement with the prediction software for the ship speeds up to 39 knots,
which is the design speed for the investigated installation. At higher speeds a
small difference is found between the results of the force integration method
and the results of the prediction software.
The method based on the momentum balance gives a good agreement up to
ship speeds of 35 knots. At higher speeds the momentum balance gives a
lower thrust than the force integration and the prediction software.
The deviations between the results of the two numerical methods are quite
significant at higher ship speeds. This phenomenon may be attributed to
numerical inaccuracies.
However, it can also be an indication that the simplification of the momentum
balance is the cause for this, as discussed in subsection 2.3.3 at page 30. It
is known, that the low pressure region along the cutwater will have a positive
contribution to the thrust. This may explain the deviations found between the
results of the force integration method based on considering the pressure and
shear stress on all solid wall cells and the ones of the momentum balance
approach, which is applied to the streamtube.
Re
l
ρv
shi p
L
wetted
µ
 =
164
Chapter 7. Analysis of a complete waterjet installation
7.4 Evaluation of required power
Evaluation of the required power is equivalent to the evaluation of the
required torque, since all calculations have been made at the same rotational
speed. The torque is obtained from the summation of the torque of the solid
wall cells of the impeller. The resulting torque is multiplied by the angular
velocity of the impeller to obtain the required power.
Figure 7.3 shows the power based on the calculations and the one obtained
from the prediction software. Results are normalised with the required power
of the full scale installation at the design speed computed with the prediction
software. Over the complete range of calculated conditions, a small over
prediction of the power can be observed. The deviation of about 2.5% is
acceptable, however. The deviation is more or less constant for all conditions,
due to the very small variation in flow rate. The increase in flow rate is only
4.5% for an increase of ship speed from 31 to 47 knots.
Figure 7.2 Comparison of thrust of waterjet installation based on WPNL
select with numerical predictions based on force integration of
solid wall forces (T
wj
) and momentum balance applied to the
streamtube (equation (2.52)).
80%
85%
90%
95%
100%
105%
110%
115%
120%
30 32 34 36 38 40 42 44 46 48
Prediction software
Integration of solid wall forces
Momentum balance
Ship speed [knots]
T
h
r
u
s
t
/
T
h
r
u
s
t
3
9
k
n
o
t
s
[

]
7.5 Analysis of vertical force on waterjet structure
Numercal analysis of a waterjet propulsion system 165
7.5 Analysis of vertical force on waterjet structure
In the preceding section, the prediction of the waterjet thrust is discussed.
While calculating the thrust of the waterjet installation, based on the
integration of the forces acting on the solid walls, it is found that there can be
a net force in the vertical direction as well. This vertical force component is in
fact a lift force acting on the waterjet structure.
This result of the numerical analysis is not in agreement with the results of
Van Terwisga [2], who claims that there is no net lift contribution. On the other
hand, Svensson [3] determined a lift force from pressure measurements in
the intake and on the hull, which can be up to 5% of the displacement of the
vessel.
Applying equation (2.35) for the vertical direction yields:
(7.5)
where L
wj
is a vertical force on either the waterjet or the hull. The vertical
components of the momentum fluxes do not contribute, since it is assumed
90.0%
92.5%
95.0%
97.5%
100.0%
102.5%
105.0%
107.5%
110.0%
30 32 34 36 38 40 42 44 46 48
Prediction software
CFD result
Figure 7.3 Comparison of required power of waterjet installation obtained
from prediction software and the one obtained from CFD
calculations
Ship speed [knots]
P
o
w
e
r
/
P
o
w
e
r
3
9
k
n
o
t
s
[

]
L
wj all ,
pz A d ⋅
A
t ube
∫
– L
wj hul l ,
+ =
166
Chapter 7. Analysis of a complete waterjet installation
that both the velocity distributions at the inlet and nozzle exit are aligned with
the horizontal direction.
A more detailed analysis of the occurring lift force coefficient, reveals that the
largest lift is found at high IVR conditions, and therefore high ship speed. The
numerical results presented in this subsection are for a waterjet installation
for very high speed vessels. The most important difference from
hydrodynamic point of view between a conventional waterjet installation and
a highspeed application is found in the geometry of the inlet duct [4].
Inlet lift coefficient
The actual lift force acting on the inlet structure is presented as a non
dimensional coefficient, according to:
(7.6)
where L
wj,all
is the calculated vertical force, ρ the density, Q the volume flow
and v
pump
the averaged axial velocity upstream of the impeller. The lift
coefficient is plotted as a function of IVR in figure 7.4.
Up to an IVR of about 1.5 the lift coefficient is indeed rather small. For high
speed applications (>60 knots) the IVR will be about 2 to 2.2. This will result
in a lift coefficient of about 0.08 to 0.10.
Equation (7.6) can be expressed in specific pump parameters Ω, D and ϕ as
well with equation (2.15) as:
(7.7)
The second factor on the right hand side is of order 1 for common waterjet
mixedflow pumps. Thus, a direct comparison between the lift force
coefficient on the inlet and the vertical force coefficient on the impeller, as a
result of nonuniformity can be made. At an IVR of 2.19, the vertical force is
about 2.85x10
3
, according to figure 6.22 on page 151. The vertical force
coefficient in the inlet for that condition is 0.081, based on figure 7.4. It is
concluded that the vertical force in the inlet is significantly larger (almost
factor 30) than the vertical component of the impeller force due to non
uniform inflow velocity distributions.
C
L
L
wj al l ,
ρQv
pump
 =
C
L
L
wj al l ,
ρΩ
2
D
4

π
4ϕ
2
 ⋅ =
7.5 Analysis of vertical force on waterjet structure
Numercal analysis of a waterjet propulsion system 167
Inlet liftthrust ratio
The lift force can also be compared to the thrust of the waterjet installation.
The thrust of a waterjet can be rewritten from equation (2.52), with aid of
equations (2.2), (2.11), (2.12) and (2.13) to:
(7.8)
where D
inlet
and D
nozzle
are two geometric parameter of the installation and w
is the wake fraction, defined in equation (2.2).
The liftthrust ratio becomes:
(7.9)
This shows that the liftthrust ratio is related to the lift coefficient, the wake
fraction and the geometry of the waterjet installation. The data for the lift
coefficient as shown in figure 7.4 is presented as liftthrust ratio in figure 7.5.
Figure 7.4 Lift coefficient C
L
as a function of IVR
0.05
0
0.05
0.1
0.15
0.2
0.25
0 0.5 1 1.5 2 2.5
Lift force coefficient
IVR []
C
L
[

]
T ρQv
pump
D
inlet
D
nozzle

\ .
 
2
1 w – ( ) IVR ⋅ –
¹ )
´ `
¦ ¹
=
L
T

C
L
D
inlet
D
nozzle

\ .
 
2
1 w – ( ) IVR ⋅ –
¹ )
´ `
¦ ¹
 =
168
Chapter 7. Analysis of a complete waterjet installation
Since the thrust decreases with increasing IVR, the liftthrust ratio shows a
steep increase at high values of IVR: a significant lift force of about 1020% of
the thrust is found. This amount of lift can not be neglected in general.
The lift force in the inlet creates a moment on the hull structure at high
speeds, which can lead to a reduced or even negative trim angle of the
vessel.
7.6 Pressure distribution on streamtube surface
7.6.1 Evaluation of momentum balance in vertical direction
The CFD analyses reveal the presence of a significant lift force in a waterjet
installation at the higher IVR conditions. Based on equation (7.5), the lift can
be the result of a net force acting on the hull surface or of a resultant force of
the pressure distribution, which acts on the streamtube surface.
These two terms are neglected in the standard momentum theory, which is in
accordance with the simplified thrust equation (2.52). As a result, the net lift
force is equal to zero for all conditions.
The contribution of the vertical forces acting on the hull surface are limited to
a region near the cutwater. It is known from CFD analyses that the pressure
in this region is very low for high IVR conditions (see for example figure 5.6).
It is expected that any possible contribution to the vertical force will be a
negative lift for high IVR. It is concluded that the lift force on the complete
5%
0%
5%
10%
15%
20%
25%
0 0.5 1 1.5 2 2.5
Liftthrust ratio
Figure 7.5 Liftthrust ratio L/T as a function of IVR
IVR []
L
/
T
[

]
7.6 Pressure distribution on streamtube surface
Numercal analysis of a waterjet propulsion system 169
waterjet structure, should originate from a pressure distribution along the
streamtube surface. A method to integrate the pressure distribution on the
streamtube surface is required to confirm this.
Accurate determination of the lift force, acting on the hull structure, is not
possible with the applied method. It is not possible to determine the
contribution of the solid wall cells near the streamtube interface accurately,
because the effect of partially included solid wall cells can not be taken into
account.
The shape of the streamtube depends on the IVR at which the waterjet
operates. The pressure distribution also depends on IVR. In order to
determine the net force, an integration of the pressure over the surface has to
be made. For this numerical integration, the surface of the streamtube is
subdivided into triangles, based on the Delaunay triangularisation method.
The net lift force can be determined, given the average pressure on each
triangle and the normal direction.
7.6.2 Calculation of vertical force on streamtube
The streamtube shape is determined for various operating conditions of the
highspeed waterjet installation, as discussed in the preceding section. An
example of the triangular mesh of the streamtube surface is shown in figure
7.6. The triangular mesh represents the streamtube surface and the
rectangular cells represent the solid wall cells of the waterjet inlet.
The resulting lift force of the streamtube pressure distribution is shown in
figure 7.7. The force has been made nondimensional according to equation
(7.6).
The force of the integrated pressure distribution clearly depends on IVR. At
values of an IVR above 1.8 the net lift force on the streamtube exceeds the
total lift force on the inlet. At these IVR conditions, the net force on the hull
(near the cutwater) will be negative. This confirms the hypothesis of the
contribution of negative lift of the wall cells near the cut water at higher values
of IVR. At an IVR of about 1.0 the net force on the cutwater region will be
positive, which is in agreement with expectations as well.
170
Chapter 7. Analysis of a complete waterjet installation
Figure 7.6 Example of triangular surface mesh of streamtube surface (in
green). Rectangular cells represent solid wall cells of waterjet
inlet in blue.
0.15
0.1
0.05
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.5 1 1.5 2 2.5
Streamtube lift coefficient
Inlet lift coefficient
Figure 7.7 Lift force coefficient of streamtube pressure distribution as
function of IVR compared to inlet lift coefficient
IVR []
C
L
[

]
7.7 Nomenclature
Numercal analysis of a waterjet propulsion system 171
7.6.3 Concluding remark
A first attempt to quantify the forces, which result from the pressure
distribution along the streamtube, has shown that there can be a significant
force in vertical direction. This force depends on the value of the IVR of the
waterjet installation. However, the actual geometry of the inlet geometry
might have an important influence on the magnitude of the lift force as well.
Further research of this is recommended before general conclusions can be
drawn.
7.7 Nomenclature
A area m
2
C
f
friction coefficient 
C
L
lift coefficient 
D diameter m
D drag force N
IVR inlet velocity ratio 
L length of flat plate m
L lift force N
p static pressure Pa
Q flow rate m
3
/s
T thrust N
v velocity m/s
w wake fraction 
z vertical direction
Greek symbols
µ
dynamic viscosity kg/ms
ρ
density kg/m
3
ϕ
flow coefficient 
Ω
angular velocity rad/s
Subscripts
inlet inlet plane
nozzle nozzle exit plane
pump pump entrance plane
wj waterjet
172
Chapter 7. Analysis of a complete waterjet installation
7.8 References
[1] Schlichting, H., ‘Boundary layer theory’, McGrawHill, New York, 1968
[2] Terwisga, T.J.C. van,’Waterjet hull interaction’, PhD thesis, Delft Uni
versity,1996
[3] Svensson, R., ‘Experience with the KaMeWa waterjet propulsion sys
tem’, AIAA conference, Paper No 891440CP, Arlington, 1989
[4] Bulten, N. & Verbeek, R.,’Design of optimal inlet duct geometry based
on operational profile’, Proceedings FAST2003 conference Vol I, ses
sion A2, pp 3540, Ischia, Italy, 2003
Numerical analysis of a waterjet propulsion system 173
Chapter 8 Concluding remarks
8.1 Conclusions
The numerical analysis of a waterjet propulsion system has led to an increase
of the knowledge of the flow phenomena occurring in the system. The basic
theory which is used to describe the flow in the system has been evaluated
as well, resulting in a reassessment of the widely applied methods. Results
have been obtained for both the flow through the inlet as well the flow through
the mixedflow pump. Typical examples are the investigation of the
streamtube shape and the time dependent forces acting on the impeller. With
aid of the numerical results, it has been possible to evaluate some general
used assumptions in waterjet propulsion theory as well.
8.1.1 Theory of thrust prediction for waterjet systems
The theory of thrust prediction is based on the method of momentum balance
with a control volume involving the streamtube. It is shown that this approach
is not completely correct. Partly because the streamtube approach does not
take into account the complete waterjet geometry and partly due to the
neglect of the contribution of the pressure on the streamtube surface.
The method of the momentum balance is derived from the theory for open
propellers, for which the usefullness is generally accepted. Direct application
of this theory to waterjet propulsion systems is not allowed because of an
open propeller is an external flow machine and a waterjet is an internal flow
machine. The thrust of an open propeller is almost entirely transferred to the
174
Chapter 8. Concluding remarks
ship through the shaft. For a waterjet installation, the thrust can for a certain
part also be transferred to the ship through the pressure and friction on the
solid surface of the inlet and the pump housing.
However, it appears that the effect of the shortcomings of the applied theory
are in general within a few percent. In practice, this is compensated for by
most waterjet manufacturers using the socalled thrustdeduction factor.
8.1.2 Numerical aspects
For waterjets with a flush inlet it is unavoidable to get a nonuniform velocity
distribution into the pump at the design point. This necessitates the
investigation of the effects on nonuniform inflow to the pump. One of the
important topics of the thesis was to investigate the origin and the
consequences of the nonuniform inflow velocity distribution to the pump of
the waterjet installation. It is found that the velocity distribution is a result of
accumulating vorticity in the flow. The vorticity is generated in the hull
boundary layer and increased in the inlet, where the flow is retarded rapidly.
For open propellers similar phenomena play a role in the creation of the so
called wake field behind a ship.
Because of the dominance of viscous flow effects only a numerical method
based on the NavierStokes equations is appropriate. Currently available
commercial Reynoldsaveraged NavierStokes (RANS) methods are well
suited for the numerical analyses of both the flow in the waterjet inlet and the
waterjet pump flow. It is demonstrated that standard twoequation turbulence
models in combination with wall functions give results with acceptable
accuracy, whilst keeping the overall calculation times and required hardware
within acceptable limits. However, the used twoequation turbulence models
do have a general shortcoming in the prediction of the pressure at a
stagnation point. This is reflected in an overprediction of drag of profile
sections.
8.1.3 Waterjet inlet flow characteristics
The calculations of the waterjet inlet flow revealed that most characteristics
are strongly related to the inlet velocity ratio (IVR=v
ship
/v
pump
). Typical
examples are:
• static pressure distribution inside the inlet
• cavitation inception at cutwater
• velocity distribution in the impeller plane
• margin against flow separation in the inlet
• shape of the inlet streamtube
• lift force on the inlet geometry
8.2 Recommendations
Analysis of mixedflow pump in a waterjet propulsion system 175
Due to variations in design ship speed and power density of the installations,
the actual design IVR can vary per vessel. For optimum performance of the
inlet, which means for example optimal cavitation inception margins and
avoidance of flow separation in the inlet, a dedicated inlet design is
recommended for each specific ship.
The shape of the streamtube upstream of the inlet can be approximated with
a semielliptical shape. The development of the flow in streamwise direction
can be characterised as a diffuser flow. The equivalent diffuser angle
depends on IVR.
The pressure distribution on the solid part of the streamtube surface creates
a net vertical force, specifically at high IVR. This lift force can be up to 20% of
the thrust of the system.
8.1.4 Waterjet mixedflow pump analyses
The numerical method is used to analyse the flow through the rotating pump
impeller as well. Calculations are based on the quasisteady multiple frames
of reference method and the fully transient moving mesh method. Differences
between the two approaches has shown to be small for performance
indicators like head, torque and efficiency.
The performance of the pump is not significantly influenced by variation of the
inflow velocity distribution. However, the pressure distribution on the rotating
impeller blades is strongly related to the inflow velocity distributions. This
results in a change of the behaviour of the inplane forces. Calculations with
uniform inflow show a radial force that is almost constant in magnitude, but
which rotates about the impeller axis with the blade passing frequency. This is
due to rotorstator interaction. In case of nonuniformity of the inflow, an
additional contribution to the mean radial force is found. The magnitude and
direction of this additional radial force depend on the amount of non
uniformity and the flow rate through the pump.
8.2 Recommendations
8.2.1 Research topics for marine propulsion systems
The general applied theory for thrust prediction of waterjets has some short
comings. Similar shortcomings might be present in the theory for the
prediction of thrust for ducted propellers and thrusters. In this respect these
type of propulsion units can be regarded as miniwaterjet installation as well.
The consequences of the net vertical force occurring in the inlet have to be
investigated in more detail. For sufficiently large forces, the behaviour of the
vessel may be influenced.
176
Chapter 8. Concluding remarks
Improvement of the thrust prediction method might give an improved insight
in the thrust deduction factor.
8.2.2 Application of RANS methods
With the currently available hardware, i.e. single linux PCs, the calculations
presented in this thesis have been made within acceptable wallclock turn
around times. This will even get better, given the fact that the performance of
computers increases continuously. It is thus recommended to start building
experience with RANS methods for marine propulsion systems, centrifugal
pumps and even complete ship hulls.
Increase of the use of RANS methods in propeller and hull design may have
an interesting consequence; the fullscale RANS calculation eliminate the use
of Reynolds scaling laws, which are currently in use to scale model scale
experimental results to fullscale.
The use of panel methods in the analysis of propellers should be evaluated.
Compared to a RANS method the required wallclock time for an analysis is
small, but the obtained accuracy, certainly at offdesign conditions, can be
rather poor. Investigations should be made to determine which level of
accuracy can be obtained with a RANS method.
A similar study can be made for ducted propellers, for which an additional
Kuttacondition is required in a panel method at the nozzle trailing edge.
Application of a RANS method for such configuration eliminates the
requirement of Kuttaconditions.
Numerical analysis of a waterjet propulsion system 177
Appendix A Stability of nonuniform
velocity distribution
In chapter 3 the velocity distribution just upstream of the impeller is
discussed. It is shown that the velocity field is strongly nonuniform at normal
waterjet operating conditions. This type of velocity distribution has to be
reproduced with the numerical method in order to obtain a correct analysis of
the complete waterjet installation.
Velocity gradients occur by definition in a nonuniform velocity distribution. If a
potential flow method is to be used, then an irrotational velocity field is
required. For a velocity field free of vorticity, velocity gradients in one direction
are compensated by gradients in the other two directions. A typical example
of a nonuniform velocity distribution in a potential flow is the flow through a
90 degree bend, where high velocity is found at the inner corner and low
velocity at the outer corner. In the first half of the bend the influence of
viscous forces on the flow will be relative small and therefore the flow can
indeed be regarded as irrotational. However, in the second part of the bend
the differences between irrotational flow and the actual, viscous flow become
apparent. In a viscous flow the high momentum fluid will move outwards in
the second part of the bend. Then it will take 30 to 50 pipe diameters, before
the velocity distribution is uniform again. In a potential flow the high speed
velocity region will vanish within a length of about 1 pipe diameter.
Ito [1] measured the pressure drop downstream of a 90 degree bend. It is
shown that the pressure drop, in addition to the one that occurs in a straight
pipe, after a bend occurs over a length of about 50 pipe diameters. This is in
accordance with the length which is required to obtain a fully developed
velocity distribution without swirl again. This phenomenon can be attributed to
178
Appendix A Stability of nonuniform velocity distribution
the development and decay of vorticity in the flow. This vorticity is transported
and redistributed in the flow, until it is dissipated into heat. In this type of flow
velocity gradients contribute to the vorticity. On the other hand, reduction of
the vorticity leads to smaller velocity gradients and thus a more uniform flow.
For an accurate description of the viscous flow through a 90 degree bend
vorticity can not be neglected. It is questionable whether the typical non
uniform waterjet velocity distribution can be represented in an irrotational
velocity field then. This velocity field is a result of boundary layer ingestion
and retardation of the flow, as discussed in section 3.4. The boundary layer
velocity profile has large velocity gradients in the direction normal to the wall.
These velocity gradients are also present in the flow in a diffuser. It is to be
expected that the axial component of the velocity gradient in the waterjet inlet
contributes to a higher vorticity. This hypothesis can be verified, analysing the
stability of the nonuniform velocity field in a potential flow method and a
RANS method.
A.1 Test case with nonuniform pipe flow
For this test the nonuniform axial velocity distribution will be prescribed at the
inlet of a circular pipe. In a potential flow method velocity components in
radial and tangential direction will be introduced to create the velocity
gradients (in these directions), necessary to ensure irrotational flow. These
velocity components are constrained by the wall boundary conditions, which
will imply constrains on the admissible inlet velocity distribution. In a viscous
flow method only velocities in the axial direction will result. As a consequence
the flow is rotational.
The length of the pipe was set to 2 pipe diameters. This will not be sufficient
to get complete uniform flow in the RANS calculation, but that is not the
object of this analysis. Several runs have been made to ensure mesh
independent results.
Figure A.1 shows the nonuniformity parameter ζ, as defined in equation
(3.1), as function of the nondimensional pipe length for the potential flow
method. Four different IVR conditions in the range from 1.68 to 2.19 have
been analysed, the velocity distributions have been shown in figures 5.13a to
5.13d.
For all conditions a very strong decay of the nonuniformity is observed. After
a pipe length of about 1 diameter the flow field is almost completely uniform.
The results of the calculations with a RANS method are presented in figure
A.2. Here a very small decay of the nonuniformity can be seen. The viscosity
causes a slow decay of the initial vorticity, what leads to smaller velocity
gradients and thus more uniform flow. In case of a circular pipe geometry,
vorticity stretching will not occur. Transformation into heat is thus the only
remaining option.
A.1 Test case with nonuniform pipe flow
Numerical analysis of a waterjet propulsion system 179
It can be concluded that there are significant differences between both
methods. Use of a potential flow method for the analysis of a waterjet
installation is therefore not possible.
0
0.05
0.1
0.15
0.2
0.25
0.3
0 0.5 1 1.5 2 2.5
IVR = 2.19
IVR = 2.03
IVR = 1.87
IVR = 1.68
Figure A.1 Development of nonuniformity ζ in potential flow for circular
tube
0
0.05
0.1
0.15
0.2
0.25
0.3
0 0.5 1 1.5 2 2.5
IVR = 2.19
IVR = 2.03
IVR = 1.87
IVR = 1.68
Figure A.2 Development of nonuniformity ζ in flow with vorticity for
circular tube
180
Appendix A Stability of nonuniform velocity distribution
A.2 References
[1] Ito, H., ‘Pressure losses in smooth pipe bends’, ASME Journal Basis
Engineering, Vol. 82, p.131, 1960
Numerical analysis of a waterjet propulsion system 181
Appendix B Fourier analyses of transient
flow calculations
Results of the Fourier analyses of the axial velocity and pressure obtained for
the transient calculations are presented in this appendix. The solution is mon
itored at six positions in the domain during the solution process. Three points
are located upstream of the impeller and three in between the rotor and the
stator. The Fourier analyses of the solution of these last three points can give
a good indication whether the solution has become periodic.
The results of the axial velocity at the monitoring points upstream of the pump
are governed by the inlet boundary condition. Fourier analyses of these
results do not provide additional information. The results of the pressure level
upstream of the pump show small fluctuations with a frequency of about 13
times the blade passing frequency. These pressure results are not used for
the evaluation of periodicity of the flow field.
The results of the Fourier analyses for design flow rate Q are presented in fig
ures 6.11 and 6.12 as well.
The Fourier analyses for the results of calculations with uniform inflow are
presented in figures B.1 to B.6. The frequencies are normalised with the shaft
frequency in all diagrams.
The results of the calculations with nonuniform velocity distributions are pre
sented in figures B.7 to B.10 for design flow rate and in figures B.11 to B.14
for 80% of the design flow rate.
Review off all results confirms the periodic behaviour of both the axial velocity
and the pressure at the three arbitrarily chosen monitoring points.
182
Appendix B Fourier analyses of transient flow calculations
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

] Pressure point 4
Pressure point 5
Pressure point 6
Figure B.1 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for 60%
of design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.2 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for 70%
of design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
Numerical analysis of a waterjet propulsion system 183
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.3 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for 80%
of design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

]
Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.4 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for 90%
of design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
184
Appendix B Fourier analyses of transient flow calculations
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.5 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for the
design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

]
Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.6 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller with uniform inflow for 110%
of design flow rate
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

]
Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
Numerical analysis of a waterjet propulsion system 185
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.7 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=1.68 at design flow
rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.8 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=1.87 at design flow
rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
186
Appendix B Fourier analyses of transient flow calculations
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.9 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=2.03 at design flow
rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.10 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=2.19 at design flow
rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
Numerical analysis of a waterjet propulsion system 187
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.11 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=1.68 at 80% of
design flow rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.12 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=1.87 at 80% of
design flow rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
188
Appendix B Fourier analyses of transient flow calculations
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.13 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=2.03 at 80% of
design flow rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
D
i
m
e
n
s
i
o
n
l
e
s
s
p
r
e
s
s
u
r
e
[

]
Pressure point 4
Pressure point 5
Pressure point 6
Figure B.14 Fourier transforms of axial velocity and pressure at monitoring
points downstream of the impeller for IVR=2.19 at 80% of
design flow rate
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0 3 6 9 12 15 18 21 24 27 30
Normalised frequency []
N
o
r
m
a
l
i
s
e
d
a
x
i
a
l
v
e
l
o
c
i
t
y
[

] Axial velocity point 4
Axial velocity point 5
Axial velocity point 6
Numerical analysis of a waterjet propulsion system 189
Summary
A waterjet propulsion system is used to propel ships, using a pump which
produces a high speed jet. A standard waterjet installation can be divided into
an inlet, a pump and a nozzle. For manoeuvring and reversing purposes an
additional steering device can be integrated into the installation. The
development of waterjet propulsion systems has made significant progress
over the last few decades. Nowadays, commercial fastferries reach
velocities of 50 knots, which is about 90 km/h.
The theory to describe waterjet propulsion systems is derived from open
propeller theory. The prediction of the thrust of a propeller is based on the
momentum balance of a streamtube control volume. This thrust is then
transferred though the shaft of the propeller to the hull of the ship. In contrast,
for a waterjet propulsion system, forces are transferred to the hull not only
through the shaft but also through the solid surface of the installation. A
critical review learns that some assumptions made for open propellers are
not valid for waterjets.
The inflow to the waterjet pump is nonuniform. This results in a blade loading
that varies during an impeller revolution. The cause and the effects of this
nonuniform inflow have been investigated. Four contributing factors are
identified for the development of a nonuniform velocity distribution just
upstream of the pump. As a first cause, the water is ingested from below the
the hull of the ship, where a boundary layer with a nonuniform velocity
distribution is present. Even at normal operating conditions, the water is
subsequently retarded in the inlet, which results in an increase of the non
uniformity. Finally, the inflow passes the bend in the inlet and the protruding
190
Summary
shaft which add to the increase in nonuniformity. It is concluded that the non
uniformity is the result of the accumulated vorticity in the flow. Due to this
vorticity, a stable velocity distribution is found, and the typical velocity
distribution is more or less independent of the actual design of the inlet.
The investigations are based on numerical analyses of the flow through the
complete waterjet installation. Selection of the numerical method is based on
the capability to capture typical flow phenomena in a waterjet installation:
high Reynolds number, timedependency, and incompressible flow in a
partially rotating frame of reference. Due to the high level of nonuniformity of
the inflow, the ability to generate and transport vorticity in the flow is an
important requirement, as well as the possibility to take into account the flow
phenomena in the tip clearance region between the rotating blades and the
stationary housing.
A Reynolds averaged NavierStokes (RANS) method is chosen to perform all
numerical analyses. The Reynoldsstresses are obtained using the two
equation kε turbulence model. This turbulence model is known to produce an
error near a stagnation point. An estimation of the influence of this error on
the prediction of thrust and torque shows that the actual deviations are
acceptable.
The numerical models of both the waterjet inlet and the mixedflow pump are
validated with available experimental data. Results of calculations of the
waterjet inlet flow are compared with measurements of static pressure along
the inlet and with the total pressure and velocity distribution at the impeller
plane. Agreement between the CFD results and the experimental data is
good for all calculated conditions. The flow phenomena in a waterjet inlet are
characterised by the inlet velocity ratio (IVR), which is the ratio of the ship
speed and the pump speed.
The shape and location of the streamtube of the ingested water is determined
with aid of a concentration scalar. This enables the visualisation of the
streamtube and the calculation of the mass averaged inflow velocity. In this
way the wake fraction of the waterjet installation is determined accurately. It is
shown that the actual shape of the streamtube depends on IVR.
The CFD calculations of the mixedflow pump are validated with experimental
data for the pump head and the shaft power. The calculations are performed
with a quasisteady multiple frame of reference (MFR) method and a fully
transient moving mesh method. Differences between predicted head and
power in both methods are small.
The fully transient moving mesh calculations with a uniform inflow velocity
distribution provide the unsteady excitation forces on the impeller due to
rotorstator interaction. It is found that the magnitude of the radial interaction
force depends on the flow rate though the pump.
The influence of the nonuniform velocity distribution to the pump is
investigated as well. The deviation in pump performance is limited to a few
Numerical analysis of a waterjet propulsion system 191
percent for the calculated conditions. The influence on radial forces is far
greater, however. An additional mean component of the radial force is found,
the magnitude and direction of which are related to the flow rate and the level
of nonuniformity. The origin of this mean force is an unbalanced torque on
the impeller blades, due to a variation of the angle of attack during a
revolution.
Both validated numerical models of the inlet and the pump are combined to
form the complete waterjet installation. Results of the calculations of the
complete unit are compared with the results of the standard waterjet
performance prediction and selection software of Wärtsilä Propulsion
Netherlands BV. Good agreement is found for the prediction of flow rate,
thrust and torque of the installation. Two methods to determine the thrust are
used: (i) the integration of the axial force component on the solid wall and (ii)
the application of a simplified version of the integral momentum balance
equation. The latter method is generally applied by ship building companies.
A clear deviation between the two methods is found for higher ship speeds.
Analysis of the net force in vertical direction reveals a significant lift force at
high speeds. It is concluded that the method based on the momentum
balance for the streamtube control volume, has some shortcomings. The
deviation increases for higher ship speeds.
The numerical results confirm the hypothesis that the simplified method to
describe waterjet installations is not correct. This can be partly attributed to
the neglect of the influence of the hull in the vicinity of the waterjet inlet and
partly to the neglect of the contributions of the pressure distribution acting on
the streamtube.
192
Summary
Numerical analysis of a waterjet propulsion system 193
Samenvatting
Een waterjet voortstuwingssysteem wordt gebruikt om schepen voort te
stuwen met behulp van een pomp, die een waterstraal met hoge snelheid
produceert. Een standaard waterjet voortstuwingssysteem is opgebouwd uit
een inlaat, een pomp en een nozzle. Een stuurdeel kan geïntegreerd worden
in de installatie om manouvreren mogelijk te maken. De ontwikkeling van
waterjet voortstuwingssystemen heeft de laatste decennia een enorme
ontwikkeling doorgemaakt. Tegenwoordig worden met commerciële fast
ferries snelheden van rond de 50 knopen bereikt. Dat komt overeen met
ongeveer 90 km/uur.
De theorie om het voortstuwingssysteem te beschrijven is afgeleid van de
theorie voor open schroeven. De voorspelling van de stuwkracht van een
schroef is gebaseerd op de impulsbalans van een stroombuis controle
volume. Deze stuwkracht wordt door de as van de schroef op het schip
overgebracht. Echter, de krachten die werken op een waterjet
voortstuwingssysteem worden behalve via de as ook via de vaste wanden
van de installatie naar het schip overgebracht. Een kritische analyse laat zien
dat sommige aannames die gemaakt zijn voor open schroeven niet geldig
zijn voor waterjets.
De instroming naar de waterjet is nietuniform. Dit heeft tot gevolg dat de
belasting van de waaierbladen varieert gedurende een omwenteling. De
oorzaak en de gevolgen van deze nietuniforme instroming zijn onderzocht.
Er zijn vier factoren geïdentificeerd die bijdragen aan de ontwikkeling van de
nietuniformiteit van de instroming vlak voor de pomp. Allereerst wordt het
water afgezogen uit het gebied onder het schip. Daar bevindt zich een
194
Samenvatting
grenslaag met een nietuniforme snelheidsverdeling. Vervolgens ondergaat
deze instroming een vertraging in de inlaat van het systeem, wat resulteert in
een toename van de nietuniformiteit. Als laatste passeert de instroming een
bocht in de inlaat en de as van de pomp, waardoor de nietuniformiteit verder
wordt vergroot. Er kan worden geconcludeerd dat de nietuniformiteit het
resultaat is van een accumulatie van vorticiteit in de instroming. Door deze
vorticiteit wordt een stabiele snelheidsverdeling gevormd en daardoor is de
typische snelheidsverdeling nagenoeg onafhankelijk van de vorm van de
inlaat.
Het onderzoek is gebaseerd op een numerieke analyse van de stroming door
de complete waterjet installatie. De eisen die worden gesteld aan de
numerieke methode worden bepaald door de aard van de stroming door een
waterjet installatie. Deze kan worden gekarakteriseerd als een
tijdsafhankelijke, nietsamendrukbare stroming met een hoog Reynolds getal,
in een systeem dat gedeeltelijk roteert. Vanwege de hoge mate van niet
uniformiteit zijn verder de productie en het transport van vorticiteit belangrijke
aspecten, evenals de mogelijkheid om de stroming in de nauwe spleet tussen
de roterende waaier en het stationaire pomphuis goed te kunnen modelleren.
Een Reynoldsgemiddelde NavierStokes (RANS) methode is gekozen voor
alle numerieke analyses. De Reynoldsspanningen worden bepaald aan de
hand van het kε turbulentie model. Het is bekend dat dit model een fout
introduceert nabij een stagnatie punt. Uit een schatting van de invloed van
deze fout op de voorspelling van stuwkracht en koppel blijkt dat de
afwijkingen acceptabel zijn.
De numerieke modellen van zowel de waterjet inlaat als ook de mixedflow
pomp zijn gevalideerd met beschikbare experimentele data. Resultaten van
berekeningen van de stroming door de waterjet inlaat zijn vergeleken met
metingen van de statische druk langs de inlaat en met de totale druk en de
snelheid in het waaiervlak. De overeenkomst tussen de CFD resultaten en de
experimentele data is goed voor alle berekende condities. De
stromingsfenomenen in een waterjet inlaat worden bepaald door de inlet
velocity ratio (IVR), wat de verhouding geeft tussen de scheepssnelheid en
de gemiddelde axiale snelheid vlak voor de pomp.
De vorm en de ligging van de stroombuis waardoor het water naar de inlaat
van de waterjet stroomt, is bepaald met behulp van een concentratiescalar.
Dit geeft de mogelijkheid om de stroombuis te visualiseren en om de massa
gemiddelde instroomsnelheid te bepalen. Op deze manier kan het
volgstroom getal (of wakefractie) van de waterjet installatie nauwkeurig
worden bepaald. Het is aangetoond dat de vorm van de stroombuis afhangt
van de waarde van IVR.
De CFD berekeningen van de mixedflow pomp zijn gevalideerd met
experimenteel bepaalde waarden van opvoerhoogte en asvermogen. De
berekeningen zijn uitgevoerd met een quasistationaire multiple frame of
reference (MFR) methode en met een volledig tijdsafhankelijke roterende
Numerical analysis of a waterjet propulsion system 195
mesh methode. Voor wat betreft de berekende waarden van opvoerhoogte
en vermogen zijn de verschillen tussen beide methodes klein.
De volledig tijdsafhankelijke roterende mesh berekeningen met uniforme
instroming geven de instationaire excitatiekrachten op de waaier als gevolg
van rotorstator interactie. De grootte van de radiale interactiekracht blijkt
afhankelijk van het debiet door de pomp.
De invloed van de nietuniforme snelheidverdeling vlak voor de pomp is ook
onderzocht. De afwijking in de pompprestaties is beperkt tot een paar
procenten voor de berekende condities. De invloed op de radiale krachten is
echter veel groter. Een additionele tijdsgemiddelde component van de radiale
kracht is aanwezig met een grootte en richting die afhangt van het debiet en
de mate van nietuniformiteit. De oorzaak van deze tijdsgemiddelde kracht is
een onbalans in het koppel van de verschillende waaier bladen, als gevolg
van de variatie in de aanstroomhoek van het blad tijdens een omwenteling.
De twee gevalideerde numerieke modellen van de inlaat en de pomp zijn
gecombineerd om een complete waterjet installatie te vormen. Resultaten
van de berekeningen van de complete unit zijn vergeleken met het standaard
waterjet performance predictie programma van Wärtsilä Propulsion
Netherlands BV. Goede overeenkomsten zijn gevonden voor de voorspelling
van het debiet door de pomp, de stuwkracht en het koppel van de installatie.
Twee methoden voor de bepaling van de stuwkracht zijn gebruikt: (i)
integratie van de axiale component van de kracht op alle vaste wanden en (ii)
de toepassing van een vereenvoudigde versie van de integrale impuls
balans. De laatste methode wordt algemeen toegepast door producenten van
schepen. Voor hoge scheepssnelheden is een duidelijke afwijking gevonden
tussen de twee methoden.
Analyse van de netto kracht in verticale richting laat zien dat er een
significante liftkracht aanwezig is bij hoge scheepssnelheden. Hieruit volgt de
conclusie dat de vereenvoudigde methode, gebaseerd op de impuls balans
voor het stroombuis controle volume, tekortkomingen heeft. De afwijking
wordt groter naarmate de scheepssnelheid toeneemt.
De numerieke resultaten bevestigen de hypothese dat de vereenvoudigde
methode voor de beschrijving van de waterjet installatie niet correct is. Dit
kan deels worden toegeschreven aan de verwaarlozing van de invloed van
de scheepswand in de buurt van de waterjet inlaat en deels aan de
verwaarlozing van de bijdragen van de drukverdeling die werken op de
stroombuis.
196
Samenvatting
Numerical analysis of a waterjet propulsion system 197
Dankwoord
Tijdens mijn afstuderen kreeg ik van prof. Bert Brouwers het advies om ‘eerst
maar eens aan het werk te gaan en dan later een keer een promotie
onderzoek te komen doen’. Dit advies heb ik opgevolgd en dit proefschrift is
nu het resultaat van de combinatie van 9 jaar werken bij Wärtsilä Propulsion
Netherlands BV en het promotie onderzoek aan de Technische Universiteit
Eindhoven.
Naast prof. Bert Brouwers wil ik ook mijn tweede promotor prof. Harry
Hoeijmakers en mijn directe begeleider, dr. Bart van Esch, bedanken voor
hun bijdragen. Ik ben me ervan bewust dat het begeleiden van een externe
promovendus niet altijd even gemakkelijk was, iets waar vooral Bart mee te
maken had.
De combinatie van het onderzoek met mijn werkzaamheden bij Wärtsilä
Propulsion Netherlands BV was niet mogelijk geweest zonder alle
studiedagen. Zelfs in tijden van drukte was het mogelijk om tijd aan het
onderzoek te besteden. Hiervoor wil ik Bram Kruyt, Hanno Schoonman en
Do Ligtelijn bedanken.
Het interessante van de combinatie van werken en promoveren is
waarschijnlijk de directe feedback van de varende waterjet installaties. Bij
het zoeken naar verbeteringen van de waterjet installaties heb ik vele
gesprekken en discussies gehad met Rob Verbeek, die waarschijnlijk de
meest waardevolle bijdragen aan dit proefschrift hebben opgeleverd. In ieder
geval hebben we de term ‘voortschrijdend inzicht’ bij het management
geïntroduceerd.
Norbert Bulten,
Rosmalen, september 2006.
198
Dankwoord
Numerical analysis of a waterjet propulsion system 199
Curriculum Vitae
29 april 1973 Born in Winterswijk, the Netherlands.
19851991 VWO at RSG Hamaland, Winterswijk
19911996 Study Mechanical Engineering at the University of Twente,
The Netherlands
Specialisation: fluid dynamics and thermal engineering
Graduation thesis on numerical analysis of a centrifugal pump
impeller with two different RANS methods
1996 Traineeship at Sulzer Innotec in Winterthur, Switzerland
1997 Researcher at University of Twente, The Netherlands.
Project for IHC Parts & Services to optimise dredger pumps
1997 Wärtsilä Propulsion Netherlands (formerly Lips propellers)
in Drunen.
1997  2002 Research engineer waterjets
2002  2006 CFD specialist at Propulsor Technology
department.
from 2006 Team leader CFD at Propulsor Technology
department
20012006 PhD study at Technische Universiteit Eindhoven, The
Netherlands
200
Curriculum Vitae
Acknowledgement: The research described in this thesis was supported by Wärtsilä Propulsion Netherlands B.V.
Cover: Michelle Tjelpa Photo: Bram Kruyt Printing: Printservice Technische Universiteit Eindhoven
Copyright © 2006 by N.W.H. Bulten, The Netherlands All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronically, meachanically, by photocopying, recording, or otherwise, without the written permission of the author.
A catalogue record is available from the Library Eindhoven University of Technology ISBN10: 9038629885 ISBN13: 9789038629889
Numerical Analysis of a Waterjet Propulsion System
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de Rector Magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor Promoties in het openbaar te verdedigen op woendag 15 november 2006 om 16.00 uur
door
Norbert Willem Herman Bulten
geboren te Winterswijk
Dit proefschrift is goedgekeurd door de promotoren: prof.dr.ir. J.J.H. Brouwers en prof.dr.ir. H.W.M. Hoeijmakers Copromotor: dr. B.P.M. van Esch
.............................. 43 Closing remark........................ 5
1...................6 Waterjet layout........... 39 2.......................................... 16 2........................... 22 General pump theory ...4 Concluding remarks ............................2...........3 1..3........3
2...1 General thrust equation ......................................................................... 42 Waterjet selection ..............1..... 11 Nomenclature ..............................................................3 Waterjet thrust .............................................................................3.. 10 Outline of this thesis .............................3........................3 Limitations in jet velocity ratio ......... 22 2............... 6 Relation of waterjet propulsion system to other turbo machinery ................. 7 Aim of the analysis............................................................................................................. 28 2.......................................................................................................... 12
Chapter 2 Waterjet propulsion theory ..........................1 Dimensionless performance parameters ............................. 44 Nomenclature ..... 34 Pump head .....3 Jet velocity ratio ..... 22 2....................................2 Pump geometry parameters ...........................8 2.........5........................................2.................. 37 2....3.............2
2......... 40 2......................................................................................... 49
Numerical analysis of a waterjet propulsion system
1
................................................................................. 46
2.................................................. 41 2.......................1..................... 26 Thrust .........1 1.....................3 Cavitation parameters .......... 27 2....2 1............................ 21 2....4 Limitation of power density .........................................6 2..........4 Summary ....5............1 Representation of nonuniform velocity distribution....................................................................1.................................................................................... 25 2................ 20 2.................. 12 References ............ 34 Overall propulsive efficiency ............................7 2.......1 Characteristic velocities in a waterjet system ........................2. 27 2...................2 Open propeller thrust ....4 Correlation with propeller performance parameters ...................................................................... 30 2............2 Limitations in specific speed ......................5...............1 Cavitation margins .....2.................... 44 References .........Table of contents
Chapter 1 Introduction .......1 Wake fraction ........4 1........................................................ 15
2..5 1......................................5
2................5.....4 2....................... 24 2............... 49
3................. 17 2.............................9
Chapter 3 Nonuniform distribution of pump entrance velocity field ........................................2 Inlet velocity ratio ........1......................................
.............3 Twodimensional representation ..............2 High Reynolds number .............................3 Review of CFD analyses on waterjet inlets.........
3...............1........3....... 52 3........................7
3.........1.........................1 Incompressibility .......................................................................................1...............63
4................................4....................89
5.................2 Eddy viscosity turbulence models ............................1 Boundary layer ingestion ...............................................1 Isolated NACA 0012 profile .............3...................4 Bend in the inlet duct . 60 Nonuniform inflow velocity distributions in other turbo machinery60 Nomenclature.........1 5................. 63 4.....................................1 Boundary conditions ...........................6
Chapter 5 Numerical analysis of waterjet inlet flow .4.......... 68 4..........................3
4............ 95 Validation with experimental data ................ 91 Numerical approach......1..........................5 Tip clearance flow .................... 75 4............................. 52 Local flow rate fluctuations......................................5 Closing remark .......3.....................................................................1................................... 65 4.5 3............1 Requirements of mathematical method ...........................4.........................................................4.2 3............... 61 References........................... 60 3............... 50 3.1................ 89 Geometry and mesh generation ..........5 4... 59 3....... 57 3..3.......................................................3 3....6 3..... 75 4.............. 66 4.....................................................................1......................................... 55 Origin of the nonuniform velocity distribution..............3.............................2 Fluid properties ....4
3..... 65 4.................3 Discretisation and solution algorithm ....................3 Time dependency .3 Obstruction of the flow due to the shaft ................2 5..............................3 Sensitivity of errors in drag on thrust and torque .......................................................... 64 4.................. 96
5..........................4.....4 Nonuniformity of impeller inflow ............................................................2 Deceleration of the flow .............................................................................. 67 Reynolds Averaged NavierStokes (RANS) flow ....2 Nondimensional representation ..................................... 70 Twodimensional test cases..............................................4
2
...... 53 Impeller velocity triangles.... 80 4.................... 82 Nomenclature.2 Cascades with NACA 65410 profiles ........................ 62
Chapter 4 Mathematical treatment ... 64 4....2 4... 67 Conservation laws. 68 4........................1..........................................4... 94 5...... 58 3...................... 95 5...........................4
4................. 86
4............1 Experimental setup ....4..............1 Reynolds averaging ..6 Final remarks ..............................................1..................................4................. 94 5........................... 85 References......................................... 57 3...........................
............4 Comparison of velocity field at impeller plane ..................... 134 6.....................5..4
Numerical analysis of a waterjet propulsion system
3
.......4
6...................4...1 Comparison of static pressure along the ramp centre line 97 5....3 Comparison of total pressure at impeller plane .......................2...........4. 118 Evaluation of wall shear stress ............................. 127 Numerical approach......................................... 135 6........................2............1 Visualisation of suction streamtube ................................. 117 5.2 7.......................................... 118 5.............................. 163 Evaluation of required power ............................... 131 6...................1 Boundary conditions .......4......................... 127
6...............5
5..... 159
7...........2 Fluid properties ....1 6.3....4.......... 161 7............................ 125 References .................... 156
6.. 149 6.............. 159 Evaluation of volume flow rate.....................7 5.......... 117 Analysis of the suction streamtube ...... 132 Validation with experimental data .6 Mesh convergence study ............. 163 7.4...............3 Impeller rotation ........... 160 Evaluation of waterjet thrust ....................4...................7 Closing remarks .............................3 Results .............................................4.......5 6..............3
6.............................................3 Rotorstator interaction forces ................................................. 125
Chapter 6 Numerical analysis of waterjet pump flow . 131 6........ 164
7.............................. 147 6............. 150 6......... 148 6..........3......... 105 5. 108 5........4............ 147 6...6 5......................................2 Comparison of cavitation inception pressure at cutwater 101 5.....4 Radial forces for nonuniform inflow ............ 155 References ......3......................4.......................8
5..............................1 Quasisteady flow calculations with the MFR method .........2 Momentum balance .............5.........4... 123 Nomenclature .......3 Generation of the numerical model.................3......5 Concluding remark ....... 144 Influence of nonuniform axial inflow .........2....................................1 Integration of solid wall forces ......3 Flow rate fluctuations in the impeller channel ... 155 Nomenclature . 132 6........................................................4 Calculation of global pump performance ..............5............ 138 6....1 7........................................2 Background of radial forces acting on the impeller ....... 161 7.6
Chapter 7 Analysis of a complete waterjet installation ..........................1 Pump performance for nonuniform inflow ........................3...........................2.......... 132 6.................................................4..................................2 Geometry and mesh generation ...........................................................3........................ 115 5............................5 Results obtained with kω turbulence model ................................2 Determination of suction streamtube shape ............................. 113 5......4.......2 Transient flow calculations with moving mesh ..............
..............173
8..... 175 8.................. 169 7... 173 8...........2
Appendix A Stability of nonuniform ....................................................5 7..................4 Waterjet mixedflow pump analyses ................................... 171 References................1............6...................1 Conclusions ........................3 Concluding remark ......................... 178 References.....................2 Numerical aspects ..............................................2..7 7..... 168 7................... 168 7.........6...........................8
Analysis of vertical force on waterjet structure.... 174 8.................................... 174 8............... 176
8................................3 Waterjet inlet flow characteristics .........6..................... 172
Chapter 8 Concluding remarks ............ 173 8.............189
Samenvatting ..................................... 171 Nomenclature.199
4
..................velocity distribution 177
A.......................................................................................1 Evaluation of momentum balance in vertical direction .............................2 Calculation of vertical force on streamtube . 180
Appendix B Fourier analyses of transient flow calculations 181
Summary ...........1........................................................ 175 Recommendations ....................2 Application of RANS methods ....... 175 8..........................................................2................2 Test case with nonuniform pipe flow....................................................1...........................1 Research topics for marine propulsion systems ...6
7...................................... 165 Pressure distribution on streamtube surface .1 Theory of thrust prediction for waterjet systems ............197
Curriculum Vitae ...............................................................................193
Dankwoord ......................................
7.....................1 A...1....................
. most readers do not realize the substantial development in high speed ship transportation. But probably.] together with a particular way of Forceing water through the Bottome or Sides of Shipps belowe the Surface or Toppe of the Water. this type of vessel had entered the market less than two decades before. 50 that he was able ‘to make Boates. However. since at that time there was a great interest in using steam to raise water and to operate fountains. the first type of waterjet propulsion was invented already 300 years earlier. 132 was granted to Thomas Toogood and James Hayes for their invention of ’Forceing Water by Bellowes [.Chapter 1 Introduction
The desire to travel faster and further is probably as old as mankind itself. and later also over sea. In 1661. English patent no. And since about a century is it possible to travel through air as well.. Achievements in automotive and aerospace technology are widely recognized. At first it was only over land. There has been an enormous development in the way people use to travel from one place to another. However. by David Ramseye [2]. However. At the end of the 20th century. fast ferry catamarans sailing at 50 knots (equivalent to about 90 km/h) were in commercial service all over the world. He stated in 1630 in English Patent No. Shippes and Barges to goe against Stronge Winde and Tyde’. Currently used stern mounted waterjets are based on principles as applied by Riva Calzoni in 1932 [1]. The considerable development in the high speed craft can be partly contributed to the application of waterjet propulsion systems. It is supposed that he had a waterjet in mind for the propulsion. they did not
Numerical analysis of a waterjet propulsion system 5
. This concept was based on a waterjet without a doubt. which may be of siguler Use and Ease in Navagacon’.
Downstream of the nozzle there is a steering device. Besides the flush mounted inlet. Figure 1. In the remainder of the thesis.1 shows a drawing of a typical waterjet installation. The latter two have an opening that is situated more or less perpendicular to the flow direction. This can be useful for quick crashstop manoeuvres.
4
3
2 1
1. with the main components labelled. it is called a booster waterjet.
1. in fastferries and high speed motor yachts. Luxury high speed motor yachts have achieved ship speeds well above 65 knots. which can deflect the jet in order to create steering and reversing forces. The ram and scoop intake will not be considered in this thesis. Kruppa et al. The ducting system upstream of the pump is called the inlet. the nozzle and the steering device. pump 3. There are also installations for the deflection of the jet possible. Introduction
manage to develop a working prototype. the pump. ram and scoop type inlets are mentioned. At the start of the 21th century the sizes of installed waterjets have increased to diameters of about 3 meter. In general the stator bowl and the nozzle are integrated in one part. whereas the flush mounted inlet opening is parallel to the flow.
6
. can be divided into four components: the inlet. The waterjet in figure 1. This is used. From 1980 onwards the use of waterjets in commercial applications really started to grow [4]. This invention and the subsequent development of the waterjet until 1980 is described in much more detail by Roy [3]. steering device Threedimensional view of a waterjet installation
The main component is the pump. which is about 120 km/h [5].1 shows a flush mounted inlet duct. which delivers the head to produce the jet at the nozzle exit. for example. This has led to installed powers of 25 MW per installation. inlet 2.1
4.Chapter 1. [6] have given an overview of the basic concepts of waterjet inlet ducting systems. If the waterjet has no steering device at all. with only the reversing option. the combination of the pump unit and the nozzle is regarded as the waterjet pump. nozzle
Figure 1.1
Waterjet layout
A sternmounted waterjet installation as used in commercial applications.
2
Relation of waterjet propulsion system to other turbo machinery
If the very early 17th century developments are neglected. So here the fluid is the distinguishing factor. waterjet propulsion is relatively new. The left side is formed by external flow machines and the right side by internal flow machines. In aerospace the propeller has been replaced by the jet engine. The two side planes of the box show the difference in type of flow. Transmission of the forces in an external flow machine can only be done through the shaft. Another important fluid property of water is its very low compressibility. This can lead to misunderstandings. This group contains. Both phenomena can be important in the selection of numerical solution methods. besides the waterjet. The box model will be used to relate the occurring phenomena in a waterjet installation to known ones in other machines.1. Internal flow machines can also transfer forces through the surrounding structure. The top plane of the box shows four installations which operate in water.2 shows a box with eight different types of apparatus. whereas the applications on the bottom plane operate in air. if the same waterjet is described as a mixedflow pump. Cavitation is a typical problem for installations operating in water. The three faces which are connected to the waterjet share a common property. The front face is formed by four installations which are designed to produce thrust. A ship propeller seems to be the most logical connection to a waterjet for a description of the propulsion system. If history is reviewed an interesting parallel can be recognised. Any thrust production by the installations at the back face (mixedflow pump. the ship propeller and the two main aeroplane propulsion systems. with the same nomenclature. Application of waterjets in marine industry shows a similar trend where the waterjet propelled vessels reach higher speeds. with the accompanying pump nomenclature. Typical parameters used in propeller
Numerical analysis of a waterjet propulsion system
7
. compressor. Figure 1. Many relations which describe the principles of waterjet propulsion are directly derived from propeller theory. like the ship propeller. what makes these methods less suitable for the analysis of a waterjet propulsion system. For further development of the installation it may be useful to look at related engineering applications. often Q is used for torque in propeller theory and for flow rate in pump theory. ventilator and mixer) is an undesirable side effect. the aeroplane jet engine and the mixedflow pump. which was necessary to reach higher speeds. For example. Numerical methods used for the analysis of compressors and other flow machinery often require a certain amount of compressibility.2 Relation of waterjet propulsion system to other turbo machinery
1.
However.2
Box model of connections of waterjet to other types of turbomachinery
theory are the thrust loading coefficient and the cavitation number [7]. Introduction
Figure 1. Moreover the concept of wake fraction. there is a very important difference between a propeller and a waterjet installation. A propeller is an external flow machine whereas the waterjet installation is mainly an internal flow system. In a waterjet installation the forces can be transferred to the vessel via the shaft and via the ship structure. This results in fluctuations of the pressure distribution on the blades and in a radial force on the shaft. which represents the difference between the free stream advance speed and actual inflow velocity. Due to this wake field the loading of a propeller blade fluctuates during a revolution. These phenomena will also be present in a waterjet. can be used to account for the effect of the hull boundary layer ingestion. This is similar to the wake field of a ship propeller. It is wellknown that the inflow velocity distribution to the waterjet impeller is strongly nonuniform.
8
. In that case a negative force will work on the transom stern and the inlet ducting. These parameters can be employed to describe the performance of a waterjet as well.Chapter 1. The thrust of a propeller will always be guided through the shaft into the ship. Therefore the choice of a propeller as a starting point for the analysis of a waterjet installation seems to be logical. In fact it is possible to have a higher thrust acting on the shaft than the net thrust of the installation [8].
As a consequence.= 1 + ( v out ⁄ v in ) P shaft (1. which is based on the required head to produce the jet velocity and the head to overcome the hydraulic losses. compressor. just like the waterjet. vout the jet velocity leaving the engine and vin the velocity of the air entering the intake.2). combustion chamber. The ratio between intake and nozzle velocity is called nozzle velocity ratio (NVR = vout/vin). Strictly spoken the mass flow in the system increases due to the addition of fuel.3)
which is often denoted as Froude efficiency.2)
· where m is the mass flow through the system. Net thrust of a turbojet engine is based on the change of momentum: · F = m ( v out – v in ) (1. the compressor is driven by the turbine. the ship speed can be used as a governing parameter for the operating point. turbine and nozzle (see for example [11]). At zero speed
Numerical analysis of a waterjet propulsion system 9
. A turbojet engine is a thrust producing internal turbomachine. n the shaft speed and D the diameter of the propeller. i. In this system the pump head curve matches the system resistance curve. since the flow is incompressible: · m = ρQ . the available set of propeller equations cannot be used for a good description of the waterjet propulsion system. The definition of the propulsive efficiency of a turbojet engine can be found in literature [11]: F ⋅ v in 2 η p = . The influence of ship speed on the operating condition is small. The turbojet engine can be divided into five major components: intake. In nondimensional notation it is called advance ratio (see for example [10]. In a waterjet a separate diesel engine or gas turbine is needed to supply the required power to the shaft.2 Relation of waterjet propulsion system to other turbo machinery
Because the ship propeller operates as an external flow machine. [10]): v ship J = nD (1. According to equation (1.1)
where vship is the ship (or advance) speed. These components include the power generating part of the jet engine.e. The theory of aeroplane jet engines may provide the missing equations to describe the performance of a waterjet system. The working point of the waterjet installation is based on the volume flow rate Q through the system. but this increase in mass flow is negligible. the thrust of a waterjet system is directly related to the volume flow rate.1. which is equal and opposite to the forward speed of the aircraft.
Chapter 1. These assumptions will be reviewed to check their validity. Nevertheless. such as flow rate. the head curve of the pump and the system resistance curve provide sufficient information to determine the volume flow rate Q through the system. Unfortunately.
10
. Typical complicating factors in the analysis of waterjets are the boundary layer ingestion and the nonuniform velocity distribution just upstream of the pump. These phenomena will increase the noise and vibrations in the installation. it should be kept into mind that cavitation and nonuniform inflow.3
Aim of the analysis
In this thesis a detailed analysis of a waterjet propulsion system is made. This will give rise to a blade loading.
1. which varies strongly with time. both the boundary layer ingestion as well as the nonuniformity of the velocity distribution are inevitable in commercial waterjet propulsion systems with a flush type of inlet. torque and thrust. two important issues in waterjet propulsion. To get a first estimation of the thrust of the system. are not dealt with in jetengine research. therefore the reciprocal value is used in literature for waterjets. At first sight this is a bit strange. this is known as jet velocity ratio µ [12]. a deterioration in cavitation behaviour and an increase of forces acting on the impeller. The aim of the analysis presented in this thesis is (i) to quantify the effects of the nonuniform inflow to the mixedflow pump and the resulting nonstationary flow in the pump on the system performance and (ii) to quantify the forces on the complete waterjet installation in both axial and vertical direction. This may lead to a decrease in system performance. Results of Computational Fluid Dynamics (CFD) calculations are used to get an impression of the flow phenomena occurring in such systems and to quantify system parameters. Introduction
the NVR becomes infinite. because normally the axial thrust in pump operation is not exploited. With the application of a numerical method some flow features are easier to determine than in a model scale test. Although the working principle of the aeroplane jet engine and the waterjet seem to be similar. The currently used theory to determine system performance includes some assumptions about the influence of the pressure distribution on the streamtube of the ingested water. like a reduced efficiency. only the average velocity of the ingested flow and the dimension of the nozzle diameter have to be known. The major problem of the impeller inlet velocity distribution is the large variation of the velocity in circumferential direction. The third type of turbomachinery which may provide part of the basic theory to describe system performance is a mixedflow pump.
Some of the underlying assumptions made will be discussed to enable assessment of these assumptions later on. Values for pump parameters in literature are based on uniform inflow. will be analysed. The chosen method for the calculation of the flow through a waterjet inlet will be validated with available experimental data in chapter 5. like volume flow. Results of measurements will be shown to give an impression of the level of nonuniformity. such as potential flow. the conclusions of the present research will be presented in chapter 8. The nature of the velocity distribution will be discussed in chapter 3. The results of the numerical analysis of the inlet will also be used to evaluate the shape of the streamtube upstream of the inlet duct. The analysis also reveals the basic principles of waterjet selection which is suitable for most of the current applications. Overall performance indicators.4
Outline of this thesis
In chapter 2 the conventional theory of waterjet propulsion systems will be discussed in detail. Euler and RANS. An important requirement is the capability to capture the effects of the nonuniform inflow to the pump. the mass flow rate is prescribed as a boundary condition. In these calculations. it is wellknown that a waterjet impeller has to operate in a nonuniform inflow velocity field.1. Chapter 6 will deal with the numerical analysis of the nonstationary flow through the mixedflow pump. These forces are strongly related to the level of nonuniformity in the flow. Results of calculations are compared with model scale measurements of the pump performance. However. Finally.4 Outline of this thesis
1. An evaluation of several methods. A more detailed analysis of the streamtube will reveal some new insights into the forces acting on the installation in vertical direction. Comparisons are made with performance prediction software of Wärtsilä Propulsion Netherlands (WPNL). This will give insight in the governing parameters of the total propulsion system. Chapter 7 will show the results of the analysis of a complete full scale waterjet propulsion system. It is concluded that the typical nonuniform velocity distributions are inevitable in waterjet installations with flush mounted inlets. since the pump is not included in the model. thrust and power. Transient calculations with both uniform and nonuniform velocity distributions will show the presence of fluctuating radial forces. Chapter 4 deals with the choice of a mathematical method to analyse the flow through the system. Determination of this streamtube enables a more detailed analysis of the momentum distribution of the ingested water. Some connections will be made with standard propeller theory to show the similarities and the differences. will be presented.
Numerical analysis of a waterjet propulsion system
11
. based on an analysis of the development of the nonuniformity in the duct upstream of the impeller.
Chapter 1.J. Proceedings RINA Waterjet Propulsion conference.F. van. 1995 [2] Ramseye. Genoa. Proceedings RINA Waterjet Propulsion conference. Proceedings FAST2003 conference Vol I. Ischia... & Sims. 1630 [3] Roy.’Waterjet hull interaction’. pp. 50. 1994 [5] Bulten. T.C. Jahrbuch der STG 62.. R. 228258... ‘Manufacture of Saltpetre. London. Italy. Proceedings SATEC’96 conference. 1996
12
. & Verbeek. pp 3540. N..’. ‘The evolution of the modern waterjet marine propulsion unit’. S. Band 1968. Propulsors for the future’. PhD thesis. &c. Östergaard.. ‘Waterjets and Propellers. Nov.5
D F J · m n NVR Pshaft Q vship vin vout
Nomenclature
propeller diameter thrust advance ratio of propeller (= vship/nD) mass flow rate rotational speed nozzle velocity ratio (NVR = vout/vin) shaft power volume flow rate advance velocity of propeller advance velocity of jetengine jet velocity of jetengine m N kg/s 1/s W m3/s m/s m/s m/s
Greek symbols
ηp µ ρ
propulsive efficiency jet velocity ratio (= 1 / NVR) fluid density
kg/m3
1. C. David. S. English patent no. London. Raising Water. 1968 [7] Terwisga.. ‘Wasserstrahlantriebe für Hochgeschwindigkeitsfahrzeuge’. a shipbuilder’s view’. 1994 [4] Warren. N. Brandt.6
References
[1] Voulon. Italy. C. Introduction
1. N.M. Propelling Vessels.’Waterjet propulsion. session A2. H.’Design of optimal inlet duct geometry based on operational profile’.. 2003 [6] Kruppa. Delft.
1972 [12] Verbeek. Society of Naval Architects and Marine Engineers. R.H..’Application of waterjets in highspeed craft’.. E. Volume II. ‘Principles of naval architecture’. van den Boom (Editor) Elsevier Science Publication. Longman Group. 1998 [11] Cohen.F. Jersey City. ‘Gas turbine theory’. G.J. H. R. 1992
Numerical analysis of a waterjet propulsion system
13
.C..’Waterjet forces and transom flange design’.6 References
[8] Verbeek. MIT press. Rogers. London..V.N..I... 1994 [9] Newman. H. 1977 [10] Lewis. Model Tests and Reality. ‘Marine hydrodynamics’. London.1. Saravanamuttoo. H. J.J. RINA waterjet propulsion conference. in Hydrodynamics: Computations.
Chapter 1. Introduction
14
Chapter 2 Waterjet propulsion theory
In this chapter the basic principles of waterjet propulsion will be discussed. The equations of the waterjet theory will be based on standard nomenclature used in the description of pump performance. Where possible, equivalent nomenclature of commonly used propeller theory will be mentioned as a reference. In the first section some specific velocities, as used in waterjet theory, will be defined. These definitions form the basis for the remainder of the chapter. In the second section, the generally applied standard parameters are defined, which are used to describe the overall pump performance. In the commonly used waterjet propulsion theory, equations for the derivation of thrust of a waterjet propulsion system are based on open propeller theory. The transition from open propellers to waterjets will be reviewed in detail, in order to reveal possible deficiencies in the waterjet theory. The equations for the waterjet thrust can be coupled to the required pump head and flow rate. This will be discussed in section 2.4. It will be shown that a certain thrust can be achieved with different combinations of flow rate and pump head. Determination of the optimal combination of flow rate and pump head is obtained with the aid of the overall propulsive efficiency. This will result in the design operating point of the pump in the waterjet installation. In some conditions, the optimal pump operating point can not be reached due to severe cavitation in the pump. This limitation in the optimization process will be discussed in section 2.5.1.
Numerical analysis of a waterjet propulsion system
15
Chapter 2. Waterjet propulsion theory
In the selection of a waterjet installation for ship propulsion the weight of the installation is an important issue. To minimize the weight of the system, the size of the waterjets is selected as small as possible. The shaft speed of the pump is then maximised. It will be shown that for a given available power the minimum required pump size depends on the ship speed. The available power is governed by the installed diesel engine or gasturbine. This dictates the selection procedure to a large extent.
2.1
Characteristic velocities in a waterjet system
In the equations for pump performance and thrust, use is made of some specific velocities. Four main velocities are distinguished and will be used throughout this thesis: 1. ship speed (vship) 2. mass averaged ingested velocity at duct inlet (vin) 3. averaged axial inflow velocity at the pump entrance (vpump) 4. averaged outlet velocity at the nozzle (vout) Figure 2.1 shows a sketch of a waterjet installation with the four velocities indicated. A nonuniform velocity distribution is sketched to indicate the development of the boundary layer along the hull surface, upstream of the inlet. This figure is also used to give an impression of the dividing streamline. By definition there will be no mass flow across this line. In three dimensions, this line is extended to a dividing streamtube. The curved part of the inlet, where the streamline ends, is denoted as inlet lip or cutwater.
vpump
vout
l
vin
z
x
vship
Figure 2.1
Characteristic velocities in waterjet propulsion system
The inlet velocity is determined at a crossflow plane just upstream of the waterjet inlet, where the influence of the waterjet is not yet noticeable. The ingested velocity distribution is massaveraged over the crosssectional shape of the streamtube to find the actual inlet velocity vin: 1 v in =  v ( z )v n dA Q
∫
(2.1)
A
16
2.1 Characteristic velocities in a waterjet system
where v(z) is the velocity distribution in the boundary layer. The four velocities are related by three parameters; wake fraction, inlet velocity ratio and jet velocity ratio. These three parameters are discussed in detail in this section.
2.1.1
Wake fraction
The water that is ingested into the waterjet inlet channel partly originates from the hull’s boundary layer. The mass averaged velocity of the ingested water (vin) is lower than the ship speed due to this boundary layer. The velocity deficit is expressed as the momentum wake fraction (w), which is defined as: v in w = 1 – v ship (2.2)
Calculation of the wake fraction is rather complex, since the crosssectional shape of the streamtube is not known a priori. Experiments have revealed that the crosssection of the streamtube has a semielliptical shape under the hull [1]. This is often simplified by a rectangular box with a width of 1.3 times the pump diameter. Some comparisons have been made with experimental results [2], [3] and it is concluded that the resulting value for the wake fraction can be determined within acceptable limits, if the rectangular box approximation is used. For a given volume flow rate through the waterjet the height of the box can be calculated once the velocity distribution in the boundary layer is known. Standard theory for a flat plate boundary layer, as described in several textbooks ([4], [5]) can be used to get a first indication of the velocity distribution. It is convenient to use a power law velocity profile for the boundary layer velocity distribution: v z n  =  δ U∞
1 
(2.3)
where v denotes the local velocity in the boundary layer at a distance z normal to the wall, U ∞ the undisturbed velocity, δ the local boundary layer thickness and n the power law index. Besides the thickness of the boundary layer δ there are also derived quantities like the displacement thickness δ1 and momentum thickness δ2 of the boundary layer.
Numerical analysis of a waterjet propulsion system
17
5)
(2.4)
where l is the wetted length. full scale waterjet installations operate at Reynolds numbers of about 109. which is 2 orders of magnitude larger. Substitution of the power law velocity distribution in the definitions of the boundary displacement thickness δ1. A typical example is the ITTC’57
18
.6) gives an expression for the friction coefficient cf(l) as function of the boundary layer thickness δ(l) and the power law exponent n. The wall friction coefficient for n=7 becomes: c f ( l ) = 0. The wall friction coefficient for a flat plate cannot be based on equation (2. For turbulent flow a value of n = 7 is often used.4) and (2.8)
where Rel is the Reynolds number based on the wetted length. With aid of the analysis of developed turbulent pipe flow.9)
Comparison with experimental data shows good agreement for Reynolds numbers between 5x105 and 107 (see [4]).9) at these high Reynolds numbers.Chapter 2.= 2 2 dl 1 .7)
Combination of equations (2. Waterjet propulsion theory
The momentum thickness can be related to the wall friction coefficient cf(l) for a flat plate: dδ 2 τw c f ( l ) ≡ .0576 ⋅ Re l
–1 ⁄ 5
(2. Several logaritmic equations for the flat plate wall friction coefficient are defined for high Reynolds numbers.ρU ∞ 2 (2. the momentum thickness δ2 and the energy thickness δ3 results in a set of the following relations: δ δ 1 = n+1 δn δ 2 = (n + 1)(n + 2) 2δn δ 3 = (n + 1)(n + 3) (2. In general. an expression for the flat plate boundary layer thickness is derived: δ
n=7
= 0. This relation gives the frictional drag of the flat plate in terms of the development of the boundary layer.6)
(2.370 ⋅ l ⋅ Re l
–1 ⁄ 5
(2.
Velocity profiles are measured on two different vessels and at different ship speeds.8). the corresponding boundary layer thickness can be calculated once the wall friction coefficient cf(l) and the power law exponent n are established. which is commonly used to extrapolate the viscous resistance component of a model scale ship to full scale dimensions.2. the wake fraction obtained from the numerical results is more accurate than the one obtained from experimental data. is modified for n = 9: δ
n=9
= 0. The actual power law exponent n is determined from velocity profile measurements by Wieghart. Based on equations (2. the average incoming velocity and thus the wake fraction can be calculated. The accuracy of the rectangular box approximation will be reviewed in chapter 5 when CFD calculations of the flow through the waterjet inlet duct are discussed.
Numerical analysis of a waterjet propulsion system
19
.6). Adjustment of these values should result in the right boundary layer thickness and in an accurate prediction of the velocity profile and the wall friction. This is in accordance with measurements of Wieghardt (see [4]). A reasonable fit of a power law profile with n = 9 and the measured values is found. whereas the determination of the wake fraction in experiments is based on an approximated shape of the streamtube. With the numerical method it is possible to visualize the actual shape of the streamtube and determine the massaveraged velocity by numerical integration. Consequently. This results in a large variation of Reynolds numbers. there is a relation between friction coefficient cf(l). as given in equation (2. boundary layer thickness δ(l) and power law exponent n. This numerical method is based on the computed shape of the streamtube. With known boundary layer thickness and volume flow through the pump.4) and (2.10 to 0.14 for a fast ferry. Full scale measurements of the hull boundary layer velocity distribution are presented by Svensson [7].270 ⋅ l ⋅ Re l
–1 ⁄ 6
(2. Results of measurements at different Reynolds numbers are presented in Schlichting [4].10)
It can be noticed that both the constant as well as the power of the Reynolds number have to be changed when the value of n is changed. For a certain Reynolds number. A typical value for the wake fraction w is 0. The equation for the boundary layer thickness. For Reynolds numbers of order 109 the power law exponent becomes 10 to 11. The logaritmic friction line gives the wall friction coefficient as function of Reynolds number.1 Characteristic velocities in a waterjet system
friction line.
2
Inlet velocity ratio
The averaged axial inflow velocity of the pump is denoted by vpump.12) is preferred since the operating range is bounded between 0 and about 2. e. If the vessel sails at design speed.3 to 1.3 to 1.2 shows a sketch of the flow phenomena at low IVR condition. this results in values of this quantity.8. Waterjet propulsion theory
2. This velocity can be written as: Q v pump = π 2 .2 Flow phenomena at low IVR
20
. This might lead to cavitation and/ or separation in the inlet at the upper side of the lip.1. at operating conditions below 1 and a value of infinite for zero ship speed.12)
At normal operating condition. IVR is used to denote the flow conditions in the waterjet inlet duct. the design IVR will be around 1.8. Use of the definition in equation (2. In this condition the stagnation point of the dividing streamline is located at the hull side of the inlet lip (or cutwater). As mentioned.5.11)
where Q is the volume flow through the pump and Dinlet the diameter at the suction side of the pump. IVR values
Increased risk for cavitation and/or separation Figure 2. IVR will be smaller than 1. At relatively low ship speed. This means that the flow is accelerated upon entering the inlet duct. where the speed is changed from the ship speed to the pump velocity.12) is used in literature as well ([2]. The reciprocal of equation (2. This velocity is an important parameter to describe the flow phenomena in the inlet.Chapter 2.g.D inlet 4 (2. during manoeuvring in harbour. IVR will be around 1. Figure 2. the inlet flow phenomena are quite different. The pump velocity is related to the ship speed through the Inlet Velocity Ratio (IVR): v ship IVR = v pump (2. [8]) and used by the ITTC.
It should be kept in mind. because each vessel has to start from zero ship speed.3
Jet velocity ratio
The velocity vout at the outlet of the waterjet nozzle.3
Flow phenomena at high IVR
of more than 2.13)
The outlet velocity is related to the incoming velocity by the jet velocity ratio µ: according to [10]: v in µ = v out (2. With a good inlet design cavitation and separation free operation is possible up to about 44 knots [9].0 are known for high speed motor yachts (>60 knots). Whether or not cavitation or separation really occurs in a practical situation. This implies a significant deceleration of the flow in the inlet. which is a commonly used design speed for fast ferries. The critical location for cavitation is located at the hull side of the lip for this condition.14)
Numerical analysis of a waterjet propulsion system
21
. The possible flow phenomena at high IVR are sketched in figure 2. that an inlet has to be designed to cope with the low IVR and the design IVR condition.3. is related to the volume flow through the pump and the diameter of the nozzle as: Q v out = π 2 .
2.D nozzle 4 (2. If this pressure gradient becomes too large.2. strongly depends on the actual geometry of the inlet duct. flow separation is likely to occur at the top side of the inlet. In this condition the stagnation point is located at the inlet side of the cutwater.1 Characteristic velocities in a waterjet system
Increased risk of separation
Increased risk of cavitation
Figure 2. The deceleration of the flow in the inlet duct leads to an adverse pressure gradient in the inlet.1.
[12]. where the overall propulsive efficiency of the waterjet system is derived.16)
where H is the head in m. see for example [11].2
General pump theory
In this section a short overview of the standard pump theory is given in order to introduce a set of parameters to describe the pump performance. The performance is expressed in terms of flow rate. Elimination of the diameter D from equations (2.
2. vpump and vout. equations (2.15) and (2.15)
where Q is the flow rate in m3/s. The relations between these velocities are defined by three ratios: wake fraction w.16) results in: ΩQ ψ 1 = . It can be shown that geometrically similar pumps have equal values for flow and head coefficient.⋅ 3⁄4 1⁄2 ( gH ) ϕ
1⁄2 3⁄4
(2. inlet velocity ratio IVR and nozzle velocity ratio µ. It will be shown that typical values are in the range of 0. In dimensionless form. If the performance of a pump for a certain size and shaft speed is known. vin.17)
22
. This forms the basis of the socalled similarity method.
2. This theory can be found in many textbooks about centrifugal pumps.5 to 0.Chapter 2.
2. Waterjet propulsion theory
The importance of the parameter µ will be shown in section 2. The head coefficient ψ of a pump is defined as: gH ψ ≡ 2 ( ΩD ) (2.1. Ω the speed of the impeller in rad/s and D the impeller diameter in m.7. head and cavitation behaviour.4
Summary
In this section four velocities are introduced. vship.15) and (2.1
Dimensionless performance parameters
Performance of a pump can be expressed in terms of a set of nondimensional parameters.5. The theory of waterjet propulsion will be based on these velocities and ratios. the flow rate through the pump is given as the flow coefficient ϕ: Q ϕ ≡ 3 ΩD (2.2.16) can be used to predict the performance for different sizes and shaft speeds.
= P shaft ΩT q (2. Q the flow rate in m3/s and H the head in m. similarity of performance is only valid in cases of both geometrically and dynamically similar internal flows.18)
where Ω is the speed of the impeller in rad/s.0). which is the product of flow rate and pressure rise. It is also found that the similarity method implies that geometrically similar pumps have equal values of specific speed: ϕ n ω = 3⁄4 ψ
1⁄2
(2.21)
The specific power is related to the flow coefficient. In this analysis viscosity is not taken into account. and the required shaft power Pshaft.19)
The value of the specific speed of a specific pump gives a good indication of its type: typical axial flow pumps have a specific speed above 2.2. The required shaft power can be expressed in a nondimensional specific power P*: P shaft * P = 3 5 ρΩ D (2. head coefficient and pump efficiency. Pump efficiency ηpump is defined as the ratio between the hydraulic power Phydr. (2. Mixedflow pumps have intermediate values for the specific speed.2 General pump theory
which leads to the definition of the specific speed of: Ω Q n ω ≡ 3⁄4 (g ⋅ H) (2.20)
where Tq is the shaft torque. Since hydraulic losses do scale differently.15).16) and (2.22)
Strictly speaking. Combination of equations (2. additional empirical relations are used to predict the effect of these losses on pump efficiency and specific power.
Numerical analysis of a waterjet propulsion system
23
.4.20) yields:
* ϕψ P = η pump
(2. P hydr ρgHQ η pump = . whereas radial flow pumps have low values of the specific speed (typically below 1.
4 shows the expected maximum pump efficiency for three flow rates. The large similarities in pump geometry lead to comparable efficiencies for different pumps with the same specific speed. In a similar way.Chapter 2. The highest efficiency is found at a specific speed of 1.– 0.⋅ 2 ψ Q so that: ( gH ) δ ≡ D ⋅ Q and: ψ δ = 1⁄2 ϕ
1⁄4 1⁄4 4 2
(2. if the rotational speed Ω is eliminated: gHD ϕ 1 = . however. Decrease in efficiency is rather slow when the specific speed is increased to values above 1.23)
(2.0.0. In general. An example of such empirical formula is given in [15]:
2 0. the specific diameter δ is found. that the specific speed is found from the expressions for flow coefficient and the head coefficient.25)
The specific speed and specific diameter are based on the same two parameters.2.95 – . based on measured performances. waterjet pumps have a specific speed around 2. namely ϕ and ψ.
24
.2
Pump geometry parameters
It is shown in the preceding section. Waterjet propulsion theory
2.05 η pump = 0. Figure 2.03. The relation between the two is represented in the so called Cordierdiagram [14].26)
where Qref is set equal to 1 m3/s in order to maintain the nondimensional representation.0. when the diameter is eliminated. The statistically attainable optimal pump efficiency can be derived from several published prediction formulas.24)
(2.125 [ log ( n ω ) ] 3 Q⁄Q ref
(2. The basic geometry of the impeller of conventional centrifugal pumps is strongly related to the specific speed of a pump. Waterjet pump designs may deviate from this empirical rule for conventional pumps due to the difference in functionality as outlined in chapter 1. which is based on experience from actual pumps.
or suction side. pump operation is still possible beyond the cavitation inception level. Therefore the criterion for the inlet suction head is based on a certain loss of pump performance (for example 1 or 3% head loss or a certain percentage of pump efficiency decrease.e.5 2 2. In general.28)
25
. defined as: NPSH R σ = H
Numerical analysis of a waterjet propulsion system
(2. see [16]). which is a pressure expressed in meters water column.5 1 1. This is denoted with the inception net positive suction head (NPSHi).1 m3/s 60% Q= 1.2 General pump theory
100% 95% 90% 85%
ηpump [−]
80% 75% 70% 65% Q= 0. based on equation (2.5 3 3.0 m3/s 55% 50% 0 0. i.26). The required net positive suction head (NPSHR) can be made nondimensional in a similar way as the head to form the suction coefficient κ: gNPSH R κ = 2 ( ΩD ) (2.5 5 Q= 10 m3/s
Pump specific speed Nω [−] Figure 2.
2. wellknown method to present the NPSH in dimensionless representation is the Thoma number. for lower NPSH levels. Based on the choice for the admissible head loss.
Achievable pump efficiencies around 90% for large pumps seem to be a reasonable estimate.2.27)
Another.5 4 4. This value will be used in the remainder of this chapter for estimates of the overall waterjet efficiency.3
Cavitation parameters
For cavitation free operation the pump requires a certain available pressure at the inlet.4 Maximum pump efficiency as function of specific speed.2. a required NPSH is defined.
where in the waterjet IVR is introduced as an additional parameter. In a similar way. defined as: Ω Q n ωs = 3⁄4 ( g ⋅ NPSH R ) (2. In order to create some extra margin to accommodate cavitation. Waterjet propulsion theory
The nondimensional parameters are related by the head coefficient.29)
The suction specific speed of a pump is more or less constant for all pump types.Chapter 2.5 for a waterjet impeller is adopted.
2. due to the infinitesimal thickness of the actuator disk. the nondimensional head can be related to the thrust coefficient of an open propeller.30)
This relation shows the fundamental difference between an open propeller and a waterjet installation. Values of about 4. For a propeller the thrust coefficient is defined as [18]: T K T = 2 4 ρn D (2.4
Correlation with propeller performance parameters
The flow coefficient ϕ of a pump can be related to the propeller advance ratio J (as given in eqn.2.= 4ΩD 8IVR (2.0 are common in commercial pumps [1417]. The required NPSH can also be related to the flow rate and the rotational speed of the impeller.1)) with substitution of equations (2.31)
The head H of a pump is related to the total pressure increase generated by the impeller according to: ∆p tot = ρgH (2.33)
It is assumed that the static pressure rise is equal to the total pressure rise.32)
In actuator disk theory the production of thrust of an open propeller equals the product of the pressure rise and the crosssectional area of the propeller: T = ∆p ⋅ A prop (2. This gives the suction specific speed of the pump nωs. a design value of 3. This results in a relation
26
.11) and (2. This parameter is needed because of the principle of internal flow of the pump compared to the external flow of the propeller.12): πv pump J ϕ = . (1. similar to the pump specific speed.
2.3 Thrust
between the head coefficient ψ of a pump and the thrust coefficient KT of a propeller: KT ( ∆p ) ⁄ ρ ψ =  = 2 2 2 3 4π n D π (2.34)
It is concluded that the QH curves of a pump are equivalent to the JKT curves of an open propeller. The main difference is caused by the used inflow velocity.
2.3 2.3.1
Thrust General thrust equation
The purpose of a propulsion installation is to produce thrust to propel a vessel. Water is accelerated in the installation, which results in a reaction force on the ship structure. The thrust can be derived from the momentum balance for an incompressible fluid [5]: F = Fs + Fb = ∂ vρ dV + v ρv ⋅ dA ∂t
∫
∫
(2.35)
V
A
The momentum balance states that the sum of all surface forces Fs and all body forces Fb acting on the spatially fixed control volume V equals the rate of change of momentum in the control volume with surface A. The surface force is defined as: Fs =
∫ ( – p I + σ ) ⋅ dA
A
(2.36)
where p is the static pressure, I the unit tensor and σ the viscous stress tensor. In the remainder of this section the steady flow situation will be analysed. As a consequence, the first term on the right hand side of equation (2.35) vanishes. Moreover, the body forces, like gravity, acting on the fluid will be neglected. In the following subsections the momentum balance will be derived for both an open propeller and a waterjet.
Numerical analysis of a waterjet propulsion system
27
Chapter 2. Waterjet propulsion theory
2.3.2
Open propeller thrust
An expression for the thrust of an open propeller is determined with equation (2.35) [6]. The propeller is treated as an actuator disk, which is a singularity modelled by a body force acting over an infinitesimal thin disk. The control volume consists of the streamtube of fluid which passes through the propeller plane area. Figure 2.5 shows a sketch of the control volume of an open propeller with the nomenclature of the velocities. Evaluation of the momentum balance is split in two parts; the contribution of the momentum fluxes and the contribution of the surface forces. The contributions of the momentum fluxes in xdirection result in a net momentum flux component in xdirection of: φ mx = ρv out A out – ρv in A in This can be rewritten, with aid of the continuity condition, as: φ mx = ρv prop A prop ( v out – v in ) The contributions of the surface forces in xdirection are defined as: F x = – T prop – (2.38)
2 2
(2.37)
∫ ( p – p∞ ) dA + ∫
A in A o ut
( p – p ∞ ) dA +
∫
A tub e
( p – p ∞ )x ⋅ dA
(2.39)
It is assumed that the pressure at the inlet (far upstream) and at the outlet (far downstream) is equal to the ambient pressure p ∞ . Moreover, the contribution of the viscous forces is neglected on the inlet and outlet area as well as on
vprop p = p∞ vout Aout
z
vin
ptube
Ain Atube p = p∞
x
Figure 2.5
Control volume for the momentum balance applied to an propeller within a streamtube
28
2.3 Thrust
the streamtube surface. Combination of equations (2.38) and (2.39) gives the final thrust equation for an open propeller, based on the momentum balance: T prop = ρA prop v prop ( v out – v in ) –
∫
A tube
( p – p ∞ )x ⋅ dA
(2.40)
where Aprop is the crosssectional area of the propeller plane, x the unit vector in xdirection and Atube the streamtube surface. The contribution of the pressure acting on the streamtube to the thrust vanishes, based on the paradox of d’Alembert, if the streamlines are aligned in xdirection far upstream and downstream. If Bernoulli’s theorem is applied along the streamlines in the part of the control volume upstream and downstream of the propeller, a second relation for the propeller thrust is found: 1 2 2 T prop = ∆p ⋅ A prop = ρA prop ⋅  ( v out – v in ) 2 (2.41)
Combination of the momentum balance and Bernoulli’s law, leads to a simple relation between the inlet and outlet velocity and the volume flow through the propeller disk (see [18]): 1 v prop =  ( v in + v out ) = v in + v ind 2 (2.42)
It can be seen that the velocity through the disk is the average of the upstream and downstream velocities. The difference between the velocity through the disk and the incoming velocity is called the induced velocity vind. Thrust loading coefficient Loading of an open propeller is often expressed by the propeller loading coefficient, defined as [18]: T prop C Tprop = 1 2  ρv ship A prop 2 (2.43)
where Aprop is the crosssectional area of the propeller disk, based on the propeller diameter. The propeller loading coefficient can be expressed in terms of the ratios as defined in section 2.1. Substitution of equation (2.41), with the inflow velocity equal to the ship speed, i.e. vin=vship, yields:
Numerical analysis of a waterjet propulsion system
29
Chapter 2. Waterjet propulsion theory
C Tprop
1 2 2  ρ ( v out – v in )A prop 2 v out 2 2 1–µ =  =  – 1 =  v in 2 1 2 µ  ρv in A prop 2
(2.44)
With µ<1 the propeller loading coefficient is thus directly related to the jet velocity ratio. The jet velocity ratio can be related to the IVR, if equation (2.42) is substituted into equation (2.12): 2v in v in 2µ IVR =  =  = v prop v in + v out µ+1 (2.45a)
It can be seen that open propellers always operate at IVR values below 1. After rearranging this equation, it is shown that the IVR is equal to Froude efficiency as given in equation (1.3): 2 2 IVR =  =  = η p 1 1 + ( v out ⁄ v in ) 1 + µ (2.45b)
Although the term IVR is not used in the theory for open propellers, it is already present as the Froude efficiency.
2.3.3
Waterjet thrust
For the determination of the thrust of a waterjet installation in general the same approach as for the open propeller is used. The control volume will be bounded by the streamtube surface on one side and the solid wall on the other side. It is assumed that the inlet and exit planes are perpendicular to the xdirection and the hull is parallel to the xaxis. Figure 2.6 shows the control volume and the contributing terms to the momentum balance. The forces acting on the waterjet structure, which are included in this control volume, are denoted as Twj,tube. It is noted that the control volume based on the streamtube of the ingested water does not take into account the part of the waterjet inlet structure at the hull side near the cutwater lip, which is excluded from the streamtube control volume. The thrust or drag on that part of the waterjet structure will be denoted will Twj,hull. At high IVR conditions a significant part of the cutwater geometry belongs to the excluded cutwater region. The subdivision of the complete waterjet inlet structure into the part, which is included in the streamtube approach, and the part which is excluded is shown in figure 2.7.
30
7 Subdivision of all solid wall cells of the waterjet installation into group belonging to streamtube control volume (left) and group of remaining cells on hull (right)
The total thrust Twj.2.3 Thrust
Aout vout p = p∞ Ahull Awj
z
x
ptube
vin Atube
Ain p = p∞
Figure 2.6
Control volume for a momentum balance on the streamtube of the ingested water of a waterjet installation
all solid wall cells
remaining hull cells
streamtube solid wall cells Figure 2.tube + T wj .hull (2.all of a waterjet is therefore: T wj . these are the fluxes at the nozzle exit surface Aout and at the plane Ain upstream of the inlet:
Numerical analysis of a waterjet propulsion system 31
.46)
Application of the momentum balance for a waterjet learns that there are two momentum flux terms that contribute to the force in xdirection.all = T wj .
50)
The contribution of the streamtube pressure can not be quantified analytically.50) is substituted in equation (2.48)
where Q is the flow rate through the waterjet installation. though there is a nonuniform velocity distribution present at the inlet plane Ain. With equations (2.tube = ρQ ( v out – v in ) –
∫
A tube
( p – p ∞ )x ⋅ dA
(2.Chapter 2. The contributions of the surface forces in xdirection are defined as: F x = – T wj . while the pressure levels at the inlet (far upstream) and at the outlet (far downstream) are equal to the ambient pressure p ∞ . when equation (2. it is assumed that volumetric forces and viscous forces can be neglected.tube –
∫ ( p – p∞ ) dA + ∫
A in A out
( p – p ∞ ) dA +
∫
A tub e
( p – p ∞ )x ⋅ dA (2.48) and (2. The thrust of the complete waterjet installation is found.49)
Similar to the open propeller. which yields: T wj . Even with numerical methods it is a very complex task to determine this value.51)
The last two terms on the righthandside are assumed to be small compared to the first term.47)
where vin is the mass averaged inflow velocity. In chapter 7 the contribution of the streamtube pressure term will be reviewed in more detail. With aid of the continuity condition.49) can be combined to get the expression for the waterjet thrust in xdirection based on the streamtube momentum balance: T wj . Contribution of this shear stress force is assumed to be negligible.hull
(2. The
32
. Waterjet propulsion theory
φ mx = ρv out A out – ρv in A in
2
2
(2.46). Effect of the viscous forces is neglected also on these two planes. due to the threedimensional shape of the streamtube surface and the dependency of the shape on IVR. since the shape of the streamtube and the pressure distribution are unknown. this becomes: φ mx = ρQ ( v out – v in ) (2.all = ρQ ( v out – v in ) –
∫
A tube
px ⋅ dA + T wj . and often neglected in waterjet propulsion literature.
The wake fraction becomes zero. i.e. the relation between jet velocity ratio and the thrust loading coefficient becomes: T 2( 1 – µ)(1 – w ) C Tnozzle = .2. The resulting simplified thrust equation for a waterjet becomes [10]: ρQ T wj = ρQ ( v out – v in ) = .3 Thrust
influence of this simplification will be addressed in more detail in chapter 7. In this way the dimensions of the complete installation are recognised more clearly. when the inflow velocity is equal to the ship speed. where the propeller diameter is used. this simplified equation can be used to explain the main theory on waterjet propulsion.( 1 – µ ) A nozzle
2
(2. This is due to the fact that a waterjet is an internal flow machine. whereas it is related to the thrust for an open propeller.
Numerical analysis of a waterjet propulsion system
33
. vin=vship.= 2 1 2 µ . This approach is more in agreement with the open propeller thrust loading coefficient. Thrust loading coefficients The thrust loading coefficient of a waterjet installation can be based on the nozzle outlet area or the pump inlet area. The resulting loading coefficient for a waterjet with undisturbed inflow yields: C Tnozzle 2(1 – µ ) = 2 µ (2.54)
w=0
Comparison with the open propeller thrust loading coefficient (equation (2. the nozzle exit area Anozzle and the jet velocity ratio µ. The waterjet thrust loading coefficient can also be based on the pump inlet diameter. This is equivalent with an open water test of a propeller with uniform inflow. This equation shows the three main parameters of a waterjet propulsion system: the volume flow rate Q through the system. The thrust loading coefficient based on nozzle exit area is discussed in [13].ρv ship A nozzle 2
2
(2. For a waterjet the ratio between the inlet and nozzle area is fixed.52)
Despite neglecting the streamtube and hull surface forces.53)
where w is the wake fraction according to equation (2.2).44)) reveals a difference between the waterjet and the open propeller. With the nozzle area as reference area.
the pump head curve. For conventional pumps the axial thrust is to be kept as low as possible.
34
.
2. Since it is an internal flow machine. In such condition a negative thrust acts on the transom stern or the inlet ducting. can be used to evaluate the performance of a waterjet installation. Thrust production is not regarded as an important performance indicator. In case of a waterjet. For the sake of simplicity.3.Chapter 2. These can be related to the mass flow for a given geometry of the waterjet installation.ρv ship A pump 2
2
(2. as used in the examples in this section. The volume flow rate through the system follows from the intersection of the required system head curve and the pump head curve. The head H of a pump represents the increase of total pressure in a pump measured in meters liquid water column as given in equation (2.55)
The thrust loading coefficient based on the pump inlet diameter shows that the IVR is introduced to describe the system performance.32). This gives the designer of waterjets another optimization option. On the other hand. In this way the standard pump performance characteristics. it can also appear that the thrust acting on the shaft will exceed the total thrust of the installation [19]. like head curve. if the values for the velocities vin.
2. mixedflow pumps have a headcurve with a negative slope in the design point to ensure a stable operating point. efficiency and cavitation behaviour. the thrust can be calculated.= 1 2 IVR ⋅ µ . In general. This mass flow through the system is related to the pump head. is assumed to be a linear function of flow rate. The pump head curve depends on the type of pump used in the waterjet system. compared to open propellers. part of the thrust can be transferred to the hull structure via the transom stern and the inlet ducting.4
Concluding remarks
In a waterjet there is no direct relation between the IVR and µ like there is for an open propeller.4
Pump head
The required head of a waterjet installation will be discussed in this section. like efficiency and head as function of the mass flow. Waterjet propulsion theory
T 2( 1 – µ)( 1 – w) C Tpump = . vpump and vout are known. For lower volume flow rates the slope may become zero or even negative.
4 Pump head
The required system head curve can be regarded as a pipe resistance curve of the waterjet installation.( 1 – ε ) + h j 2g 2g
2 2
(2. the elevation hj can be neglected relative to the other contributions in equation (2. All contributions together give the equation for the required system head HR: HR v out v in = .58)
The difference between the two methods of averaging is less than 1% for a powerlaw exponent of n=9.56)
where φ is the nozzle loss coefficient. which will require some more pump head. therefore the system performance is coupled to the ship speed. Finally. In general.( 1 + φ ) – .8 shows an example of a pump head diagram with a pump head and efficiency curve and two system lines for a constant ship speed and different nozzle sizes. Additional head is required to overcome the hydraulic losses in the inlet and the nozzle. the average ingested velocity vin should be based on a mass averaged dynamic pressure term:
2 1 v˜ = . The acceleration of the fluid in the nozzle requires a certain pressure difference.1).57)
A
whereas vin in equation (2. At constant ship speed.2. Equation (2. The error will be even smaller if the water is ingested from the undisturbed fluid.56). ε the inlet loss coefficient and hj the nozzle elevation above the waterline. the required system head HR can be approximated as a quadratic function of the flow rate Q. The assumption of constant
Numerical analysis of a waterjet propulsion system 35
.= 2 (n + 3)(n + 1) v in
(2. which is beneficial for the head requirement. In general. the waterjet nozzle may be positioned above the waterline. The elevation of the nozzle is limited by the selfpriming requirement of the waterjet installation. Figure 2. Strictly speaking. The difference between the two methods can be expressed in the powerlaw exponent.56) shows a positive contribution from the incoming velocity.56) is based on the mass averaged velocity as given in equation (2. assumed that the water is ingested completely out of the boundary layer:
2 2 v˜ (n + 2) in . the introduced deviation is compensated for in the determination of the loss coefficient. the energy of the ingested fluid can be used partly. However.v ( z ) v n dA in Q
∫
1⁄2
(2. The slope of this quadratic curve depends on the nozzle diameter.
2 0.
8 Pump_head 7 System_head_nozzle1 6 5 4 System_head_nozzle2 Work point Pump efficiency
100% 90% 80% 70% 60% 50% 40% 30%
3 2 1 0 1 2 0 0. An increase of the nozzle diameter.4
INCREASING NOZZLE DIAMETER
20% 10% 0%
ϕ [−] Figure 2. The effect of the ship speed on the volume flow through the installation is shown in figure 2.9. Waterjet propulsion theory
ship speed is a hypothetical condition.In this pump head diagram. that the increase in volume flow between 20 and 40 knots is only 6%. It can be observed from this diagram.8 Pump head diagram for different nozzle sizes and constant ship speed
36
ηpump [−]
ψ [−]
. The nozzle size is kept constant in this figure.4 0.8 1 1. This increase in velocity through the impeller results in a small change of the pump operating point. This leads to a lower required head. results in a lower nozzle velocity for constant volume flow. the ship speed will depend on the delivered thrust.Chapter 2.6 0. which is in turn related to the flow rate. The system lines are based on different nozzle diameters. The work point of the pump can be controlled by the size of the nozzle exit area. since in actual situations. It is concluded that a waterjet installation can operate in a relatively small range of flow rates. 30 and 40 knots are plotted.2 1. the system lines of ship speeds of 20.
2.9
Pump head diagram for different ship speeds
2. a low pressure region at the rear of the vessel is created.4
Figure 2. If the propulsion system is regarded as a black box.6
0%
ϕ [−]
0.2
1. The overall propulsive efficiency ηd of this black box is then based on the bare hull resistance Rbh of a vessel [18]: R bh ⋅ v ship η d = P shaft (2. which results in an increased drag of the vessel.59)
where Rbh is the bare hull ship resistance and Pshaft the power at the waterjet shaft. the resistance of a ship with an active propeller is found to be different from the bare hull resistance.4 0.60)
Numerical analysis of a waterjet propulsion system
ηpump [−]
37
5
work point
70%
.8
1
1. then engine power Pshaft is input and thrust T at a certain ship speed is output.2 0. In conventional naval architecture theory. Due to the action of the propeller.5
Overall propulsive efficiency
This section deals with the influence of the parameter µ on the overall propulsive efficiency.5 Overall propulsive efficiency
8
100%
Pump_head
7
system_head_20 knots system_head_30 knots system_head_40 knots
90%
6
80%
Pump efficiency
ψ [−]
4
60%
3
50%
2
40%
1
30%
0
INCREASING SHIP SPEED
20%
1
10%
2 0 0. The difference between the bare hull resistance Rbh and the required thrust T at a certain ship speed is expressed in terms of the thrust deduction factor t according to: R bh = ( 1 – t ) ⋅ T (2.
90) and outlet loss coefficient is φ=0.62) shows that the overall propulsive efficiency is mainly a function of the jet velocity ratio µ.61)
In the next step.75 depending on the inlet loss coefficient ε. In waterjet propulsion theory. This efficiency is defined already as Froude efficiency in equation (2. which results in a stable working point.Chapter 2. Substitution of equations (2. which is slightly below the best efficiency point. After rearranging of all variables.65 to 0.
38
. equations (2. if the jet velocity ratio is in the range of 0. Figure 2.56) are substituted into equation (2.61). Also plotted is the ideal efficiency.12. Thrust deduction is set to t=0.52) and (2. since the hull efficiency ηhull and the pump efficiency ηpump as well as the inlet and nozzle loss coefficients may be regarded as constant values in a first approximation. Waterjet propulsion theory
For a propeller the thrust deduction factor is always positive. pump efficiency is 90% (ηpump = 0. when the inflow conditions show some variation. the thrust deduction factor t can be used to account for the effects of (i) the neglected surface forces such as the force on the streamtube and the force on the region aft of the waterjet inlet and (ii) a change in the pressure distribution along the hull. the equation for overall propulsive efficiency becomes: (1 – t) 2µ ( 1 – µ ) η d = .η pump (1 – w) ρgHQ (2. (2.62)
Eqn (2.63) (2. which leads to a higher ship resistance due to the action of the propeller and therefore a higher required thrust.20) and (2.59) gives: T ⋅ v in (1 – t) η d = . wake fraction to w=0. In general the design point is chosen at a jet velocity ratio.10 shows the overall propulsive efficiency for three inlet loss coefficients.⋅ η pump ⋅ 2 (1 – w) (1 + φ) – µ (1 – ε) where the first term is denoted as hull efficiency: (1 – t) η hull = (1 – w) (2. This approach is used by Van Terwisga [2].2).60) in equation (2.14).45) for an open propeller. where all losses are neglected. where a jet thrust deduction factor tj and a resistance increment factor 1+r are introduced. It is obvious that the optimum propulsive efficiency can be obtained. This part of the curve is relatively flat. (2.02.02.
Due to selfpriming constrains of the pump.6
ideal case
. this elevation is negligible in most cases.4 0.2 0.1
Cavitation margins
The waterjet pump needs a certain level of the pressure at the suction side of the pump in order to prevent cavitation.4 0.8 0.20 . pump efficiency ηpump=90%. The NPSHA can be expressed as function of the ship speed: p ∞ – p v v ship p ∞ – p v v in 2 NPSH A = .3 0. nozzle loss coefficient φ=0.8 0.65)
Numerical analysis of a waterjet propulsion system
39
.2. ε = 0.2. Pump operation is allowed as long as the available NPSH exceeds the required NPSH.+ .7 0. ε = 0.5. ε = 0.10 .( 1 – ε ) ( 1 – w ) – h j ρg ρg 2g 2g
2 2
(2.5 0. which is introduced in section 2.02
2.1 0.02.1 0 0 0.6 0.2 0. The pump will perform well as long as the required inlet suction head (NPSHR) is below the available inlet suction head (NPSHA): NPSH R ≤ NPSH A (2. For a waterjet installation the available suction head is determined by the waterjet operating point. The available suction head is the total head at the inlet of the pump minus the vapour pressure of the liquid.3 0.7 0.9 0.3.+ .( 1 – ε ) = .10 Overall propulsive efficiency ηd as function of jet velocity ratio µ.64)
where hj represents the elevation of the pump above the waterline.30
increasing inlet loss coefficient ε
ηd [−]
0. wake fraction w=0.5 Overall propulsive efficiency
1 0. This required pressure is expressed in the required net positive suction head (NPSHR).9 1
Jet velocity ratio µ [−] Figure 2.12. Thrust deduction factor t=0.5 0.
wake fraction w = 0.5 1 0. with the requirement of equation (2.64).5.5.5 0 20 25 30 35 40 45
increasing jet velocity ratio µ allowable specific speed nω
50 55 60 65 70 75
Ship speed vship [knots] Figure 2.Chapter 2.5 .
5 4.67)
This equation shows that the allowable specific speed of the pump will be limited for constant jet velocity ratio µ.
Pump specific speed nω [−]
4 3.20.66). Inlet loss ε = 0.11 for different values of the jet velocity ratio.
40
.11 Maximum allowable pump specific speed as a function of ship speed for various jet velocity ratios. This phenomenon is illustrated in figure 2. Waterjet propulsion theory
Combination of equations (2. µ=0. ≤ n ωs  H H (2.29).5 3 2.7 0 . .6 5 . the wake fraction w and the inlet and outlet loss coefficients ε and φ. a function of the ship speed and the jet velocity ratio is found for the maximum allowable specific speed: 2 ( p∞ – pv ) . outlet loss φ = 0. µ=0.56) and available suction head (2.5 2 1.65) yields: NPSH R 3 ⁄ 4 NPSH A 3 ⁄ 4 n ω = n ωs . when the ship speed increases.66)
2. Given values for the suction specific speed nωs.18) and (2.02.2
Limitations in specific speed
The expressions for pump head (2.12 and suction specific speed nωs = 3.+ ( 1 – ε ) 2 2 ρv ship ( 1 – w ) n ω ≤ n ωs ⋅ 1+φ  – ( 1 – ε ) 2 µ
3⁄4
(2. with negligible pump elevation. can be substituted into equation (2.
2. as shown in figure 2.12
Minimum allowable jet velocity ratio as a function of ship speed for various pump specific speeds.67).7 is possible up to 35 knots for a pump with a specific speed of 4. however.8 0.3
Limitations in jet velocity ratio
In practice. to cover the complete speed range waterjet manufacturers use a set of standard pumps with different specific speeds.5. Since this is undesirable.7 0. the range of possible jet velocity ratios can be determined for a given design ship speed.12 shows the minimum allowable jet velocity ratio µ for a number of specific pump speeds nω.78 at 65 knots.
Figure 2. Figure 2.0 requires a minimum jet velocity ratio of 0. a pump with lower specific speed is to be used.4 0.6 0. 0 .0 can be used at 65 knots for this jet velocity ratio.1 0 20 25 30 35 40 45 50 55 60 65 70 75
increasing pump specific speed nω
.
nω=3.5. wake fraction w = 0. A pump with a specific speed of 3. nω=4.5 0.10. A jet velocity ratio of 0.5 Overall propulsive efficiency
For a given design speed and a chosen jet velocity ratio the maximum allowable pump specific speed can be determined. Inlet loss ε = 0.9
allowable jet velocity ratio µ
Jet velocity ratio µ [−]
0.2 0. A pump with a specific speed of 4. This condition may be at the right side of the optimum propulsive efficiency curve.02.3 0.2. for a certain available pump type.0) compared to an axial flow type pump. It is also shown that the range of allowable jet velocity ratios at very high ship speeds (>60 knots) increases significantly for a radialflow type pump (nω=2.
Numerical analysis of a waterjet propulsion system
41
.
1 0. On the other hand.0. This diagram shows that the optimal jet velocity ratio µ can be selected over a large range of ship speeds with a limited number of different pumps.12 and suction specific speed nωs = 3. The allowable jet velocity ratio for given specific speed can be determined after rearranging equation (2. outlet loss φ = 0. Ship speed vship [knots]
. with a known specific speed.20.
Waterjet propulsion theory
2. it is found that the allowable tip speed increases with ship speed.22)): P shaft * 3 . Since the available NPSH is a function of the ship speed.12. This gives a relation between the required NPSH and the tip speed.= ρP ( ΩD ) 2 D with (ΩD) twice the tip speed of the impeller: 1 v tip = πnD = . nondimensional flow coefficient ϕ = 0.5. the allowable power density P/D2 increases also with increasing ship speed. This can be achieved with highly loaded.2. the minimum allowable impeller diameter as function of the ship speed can be determined
10000
Power density P/D2 [kW/m2]
9000 8000 7000 6000 5000 4000 3000 2000 1000 0 20 25
Maximum power density
OPERATION ALLOWED
30
35
40
45
50
55
60
65
70
75
Ship speed vship [knots] Figure 2. Effect of the ship speed on the allowable power density is shown in figure 2.69) (2. small installations.
42
. Consequently.68)
Cavitation behaviour of different pump sizes with the same specific speed can be compared with the net positive suction head coefficient κ.0 and suction specific speed nωs = 3. specific speed ns = 3.13. This criterion can be expressed in terms of the power density P/D2 given in terms of the specific power P* (according to equation (2.13 Maximum allowable power density as function of ship speed. as defined in equation (2. Consequently.Chapter 2. Inlet loss ε = 0.5.4
Limitation of power density
Weight reduction of a waterjet installation is an important issue.20.ΩD 2 (2.27). It should be obvious that an increase of the impeller diameter reduces the power density for given engine power. wake fraction w = 0.
The larger waterjets. The resistance at the hump speed is about equal to the maximum thrust of the small waterjet. etc. This means that the vessel has no extra thrust available to accelerate. at which the local maximum resistance occurs. In many cases waterjet installations have multiple operating points. the ship speed and the power applied to the pump.2. The hatched area represents the noncontinuous operational region. The speed. It is found that the minimum allowable size depends on the cavitation behaviour of the pump. The upper line represents the thrust breakdown line and the lower line of this area is denoted as the 1% cavitation line.14 shows the output of a waterjet performance prediction program as used at the authors’ company for three different sized waterjets. To get a good balance between cavitation performance and weight of the installation. The maximum speed will not exceed 25 knots in this case. For a realistic comparison of various installations with different sizes over the complete range of ship speeds. have sufficient margin up to the design speed. The thrust is kept constant and the size is changed to show the effect of power density on the cavitation margins. This requirement may lead to a larger waterjet than necessary for the design operating point. Thrust breakdown occurs when the mass flow through the system collapses due to extreme cavitation.6
Waterjet selection
Up to this point. such program should be used in order to take the actual values for the loss coefficients and the efficiencies into account.6 Waterjet selection
from the power density. have been considered as constants. If the waterjets are applied for example in large fast ferries or planing hulls.. the entire analysis has been based on optimal performance at one single design operating condition. In the same figure an indication of a typical resistance curve of a fast ferry is plotted. In this region severe cavitation in the waterjet will be present.
Numerical analysis of a waterjet propulsion system
43
. the resistance line is quite different from the one of a regular displacement vessel. Implementation of all of these dependencies will result in a complex waterjet performance prediction program. The increased resistance at the hump speed of 25 knots can be noticed. inlet losses. The waterjet installation has to provide sufficient thrust to exceed the resistance at the hump speed. Figure 2. In actual installations all these parameters depend on the ship speed and/or the flow rate Q through the installation. In all shown examples. the midsize jet will be selected for this application. is denoted as hump speed. The effect of waterjet size is obvious from this figure. pump efficiency ηpump. with lower power densities. values for the wake fraction µ. which is only 55% of the design speed. There is an additional resistance at a speed range between about 20 and 30 knots.
2.
14 Maximum thrust curves for three different waterjet sizes and resistance curves for fast ferry. This inflow might have an influence on the empirical values or the pump performance itself. where the complete waterjet installation is analysed numerically.
2. with variations in radial and tangential direction.7
Closing remark
All equations and empirical values in this chapter are based on a uniform inflow to the pump. Effects of the nonuniformity on pump performance will be reviewed in more detail in chapter 7. Waterjet propulsion theory
180% 160% 140% 120%
3000 kW/m2 5000 kW/m2
Thrust [%]
100% 80% 60% 40% 20% 0% 0 5 10 15 20 25 30 35
7000 kW/m2
P/D2 kW/m^2 =7000 P/D^2 = 5000 2 kW/m kW/m^2 P/D^2 = 3000
P/D^2 = 7000 kW/m^2
P/D2 =5000
40 45 50
Resitance
Ship speed vship [knots] Figure 2. the inflow velocity field of a waterjet shows a strong nonuniform distribution. However.8
A cf(l) D F g H hj IVR l NPSH
44
Nomenclature
area local wall friction coefficient diameter force gravitational acceleration pump head nozzle elevation inlet velocity ratio (vship/vpump) wetted length Nett positive suction head m2 m N m/s2 m m m m
.
2.Chapter 2.
2.8 Nomenclature
nω nωs n n P P* p Q R Rel T t U∞ v w z
pump specific speed suction specific speed propeller/pump shaft speed boundary layer power law exponent power specific power pressure volume flow rate resistance Reynolds number based on length thrust thrust deduction factor undisturbed free stream velocity velocity momentum wake fraction distance normal to the wall specific pump diameter boundary layer thickness boundary layer displacement thickness boundary layer momentum thickness boundary layer energy thickness inlet loss coefficient momentum flux nozzle loss coefficient efficiency flow coefficient net positive suction head coefficient jet velocity ratio fluid density Thoma number wall shear stress angular velocity head coefficient inflow plane
1/s W N/m2 m3/s N N m/s m/s m m m m m N kg/m3 N/m2 rad/s 
Greek symbols
δ δ δ1 δ2 δ3 ε φ φ η ϕ κ µ ρ σ τw Ω ψ
Subscripts 1
Numerical analysis of a waterjet propulsion system
45
.
& Walker.’Waterjet hull interaction’. pp. John Wiley & Sons.all wj. 1999 [9] Bulten.J. in Hydrodynamics: Computations..Chapter 2. Purnell. Delft University. S.hull wj. Stricker. R.B. ‘Analysis of hull boundary layer velocity distributions with and without active waterjet inlets’.J. McGrawHill.L. MIT press. H.G. 1968 [5] Fox. R.. PhD thesis.. pp 3540. London. Gowing. 853868.G.N. R.’Application of waterjets in highspeed craft’. J. 1998 [8] Seil.J.J.’Trial result including wake measurements from the world’s largest waterjet installation’.J. ‘Boundary layer theory’. H. Chesnakas.. 1977 [7] Svensson. 1998 [2] Terwisga. RINA Waterjet IV conference.C. Proceedings RINA Waterjet Propulsion II conference.W.J. R. J. 1985 [6] Newman. session A2.. New York. M. van den Boom (Editor) Elsevier Science Publication. Cambridge.. Seattle. Proceedings FAST’99 conference. J. 2003 [10] Verbeek.. Proceedings FAST2003 conference Vol I. G. & Grossi. 2004 [4] Schlichting. New York.. ‘Marine hydrodynamics’. Proceedings RINA Waterjet Propulsion II conference.T. 1992
46
. Becnel..’Design of optimal inlet duct geometry based on operational profile’. A. Amsterdam.1996 [3] Wilson. & Verbeek. ‘Development of waterjet inlets for 100 knots’.tube
available inflow induced outlet / nozzle propeller required streamtube complete waterjet structure excluded part of waterjet streamtube control volume included part of waterjet streamtube control volume
2.9
References
[1] Roberts.’Introduction to fluid mechanics’.. G. van. L. Amsterdam.. Waterjet propulsion theory
A in ind out prop R tube wj.. Third Edition. N. J.. Model Tests and Reality.’Boundary layer ingestion in flush waterjet intakes’. T. Ischia. Italy. A. & McDonald. C...
pp.. design and application’. Society of Naval Architects and Marine Engineers. New York. London. G. Published by Nijhuis Pompen. Seattle. D. 1998 [19] Verbeek.. Würzburg. ‘Fluid mechanics of turbomachinery’.J. theory. M. 1999 [17] Os. R.2. 879882. design and operation of rotodynamic pumps’. J.. 1994
Numerical analysis of a waterjet propulsion system
47
.’Strömungsmachinen (Aufbau und Wirkungsweise)’. 2005 [16] Gülich. John Wiley & Sons. ‘Centrifugal and axial pumps. Twente University.’On the flow and cavitation inception of mixedflow impellers’. 1997 [18] Lewis. A. E. Vogel verlag. ‘Kreiselpumpen’. van. Dover..J. Nijhuis. Jersey City. 1977 [15] Arnold. 1999 [14] Bohl. G. Volume II. RINA waterjet propulsion conference.J... Proceedings FAST’99 conference.’Waterjet forces and transom flange design’. ‘A universal parametric model for waterjet performance’.. W.9 References
[11] Stepanoff. 1957 [12] Wislicenus. Springer Verlag.F. J..V. PhD thesis. New York.F. ‘Selection. 1965 [13] MacPherson. ‘Principles of naval architecture’.M. first edition. Berlin..
Waterjet propulsion theory
48
.Chapter 2.
It is known that in normal operating conditions the inflow velocity in waterjet pumps is far from uniform.
Numerical analysis of a waterjet propulsion system
49
. Both methods will be discussed in this section as well. The distributions can be expressed as nondimensional parameters or with a twodimensional representation. the nonuniformity should be kept minimal from a hydrodynamic point of view. In the following subsection some methods will be presented to represent the nonuniform pump inflow velocity distribution.4 it will be shown that the major contributions are unavoidable in waterjet applications with flush type inlets. With such representation it is possible to derive an estimation of the flow rate fluctuations through an impeller channel and the variations of the inflow angle at the leading edge of the blade.Chapter 3 Nonuniform distribution of pump entrance velocity field
The theory as presented in chapter 2 is based on experience with pumps with uniform inflow. Obviously. The first developers of waterjets have accepted this phenomenon. In section 3. Therefore. The distributions are derived from experimental results.1
Representation of nonuniform velocity distribution
In order to get an impression of the type of nonuniform velocity distributions discussed here. most probably since it has clear similarities with a shippropeller wake field. a thorough analysis of the contributing factors to the nonuniformity is made. typical distributions will be shown in this section.
3.
In the model scale testrig the actual waterjet pump is not included.1. The shape of the edge is selected after an extensive test procedure.12). Reynolds number for these conditions is based on the diameter of the inlet D and the averaged pump velocity vpump. The model scale inlet has an inlet diameter of 150 mm. The condition with an IVR of 1. This condition can occur in high speed motor yacht applications (> 60 knots).2 shows the axial velocity distributions for two inlet velocity ratios. It should be noted that the shaft
pitottube vpump
Z vtunnel x pref
z y
Figure 3.Chapter 3. the velocity distribution is measured with a 3 hole Pitottube. derived from the measurements. is included in the test setup. as defined in equation (2. During the tests the growth of the boundary layer thickness and the smoothness of the profiles was evaluated.68 represents a normal cruising speed of a fast ferry. The figure on the right shows the distribution at very high IVR. the impeller drive shaft.1
Sketch of test setup with inlet mounted on top of cavitation tunnel
50
.1. Nonuniform distribution of pump entrance velocity field
3. Though. Upstream of the inlet an serrated edge is applied to thicken the natural tunnel wall boundary layer. At a crosssectional plane just upstream of the pump. The mass flow through the inlet is adjusted to get the desired IVR values. which will be denoted as impeller plane. Figure 3. with a diameter of 22 mm. Experiments have been carried out with a constant tunnel speed vtunnel of 8 m/s. A sketch of the test setup is shown in figure 3. which is mounted on the Tom Fink cavitation tunnel [1]. The tunnel has a square crosssection of 600 x 600 mm.1
Experimental setup
Measurements have been carried out on a model scale inlet. The measured conditions are listed in the table below. The tunnel vtunnel is used to represent the ship speed vship.
Moreover.
z y
Figure 3.00 8.19.00 8. The measurements in the cavitation tunnel indicate that the level of nonuniformity is almost independent of the actual inlet geometry. Measurements in a windtunnel with systematic changes of the geometry confirm this behaviour [2].
51
Numerical analysis of a waterjet propulsion system
.68 1.91 x 105 5.03 2. which results in a symmetric velocity profile. right high IVR of 2.94 3.00 8.2
Experimentally determined nonuniform velocity distributions for 2 IVR conditions.68.48 x 105
diameter (d=22 mm) is about half the diameter of the non measured (blank) region.28 3.65 Reinl [] 7. Based on a survey of the available (confidential) experimental and numerical data at the authors’ company. It is concluded that the nonuniformity is strongly related to the inlet velocity ratio and only weakly related to the precise shape of the inlet. Left: medium IVR of 1.19
Parameters of conditions of the measured velocity distributions Vtunnel [m/s] 8. Similar velocity distributions are found for another inlet geometry in this testrig.1 Representation of nonuniform velocity distribution
Table 3.3.42 x 105 5. it is concluded that within the design space for commercial applications all possible inlet geometries show more or less the same type of velocity distribution.76 4. it should be taken into account that the shaft was not rotating during experiments.87 2.1 IVR [] 1.00 Vpump [m/s] 4.14 x 105 6.
θ ) = a 0 ( r ) +
∑ ( an ( r ) cos ( nθ ) )
n=1
(3. Results of the two dimensional velocity field description are shown in figure 3. However. for a detailed analysis of the timedependent effects of nonuniformity on pump performance. this quantification is not sufficient.1.3
Twodimensional representation
Hu&Zangeneh [4] use a circumferentially averaged velocity distribution to investigate the effects of the nonuniform velocity distribution on the waterjet pump performance.2)
where the coefficients an(r) are taken as quadratic functions of the radius and m is the number of harmonics. Nonuniform distribution of pump entrance velocity field
3.2
Nondimensional representation
For easy comparison of different inlet geometries and operating conditions. The quadratic functions an(r) are based on coefficients at five different radii.Chapter 3.1)
where v is the local axial velocity and vpump the average axial velocity.3 for the condition IVR=2. of the inflow field in front of the impeller is required to capture the time varying phenomena. For the present Fourier approximation 4 harmonics have been used. At least a two dimensional description.1. Such a description of the axial velocity distribution can be obtained if the velocity field is approximated as a Fourier series:
m
v x ( r. A similar agreement is found at lower IVR. The relation between nonuniformity and IVR can be assessed with this parameter.
3. the level of nonuniformity is expressed as a single value ζ [3]: 1 ζ ≡ Q
∫
( v – v pump ) dA
2
(3. based on the radius r and the angle θ.
52
.19 together with the measured values. Agreement between measurements and the numerical approximation is satisfactory for all radii along the complete circumference.
8
0. r1 the hub radius and r2 the tip radius at the impeller inlet. The amplitude of the fluctuation is not only dependent on the level of nonuniformity of the velocity distribution but also on the number of impeller blades.4 Measurement r=20 mm Measurement r=40 mm Measurement r=55 mm Measurement r=65 mm Measurement r=70 mm 0 45 90 135 180 225 Fourier r=20 mm Fourier r=40 mm Fourier r=55 mm Fourier r=65 mm Fourier r=70 mm 270 315 360
0.19
3.2 Local flow rate fluctuations
1.3. Qbb the volume flow rate in an impeller channel.2
vx(r. An estimation of the channel flow rate as function of the rotor position can be calculated with:
π θ + N r2
Q bb =
∫
π θ – N
∫ vx ( r.4
1. with N the number of impeller blades.3 Comparison of axial velocity derived from measurements with the velocities based on Fourier series for velocity distribution at IVR = 2.2
0
Angle θ [degrees] Figure 3.6
1.2
Local flow rate fluctuations
It is expected that the nonuniform inflow velocity distribution will result in a varying flow rate through the channels between the impeller blades during a revolution.3)
where the axial velocity vx is integrated over the impeller passage inlet area.6
0. θ ) ⋅ r dr dθ
r1
(3.
Numerical analysis of a waterjet propulsion system
53
.θ)/vpump []
1
0.
The flow rate deficit is 20% at medium IVR and 35% at high IVR.4.
In figures 3.4 and 3.5. In these figures the channel flow rate Qbb is normalised with the averaged flow rate per channel (Q/N). as shown in figure 3.3.68. which increases to 46% for the high IVR condition. As expected. For impellers with only three blades the amplitude of the variation is smaller. Nonuniform distribution of pump entrance velocity field
150% 140% 130%
Qbb/(Q/N) [%]
120% 110% 100% 90%
z
2 blades
80% 70% 60% 50% 0 45 90 135 180 225 270 315 360
y
6 blades θ
2 bladed impeller 3 bladed impeller 4 bladed impeller 6 bladed impeller
Angle θ [degrees] Figure 3.5 the normalised local flow rate estimate is presented for medium and high IVR and for impellers with different number of blades.
54
.4 Local flow rate estimate as a function of impeller channel position for medium IVR of 1. the amplitude increases with increasing IVR. using data of figure 3.Chapter 3. The six bladed impeller shows a flow rate deficit of 30% for the medium IVR in figure 3. The actual local flow rate through an impeller channel for different IVR conditions will be determined in chapter 6.
3. The design inlet blade angle is based on the inlet velocity triangle with uniform flow and without prerotation. Moreover the cavitation behaviour of the impeller will depend on the fluctuations of the inflow angle.4)
vx
βdesign
Ωr Figure 3.3. as shown in figure 3.3 Impeller velocity triangles
150% 140% 130%
Qbb/(Q/N) [%]
120% 110% 100% 90%
z
2 blades
y
80% 70% 60% 50% 0 45 90 135 180 225 270 315 360 2 bladed impeller 3 bladed impeller 4 bladed impeller
θ
6 blades
6 bladed impeller
Angle θ [degrees] Figure 3.3
Impeller velocity triangles
The axial velocity distribution can be used to derive the fluctuations in inlet flow angles.19. A lower inlet flow angle will lead to higher blade loading in general. using data of figure 3.5 Local flow rate estimate as a function of impeller channel position for high IVR of 2.6 Velocity triangle of inlet flow angle.3.
55
Numerical analysis of a waterjet propulsion system
.6: vx β design = atan  Ωr (3. This angle determines for a great deal the loading of the impeller blade.
Figure 3. The incidence angle is the difference between the design angle and the actual inflow angle. This simplification is allowed.θ) is the local axial velocity. θ ) β actual = atan  Ωr (3. Typical crossflow plane velocities are about 5% of the tip speed of the impeller.7 Estimation of the incidence angle as function of impeller channel position for the velocity distribution at IVR = 2.
15
10
Incidence angle [degrees]
5
0 0 45 90 135 180 225 270 315 360
5
10
radius = 20 mm radius = 40 mm radius = 55 mm radius = 65 mm radius = 70 mm
15
20
Angle θ [degrees] Figure 3. since the tangential velocities are small compared to the tangential velocity component Ωr of the impeller.7 shows the estimation of the incidence angle for high IVR. It is to be expected that such amplitude of inflow angles will result in significant variation of the impeller blade loading during a revolution of the blade.Chapter 3. Nonuniform distribution of pump entrance velocity field
The actual inflow angle will vary due to the axial velocity variations: v x ( r. Note that crossflow plane velocity components are neglected in this approach.19
56
. The deviations vary from +10 degrees to 10 degrees at the outer radii.5)
where vx(r.
the Reynolds number becomes 1. 2.3. obstruction of the flow by shaft.8
Phenomena which contribute to development of a nonuniform inflow velocity distribution to the pump. 4. With a wetted length of 80 m. boundary layer ingestion deceleration of the flow obstruction of the flow due to the shaft bend in the inlet duct
A sketch of the phenomena is presented in figure 3. boundary layer ingestion. then all the water is taken from the hull boundary layer.8.
3. 2. bend in duct. 4. 3. The ingestion of boundary layer water can be expressed with the ratio between the suction depth h and the boundary layer thickness δ. 3. as shown in equation (2. The above items are discussed in more detail in the remainder of this section.1
Boundary layer ingestion
Waterjets with flush mounted inlet ducts ingest water from the boundary layer below the hull.
Numerical analysis of a waterjet propulsion system
57
.4 Origin of the nonuniform velocity distribution
3.4.
3
4
1
h
2
Figure 3. The contribution of the boundary layer ingestion to the nonuniformity depends on the amount of water that is ingested from the boundary layer. with a power n of about 10.3). 1. If this ratio (h/δ) is smaller than 1.4
Origin of the nonuniform velocity distribution
The development of the nonuniform velocity distribution can be explained with basic fluid dynamics theory. There are four phenomena which contribute to the nonuniformity of the velocity distribution: 1. A typical ship speed for a waterjet propelled vessel is 40 knots (about 20 m/s). deceleration of the flow.6 x 109. At these high Reynolds numbers the boundary layer thickness can be approximated with a power law velocity distribution.
If the suction depth h exceeds the boundary layer thickness. The retardation of the flow can be regarded as a (subsonic) diffuser flow
58
.2
Deceleration of the flow
Waterjets operate in IVR conditions of 1. As mentioned before. As a consequence. The ratio h/δ decreases with ship speed for all ship lengths.+ δ 1 = . The averaged velocity of the water just in front of the pump is thus smaller than the ship speed. there is still a significant retardation of the flow in the inlet duct.9. waterjet manufacturers and ship model basins use a simplified method. for longer vessels. then the displacement thickness can be used to calculate the suction depth h: Q Q δ h = .4. Since the velocity profile has the largest gradients near the wall. vship the undisturbed velocity and λD the assumed width of the rectangular box with λ equal to 1. The flow rate Q is adjusted for each ship speed. in which the shape of the streamtube is approximated as a rectangular box with a width of 1. the nonuniformity will increase with increasing IVR as well.+ v ship λD v ship λD n + 1 (h > δ) (3. to take the effect of increasing flow rate with increasing ship speed into account.90 at 45 knots for the short vessel. It can be seen that the boundary layer ingestion ratio h/δ is about 0.⋅ . the suction height h will reduce. Even if the velocity in front of the pump is compared with the mass averaged incoming velocity.3 times the inlet diameter.6)
where Q is the volume flow. where the boundary layer ingestion ratio h/δ is shown for three different vessel lengths.50 when the length is doubled. The ratio is reduced to 0.3 to 1.Chapter 3. the suction depth h is calculated with: Q n + 1 1 ⁄ n n ⁄ (n + 1) h = .
3. This effect is shown more clearly in figure 3.10) for n=9) show that the length of the vessel has a major influence on the development of the boundary layer. Nonuniform distribution of pump entrance velocity field
The suction depth h depends on the total ingested volume flow rate and the assumed shape of the streamtube. With higher incoming velocity and constant suction box width.8 in general. Decrease of the boundary layer ingestion ratio h/δ is coupled to an increase of IVR.⋅ δ v ship λD n (h < δ) (3. which is smaller due to boundary layer development.7)
The relations for the boundary layer thickness (for example equation (2. this results in increased nonuniformity.3. When the suction depth h is smaller than the boundary layer thickness.
the level of nonuniformity will even increase. which have negative diffusor angles.9 Boundary layer ingestion ratio h/δ for three different vessel lengths l as function of ship speed vship with corresponding IVR.5
Ship speed [knots] Figure 3. In contrast. The standard diffusor theory assumes a uniform inflow at the entrance of the diffusor. Therefore. the increase of nonuniformity is related to the value of IVR. it corresponds to a lower n value in a power law profile).00
1.50
2
1. The velocity profiles in convergent channels. the velocity profile becomes less uniform in divergent channels (or in other words.50
1
0. retardation of the flow increases the level of nonuniformity in the velocity distribution.75
1. If the pump shaft is equipped with a stationary sleeve. However in a waterjet the boundary layer creates a nonuniform distribution at the beginning of the diffusor.4 Origin of the nonuniform velocity distribution
1.3
Obstruction of the flow due to the shaft
The impeller shaft forms an obstruction of the flow in the inlet.4.75
0.25
0. According to the theory of standard diffusors.3. When a velocity profile is nonuniform at the entrance a diffusor. Consequently. become more uniform.75
1.00
0. Hu & Zangeneh [7] and Seil [8]
Numerical analysis of a waterjet propulsion system 59
IVR []
.
phenomenon.
3. comparable to a propeller shaft in a ship. This phenomenon is explained by Betz [6] with a simple example.25
1. Nikuradse measured the velocity profiles in convergent and divergent channels with diffusor angles between 4° and +8° (shown in Schlichting [5]). A narrow wake region with low velocity is found near the shaft.5
h/δ []
0. then it can lead to unstable vortex shedding. the velocity profile in a conventional diffusor depends on the diffusor angle.25
Wetted length 50 m Wetted length 75 m Wetted length 100 m IVR 25 30 35 40 45 50
0.
For example the work of Van Manen [9] in the 1950s can be mentioned.4.5
Closing remark
The major contributions to the nonuniformity are caused by the IVR related phenomena of boundary layer ingestion and retardation of the flow. the length of the cylindrical pipe between the bend and the pump is in general smaller than the inlet diameter.4. This nonuniformity will restore to a certain extent downstream of the bend. Nonuniform inflow to mixedflow or centrifugal pumps may occur if the pump is mounted close to an upstream disturbance. so a stable flow is found in the vicinity of the shaft. which might have a positive influence on the velocity distribution. For an acceptable inflow to the pump. The convex shape of the first part of the roof of the inlet may be regarded as a bend as well. This range is significantly smaller than the expected variations for the waterjet pump. In a waterjet however. so the effect of this curvature can be neglected in the analysis and is therefore not marked in figure 3. As a result part of the created nonuniformity due to the bend remains in the velocity distribution. a substantial nonuniform flow will exist. most manufacturers prescribe a minimum
60
. like a pipe bend.
3. as shown in figure 3.
3. They show that the flow stabilizes due to rotation of the shaft. This may be attributed to its close relation to the conventional ship propeller. Variations in effective inflow angle for the tested propeller were between 1° and +4°.
3.5
Nonuniform inflow velocity distributions in other turbo machinery
In the literature relatively little attention has been paid so far to the effects of nonuniform inflow to the waterjet pump. As long as waterjets are operated at IVR values above 1. This wake field is determined by the shape of the vessel. A lot of research has been done on the effects of nonuniform inflow to a ship propeller already. Nonuniform distribution of pump entrance velocity field
investigated the effect of the rotation of the shaft.8.4
Bend in the inlet duct
The fourth contributing factor to the nonuniform velocity distribution is the bend in the inlet duct. so the propulsion system designer has to cope with it. In general a waterjet installation does not have a stationary sleeve around the shaft.7.Chapter 3. Inside the bend a variation in velocity is found between the inner and outer part of the bend. which also operates in a nonuniform wake field in general.5. The radius of curvature is about six times larger than the radius of curvature of the bend.
6
A D h IVR N n Q Qbb r v
Nomenclature
area diameter suction depth inlet velocity ratio (vship/vpump) number of impeller blades boundary layer power law value flow rate channel flow rate radius velocity m2 m m m3/s m m/s
Greek symbols
β δ δ1 θ λ ζ Ω
blade angle boundary layer thickness boundary layer displacement thickness angle suction tube width factor nonuniformity angular velocity
rad m m rad rad/s
Numerical analysis of a waterjet propulsion system
61
. Cooper [11] shows the calculated flow pattern at the exit of a radial inlet passageway for such a pump. Another typical inflow is found for pumps with a sump. The requirement of long straight suction pipes can not always be met. A maximum velocity of 1.3. Additional acceleration can be applied for further improvement. as used in cooling water pump applications in power stations. like the flow into the first stage of a boiler feed pump.75 times the average inflow velocity is found for this example. Similar flow phenomena will occur in double suction pumps and inline pumps. The use of an acceleration nozzle just upstream of the pump is also recommended to enhance the uniformity of the inflow for certain types of pumps. Examples can be found in a pump handbook of Sulzer [10]. typically several pipe diameters long.6 Nomenclature
required length of straight pipe on the suction side of the pump. Here a suitable design of the inflow and inlet chamber is to obtain an acceptable inflow pattern. This level of nonuniformity seems to be more in line with that found in waterjet applications.
3.
2001 [2] Bulten. & Walker. R.. N. van.‘Influence of boundary layer ingestion on waterjet performance parameters at high ship speeds’. ‘Investigations of 3D turbulent flow inside and around a waterjet intake duct under different operating conditions’. ‘Sulzer centrifugal pump handbook’. pp 883892. 1951 [10] Sulzer Brothers Ltd. Oxford. 1966 [7] Hu. 1996
62
.H. PhD thesis. M. P.. ‘Introduction to the theory of flow machines’.J. Proceedings of the 5th international conference on Fast Sea Transportation. Zangeneh. P. 1989 [11] Cooper. pp. N. 654664.’Boundary layer theory’. Gothenburg.. 396404. London. Proceedings Waterjet Propulsion III conference. Seattle. Nonuniform distribution of pump entrance velocity field
Subscripts x axial direction
3.. G. 1968 [6] Betz.W. Amsterdam.7
References
[1] Brandner.. 1998 [4] Hu. & Zangeneh. New York. Gothenburg. ASME Journal of Fluids Engineering. G.W.’CFD calculation of the flow through a waterjet pump’.’Perspective: the new face of R&D . Mc GrawHill. ’Invloed van de ongelijkmatigheid van het snelheidsveld op het ontwerp van scheepsschroeven’. P.’The effect of the shaft.J. Vol 118... P. H. 2001 [5] Schlichting. ‘A waterjet test loop for the Tom Fink cavitation tunnel’. Proceedings Waterjet Propulsion III conference. Elsevier Applied Science. Pergamon Press. & Bulten.’Recent developments in waterjet design’. 1999 [3] Verbeek. J. Proceedings RINA Waterjet Propulsion II conference. ASME Journal of Fluids Engineering.H.. 1999 [8] Seil.. A.D. 2001 [9] Manen. M.Chapter 3. Vol 121. Proceedings Waterjet Propulsion III conference. pp. Gothenburg.A case study of the pump industry’. shaft rotation and scale on the flow in waterjet inlets’. Wageningen.
however these quantities will be determined by numerical simulations.1
Requirements of mathematical method
In this section the requirements for the mathematical method will be formulated. pressure and forces can be measured. A numerical method is based on a mathematical model of the physics of the flow.Chapter 4 Mathematical treatment
To get more insight in the behaviour of the flow through a waterjet installation. experiments can be done and calculations can be made.1 to get an impression of the feasible simplifications without major loss of accuracy. often complemented with empirical relations. The goal of the mathematical analysis is the prediction of the flow through a waterjet installation. The numerical method is obtained by discretisation of the governing equations. This analysis will result in a selection of the best suitable mathematical method. A review of the important flow features is given in section 4. it is necessary to make some simplifications.
4. The chosen method will be described in more detail in section 4. Based on the different requirements for the prediction three different mathematical methods will be evaluated: the
Numerical analysis of a waterjet propulsion system
63
. Some known weaknesses of the mathematical method will be discussed in more detail to get an indication of the obtainable accuracy. With an experimental setup of a waterjet inlet velocity. Before these equation can be solved numerically for the application of interest. It constitutes a set of governing equations derived from first principles. In this thesis.3. The present type of flow can be described by the NavierStokes equations.
2. This gives the opportunity to use a single numerical technique for subsonic. The resulting Mach number is thus sufficiently small to treat the flow as incompressible.2
High Reynolds number
For full scale waterjet installations the typical Reynolds numbers for the inlet and the impeller are both very large. for example effects of viscosity. such as flow separation.
4. whereas the other two methods do not include terms due to viscosity and heat conduction. the speed of sound reduces significantly.1. It is noted that the speed of sound is based on a noncavitating flow.e. which results in amongst others travelling pressure waves through the domain. as discussed in section 1. except in infinitesimal regions such as vortex sheets and vortex filaments. which is about 1450 m/s. These methods are commonly called time marching methods.Chapter 4. pressure fluctuations are instantaneous in the whole domain. µ the dynamic viscosity. Nevertheless it can be mentioned that this requirement represents one of the major differences between pumps and compressors. supersonic or mixed subsonicsupersonic (transonic) flow problems [1]. mathematical methods for compressor analysis cannot be used directly for incompressible flow. In potential flow methods. In an incompressible flow. the model based on the Euler flow and the model based on the Reynolds averaged NavierStokes (RANS) equations. i. The maximum velocity in a full scale waterjet installation is about 50 m/s. For compressible flow the Euler equations are hyperbolic in time for any Mach number. some form of artificial compressibility has to be introduced in the compressible flow solver in order to analyse an incompressible flow. All three mentioned mathematical methods can satisfy this requirement. In these methods compressibility is employed. which results in a compressible flow behaviour. They require. which is much smaller than the speed of sound in water. RANS methods are also denoted as viscous methods. Mathematical treatment
potential flow model. i. the flow is irrotational. For
64
. to be modelled.1)
where ρ is the density.
4. vtip is the tip speed of the impeller blade and Dinl the diameter of the inlet at the pump suction side. In a cavitating flow. For the impeller the Reynolds number is defined as: ρv tip D inl Re imp = µ (4.1
Incompressibility
The first requirement for the mathematical method is the capability of handling incompressible flow. the vorticity equals zero everywhere.1. Reconditioning.e.
the Reynolds number Reinl will be about 107. For an inlet diameter of 1. In order to represent the flow phenomena inside the inlet best. In the stationary frame of reference the frequency of the fluctuations due to rotorstator interaction is related to the number of impeller blades. These values still exceed 5x105.1. It is shown that the velocity field is strongly nonuniform even at normal waterjet operating conditions. The frequency of the interaction phenomena between the rotor and stator blades depends on the number of stator blades and the shaft frequency.4. at least at design conditions. The actual choice is not so critical. In the rotating frame of reference.1 on page 51. viscous effects play a limited role. This velocity distribution has to be reproduced by the mathematical method in order to obtain a correct prediction of the inlet flow phenomena.
4. This can be either the ship speed or the pump speed (as defined in equation (2. which is 6 times smaller than the impeller Reynolds number.1.3
Time dependency
There are two different reasons for the flow to be timedependent in a waterjet installation: (i) the nonuniformity of the flow at the impeller entrance results in a time varying onset flow of the rotating blades and (ii) the interaction between moving rotor and stationary stator blades at the impeller outlet is unsteady. It is to be expected that. The effect of the nonuniform inflow will give a steady component with superposed fluctuations with the blade passing frequencies.2)
For the same typical configuration as above. For an analysis of the isolated pump.11)). the Reynolds number for the inlet is defined as: ρv pump D inl Re inl = µ (4.
4.
Numerical analysis of a waterjet propulsion system
65
.4
Nonuniformity of impeller inflow
In chapter 3 the velocity distribution upstream of the impeller is discussed. Calculation of the Reynolds number that characterises the flow in the inlet requires another typical velocity. phenomena associated with the first reason have a frequency related to the shaft frequency. the nonuniform inflow to the impeller has to be implemented as an inflow boundary condition.1 Requirements of mathematical method
typical waterjet applications the tip speed can be about 50 m/s. Model scale experiments are performed at lower Reynolds numbers. since both values do not differ by more than a factor of two. which can be regarded as high Reynolds number flow.2 m this gives a Reynolds number Reimp of 6x107. as shown in table 3.
Chapter 4. Mathematical treatment
The development of the velocity distribution in the inlet is attributed to the ingestion of the boundary layer and to the deceleration of the flow. Generation of vorticity plays an important role in these flow phenomena. Application of a mathematical method based on the potential flow assumption or an inviscid flow model is therefore not suitable for the analysis of inlet flow phenomena. Potential flow analyses of mixedflow pumps have been made by Van Esch [2] and Van Os [3]. These calculations are based on a uniform inflow velocity distribution. Implementation of a nonuniform velocity distribution introduces velocity gradients for the axial velocity component. The constraint of irrotational flow results in additional velocity gradients in the directions perpendicular to the axial inflow direction. The cross components of the inflow velocity distribution have such a dominating influence on the overall development of the nonuniformity, that the velocity distribution is almost uniform within about one diameter pipe length. The analysis of the stability of the nonuniform velocity distribution is discussed in more detail in appendix A. It is concluded that for the analysis of both a waterjet inlet as well as a pump with nonuniform inflow a suitable mathematical method should take the presence of vorticity into account.
4.1.5
Tip clearance flow
In general, waterjets are equipped with unshrouded mixedflow or axial impellers. Unshrouded impellers have a small clearance between the blade tips and the stationary housing. This housing is called the seatring. The pressure difference between the pressure and suction side of the blade causes some leakage flow through the clearance. This leakage flow should be kept as low as possible to maintain a high efficiency of the pump. The distance between the impeller tip and the seatring is therefore very small, about 12% of the diameter. On the other hand, the velocity difference between the rotating blade tip and the stationary seatring can be about 50 m/ s, which leads to large velocity gradients in the clearance. As a result of the available pressure difference between pressure and suction side of the blade and the occurring viscous losses in the clearance a certain flow will establish. Effects of viscosity can not be neglected, if this flow is to be determined. Second effect of the viscosity is the development of boundary layers on the blade tip and the seatring. This is confirmed by Kunz et al. [4] in the investigation of tip clearance phenomena in an axial compressor cascade with an Euler and a NavierStokes method. It is found that the NavierStokes method shows better agreement with experimental data than the Euler method. Prediction of the tip clearance mass flow rate is presumed to be more accurate with the NavierStokes method.
66
4.2 Conservation laws
The flow in a mixedflow pump with various tip clearances has been analysed by Goto [5]. For this analysis he uses the unsteady Reynoldsaveraged NavierStokes equations in a rotational frame of reference. Results of the calculations show reasonably good agreement with the experimental data. Moreover a variety of jetwake flow patterns at the exit of the impeller are predicted well.
4.1.6
Final remarks
Based on the listed requirements and the capabilities of the different models, the choice for a viscous flow model is justified. In the next two chapters, results of a detailed validation study of numerical simulations for the waterjet inlet and the mixedflow pump are presented. It is acknowledged that a viscous (i.e. RANS) flow method requires significantly more computational resources than an Euler method and certainly much more than a potential flow method. Cavitation will not be taken into account in the analyses. Since cavitation models are presently developed for most commercial RANS codes, is it to be expected that application of methods with some form of cavitation model will be feasible in the near future. The calculations presented in this study are carried out with the commercial CFD method StarCD. This method is based on a finite volume numerical method. Both tetrahedral and hexahedral cell types can be used for the mesh. In this study all meshes are generated with hexahedral cells only.
4.2
Conservation laws
To describe the flow phenomena in a waterjet installation, two conservation laws are used. These are the laws for conservation of mass and that of momentum. Conservation of mass is also denoted as the continuity equation (see for example [6]): ∂ρ + ∇ ⋅ ρv = 0 ∂t where ρ is the density of the fluid and v the velocity. In many cases the flow can be assumed to be incompressible. This is allowed, whenever variations in density are small. These variations in density are caused by variations in the pressure. Since in the present applications the velocities are much smaller than the speed of sound in the water, which is about 1450 m/s, the variation in density will be negligible. For an incompressible flow the continuity equation (4.3) reduces to: ∇⋅v = 0
Numerical analysis of a waterjet propulsion system 67
(4.3)
(4.4)
Chapter 4. Mathematical treatment
Conservation of momentum is given by (see also [6]):
2 1 ∂ ρ v + ρ ( v ⋅ ∇ )v = – ∇p + µ∇ v +  µ∇ ( ∇ ⋅ v ) + ρg 3 ∂t
(4.5)
where p is the pressure, µ is the dynamic viscosity, which is assumed to be constant in the derivation of equation (4.5) and g is the acceleration of gravity. This set of equations is known as the NavierStokes equation. For an incompressible flow, equation (4.4) can be substituted into equation (4.5). This results in the NavierStokes equation for incompressible flow: ∂ ρ v + ρ ( v ⋅ ∇ )v = – ∇p + µ ∇2v + ρg ∂t (4.6)
The NavierStokes equation can be solved numerically directly without further assumptions. This requires direct numerical simulations (DNS) to obtain the timeaccurate solution of equations (4.4) and (4.6) on a grid that is sufficiently fine to resolve all flow details. Such a method is not suitable for practical engineering analyses, however. In the following section an approach will be discussed, which enables numerical solution of the flow field, within a practical context.
4.3
Reynolds Averaged NavierStokes (RANS) flow
In order to convert the NavierStokes equations into a set of equations, that can be solved numerically for general engineering applications, the concept of splitting the flow variables in a mean and a fluctuating part is employed. Substitution of this decomposition in the NavierStokes equation and time averaging the continuity and the NavierStokes equations, results in a set of equations for the meanflow field variables. The concept of time averaging was introduced by Reynolds in 1895. The resulting equation is therefore called the Reynoldsaveraged NavierStokes (RANS) equation.
4.3.1
Reynolds averaging
The principle of Reynolds averaging is based on a decomposition of the variables in a time averaged value and a fluctuating part: v i = v i + v i' (4.7)
68
4.3 Reynolds Averaged NavierStokes (RANS) flow
The time averaged variable is defined as:
t+T
1 v i = lim T → t 1T
∫
t
v i dt
(4.8)
where t1 has to be larger than the time scale of the smallest fluctuations. Therefore,
t+T
1 v i' = lim T → t 1T
∫
t
[ v i – v i ] dt = 0
(4.9)
The average of the fluctuating part is zero by definition. Nevertheless, if the product of two variables is considered, not all fluctuating terms vanish when they are correlated, thus: v i v j = v i v j + v i'v j' (4.10)
Substitution of the decomposed terms in the NavierStokes equation for the velocity and pressure and application of the method of Reynolds averaging gives: ∂ ρ v + ρ ( v ⋅ ∇ ) v = – ∇p + µ ∇2v + ρg – ∇ ⋅ ρ ( v'v' ) ∂t (4.11)
The term ρ ( v'v' ) is called the Reynolds stress term, where ( v'v' ) is a diadic product defined by ( v'v' ) ij = v i'v j' . This tensor contains the correlations of the fluctuating terms of the velocity components. The term with the Reynolds stress can be treated in several ways. Boussinesq proposed a closure hypothesis for the Reynolds stress term. In index notation the closure is defined as: ∂v i ∂v j 2 – ρv i'v j' = µ T + –  ρkδ ij ∂ x j ∂ x i 3 (4.12)
which is referred to as an eddy viscosity model with µT the turbulent or eddy viscosity and k the turbulent kinetic energy defined as: 1 1 k =  ( v i'v i' ) =  ( v x'v x' + v y'v y' + v z'v z' ) 2 2 (4.13)
Turbulence models are used to obtain a value for the eddy viscosity. An alternative approach is to derive from the original timedependent NavierNumerical analysis of a waterjet propulsion system 69
Substitution of equation (4.
70
. the eddy viscosity µT is based on the mixing length concept. The available eddy viscosity turbulence models can be divided into 3 groups: 1.12) into the Reynolds averaged NavierStokes equation (4. This mixing length is based on algebraic relations. mixing length or algebraic models or zeroequation models 2.11) gives:
T 2 ∂ ρ v + ρ ( v ⋅ ∇ )v = – ∇p + ∇ ⋅ ( µ + µ T ) ( ∇v + ( ∇v ) ) + ρg – ∇ . the Reynolds stresses are known and consequently the Reynoldsaveraged NavierStokes equations can be solved.ρk – ρgz 3 (4. becomes:
T ∂ ρ v + ρ ( v ⋅ ∇ )v = – ∇p∗ + ∇ ⋅ µ eff ( ∇v + ( ∇v ) ) ∂t
(4.14)
The next step is the combination of the pressure term. These equations require another group of closure relations before the equations can be solved numerically. the turbulent kinetic energy term and the gravitational term.
4. The eddy viscosity for an algebraic model is defined as:
2 µ T = ρl mix dv dy
(4. Once this eddy viscosity is known. however. The resulting NavierStokes equation with the Reynolds stress terms included.16)
where the molecular viscosity µ and turbulent viscosity µT have been combined into the effective viscosity µeff.ρk 3 ∂t
(4.3. twoequation models In a zeroequation model.Chapter 4.2
Eddy viscosity turbulence models
Eddy viscosity turbulence models are used to determine a value for the eddy viscosity µT. Such a model is called a Reynolds stress turbulence model. Mathematical treatment
Stokes equations a transport equation for each of the Reynolds stress components. The modified pressure is defined as: 2 p∗ = p + .17)
where lmix is Prandtl’s mixing length.15)
where it is assumed that gravity is directed in the zdirection. oneequation models 3.
ρv i'v i'v j' + p'v j' = – 2 σk ∂ xj where σk is a dimensionless closure coefficient. (4. This model is thus an extension of the mixing length model. turbulent velocity fluctuations and pressure fluctuations.20)
where the terms on the right hand side represent production. Two wellknown examples are the dissipation ε for the kε model and the dissipation per unit turbulence kinetic energy ω for the kω model.19)
with Cµ a dimensionless constant.4. The latter two require a closure term in order to enable the transport equation to be solved.21)
Numerical analysis of a waterjet propulsion system
71
. diffusion and dissipation respectively. diffusion by viscosity. The diffusion term has three components. The exact transport equation for the turbulent kinetic energy can be derived from the NavierStokes equation. The transport equation is given for example by Wilcox [7]: ρ ∂v i ∂ ∂k ∂k ∂k 1 µ = – ρv i'v j' + – . kε turbulence model The eddy viscosity in the kε turbulence model is defined as: k µ T = ρC µ ε
2
(4. The standard kε turbulence model will be discussed in more detail. The eddy viscosity for a oneequation model is defined as: µ T = ρk
1⁄2
l mix
(4. The turbulent transport and pressure diffusion terms are modelled by: µ T ∂k 1 .18)
which uses a transport equation for k and an algebraic relation for the mixing length lmix. This model was first presented by Jones and Launder in 1972. Twoequation models use a transport equation for both the turbulent energy k as well as the turbulent length scale or an equivalent.3 Reynolds Averaged NavierStokes (RANS) flow
The oneequation model uses a transport equation for the turbulent kinetic energy k and an algebraic relation for the mixing length scale.ρv 'v 'v ' – p'v j' – ρε + ρv j ∂t ∂ xj ∂ x j ∂ xj ∂ x j 2 i i j (4.
Cε2 and σε. The exact equation for ε has a number of unknown double and triple products.24)
with dimensionless closure coefficients Cε1.Chapter 4. The entire equation for ε can also be regarded as a model in a similar form as the transport equation for k (see [8]). The kω model uses the dissipation per unit turbulence kinetic energy ω. Similar to equation (4. Table 4.+ – C ε2 ρ k ∂ xj ∂t ∂ xj σε ∂ xj k
2
(4.23)
The exact equation for the dissipation ε can be derived from the NavierStokes equation.22)
where the production term of the turbulent kinetic energy Pk is defined as: P k = – ρv i'v j' ∂v i ∂ xj (4. + ρv j = Pk + – ρε ∂t ∂ xj σ k ∂ xj ∂ xj (4. diffusion and dissipation term can be recognized on the righthandside. [10]). whereas the dissipation ε is used in the latter. Mathematical treatment
The final transport equation for k becomes: ρ µ T ∂k ∂k ∂k ∂ µ + .22) a production. The final equation becomes: ρ µ T ∂ε ε ∂ ε ∂ε ∂ε µ + . + ρv j = C ε1 P k .09 Values for kε turbulence model closure coefficients C ε1 1. are listed in the table below.1 Cµ 0.44 Cε 2 1.22
A value of 1. so a number of closure terms are still required.3 for the constant σε can be found in literature as well ([7]. These values are found in the manual [9]. as used in the CFD method employed in the present study.0 σε 1. Further development of this model has led to
72
. kω turbulence model The kω turbulence model has a similar set of equations as the kε turbulence model. The basis of the kω model was postulated by Kolmogorov. The kε turbulence model contains five closure coefficients. but it requires a considerable amount of algebra to arrive at the final equation as given by Wilcox [7]. The values for the coefficients of the kε turbulence model.92 σk 1.
⋅ .22) except for the dissipation term and the closure coefficients: ρ ∂k ∂k ∂k ∂ ( µ + σ∗ µ T ) + ρv j = Pk + – β∗ ρkω ∂t ∂ xj ∂ xj ∂ xj (4. The transport equation for the dissipation per unit turbulence kinetic energy ω is also adapted in a similar way.28)
where u τ =
τ w ⁄ ρ and ∆yp is the distance from the wall to the near wall cell
centre.+ ( µ + σµ T ) – βρω + ρv j k ∂ xj ∂t ∂ xj ∂ xj (4.
Numerical analysis of a waterjet propulsion system 73
.4. The eddy viscosity hypothesis for the kω turbulence model is defined as: k µ T = ρ ω (4.26)
with the production term Pk according to equation (4. Note that the dimensionless distance y+ is a kind of Reynolds number.3 Reynolds Averaged NavierStokes (RANS) flow
the following set of equations.27)
where α.= ν ν ρ
(4. defined as: y
+
Values for kω turbulence model closure coefficients β 3/40 β* 9/100 σ 1/2 σ∗ 1/2
∆y p u τ ∆y p τ w = .2 Table 4. The wall functions employ special algebraic formulas for the representation of the distribution of the velocity and turbulence within the part of the boundary layer closest to the wall.25)
The equation for the turbulent kinetic energy is similar to equation (4.2 α 5/9 Wall functions Use of one of the many available high Reynolds turbulence models implies the application of wall functions to replace the noslip boundary condition. as shown by Wilcox [7]: ρ ω ∂ 2 ∂ω ∂ω ∂ω = αP k .23) and the closure coefficients β* and σ*. The values for all five closure coefficients of the kω turbulence model are shown in table 4. β and σ are three more closure coefficients. The dimensionless distance from the wall to the cell centre of the first cell is generally denoted as y+ and is. This is necessary to relax the requirement of grid resolution in the boundary layer. according to [10].
34)
74
.ln ( y ) + C κ
(4.ln ( Ey m ) = 0 κ
(4.28) will lead to numerical problems at points where the flow is about to separate. the u+ is calculated from: u
+ + 1 = .30)
where y m satisfies the equation:
+ + 1 y m – . This results in a value for C of 5. according to the manual [9].9 and 5.4187 and C a constant with a value between 4. The empirical constants of equation (4. The default value for the empirical constant E in the CFD method used is 9.⋅ C µ k µ
(4. Mathematical treatment
based on the distance from the wall and the velocity near to the wall to the power 1/2.29)
+
For y+ values larger than y m . For y+ values smaller than y m the dimensionless velocity u+ is calculated from: u
+ +
= y
+
(4. known also as loglaw [11]: u
+
1 + = . which results in an y+ of zero.32)
where u is the tangential fluid velocity and uw the velocity of the wall.ln ( Ey ) κ +
(4.5 ([12]). For accurate application of the wall function the y+ value should be in the range of 30 to 100.25.0.31)
with κ the von Kármán constant and an empirical constant E. At such locations the wall shear stress τw becomes zero. The dimensionless velocity u+ is defined as: u
+
( u – uw ) ( u – uw ) = .= uτ τw ρ
(4.30) can be coupled to the universal logarithmic velocity distribution. This is solved with the modified definition for y+ based on the turbulent kinetic energy k: y
+
ρ∆y p 1⁄4 = . The definition of y+ according to equation (4.Chapter 4.33)
with κ the von Kármán constant equal to 0.
The first region is meshed with an Ogrid around the profile. The second type is suitable for an unbounded region. The number of cells around the profile and in the direction perpendicular to the profile have been varied.
4. The number of cells in the Ogrid around the profile has been varied from 150 to 330 cells. For all configurations experimental data for the lift and drag is available.1
Isolated NACA 0012 profile
The mesh for the numerical analysis of the flow along the NACA 0012 profile is shown in figure 4. which eliminates the problem.1. The other two regions can be distinguished more clearly. The number of cell in normal direction in the first region is kept constant to keep a constant y+ value of about 110. The flow along isolated profiles as well as profiles in a cascade have been calculated. A rectangular box is placed around the first region to make a transition from the Ogrid to an Hgrid. In the normal direction the number of cells is increased from 32 to 48. the flow around a number of well known NACA profiles has been analysed. This gives good control of the quality of the boundary layer cells along the surface. which allows nonmatching cells at both sides of the interface. In this approach the mesh around the profiles is identical for all calculated conditions.
4. This is in accordance with the requirements for the use of the wall functions. The domain is divided in three subregions.4. For sufficiently large numerical domains both types will
Numerical analysis of a waterjet propulsion system 75
.4. At the upper and lower boundary of the domain two types of boundary conditions can be applied: (i) wall boundary conditions or (ii) constant pressure boundary conditions. Regions 1 and 2 can be rotated to obtain the desired angle of attack for the profile. The chord length of the profile is 600 mm. The Ogrid region closest to the profile surface can not be recognised well in the mesh plot due to the large number of mesh lines.4
Twodimensional test cases
In order to get an indication of the obtainable accuracy of both the mathematical and the numerical methods.4 Twodimensional test cases
It is assumed that there will always be turbulent fluctuations at a flow separation point. Mesh dependency studies have been carried out to evaluate the variation in lift and drag prediction for an angle of attack of 4 degrees. At the interface between the second and third domain an arbitrary coupling method is employed. The first type can be used if the experimental data is obtained from wind tunnel tests. The third domain extends the numerical domain to either the tunnel walls in the experiments or to a distance to impose the farfield boundary conditions. In this second region also an Ogrid type of mesh is applied.
Convergence behaviour of one of the calculated conditions is shown in figure 4.2.
give comparable results. All calculations have been carried out with the standard kε turbulence model and employing wall functions.
6
76
. mass and turbulent kinetic energy equations. The kε model turbulence equations are discretised with a first order upwind differencing scheme. Solution is based on a second order MARS (= Monotone Advection and Reconstruction Scheme) discretisation scheme for the momentum equations. At the inlet boundary a uniform velocity distribution of 10 m/s is prescribed and the constant density of water is used. This second order method is least sensitive to the mesh structure and skewness [9].Chapter 4. Mathematical treatment
Figure 4.01% and the length scale is set to a small fraction of the tunnel height. Turbulence intensity is set to 0. The Reynolds number for the calculations becomes 6 ⋅ 10 . The convergence criterion for all calculations is set to 104 for the momentum.1
Plot of a part of the mesh as used in calculations of isolated NACA 0012 profile.
0E04
turbulent energy
1. The experimental data is taken from Abbott & von Doenhoff [13].3 shows a comparison of the calculated and measured lift and drag for the isolated NACA 0012 profiles.ρv A 2 D c d = 1 2 . v the freestream velocity and A the surface area of the wing.0E03
1. Compressibility
The numerical values for lift and drag are based on the integrated pressure and shear forces acting on the profile surface.0E+00 Mass residu Turbulent energy residu Epsilon residu 1.2 Plot of convergence of a calculation of NACA 0012 profile at angle of attack of 4 degrees for mesh with 222 cells around profile and 40 cells in normal direction. The dimensionless lift and drag coefficients are defined as: L c l = 1 2 .36)
where ρ is the density. L is the lift force and D the drag force.0E05
mass u velocity v velocity
1.0E01
Residual
1.e.
Numerical analysis of a waterjet propulsion system
77
. i.
6
are used.
Figure 4.0E+01 Uvelocity residu Vvelocity residu 1.ρv A 2 (4.35)
(4. here the chord length times the width in span wise direction.4. where the results of the measurements with smooth profiles at a Reynolds number of 6 ⋅ 10 has been negligible for the tested conditions.4 Twodimensional test cases
1.0E06 0 50 100 150 200 250
Iteration number Figure 4.0E02
dissipation
1.
as described by Moore & Moore [14]. Mesh is based on 222 cells around profile and 32 in normal direction
Agreement is good for the lift up to an angle of attack of about 6 degrees.CFD 0.05
0. The error in prediction of drag might be related to an error in the production term of the turbulence model at the stagnation point.02
0.EXP cl .3 for the lift prediction and in table 4.4 for the drag prediction at an angle of attack of 4 degrees.6
0.8 cd . The results of the finer meshes do not show a reduction of the deviation with the experimental data. The relative error increases from 33% at 0 degrees angle of attack to 55% at an angle of attack of 4 degrees. The comparison of calculated and measured drag shows a clear overprediction. Mathematical treatment
1 cl .4
0.CFD
0.03
0.0x106. The results of the mesh sensitivity study are shown in table 4.01
0 0 1 2 3 4 5 6 7 8 9
0
Section angle of attack [degrees] Figure 4. The relative difference in lift coefficient between the minimum and maximum lift is about 3%. For the drag coefficient a variation of about 9% is found.04
0.Chapter 4.EXP cd .
78
cd []
cl []
.2
0. Reynolds number of experiments and calculations is 6. The deviation between the measurements and the calculations continuously increases with larger angles of attack.3 Comparison of measured and calculated lift an drag for NACA 0012 profile.
4386 0. Calculations are carried out with a mesh with 222 cells around the profile and 32 in normal direction.4206 cd [] 0.011495
cd [] 150 cells 222 cells 330 cells
The sensitivity of the turbulence model is evaluated by a variation of the turbulence intensity of the free stream flow at the inlet boundary condition.011324 0.4366 0.014980
Turbulence intensity [%] 0. Columns show number of cells in normal direction and rows show cells in Ogrid around the profile 32 cells 0.4.014147 0.010784 40 cells 0. The level of the turbulence intensity at the inlet boundary condition is increased from 0.4289 48 cells 0.4302
cl [] 150 cells 222 cells 330 cells Table 4. Moreover the deviation between experimental data and calculations increases significantly when the level of turbulence intensity at the inlet boundary increases.011332 0.4375 0.4242 0.01% to 1.010522 0.5 Lift and drag coefficients for calculations with variation of input values for turbulence intensity cl [] 0.010988 0.010522 0.00 %
The test calculations with the NACA 0012 profile show that the error in prediction of profile drag remains after some mesh refinement steps.010754 0.4384 0.5. It should be noted that the low turbulence levels as used in the experiments are not representative for the inflow to the waterjet pump.012713 0. whereas the change in drag is about 42%.4287 0.4291 40 cells 0. Columns show number of cells in normal direction and rows show cells in Ogrid around the profile 32 cells 0.011493 0.4418 0.011243 48 cells 0.50 % 1.4 Twodimensional test cases
Table 4.10 % 0.4407 0.4384 0.01 % 0.0%.
Numerical analysis of a waterjet propulsion system 79
. Variation in lift coefficient is 4%.3
Lift coefficient for mesh convergence study. Table 4. The results for the prediction of lift and drag coefficient are presented in table 4.4
Drag coefficient for mesh convergence study.
4
Geometry of NACA 65410 cascade.0. blade angle β = 20 degrees. These are the solidity and the blade angle. also the flow around profiles in a cascade have been analysed. An example of the cascade is shown in figure 4. Mathematical treatment
It is concluded that with the currently used cell sizes. The mesh is created in a similar way as for the isolated NACA 0012 profile.
σ
β
α
Figure 4. Cascade solidity σ = 1. which is comparable with the sizes to be used in the three dimensional pump mesh. discretisation scheme and convergence criteria are identical to the calculations for the NACA 0012 profile. The designation of the profile is based on a design lift coefficient of 0.0 mean line (see for example [13]). Turbulence model.4.4 and a maximum thickness of 10%.0 and 1. The basic camber line used is the a =1.2
Cascades with NACA 65410 profiles
Apart from calculations on isolated profiles. a significant deviation between calculated and measured profile drag will remain.5.
4.
80
. Periodic boundaries are applied to simulate a cascade with an infinite number of profiles. Extensive experimental data is available on tests with NACA 65 compressor blade profiles [15]. Data were reported for NACA 65 profiles with various camber lines and a maximum thickness of 10% of the chord. The blade angle is defined as the angle between the profile base line and the line connecting all leading edges. Cascades of profiles can be described with two additional parameters.Chapter 4.Calculations have been carried out for a NACA 65 410 profile with a blade angle of 20 degrees and a solidity of 1. The solidity is a measure for the distance between two profiles in relation to the chord length.4.
The lift is predicted quite well for most conditions.5 shows the nondimensional pressure distribution Cp for the cascade with a solidity of 1.5 degrees.5 to 12.5
0. The angle of attack varies from 0. ρ the density and v ∞ the freestream velocity. p ∞ the reference pressure.5
Cp []
0 0 20 40 60 80 100
0 0 20 40 60 80 100
0.5 degrees Cp []
α = 12. The deviations
1
1
α = 0.37)
where p is the static pressure.5 degrees
0.5
1 Experimental values CFD results 1.5 degrees Cp []
0 0 20 40 60 80 100
Cp []
0 0 20 40 60 80 100
0.4. blade angle = 20 degrees. Agreement between calculations and measurements is acceptable for most pressure taps for the different conditions.5 0.5
0.0 and a blade angle of 20 degree.5
1 Experimental values CFD results 1. Lift and drag are derived from the CFD results in the same manner as for the NACA 0012 profile. These results are compared with the experimental data in figure 4.5
Chord length [%]
1
Chord length [%]
1
α = 8.5
Chord length [%] Chord length [%] Figure 4.0.4 Twodimensional test cases
Figure 4. The dimensionless pressure Cp coefficient is defined as: p – p∞ C p = 1 2 .
Numerical analysis of a waterjet propulsion system 81
.5
0.6.5 degrees
0.5 Pressure coefficient distribution along surface of NACA 65_410 profile for different angles of attack.ρv ∞ 2 (4. Cascade solidity = 1.5
1 Experimental values CFD results 1.5
α = 4.5
1 Experimental values CFD results 1.
4.1
0.6 Comparison of measured and calculated lift and drag for NACA 65_410 profiles in cascade with solidity of 1. Lift and drag are transformed into axial and tangential forces with: F ax = cos φ ⋅ L – sin φ ⋅ D F tan = sin φ ⋅ L + cos φ ⋅ D (4.4
0.06
0.01
0.3
0.39)
82
cd []
cl []
.3
Sensitivity of errors in drag on thrust and torque
Torque and thrust of a waterjet impeller are related to the tangential and axial force experienced by the blade sections. In the next subsection the consequences of an error in lift or drag on the prediction of thrust and torque are discussed.08
0.1
0. if the blade profile drag is overpredicted.0 and blade angle of 20 degrees. L the lift force and D the drag force.03
0. It can be concluded that the trend of overprediction of drag occurs for both isolated profiles as well as for profiles in a cascade. The inflow angle φ is related to the blade angle β and the angle of attack α of the flow with respect to the chord line: φ = β–α (4.6 cd . This is in agreement with the results of the isolated NACA 0012 profile.CFD cd .2
0. there will be an effect on the prediction of the torque and the thrust.02
Angle of attack [degrees] Figure 4.5
0.8 cl .01
0
0
0.
between the measured and calculated drag vary between about 10% and 50%.exp 0. Mathematical treatment
0.04
0.05
0.38)
where φ is the inflow angle.2 4 0 4 8 12 16 20
0.02
0.Chapter 4. These forces can be derived from the lift and drag of these profiles.exp 0.07
0. Consequently. In both cases the lift is predicted much better.CFD
0.7 cl .4.
L/D is the lift over drag ratio.42)
where φ is the inflow angle. The relative error for the axial force can be calculated with: F ax_CFD ε .7 shows a sketch of the forces acting on a profile.41)
This can be up to about 50%.40)
where ε represents the relative overprediction of the profile drag.4 Twodimensional test cases
The drag prediction based on CFD calculations can be expressed as: D CFD = D exact ( 1 + ε ) (4.40) in equation (4. Figure 4.⋅ .= 1 – F ax_exact L exact 1 .– 1 tan φ D exact (4.+ 1 D exact (4. It is assumed that: L CFD = L exact (4.4.These forces can be related to the exact solutions to determine the resulting relative error in axial and tangential direction. ∆torque
∆d r ag
φ α β
∆thrust
Figure 4. From this sketch the effect of a higher drag can already be recognised.38) gives an expression for the axial and tangential force prediction based on CFD results.43)
Numerical analysis of a waterjet propulsion system
83
.7
Sketch of forces acting on profile
Substitution of equation (4.= 1 + L exact F tan_exact tan φ ⋅ . The relative error in tangential direction yields: F tan_CFD ε .
ε=0.20 Tangential force error .6. The error in tangential force can be about 2. the data for the NACA 65410 cascade will be used. The estimated error on thrust will be less than 1% for most cases. For cascades with smaller blade angles β.ε=0.8. It is expected that the calculation of pump head shows the same behaviour. Calculation of thrust seems to be insensitive to an error in drag.20 Axial force error . The relative errors in axial and tangential force are plotted as function of the angle of attack in figure 4.34). From this diagram is becomes clear that the effect on axial force is very small even for large overprediction of drag. The drag overprediction is set to 20%. On the other hand the error in tangential force remains significant for realistic values of the angle of attack α and drag error factor ε. Data is based on NACA 65410 cascade with solidity of 1.ε=0.
12% 11% 10% 9% 8% Tangential force error .40 Tangential force error . based on the similarity between propeller thrust and pump head as shown in equation (2.ε=0.0 and blade angle of 20 degrees.ε=0.60
FCFD/Fexact []
7% 6% 5% 4% 3% 2% 1% 0% 0 1% 2% 2 4 6 8 10 12 14 16 18
Figure 4.8
Angle of attack [degrees] Relative errors in tangential and axial forces due to overprediction of profile drag.60 Axial force error .0%. It is to be expected that this effect will be noticeable in both impeller and propeller torque calculations. 40% and 60%.40 Axial force error .
84
. The lift over drag ratio as function of the angle of attack is derived from figure 4.ε=0.Chapter 4. Mathematical treatment
To get an indication of the influence of the error in drag prediction on the axial and tangential force.0 to 8. the error in tangential force will increase even more.
dissipation rate
degrees degrees m2/s3 degrees kg/ms kg/m3 N/m2 1/s
axial direction
Numerical analysis of a waterjet propulsion system
85
.4.5 Nomenclature
4.5
A cd cl Cp D F g k L lmix P p Re t u+ v
Nomenclature
area drag coefficient lift coefficient pressure coefficient drag force gravitational acceleration turbulent kinetic energy lift force Prandtl’s mixing length production term of turbulence pressure Reynolds number time dimensionless velocity velocity distance from wall dimensionless wall distance coordinate in vertical direction m2 N N m/s2 m2/s2 N m m3/s2 N/m2 s m/s m m
∆yp
y+ z Greek symbols
α β ε ε φ κ µ ρ τw ω
Subscripts ax
angle of attack blade angle dissipation relative error in drag prediction inflow angle Von Kármán constant dynamic viscosity fluid density wall shear stress vorticity.
Mathematical treatment
CFD eff exact i. New York. pp 373382. A. ‘Investigation of tip clearance phenomena in an axial compressor cascade using Euler and NavierStokes procedures’. Vol 115. M.’Simulation of threedimensional unsteady flow in hydraulic pumps’.M. PhD thesis.y. Vol 114..J. 2000 [2] Van Esch.. 1993 [8] Ferziger. pp 453467.F.. A. 1985 [7] Wilcox.j imp inl T tan tip w x. Second edition. New York.P..’On the flow and cavitation inception of mixedflow impellers’. Basson. Berlin.. A. Peric..6
References
[1] Aungier. R. Glendale.. R. Griffin Printing. John Wiley & Sons. 1999
86
.H. University of Twente. M. 1998 [4] Kunz.Chapter 4.T.. 1997 [3] Van Os.. ‘Computational methods for fluid dynamics’. University of Twente. R. Journal of Turbomachinery.’Turbulence modeling for CFD’. B. ASME Press.C.z Superscripts v’ v
based on CFD results effective (=laminar + turbulent) based on exact formulations directions impeller inlet turbulent tangential direction impeller blade tip wall carthesian coordinate system directions
fluctuating part time averaged value
4. McDonald.W. 1993 [5] Goto.’Centrifugal compressors’.. B. 1992 [6] Fox..’Study of internal flows in a mixedflow pump impeller at various tip clearances using threedimensional viscous flow computations’. Springer. Third Edition. PhD thesis.H. Lakshminarayana.’Introduction to fluid mechanics’.H. Journal of Turbomachinery. D.. J.
I... 1958 [14] Moore. ‘An introduction to Computational Fluid Dynamics’..R.M. ‘Boundary layer theory’.. 1958
Numerical analysis of a waterjet propulsion system
87
.. 1974 [13] Abbott. ‘Controlling overproduction of turbulence in twoequation models by limiting the anisotropy of the Reynolds normal stresses’. A.4. Dover Publications. Smith. J.’Theory of wing sections’.. J.H. Langley Aeronautical Laboratory.G. W. Von Doenhoff. 1997 ASME Fluids Engineering Division Summer Meeting. Essex.. L. A. T.J.K. 1968 [12] Cebeci. H.O. New York. Felix.E. New York.. ‘StarCD methodology. NACA report 1368....150’. McGrawHill. Erwin. . Langley Field. 1997 [15] Emery. ‘Systematic twodimensional cascade tests of NACA 65series compressor blades at low speeds’.C. New York. H. Academic press... ‘Analysis of turbulent boundary layers’. A. Malalasekera. J. version 3. 1995 [11] Schlichting. Herrig. Longman Scientific & Technical. J.R. 2001 [10] Versteeg. Inc. Moore.6 References
[9] Computational Dynamics Limited.
Mathematical treatment
88
.Chapter 4.
During the validation process.
5.
Numerical analysis of a waterjet propulsion system
89
. but computed results were not validated with measurements. The CFD results can be used to visualise the flow behaviour in more detail.Chapter 5 Numerical analysis of waterjet inlet flow
In this chapter the flow through the waterjet inlet will be analysed in more detail. First a critical review of several published analyses of the flow through waterjet inlets will be given. For the calculations an Euler method was employed. [1]. This is caused by the neglect of the boundary layer velocity profile at the inlet and absence of viscous effects in the method. In order to validate the computational method the CFD results are compared with experimental data for a model scale waterjet inlet duct. the effects of the confinement of the flow due to the cavitation tunnel walls are also addressed.1
Review of CFD analyses on waterjet inlets
Calculations for a three dimensional waterjet inlet have been reported by Førde et al. it is also interesting to quantify wall friction and to determine the shape of the dividing streamtube. Besides pressure and velocity distribution. which provides additional information about the suitable numerical approach for the inlet flow CFD analysis. The presented velocity distribution at the impeller plane is not in agreement with a typical flush type waterjet inlet. Discussion of the application of the numerical method is divided into two parts: (i) the mesh generation for the three dimensional inlet geometry based on the experimental setup for the finite volume analysis and (ii) the selection of boundary conditions.
This nonuniformity vanishes towards the impeller plane.
90
. This is due to the neglect of viscous losses and due to small deviations in the calculated velocity field near the stagnation point at the cutwater. It is also found that the level of nonuniformity of the velocity profile at the bend increases with increasing IVR. the geometry of the pump is included. The effect of the impeller is modelled with an actuator disk. Results of both calculations were compared with experimental data. Calculations were made for a two and for a threedimensional case. The experimental data is based on measurements of a threedimensional inlet on a windtunnel. A small deviation of the angle of attack at the cutwater can lead to relatively large differences in the prediction of the static pressure distribution. whereas large deviations are found for the results of the calculations of the twodimensional geometry.Chapter 5. Comparison of the calculated pressure distribution along the ramp and the cutwater (or inlet lip) with experimental data shows large deviations. This transition is not taken into account in a twodimensional analysis either. A twodimensional model of a waterjet inlet is used for these analyses. Another example of the use of viscous methods for the calculation of threedimensional inlet flows was reported by Seil et al. Numerical analysis of waterjet inlet flow
necessary for describing the development of the boundary layer in the inlet duct. This behaviour may be a result of the implementation of the actuator disk. [6]. mounted on top of a windtunnel. Pylkkänen ([3]. [5]. A complete hull is included in the computational domain. Yang et al. The calculated pressure distribution along the inlet ramp centre line shows agreement with the measurements for the threedimensional analysis. however. A potential flow method has been applied by Van Terwisga [2]. Therefore it is expected that the pressure distribution in this part of the inlet will show a deviation from the threedimensional case as well. Differences between calculations and measurements of pressure are about 11 to 15%. [4]) presented results of CFD calculations obtained from a RANS code. the rectangular crosssection transitions to a circular crosssection in general. In these calculations. and therefore the pressure distribution at the lip section will be different. Agreement between measurements and CFD results is poor for the pressure distribution along the ramp and lip centre lines. Calculated results show that the location of the stagnation point at the cutwater depends on the IVR value. It is known that an actual threedimensional waterjet inlet ingests water from a region that is wider than the inlet itself. The necessity of considering the threedimensional geometry is shown by Van der Vorst et al. [7] have also calculated the flow around a waterjet inlet with a viscous flow method. The pressure measurements were performed for a model scale waterjet inlet. where the actuator disk is located. In a twodimensional situation this phenomenon cannot be reproduced. In the second part of the inlet.
it is very difficult to determine the exact operating conditions of the waterjet installation. all governing parameters of the operating condition can be measured accurately.1. This limits the use of measurement data obtained from full scale waterjet installations.2 Geometry and mesh generation
Experimental data from windtunnel tests has been used for comparison with calculations of the viscous flow through a threedimensional inlet geometry [8]. in accordance with normal waterjet applications.
5. the typical nonuniform velocity distribution in the impeller plane is reproduced well for the conditions considered. In a test setup with an inlet mounted on top of a cavitation tunnel. Some typical problems of model scale testing. The calculations are carried out to reproduce the flow phenomena of the measured conditions. If a wind tunnel is used for measurements instead of a cavitation tunnel. This data can be obtained from cavitation tunnel experiments. the air density has to be monitored simultaneously in order to be able to determine the flow rate accurately. The development described above has resulted in a widespread use of threedimensional viscous flow calculations for the analysis of the flow phenomena in waterjet inlets. like Reynolds scaling effects. These measurements are used for validation purposes in this chapter. However.2
Geometry and mesh generation
The analysis presented in this chapter is based on the geometry of the inlet as used in the experimental setup. This method optimizes the twodimensional symmetryplane geometry of the inlet. Relatively small variations in flow rate will result in a deviation of the IVR. Measurements on actual waterjet installations can provide validation data as well. artificial boundary layer thickening and confinement of the flow by tunnel walls are eliminated in these measurements. Moreover. Another important aspect in the experimental setup is the capability to create an incoming hull boundary layer of sufficient thickness. Validation of the CFD method requires an accurate set of experimental data. which is then extended in the third direction. for example. [10] have presented an optimization algorithm for waterjet inlet geometries. as described in subsection 3. Agreement between measurements and calculations is good for the static pressure distribution along the ramp for a range of IVR values.
Numerical analysis of a waterjet propulsion system
91
.59. Both the twodimensional as well as the threedimensional geometries are analysed using a viscous flow method. Hu and Zangeneh [9]. Accurate measurements of the static pressure at the ramp and the velocity distribution at the impeller plane were obtained in the Tom Fink cavitation tunnel [11]. A comparison between the calculated and measured velocity distribution is shown for an IVR of 1. On the other hand.5.1. these calculations are not always validated with measurements of static pressure and velocity.
Based on the geometric parameters the threedimensional shape of the inlet is calculated. The topology of the blocks is kept identical for all waterjet inlets. in order
Figure 5. the inlet angle.1
Inlet geometry with main parameters and specific nomenclature listed
The geometry of a waterjet inlet duct can be described by a number of parameters.g. It is therefore convenient to develop a fully parametric threedimensional geometry and mesh generator. A thin layer of cells is created at the walls of the inlet.1 shows a sketch of a twodimensional inlet geometry with the main parameters. e.2
92
Wire frame plot of block structure of the inlet mesh
. based on a list of geometric parameters. Figure 5. Numerical analysis of waterjet inlet flow
CYLINDRICAL LENGTH TRANSITION LENGTH
INLET SHAFT DIAMETER DIAMETER
SHAFT DIAMETER
TRANSITION ANGLE
RAMP RADIUS
RAMP TANGENCY POINT
INLET ANGLE BENDING RADIUS
CUTWATER
Figure 5. From this input data the block definitions are created. Figure 5.Chapter 5.2 shows a typical output of the used block structure. This geometrical data is used in the CFD preprocessor. the radius of curvature of the bend and the shape of the cutwater. the pump inlet diameter.
5.2 Geometry and mesh generation
Figure 5.3a
Final mesh of waterjet inlet (half of the complete domain)
Figure 5.3b
Detail of cutwater mesh with local refinement and regular mesh coupling on the block boundaries (left) and mesh around the (stationary) shaft
to get high quality cells in the boundary layer. The averaged y+ value of the cells near the wall is about 60 for this mesh. Near the socalled cutwater (or inlet lip), a local refinement of the mesh is applied to capture the gradients of the flow field better. The final mesh is shown in figure 5.3a and in figure 5.3b a detail of the cutwater with local refinement and the region around the shaft is shown. In case of computations
Numerical analysis of a waterjet propulsion system
93
Chapter 5. Numerical analysis of waterjet inlet flow
with a stationary shaft or without a shaft, the computational domain is restricted to half the geometry for reasons of symmetry with respect to the vertical plane through the centreline of the pump.
5.3 5.3.1
Numerical approach Boundary conditions
All cell faces at the boundary of the computational domain require some type of boundary condition. At the inlet of the computational domain an inlet type boundary condition is applied. This type of boundary condition requires a prescription of the velocities in all three directions and values for the turbulence intensity and the length scale, if a turbulence model is used. The velocity profile, which represents the hull boundary layer, is input through a userroutine. In this routine a powerlaw velocity profile is calculated for all cells that are located in the boundary layer. The presented calculations are based on a boundary layer profile with a powerlaw value n=7 and a thickness δ of 0.3D, with D the diameter of the inlet. An undisturbed uniform velocity is prescribed in the remainder of the cells. The turbulence intensity is set to 2.0% and the length scale to 0.05 m, which is equivalent to about 8% of the tunnel inlet hydraulic diameter. At the impeller plane the total mass flow leaving the domain through the inlet duct is imposed. This is established with a fixed flow outlet boundary condition. This type of condition allows for a nonuniform velocity distribution over the surface. The pressure distribution in the outflow plane is part of the solution as well. For the outflow plane of the cavitation tunnel a constant pressure boundary condition is used. It is assumed that the static pressure is uniform at large enough distance from the waterjet inlet. The resulting velocity distribution will be nonuniform, however. The side plane and the bottom plane of the domain are placed at the location of the cavitation tunnel walls. For these walls, the slip condition is applied as the wall boundary condition. The mesh near these walls can be made relatively coarse, since the boundary layer is not resolved. With this boundary condition the normal velocity is set to zero, which simulates the effect of the wall on the flow. The effects of the development of the natural tunnel wall boundary layer, characterised by the displacement and momentum thickness, are neglected. The effects of the actual boundary layer in the cavitation tunnel are analysed during the tests and it is concluded that the effect of blockage was less than 0.2%. For conventional waterjet inlet CFD analyses the tunnel walls are not taken into account. The side and bottom planes of the domain are modelled as
94
5.3 Numerical approach
constant pressure planes. Using pressure boundary conditions, additional inflow of water is allowed. The difference between wall boundary conditions and constant pressure boundary conditions is investigated later on in this thesis. Calculations for a half model geometry require symmetry conditions at the symmetry plane. If the complete model is analysed with shaft rotation, an additional wall boundary condition is applied to the shaft surface in order to model the rotation.
5.3.2
Fluid properties
Selection of the fluid properties for the waterjet inlet analysis is straightforward. As discussed in the previous chapter, the flow can be considered as incompressible, which results in a constant density. All model scale calculations have been carried out with a density ρ of 1000 kg/m3. The dynamic fluid viscosity µ is set to 0.001 kg/ms. Turbulent flow behaviour is modelled with a turbulence model. Though the CFD method provides several different turbulence models with different levels of complexity, the well established highReynolds number flow kε turbulence model is applied for all calculations. This also implies the use of wall functions to impose the noslip boundary condition. It is acknowledged that the standard kε turbulence model has a moderate performance for some types of flow. These are (i) some external unconfined flows, (ii) flows with large extra strains (e.g. the flow in curved boundary layers, swirling flows), (iii) rotating flows and (iv) fully developed flows in noncircular ducts, see [12]. Results from the validation process will show whether the choice of this turbulence model for the present flow is acceptable. The effect of gravity is accounted for in the analyses, but it should be mentioned that this is only an additional postprocessing feature, since the density is constant.
5.3.3
Discretisation and solution algorithm
Solution of the partial differential equations requires a discretisation scheme. As for the turbulence models, several methods are provided within the CFD method used. Here all calculations have been performed employing the second order MARS scheme (short for Monotone Advection and Reconstruction Scheme) for the momentum equations. This second order method is least sensitive to the mesh structure and skewness [13]. The kε model turbulence equations are discretised with a first order upwind differencing (UD) scheme. The coupling between the velocity and pressure field is resolved with an iterative solution strategy based on the SIMPLE pressurecorrection method.
Numerical analysis of a waterjet propulsion system 95
Chapter 5. Numerical analysis of waterjet inlet flow
In this algorithm, originally put forward by Patankar and Spalding [14], the convective fluxes are evaluated from an estimated velocity field. Furthermore, an estimation for the pressure distribution is used to solve the momentum equations. The continuity equation then yields a pressure correction equation, which yields a pressure correction field. This pressure correction term is used in turn to update the estimated velocity and pressure field. This process is iterated until the velocity and pressure fields are converged. The system of discretised partial differential equations is solved with an Algebraic MultiGrid (AMG) algorithm. Figure 5.4 shows the convergence behaviour of a calculation for an IVR of 1.87. In order to accelerate the convergence, first 125 iterations are carried out with the first order UD scheme for the velocity components. The restart with MARS discretisation causes the step in the residuals of the momentum and the mass. The convergence criterion is set to 104 for the momentum, mass and turbulent kinetic energy equations.
1.0E+01 Uvelocity Vvelocity 1.0E+00 Wvelocity Mass Turbulent energy Dissipation 1.0E01
Residual
1.0E02
dissipation
1.0E03
1.0E04
mass turbulent energy u velocity w velocity
250 300
1.0E05
v velocity
1.0E06 0 50 100 150 200
Iteration number Figure 5.4 Plot of convergence of a calculation of waterjet inlet flow phenomena for IVR = 1.87.
5.4
Validation with experimental data
The experimental data used in this chapter is measured at the Tom Fink cavitation tunnel [11] as described in subsection 3.1.1. In this test program measurements are made on two different inlet geometries. The experimental program consists among others of static pressure measurements along the
96
The locations are determined by the distance along the ramp centre line from the impeller plane towards the entrance of the inlet. Table 5.00 vpump [m/s] 7. and the pressure downstream of the inlet vary with the value of IVR value (see equation (2.ρv tunnel 2 (5.30 x 105 8. is located at a distance of 1000 mm for this model scale inlet.1 Parameters of measured conditions for the static pressure distributions.4.50 1. since the velocity.
5.42 x 105 5. The Reynolds number Reinl is defined in equation (4.65 Reinl [] 11.00 8. The entrance of the inlet.00 8.1 shows the conditions.33 4. total pressure measurements at the plane just upstream of the impeller.94 3.03 2. IVR [] (=vtunnel/vpump) 1.28 3.00 8. In the following sections.48 6.29 1. The static pressure is made nondimensional using the density ρ and the tunnel speed vtunnel: p – p ref C p = 1 2 .4 Validation with experimental data
ramp centre line. cavitation inception observations at the cutwater and visualisation of streamlines.70 1.21 1.12)).1)
The reference static pressure pref and the tunnel speed are taken at a location upstream of the inlet.1
Comparison of static pressure along the ramp centre line
The static pressure is measured at the ramp centre line at 12 different locations. sometimes denoted as ramp tangency point.00 8.19 vtunnel [m/s] 8.91 x 105 5.2).07 1.00 8.48 x 105
Numerical analysis of a waterjet propulsion system
97
.06 x 105 6. The static pressure is measured for eight different IVR values.00 8.00 x 105 7.21 x 105 9.00 8.5.61 6.71 4.20 5. used for the measurements. Dinlet = 150 mm.92 x 105 9. Table 5.87 2. measured data will be used for comparisons with CFD results.
3
0.19
0.87 IVR = 2.21 IVR = 1.
98
.3 0 200 400 600 800 1000 1200
Distance along ramp [mm]
0. The data for different values of IVR are divided in 2 groups for improved visibility.1
0.5 Comparison of measured and calculated pressure coefficient Cp along the ramp centre line.3 0 200 400 600 800 1000 1200
Distance along ramp [mm] Figure 5. Numerical analysis of waterjet inlet flow
0.3
0.1
0
0.50 IVR = 1.2
0.4
Experiments CFD IVR = 1.29 IVR = 1.70 IVR = 2.2
0.07 IVR = 1.5
0.1
0. Values are given as a function of the distance from the impeller plane.2
Cp []
0.Chapter 5.5
0.4
Experiments CFD IVR = 1.2
Cp []
0.1
0
0.03
0.
The movement of the stagnation point. As a consequence.6 shows the pressure distribution in the symmetry plane of the configuration for an IVR of 1. This influence is strongest at low IVR since the velocity in the inlet duct is then highest. This phenomenon is reported before by Seil [6]. From the pressure distributions can be seen that the location of the stagnation point at the cutwater changes with variation of IVR.29 the minimum pressure is found at a location with positive s coordinate.5. the location of minimum pressure is found at a negative s coordinate.
Numerical analysis of a waterjet propulsion system
99
. The static pressure at the impeller plane increases with increasing IVR (due to decreasing vpump) according to the expectations. On the left part of the diagram the negative s coordinates represent the lower part of the cutwater. This is due to increased nonuniformity and hydraulic losses. Analysis of the pressure in 5. The locations of the minimum pressure can be divided in two groups. The two diagrams present the data for alternate values of IVR. when the location of the minimum pressure is at the lower side (or hull side) of the inlet for high IVR conditions.03 (bottom).5. It is shown in more detail in figure 5. The influence of IVR on the results can be recognised clearly. For IVR conditions up to 1. and consequently the location of the minimum pressure. Values of pressure coefficient Cp along the ramp centre line are given as function of the distance from the impeller plane. Clear differences in pressure can be recognized at the cutwater and in the bend of the inlet.7. The cavitation inception results will be compared with CFD results in the next section.07 (top) and 2. Both figures show a good agreement between measurements and CFD calculations along the entire ramp section. this variation is dependent on IVR.5. The stagnation point moves from a negative s coordinate in the positive direction for increasing IVR.7 learns that the static pressure at the tunnel outlet plane is not equal to the reference pressure at the inlet plane and moreover. It can be observed that the effect of IVR on the pressure distribution is not restricted to the ramp. The discontinuity in calculated pressure between 350 mm and 400 mm is due to the presence of the shaft.4 Validation with experimental data
The comparison with CFD results is shown in figure 5. the location of minimum value of the pressure also changes. For higher IVR. Figure 5. whereas the positive s coordinates represent the upper part of the cutwater. However. This deviation in static pressure might influence the experimental determination of the cavitation inception pressure. where the pressure distribution along the cutwater is presented for all calculated IVR conditions. The streamtube analysis will be given in section 5. is related to the change in the shape of the dividing streamtube with varying IVR. at high IVR values the static pressure is more or less constant.
03 (bottom).Chapter 5. Numerical analysis of waterjet inlet flow
Figure 5.07 (top) and IVR of 2.6
Pressure coefficient Cp at symmetry plane for IVR of 1.
100
.
The cavitation inception pressure is presented in nondimensional form. For four IVR values the inception pressure is determined by visual observation.5
1
IVR = 1.07
Cp []
0.7 Calculated pressure coefficient Cp along the cutwater for different IVR values.ρv pump 2 (5.07 IVR = 1.03 IVR = 2. With this procedure only cavities at the lower side of the cutwater could be observed.
5.2)
where pv is the vapour pressure of the fluid.19
0
IVR=1. Negative s coordinate represents lower side of cutwater and positive coordinate represents upper side. These are typical for high IVR conditions. During the tests the reference pressure pref in the tunnel is gradually reduced until a small cavity is observed. h the height correction between the cutwater and the centre plane of the cavitation tunnel and vpump the average axial inflow velocity of the pump.5 100
80
60
40
20
0
20
40
60
80
100
Distance s along cutwater [mm] Figure 5.5
IVR = 1.50 IVR = 1.4 Validation with experimental data
1.2
Comparison of cavitation inception pressure at cutwater
Cavitation below the cutwater occurs at high IVR conditions. as shown in figure 3.
Numerical analysis of a waterjet propulsion system
101
.4.70
s>0 s<0 s=0
IVR = 1.29 IVR = 1. The measuring point for the reference pressure pref is located upstream of the inlet at half the tunnel height.07
0.5
IVR=2.5
1
IVR=1.1 on page 50. The comparison of the experimental results and the CFD results is shown in figure 5.19
1.87 IVR = 2.8. according to: p ref – p v – ρgh σ v_pump = 1 2 .5.19
2
2.21
IVR=2.
At a certain IVR the dividing streamline is optimally aligned with the cutwater geometry. With the current σ definition the inception behaviour of different inlet geometries at high IVR can be evaluated better.4
IVR [] Figure 5. this results in
102
. The mass flow which enters the tunnel is split into a part which leaves the domain through the waterjet inlet and a part which leaves through the tunnel exit section.4 1. i. Agreement between experiments and CFD calculations is good for all four tested conditions. whereas the ship speed is used for Cp.8 2 2.6 1. As long as the pump exit is open the mass flow rate at the exit of the tunnel section is lower than at the inlet.2 2.
7
6
σv_pump []
5
4
3
2
Measurements 1 CFD calculations 0 1 1. Figure 5. which results in a minimum value for σv_pump.e. The calculated conditions at lower IVR values show the expected behaviour.Chapter 5. Numerical analysis of waterjet inlet flow
It should be noted that the pump inflow velocity is used in the definition of σv_pump. Further decrease of IVR results in the point of minimum pressure.7 shows that the pressure in this region is not constant and depends on IVR. moving to the inner side of the inlet.2 1. Because the tunnel crosssection is constant. the cavitation inception point. This is due to the effect of tunnel walls and conservation of mass in the complete system.8 Comparison of measured and calculated cavitation inception at cutwater
Effect of tunnel walls At normal waterjet operating conditions cavitation inception occurs below the cutwater in general.
for a first indication of the effect of the confinement of the tunnel
0. hydraulic losses in the tunnel are not taken into account in the estimation. The pressure increase as a function of IVR can be estimated as:
2 2A pump A pump ∆C p = .3) neglects the effect of the displacement thickness of the boundary layer on the tunnel walls in the calculation of the mass fluxes. However.3)
0 1 1.1 0. Moreover.05 0.06 0.3)
where Apump is the crosssectional area at the impeller plane and Atunnel the crosssectional area of the tunnel. The effect of viscous losses is neglected in this estimation. This simplification is assumed to be justified.09 0. The diameter of the crosssectional area at the impeller plane is 150 mm and the cavitation tunnel has a square section of 600 x 600 mm.0 to 2% for IVR = 2.02
CFD
0.4 1.9
IVR [] Dimensionless pressure increase in tunnel section based on CFD results and analytical formula
Numerical analysis of a waterjet propulsion system
103
. These hydraulic losses will reduce the actual pressure increase in the tunnel. The analytical pressure difference estimate will increase with less than one percent in this case.8 2 2. which is equivalent to a lower IVR.01
Equation (5.4 Validation with experimental data
lower velocities and consequently a higher static pressure. Inclusion of this displacement thickness will result in a slightly lower mass flow entering the tunnel section. Figure 5.6 1.9 shows the pressure difference ∆Cp based on the CFD results and the theoretical value.– .2 2. IVR ⋅ A tunnel IVR ⋅ A tunnel
(5. Agreement between the analytical value and the numerical result is good over the complete IVR range.4
Figure 5.07
∆Cp []
0. Equation (5.5.03 0.08 0.0.04 0. The velocity at the outlet of the tunnel shows a drop of about 5% for an IVR = 1.2 1.
wall 0. The deviations between the two series of calculations are large. the cavitation inception diagram of the inlet is matched with the available ambient pressure. This is in accordance with the expectations.2 Inlet geometry
Cp []
0.1
0. due to the wall confinement. Effect of the tunnel walls on static pressure distribution along the ramp is shown in figure 5.4 IVR = 1.11.07 .3). is found in the tunnel downstream of the cutwater. the deviations are larger for the low IVR condition.03 . For a cavitation free design. Numerical analysis of waterjet inlet flow
0.3 IVR = 2. It is to be expected that the largest influence of the tunnel pressure increase. Further downstream in the inlet duct the pressure distribution is similar for both configurations. Clear deviations between the calculation with and without the walls can be seen in the first part of the inlet. Moreover.1
0
0. These are the same conditions as shown in the symmetry plane pressure distributions of figure 5. based on experiments or calculations with tunnel
104
.wall 0. The effect of the tunnel walls on cavitation inception pressure is shown in figure 5. In this region the pressure is influenced most by the local tunnel pressure.3) can be used as a first estimate. cavitation free operation is possible.no wall 0. The cavitation inception diagram represents the required pressure to avoid cavitation. In actual conditions at open sea the equivalent cross sectional area goes to infinity and consequently the pressure difference tends to zero. Figure 5.no wall IVR = 2.10 Comparison of calculations of static pressure distribution along ramp for configuration with and without tunnel walls
walls on the pressure distribution.2
0. As long as the required pressure is lower than the available ambient pressure.11 shows that the required pressure.07 .5 IVR = 1. equation (5. This is in line with equation (5. This is the location where the cavitation inception occurs at medium and at high IVR.Chapter 5.03 .3 0 200 400 600 800 1000 1200
Distance along ramp [mm] Figure 5.10 for two IVR values.6.
4
IVR [] Figure 5. The actual required pressure is higher. The results are made nondimensional with the tunnel speed in a similar way as the static pressure.19. i. This effect should always be taken into consideration when interpreting cavitation inception results in a test setup with finite dimensions. The lowest IVR is a normal operating condition of 1. just upstream of the impeller. In the tangential direction steps of 10 degrees are made between the measurement locations.2 1.03.12d show the comparison between the measured and the calculated total pressure distribution for four different IVR conditions.
5.4 1.87 and 2.12a to 5.5.2 2.11 Comparison of calculations of cavitation inception pressure for configuration with and without tunnel walls
walls is too optimistic for medium and high IVR. Figures 5.68 and the highest IVR is a very high speed condition of 2.e.6 1. The other two IVR conditions are in between with values of 1.3
Comparison of total pressure at impeller plane
Good agreement between measurements and calculations is shown in the previous subsections for the static pressure along the ramp and the cavitation inception that occurs at the cutwater. In this subsection the total pressure distribution in the impeller plane is evaluated.4 Validation with experimental data
7
6
5
4
σ []
3
2
Measurements 1 CFD with tunnel walls CFD without tunnel walls 0 1 1. Total pressure measurements have been made with a pitotrake positioned at different radii.4.:
Numerical analysis of a waterjet propulsion system
105
. The impeller plane is defined as the crosssectional area at the end of the inlet.8 2 2. as shown in figure 3.1. as can be seen from the calculations without tunnel walls.
Figure 5.ρv tunnel 2
(5.12a Comparison of measured (left) and calculated total pressure distribution in impeller plane for IVR = 1.e.Chapter 5. i. In the region around 12 o’clock.12b Comparison of measured (left) and calculated total pressure distribution in impeller plane for IVR = 1.68
Figure 5.4)
The experimental data shows totalpressure losses up to 65% at the highest pump speeds (low IVR) in the wall boundary layer. the region affected by the flow around the stationary shaft losses are found up to 50%.87
106
. Numerical analysis of waterjet inlet flow
p tot – p ref C ptot = 1 2 .
This means an increase of hydraulic losses with increasing IVR. measuring with a total pressure tube close to the wall will be troublesome. Comparison of the four IVR conditions shows that in the region affected by the presence of the shaft.4 Validation with experimental data
Figure 5.12c Comparison of measured (left) and calculated total pressure distribution in impeller plane for IVR = 2.19 For all four conditions the qualitative agreement between the measurements and the calculations is good. However.03
Figure 5. The typical distribution and the effect of the shaft is reproduced well within the CFD model.
Numerical analysis of a waterjet propulsion system 107
. a decrease of the total pressure level with increasing IVR takes place. This increase is in accordance with the expectations and this is partly due to the increased retardation of the flow as discussed in chapter 3.5.12d Comparison of measured (left) and calculated total pressure distribution in impeller plane for IVR = 2. It appears that the computed boundary layer is thinner than in the experiment.
13a Comparison of axial velocity derived from measured total pressure (left) and calculated axial velocity distribution in impeller plane for IVR = 1.13a to 5.4.4
Comparison of velocity field at impeller plane
Total pressure measurements of the previous subsection are used to derive an axial velocity field.68
108
.ρ 2 (5.Chapter 5. The axial velocity is derived from the experimental data according to: v axial = p tot – p stat 1 .
Figure 5. moreover. The comparison between the value of the axial velocity derived from the measured total pressure and the calculated axial velocity component is shown in figures 5. In this derivation a constant static pressure over the crosssection is assumed. Numerical analysis of waterjet inlet flow
5. the influence of the inplane velocity components is neglected.13d in a similar way as for the total pressure distribution.5)
Integration of the axial velocity over the impeller plane showed that the flow rate was predicted within 2.5% compared to the measured flow rate. The axial velocity is normalised with the averaged axial velocity.
4 Validation with experimental data
Figure 5.5.13b Comparison of axial velocity derived from measured total pressure (left) and calculated axial velocity distribution in impeller plane for IVR = 1.13c Comparison of axial velocity derived from measured total pressure (left) and calculated axial velocity distribution in impeller plane for IVR = 2.03
Numerical analysis of a waterjet propulsion system
109
.87
Figure 5.
1).
110
.6)
with vEXP based on the two dimensional Fourier representation of the measured data. The location and magnitude of minimum and maximum axial velocity is predicted well with the numerical method.Chapter 5.19 Qualitative agreement between experimental data and calculated velocity field is good for all four IVR conditions.13d Comparison of axial velocity derived from measured total pressure (left) and calculated axial velocity distribution in impeller plane for IVR = 2.19. Numerical analysis of waterjet inlet flow
Figure 5.14 shows the relative difference for the both low IVR of 1. Therefore. Figure 5. The deviation between the two is used to quantify the relative error of the calculated axial velocity. It is concluded that the CFD analysis of the inlet flow reproduces the typical nonuniform velocity distribution well. The deviations are below 15% for a significant part of the crosssectional area. The numerical results of the velocity distribution are compared to the representation of the measured data by Fourierseries (see section 3. The relative difference is defined as: v CFD – v EXP ∆ = v aver (5. the CFD method employed in the present investigation seems suitable for the investigation of the effect of nonuniform inflow into the mixedflow waterjet pump.68 and the high IVR of 2.
5.4 Validation with experimental data
Figure 5.e.68 (left) and IVR = 2. Figure 5.
0. The small deviations are negligible. Both the results of calculations with wall and pressure boundary conditions have been used.14
Relative difference between experimental data and calculated axial velocity distribution at impeller plane for IVR = 1.2
0.05
Pressure boundaries
0 0. The nonuniformity can be represented in a single value.3
IVR [] Figure 5.9 1.5 1.7 1. The relation between the IVR and the nonuniformity is shown clearly for the two types of boundary conditions.15 Nonuniformity parameter ζ in impeller plane as function of IVR
Numerical analysis of a waterjet propulsion system
111
.15 shows the nonuniformity parameter ζ as a function of IVR.25
0.19
The velocity distributions show a clear increase of nonuniformity with increasing IVR.1).i.4
0.1 2.3 1.9 2.3
0. if calculated according to equation (3.1 1.35
Nonuniformity ζ []
0.15
0.1
Wall boundaries
0. decreasing vpump.
In the region in the bend above the shaft a very low velocity magnitude is observed.16. low values of vpump. At low IVR conditions the effect of the bend on the velocity distribution can be recognised. Here flow separation is likely to occur at sufficiently high IVR. Numerical analysis of waterjet inlet flow
Figure 5. i. is shown in figure 5.e. The method to determine boundary layer separation will be discussed in section 5.16
Distribution of magnitude of velocity in symmetry plane for IVR 1. The distribution of the magnitude of the velocity in the symmetry plane for two different values of IVR. This effect is less pronounced at high IVR conditions.6.Chapter 5. For such conditions there may even be a region of flow recirculation upstream of the impeller plane. At the inner corner of the inlet the velocity reaches a maximum value which reduces further downstream.03 (bottom)
112
. where additional flow phenomena in a waterjet inlet are reviewed.13d shows a region with very low axial velocity in the upper part of the crosssectional area.
Figure 5.07 (top) and 2.
17 shows the static pressure distribution along the ramp as calculated with the kω turbulence model.70 IVR = 2.29 IVR = 1. where the pressure predicted by the method employing the kω model gives higher values than the ones predicted by the method using the kε model.03
0.17 Comparison of measured and calculated static pressure along ramp centre line obtained with kω turbulence model
Numerical analysis of a waterjet propulsion system
113
.2
Cp []
0. Similar results have been found for intermediate IVR conditions.3
0.1
0
0. the level of accuracy or the absence of accuracy of a CFD calculation is attributed to the turbulence model used.07 IVR = 1. Though an indepth study of effects of turbulence models on the flow in waterjets does not fit within the scope of this thesis. results of calculations with the kω turbulence model will be shown for comparison with results of the kε turbulence model. Often. This suggest that flow separation occurs later for the method employing the kw model. The main differences are in the region downstream of the shaft. Comparison of static pressure along ramp centre line Figure 5.5
0.1
0. as shown in figure 5. The kω turbulence model is used widely just like the kε turbulence model.4
Experiments CFD IVR = 1.4. The accuracy of the calculations is comparable to the calculations with the kε turbulence model.5. Comparison of the experimental data with the results of calculations employing the kω turbulence model can show whether the result is sensitive to the choice of a particular turbulence model.5
Results obtained with kω turbulence model
The use of turbulence models in CFD calculations always provides reason for discussion.3 0 200 400 600 800 1000 1200
Distance along ramp [mm] Figure 5.4 Validation with experimental data
5.
0. Agreement between calculations and experimental data is good for all four presented conditions.2
0.5.
but also on the prediction of the velocity field. Numerical analysis of waterjet inlet flow
Comparison of velocity field at impeller plane The assessment of the performance of the kω turbulence model is not only based on the prediction of static pressure.19 for both conditions. The relative difference between calculations and data derived from the measurements is shown in figure 5.14. Figures 5. Agreement is good for both conditions.10%.Chapter 5.18a Comparison of axial velocity derived from measurements (left) and calculated axial velocity distribution with kω turbulence model in impeller plane for IVR = 1.18b
Comparison of axial velocity derived from measurements (left) and calculated axial velocity distribution with kω turbulence model in impeller plane for IVR= 2. see figure 5.18a and b show the comparison of the velocity distribution at the impeller plane for two IVR conditions.
Figure 5. The majority of the crosssectional area has a difference below +/.19
114
.68
Figure 5. Quantitative agreement seems to be slightly better with the kω turbulence model than with the kε model.
6
Mesh convergence study
The mesh as used in this study has been made coarser and finer to evaluate to mesh convergence of the applied mesh. Currently.4 Validation with experimental data
Figure 5.87 along the cutwater for the three different meshes.4.7 will be reviewed for the different meshes.
Numerical analysis of a waterjet propulsion system
115
. The number of cells of the original mesh. The fine mesh is derived from the original mesh by doubling all block divisions in all three directions. but it should be noted that this reduction is achieved mainly in the region of the cutwater. which are acceptable during the design phase. In the direction along the cutwater the number of cells is reduced by a factor of two as well for the blocks near the cutwater. the block divisions of the blocks near the symmetry plane of the original mesh are reduced by a factor of two in the direction perpendicular to the symmetry plane.5. It should be noted. the default number of cells has not been increased. Nevertheless it is now possible to run the refined mesh.19
Relative difference between experimental data and calculated axial velocity distribution with kω turbulence model in impeller plane for IVR = 1. which has about 1. Figure 5. because the largest gradient are present in this region.5 million cells.000. of about 200. The pressure distribution along the cutwater. The number of cells in the normal direction of the extrusion layer has been kept constant to remain at the same y+ values.000. For the coarse mesh.20 shows the pressure coefficient for IVR = 1. that the original medium mesh size has been developed about 5 years ago. were governed by the hardware constrains of that time. In order to have calculation times.68 (left) and IVR = 2. The reduction of cells from the original mesh to the coarse mesh is about 30. as shown in figure 5.19
5. the tools for generation of inlet meshes are used in the design procedure for waterjet inlet geometries.
2
Fine mesh (1.4 0.6
s=0
0.2 0. located between s range of 10 and 0.4
s>0
0.6
s>0 s=0 s<0
Medium mesh (206.20 Calculated pressure coefficient Cp along the cutwater for different mesh sizes for IVR of 1.2 0 0.4 0.Chapter 5.596 cells)
1
Medium mesh (206.596 cells)
0.8 0.448 cells)
Cp []
0 0.6 0.448 cells)
0.8 1 1. Negative s coordinate represents lower side of cutwater and positive coordinate represents upper side. Top figure shows s range of 100 to 100 and bottom figure shows region with location of minimum pressure.4
Cp []
0.2 0.884 cells) Fine mesh (1.
116
.564.884 cells)
0.6 0.2
Coarse mesh (175.2 10 9 8 7 6 5 4 3 2 1 0
Distance s along cutwater [mm] Figure 5.87.8 0.564.8 1 100
s<0
80
60
40
20
0
20
40
60
80
100
Distance s along cutwater [mm]
1
Coarse mesh (175. Numerical analysis of waterjet inlet flow
1.
this in order to avoid possible effects of the confinement on the flow of the presence of the cavitation tunnel walls. Calculations of full scale waterjet inlets only require a refinement of the cells near the solid walls in order to keep acceptable y+ values.4. can be taken into account for example. and therefore the kε model is used in the remainder of the analyses.5 Analysis of the suction streamtube
The region with the location of the minimum pressure is shown in detail in the second diagram.
Numerical analysis of a waterjet propulsion system
117
.7
Closing remarks
Agreement between the experimental data and the results of the calculations is quite good.5
Analysis of the suction streamtube
The preceding section covers the comparison of measured and calculated results. It is not an aim to benchmark turbulence models in this study. In this way. The difference is found along a small part of the cutwater region of about 2 mm. Typical quantities which can be measured in experimental facilities are total and static pressure. Velocities can be derived from this data afterwards. due to an increase in the length of a vessel. This is the capability to calculate the flow for geometries according to full scale dimensions with full scale boundary conditions. Examples include the calculation of shear stresses along the inlet surface and the determination of the suction streamtube. which is 1. The thickness of the actual hull boundary layer can be applied at the inflow boundary condition. One aspect of the use of RANS methods has not been addressed yet. the effect of a thicker boundary layer.5. The CFD results used in this section are the results obtained for the waterjet inlet without tunnel walls. Increase of the number of cells with a factor of 8 results in a difference of the prediction of minimum pressure ∆cp=0. The markers in this figure represent the corner points of the cells. The difference in pressure distribution along the cutwater between the original mesh and the fine mesh is regarded to be representative for the complete mesh. Since the deviations between the two meshes are limited to a small region near the cutwater and the magnitude of the difference is small. despite the mentioned deficits of the kε turbulence model.
5.
5.3% of the inlet diameter.06 for this condition. Performance of the kω turbulence is comparable. The CFD results provide a wide range of additional postprocessing capabilities to get more insight into the behaviour of the flow. is it concluded that the cell sizes of the original mesh are suitable for the presented study.
Setting the concentration factor to 100% in the impeller plane allows for the determination of the complete streamtube with a upstream tracing method.Chapter 5.21
Visualisation of threedimensional suction streamtube for IVR=1. In this approach the flow field is reversed and frozen. This will give an indication of the risk of the onset of boundary layer separation in the inlet duct.21 from different view angles.2
Determination of suction streamtube shape
The shape of the suction streamtube at the inlet boundary of the domain can be used to calculate the average ingested velocity vin. The shape of the streamtube is not known in advance. An example of the threedimensional streamtube visualisation is shown in figure 5.5.5. which is allowed in a steady flow problem. The shape of the streamtube is derived from an isosurface plot of the concentration. In fact. Determination of the crosssectional area of the streamtube at several locations upstream of the impeller plane provides information about the diffusor effect in the streamtube. the solution of the scalar only takes a few iterations for the complete numerical domain. This concentration scalar can be implemented at inlet or pressure boundaries and it can be used as a weight function for further analysis.
118
. only at the impeller plane the streamtube is known to coincide with the impeller plane boundary.1
Visualisation of suction streamtube
The CFD method provides a way to introduce a concentration scalar as a passive traces in the flow field. Numerical analysis of waterjet inlet flow
5.29
5. With the frozen velocity field. The values of the crosssectional areas can be used to calculate the equivalent diffusor angles.
Figure 5. In this way a clear representation of the streamtube interface can be obtained. This velocity is used in the calculation of the wake fraction w and the thrust of the installation.
In the calculations a powerlaw exponent n=7 and a boundary layer thickness of 0.70 IVR 2.1
0.2 0. These two parameters are determined by employing a leastsquare fit of the data determined from the computed streamtube surface crosssection at the inlet plane of the computational domain.0 1.5 Analysis of the suction streamtube
Crosssectional shape of streamtube at inlet boundary The shape of the suction streamtube at the inlet boundary of the numerical domain can be approximated with a semielliptical shape.6 0.03 .29 .0 0. The boundary of the streamtube can be approximated by the ellipse: y z  2 +  2 = 1 w 0 h 0 (5.03 IVR 1.
0
0.3 IVR 1.4 IVR 1. Once the concentration scalar is
Numerical analysis of a waterjet propulsion system
119
.29 0.elliptic fit 0.7)
where 2w0 and h0 are the maximum width and maximum height resp.22 Computed semielliptical shape of suction streamtube at inlet of numerical domain for various IVR conditions
An efficient method to determine the inflow velocity is based on the integration of the streamtube velocity.elliptic fit IVR 1.elliptic fit 0.22 for a range of IVR values.07 . An example is given in figure 5.70 .elliptic fit IVR 2.4 0.2
z/D []
0.07 IVR 1.8 1. with D the inlet diameter.2
y/D [] Figure 5.3D. is used.5 IVR 1.5.6 0. The elliptical curve fit can be used to determine the mass averaged inflow velocity of the ingested fluid out of the boundary layer.
3D.6D and 1.9 D
0.3D is an empirical factor. rectangular box method and direct integration of CFD results. The rectangular box approach is also plotted for three different widths of the box.9D.08
0. The other two curves for the rectangular box.Chapter 5.3 D Rectangular box .16
0.
120
.1.9 1.6 D Rectangular box .3
IVR [] Figure 5.1.06
0.12
0.23 shows the calculated wake fraction w (as defined in equation (2. gives an underestimation of the wake fraction of about 20%.1
0.8)
2
A in flow v in = ρ ( c ⋅ v x ) dA
In this equation the concentration scalar is denoted by c and x is the direction normal to the inflow area Ainflow.9 2. Figure 5.2)) based on the elliptical streamtube shape and on direct integration of the CFD results. which has been derived in the past to obtain good correlations with the actual sailing fleet.5 1. based on 1.
0. the mass averaged inflow velocity can be determined by simple integration:
∫ ∫
A in flow
ρ ( c ⋅ v x ) dA (5. show that a width depending on IVR is required to obtain a good fit.1 1.23 Calculated wake fraction w based on elliptical fit. Numerical analysis of waterjet inlet flow
available.1.04
Ellips curve fit Direct integration Rectangular box .7 1.3 1. The method.1 2.
The method of direct integration and the method of the elliptical curve fit give more or less comparable results for the wake prediction.02
0 0. based on a rectangular streamtube with a width of 1. It is acknowledged that the value of 1.14
Wake fraction w []
0.
in the range 1 to 9.
Numerical analysis of a waterjet propulsion system
121
. The crosssectional area has been determined at 9 different stations.24. The parameter s is the length along the waterjet inlet contour in the plane of symmetry. The area is determined by integration of the concentration factor over the whole plane: A tube =
∫ c dA
An
(5.
Figure 5.5 Analysis of the suction streamtube
Crosssectional area of streamtube In the preceding section the crosssectional shape of the streamtube at the inflow boundary has been discussed.
Figure 5.24
Location of streamtube crosssections. Here the development of the crosssectional area of the streamtube will be analysed in more detail. The distances are calculated from the first plane to the throat area.5. The crosssections are taken perpendicular to the ramp surface from the inlet boundary to the cutwater. This area is denoted as intake throat area [2]. as shown in figure 5.9)
where n is the integration plane index number.25 shows the development of the crosssectional area of the streamtube. The areas are normalised with the streamtube area at plane 9.
2
Normalised area A/A9[]
2
1
3
4
5
6
7
8
9
0. This diffusor angle is based on the equivalent streamtube diameter and the distance dl between the different crosssections: Dn + 1 – Dn α diff = atan  dl (5. Numerical analysis of waterjet inlet flow
1.70 0. near the inflow plane.
Far upstream. which is due to an acceleration of the flow.10)
The equivalent diameter is based on a circular section with area identical to the area of the streamtube crosssection.19 0 0 1 2 3 4 5 6 7 8
Dimensionless distance s/D [] Figure 5. The numbers refer to the integration plane index numbers.4 IVR = 1.87 IVR = 2.29 IVR = 1.e. can be recognised as a small region of reduced area.6 IVR = 1.21 IVR = 1.07 0. at crosssection numbers 4 and 5.
122
.24. The influence of the ramp curvature.Chapter 5.8
0.25 Streamtube crosssectional area normalised by A9. The development of the streamtube in streamwise direction can also be expressed in an equivalent diffuser angle αdiff. With this curve fit the diffuser angle αdiff can be derived in small steps of the length dl.03 IVR = 2. a constant crosssectional area is found for each IVR condition. The typical retardation of the flow at higher IVR (i.2 IVR = 1. lower vpump) conditions is reflected in a steep increase of the streamtube area when approaching the inlet lip. as shown in figure 5.26 for various IVR values. The variation of the equivalent diameter along the streamtube in streamwise direction has been fitted with a fifth order polynomial. The development of the diffusor angle along the streamtube up to the cutwater is shown in figure 5.50 IVR = 1.
flow separation is likely to occur in normal circular diffusers [15].50 IVR = 1.07 10 IVR = 1.1.87 and higher.5 in subsection 5. Flow separation will lead to increased nonuniformity of the flow in the impeller plane and higher hydraulic losses.5).03 IVR = 2.
5.70 IVR = 1. At such high diffuser angles. the diffuser angle exceeds 8 degrees.87 6 IVR = 2. all conditions show a small negative value of αdiff.26 Equivalent diffuser angle αdiff of streamtube
At the ramp curvature (around s/D = 3.4. Occurrence of this large pressure gradient at high IVR (i. The occurrence of flow separation will be discussed in more detail in the following subsection.6
Evaluation of wall shear stress
Evaluation of the wall shear stress in the inlet duct can provide information about the probability of flow separation in the inlet.21 IVR = 1. Boundary layer separation will occur inside the inlet duct at the ramp side. Determination of a possible region of separated flow is based on the magnitude of the wall shear stress along the duct.29 IVR = 1.e. because of the strong adverse pressure gradient acting on the flow.19
αdiff [degrees]
8
4
2
0
2
4 0 1 2 3 4 5 6 7 8
Dimensionless distance s/D [] Figure 5.5. At IVR conditions of 1. which represents the contraction of the streamtube.6 Evaluation of wall shear stress
12 IVR = 1. It is assumed that the onset of separation occurs at the location where the wall shear stress
Numerical analysis of a waterjet propulsion system
123
. This is a deterioration of the performance of the whole propulsion system and should be avoided in the range of normal operating conditions. low vpump) has already been demonstrated in figure 5.
7 1.3 1. According to the wall shear stress criterion. The wall shear stress component in axial direction will change sign in case of boundary layer separation in this region.02
0.27 shows the result of the evaluation of the minimum wall shear stress at the waterjet inlet duct part. For the standard flushtype inlet geometries.3
0. The dimensionless wall shear stress is denoted as the dimensionless friction velocity vf: vf = τw 2 ρv pump (5.24). Figure 5.1 1.Chapter 5.
0. possible flow separation is located inside the duct near station 9 (as shown in figure 5.9 1. For a practical inlet design. With a dedicated inlet geometry design it is possible to avoid flow separation in the inlet for all operating conditions.9 2. Numerical analysis of waterjet inlet flow
reduces to a zero value [16].01
0 0.26).03
IVR [] Figure 5. Results of calculations with both wall as well as pressure boundaries are used in this evaluation. flow separation will occur for IVR values higher than 1. For the detection of flow separation. flow separation should not occur at normal operating conditions.1 2.27 Dimensionless wall shear stress as function of IVR
124
. the friction velocity is multiplied with the sign of the axial wall shear stress.5 1. The difference between the results of both series of calculations is negligible.75.11)
The nondimensional representation can be used to get a more comprehensive comparison of calculations for varying pump velocity. This is in accordance with the maximum allowable diffuser angle criterion of 8 degrees (as shown in figure 5.02
Friction velocity vf []
0.03
Wall boundaries Pressure boundaries
0.01
0.
J.C. N. 1994
Numerical analysis of a waterjet propulsion system
125
. VTT Manufacturing Technology. Technical report VTT VAL B11.7
A c Cp g h h0 IVR p p0 pv v vf w0 y. 6989. Trondheim. Kubberud. Technical University of Delft.’Waterjethull interaction’.. ‘Computational Fluid Dynamics applied to high speed craft with special attention to water intake for water jets’. T.. Maritime Research.. Ørbekk. PhD thesis. E..’Test cases of application of CFD code to predict waterjet inlet flows’.8
References
[1] Førde.7 Nomenclature
5. pp. Espoo. 1991 [2] Terwisga. 1996 [3] Pylkkänen.5. van. FAST’91 conference. J.z
Nomenclature
area concentration pressure coefficient gravitational acceleration height from tunnel centre line to cutwater maximum suction depth inlet velocity ratio (vship/vpump) static pressure static ambient pressure in tunnel vapour pressure velocity friction velocity maximum suction half width coordinates fluid density cavitation inception pressure wall shear stress normal direction based on values just upstream of the impeller based on tunnel values axial direction m2 m/s2 m m N/m2 N/m2 N/m2 m/s m m kg/m3 N/m2
Greek symbols
ρ σ τw
Subscripts n pump tunnel x
5. M.V.
T. Y.J. 13th Fast Ferry International Conference.. Fletcher. ’Introduction to fluid mechanics’. 1985
126
. Vol 121. Proceedings of the 5th international conference on Fast Sea Transportation.J. R.‘Influence of boundary layer ingestion on waterjet performance parameters at high ship speeds’. van der..Chapter 5. pp 883892. McGrawHill. New York. [16] Fox. H.J.. ‘The application of computational fluid dynamics to practical waterjet propulsion system design and analysis’. Int.. D. 1968.W.. Seattle. M. Malalasekera. Seattle.150’.J. International conference on ship and marine research.. New York.V. M. pp 13911401.J.M..’CFD analysis and iterative design of waterjet inlet’. P.A. 1999 [11] Brandner P.. Zangeneh. J. Longman Scientific & Technical. Numerical analysis of waterjet inlet flow
[4] Pylkkänen.V.. Vol 15. G.W. ‘A method for automatic optimisation of the intake duct geometry of marine waterjets’. Lee. McDonald. Verbeek. H.. 1999 [10] Hu. Third edition.J. John Wiley & Sons... R.K.. N. ‘An introduction to Computational Fluid Dynamics’. 1999 [9] Hu. 1995 [7] Yang. Walker. Doctors.. Proceedings Waterjet Propulsion III conference. 396404. ‘StarCD methodology..S. C. A. Zangeneh. H. L. pp 843851. Heat Mass Transfer.. ‘Boundary layer theory’. ‘A numerical analysis of the flow around the waterjet inlet’. W. H. ‘A calculation procedure for heat. A. Rome. pp. Proceedings of the 5th international conference on Fast Sea Transportation.. J. LubeckTravemunde. Proceedings of the 3rd international conference on Fast Sea Transportation.. P.. LubeckTravemunde. 2001 [14] Patankar.. Singapore. 1995 [13] Computational Dynamics Limited. Hendriks. pp. G.. mass and momentum transfer in threedimensional parabolic flows’. 1995 [8] Bulten.’A waterjet test loop for the Tom Fink cavitation tunnel’. ‘Investigations of 3D turbulent flow inside and around a waterjet intake duct under different operating conditions’. Gothenburg. 17871806 [15] Schlichting. Spalding.. Proceedings of the 3rd international conference on Fast Sea Transportation.A. ‘Design considerations of waterjet propulsion systems’..B.H.K. C. Essex. S. pp 13791390. 1994 [5] Vorst. 2001 [12] Versteeg. 1997 [6] Seil. version 3. ASME Journal of Fluids Engineering. Choi.A.B.
In section 6. The geometry and mesh generation procedure will be discussed in section 6. are used as input velocity distributions. This data is used for the prediction of the performance of the full scale installations. i. This can be partly attributed to the applied boundary condition of constant pressure at the outlet part of the numerical
Numerical analysis of a waterjet propulsion system
127
.1. The experimental data is measured during the development of the waterjet pump and is only available at the authors’ company. the choice of boundary conditions and the options for implementation of impeller rotation are reviewed.Chapter 6 Numerical analysis of waterjet pump flow
This chapter deals with the numerical analysis of the mixedflow waterjet pump. The axial velocity distributions as shown in the previous chapters. reducing the calculation to computing the flow through a single impeller channel showed poor agreement with experimental data. Its effect on the radial loading of the impeller is treated as well. in section 6.2. results of the CFD calculations with uniform inflow are compared with available experimental data to get an indication of the accuracy of the numerical method used. Preliminary calculations assuming periodic flow.1
Geometry and mesh generation
The numerical domain for the analysis of the flow through the mixedflow pump includes the complete impeller and the stator bowl with guide vanes. the stator bowl and a straight suction pipe.
6.3. The influence of a nonuniform inflow on the performance is presented in section 6. Subsequently. The numerical domain encloses the impeller.e.4.
Figure 6. Figure 6. Numerical analysis of waterjet pump flow
domain. The choice to model the complete impeller and stator enables the use of a constant pressure boundary condition at the outlet.
nozzle
inlet
sliding interfaces Figure 6.2 shows the blade geometry for two blades.1 Sketch of rotorstator pump configuration
The geometry of the impeller is based on seven blade profile sections equally spaced in radial direction and the blade root fillet geometry.Chapter 6. based on the streamline profile sections.1.2
Blade geometry based on streamline profile sections
128
. Upstream of the pump a straight suction pipe is incorporated in the numerical domain. The blade profiles provide detailed information of the shape of the blade sections near the leading and trailing edges. A sketch of the rotorstator configuration is shown in figure 6. The two dotted lines indicate the locations of the sliding interfaces between the stationary and the rotating domains in the mesh.
water can flow over the tip from pressure to the suction side of the blade. Treatment of the tip clearance is in agreement with the findings of several studies of turbo machinery addressing tip clearance flow phenomena [14]. Figure 6. At the seatring. The layer of extruded cells fills the tip region between the impeller blades and the seatring. In this region high gradients occur due to the seatring being stationary with respect to the rotating unschrouded impeller.6.
Figure 6.3
Mesh topology for the impeller of the mixedflow pump in a plane at halfspan (top) and in the direction from hub to tip
Numerical analysis of a waterjet propulsion system
129
. which follow the guide vanes curvature. The shaft is included in the model at the suction side of the impeller. In this way. Thickness of the guide vanes is taken into account in the model. like in a real waterjet installation. The stator bowl is meshed with another group of hexagonal cells.1 Geometry and mesh generation
Around the crosssections of the blades. an Ogrid is used to ensure good orthogonality of the boundary layer cells along the impeller surface. The remainder of the volume between the impeller blades is filled with additional hexagonal cells. very fine cells are created from an extrusion layer.3 shows the topology of the mesh between two impeller blades in a plane at halfspan and the block topology from hub to tip. These cells are also used in the region near the hub surface.
The total number of cells is about 950.5
Complete mesh of the mixedflow pump with about 950.4
Surface plot of impeller blades. the hub and the shaft.
nozzle
inlet
Figure 6.Chapter 6.000 for the complete mesh. the stator blades. Numerical analysis of waterjet pump flow
Figure 6.4 shows the surface plot of the impeller blades. stator blades. hub and shaft
stator nozzle
rotor
inlet
Figure 6.5. The final mesh of the complete mixedflow pump is presented in figure 6.000 cells (top) and surface mesh of the impeller
130
.
2 Numerical approach
This mesh shows the straight suction pipe. The velocity distribution in the plane can be either constant or nonuniform. The default wall boundary condition assumes zero velocity of the wall in normal and tangential direction (noslip) in the computational frame of reference. The two measured nonuniform velocity distributions as shown in figure 3. It was already mentioned in section 1.2 that the waterjet mixedflow pump belongs to the group of internal flow machines. first the applied boundary conditions are discussed.
6.2.
Numerical analysis of a waterjet propulsion system
131
. an inlet type boundary condition is applied. This requires the prescription of the three velocity components and the properties of the turbulence model. followed by the description of the fluid properties.3 and (ii) rotating walls in the stationary frame of reference. as discussed in section 6.1
Boundary conditions
The numerical domain of the mixedflow pump is bounded by a number of surfaces at which different types of boundary conditions are imposed. At the nozzle outlet plane two types of boundary conditions are available: normal outlet boundary condition or a prescribed static pressure condition.2
Numerical approach
The description of the numerical approach is divided into three parts. The computational method enforces conservation of mass and the resulting velocity distribution can be nonuniform. the implementation of the impeller rotation is discussed. Special attention is required in turbo machinery calculations for two types of wall boundary conditions: (i) stationary walls in a rotating frame of reference. Finally. The normal outlet condition only prescribes the mass flow rate.2 can be implemented using a Fourierseries approximation.6. The pressure condition is used with a prescribed constant value of the static pressure over the whole outlet plane in general. The actual distribution of the axial velocity may be nonuniform over the exit plane. The actual behaviour of the waterjet pump can be modelled with a value of the static pressure equal to the ambient pressure at the nozzle exit.2. the impeller and the stator with the nozzle.
6. A wall boundary condition can be applied to the remaining boundaries. to a certain extent in the detailed view of the impeller surface mesh. The block topology of the mesh along the fillet can be recognized. The surface mesh of the impeller is shown as well. In the inflow plane.
32)). a rotating impeller part and a stationary stator bowl part.4
Pump head
Calculation of global pump performance
By definition. or (ii) a fully transient method with a mesh moving with the impeller and with sliding interfaces with the stationary part of the mesh. Fully transient flow calculations use the PISO (Pressure Implicit with Splitting of Operators) algorithm to couple the momentum and pressure terms.2
Fluid properties
The fluid properties are specified similar to the ones specified for the calculations of the flow through the waterjet inlet as discussed in section 5. This method is identical to the method used for the waterjet inlet flow calculations. pump head is based on the difference in total pressure upstream and downstream of the pump (see (2. Numerical analysis of waterjet pump flow
6. These are implemented by additional source terms in the momentum equations.2.
6. which can be used with a steadyflow solution method. (32 Hz).2. This algorithm was originally developed for the noniterative computation of unsteady compressible flow [5].3
Impeller rotation
The basic idea is to split the numerical domain into three different parts: stationary inlet part. All calculations presented in this chapter have been made for an impeller rotational speed of 1920 RPM. Reynolds numbers based on inlet pump diameter and rotational speed (according to equation (4. Turbulence is modelled with the kε turbulence model. In case of a moving mesh with a sliding interface. The MFR method is a relatively fast method. which automatically rotates the mesh after each timestep and connects the rotating and stationary domains.Chapter 6.3.1)) are in excess of 10 . a fully transient flow solution method is required. The density and viscosity are taken constant for the fluid.2. The CFD method provides a routine. These values can be derived from the CFD results by integration of the total pressure over a cross
132
.
7
6. which may be seen as an extension of the SIMPLE algorithm with additional corrector steps. It involves one predictor step and two or more corrector steps. Rotation of the impeller can be accounted for in two different ways: (i) a quasisteady approach with a multipleframesofreference (MFR) method. Note that full scale waterjet installations normally operate at Reynolds numbers which are 3 to 5 times larger. In a rotating frame of reference rotation of the impeller results in centrifugal and Coriolis forces on the fluid elements.2.
and an estimated value Haa. It follows from the energy balance. the static pressure is only measured at the outer radius and subsequently averaged. and that there are no significant velocity components in the tangential and radial direction (i. that mass averaged quantities are to be used to determine the total pressure: 1 p tot = .3)
A
where A represents the crosssectional area. but not with the area averaged method. based on a mass average.+ . The actual mass averaged head Hma is calculated as: p tot .1H aa = . swirl).o ut – p tot .6. The areaaveraged head becomes: p stat .in Q 2 1 .– .
Numerical analysis of a waterjet propulsion system
133
. . based on area averages.2 Numerical approach
sectional area upstream and downstream of the impeller.4)
There are therefore two expressions for the head of the pump: the correct value Hma. The calculation of the head from the CFD calculations can also be based on the static pressure distribution and the flow rate.out – p stat . that in actual experiments.p stat dA A
∫
(6.in H ma = ρg (6.e. It is acknowledged. Static pressure is measured at the inlet and outlet pipe circumference through pressure taps.1)
A
where A denoting the crosssectional plane. The nonuniformity of the axial inflow velocity is taken into account with the mass averaged method correctly. Estimation of the dynamic pressure contribution is based on the inlet and outlet pipe diameter and the volume flow through the pump. The areaaveraged static pressure is calculated according to: 1 p stat = .2)
In an experimental setup it is common practice to divide the total pressure into a static and a dynamic pressure component. 2 ρg 2g A 2 A
out in
(6.ρp tot v n dA ρQ
∫
(6. This difference is neglected in the analysis. Within this approach it is assumed that the velocity and pressure distributions are uniform over the measurement planes.
The experimental data include the head and torque curves as a function of flow rate. The measured torque can be corrected for additional friction losses in seals and bearings. In the CFD calculations torque can be determined in two ways. This force will be a result of the shear forces acting on the surface of the socalled seatring. The equation for the moment of momentum is given by (see for example [6]):
T shaft =
∂ ∂t
∫ r × v ρdV + ∫ r × vρv ⋅ dA – ∫ r × τw
CV CS CS
(6. r is the distance from the axis of rotation.6) can be limited to the surfaces through which the flow enters and leaves the control volume.7)
6. vn the normal velocity at control surfaces and τw. shaft torque can be measured with strain gauges on the shaft. The first method is based on a direct summation of wall forces acting on the impeller and shaft surface.t
CS CS
(6. Numerical analysis of waterjet pump flow
Shaft torque In an experimental setup. The pump performance data is used in the waterjet performance prediction software of Wärtsilä Propulsion Netherlands
134
.6)
where vt is the circumferential velocity component.Chapter 6.t the tangential shear force at the surfaces of the stationary housing. In general turbo machinery applications.5)
where CV represents the control volume and CS all surfaces of the control volume. v is the velocity and τw the viscous stress tensor.t
(6. the axial component of this vector equation has to be evaluated only. Evaluation of the first term on the right hand side of equation (6. since the normal velocity vanishes at wall surfaces. For steady flow conditions the equation for moment of momentum reduces to: T shaft =
∫ ρrvt vn dA – ∫ rτw . Extensive series of measurements have been made on a modelscale pump at the authors’ company. The actual equation for the determination of the torque within the numerical method becomes: T shaft =
∫
A o ut
ρrv t v n dA –
∫ ρrvt vn dA – ∫
A in A seatring
rτ w . The second method is based on a momentum balance in circumferential direction.3
Validation with experimental data
The numerical method used to compute the flow in the waterjet mixedflow pump is validated with the aid of experimental data of pump performance.
The differences between the two methods at low flow rates can be attributed to a nonuniform velocity distribution at the outlet. This is similar to the torque coefficient Kq used for propellers. Torque is presented as a nondimensional quantity as well.
Numerical analysis of a waterjet propulsion system
135
.2) and (6. The presented head curves are normalised with a constant given flow rate and pump head of the design condition.20). Head curve First the head curve based on the CFD calculations will be compared with the experimental data. The numerical results are used to express the differences between mass averaged and area averaged determination of the head. The calculated head is based on equations (6. based on equation (2. These calculations are based on a pump configuration with a nozzle.3 Validation with experimental data
BV. Flow rate and head are expressed as nondimensional quantities with equations (2. The measured head will give the correct value.6 shows the dimensionless head curves from the measurements and the calculations.
6. The nondimensional torque is defined as:
* T T = 2 5 ρΩ D
(6. Differences between the two methods to determine the head from the CFD results are limited to the low flowrate conditions. head and torque. Near the design point the deviations between the two approaches are negligible.16) respectively.15) and (2. which has ambient pressure at the exit area. The measurements are made in a closedloop system however.3. which are currently in service.4). Agreement between the calculations and the experimental data is good over a large range of flow coefficients. when the static measurement locations are located sufficiently far upstream and downstream of the pump to ensure uniform velocity distributions at those locations. corresponding with the mass averaged numerical prediction. Figure 6.8)
Pump efficiency is derived from the values for the flow rate.6.1
Quasisteady flow calculations with the MFR method
Results of CFD calculations with the method using the quasisteady multiple frame of reference approach are presented in this section. The performance prediction software has been used for the prediction of all full scale installations.
6 Comparison of measured and calculated head curves.2) and (6. where an underestimation is observed at low flow rates and an overestimation at high flow rates. The currently found deviations between measurements and calculations are of the same order or smaller. Differences between the results of the two numerical methods are negligible for all conditions. Numerical analysis of waterjet pump flow
200% 180% 160%
H/Hbep [%]
140% 120% 100% 80% 60% 40% 20% 0% 50% CFD_H_massaveraged CFD .Chapter 6.
136
. Deviations of 5% in head prediction are also presented for a complete stage of a boiler feed pump [8]. The differences are limited to a few percent at most. Agreement between the numerical results and the measured data is acceptable over the whole range of analysed conditions. however. It can be concluded that the prediction of head with the CFD method is sufficiently accurate for further analysis of the complete waterjet installation in chapter 7. Calculation of torque is based on the integration of the wall forces and on the moment of momentum balance (equation (6. The magnitude of the differences appears to be related to the flow rate. based on equations (6.4).7)).
Gülich et al [7] show the obtained accuracy of calculated pump head for a large number of different pumps with a RANS method. in the range considered. Comparisons with measured data show deviations up to about 4% for mixedflow type pumps.7. Shaft torque A comparison of the calculated and measured shaft torque is shown in figure 6.H_areaaveraged Measurements
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.
The deviation between the calculated and the measured efficiency is about 1% near the design flow rate. Figure 6.7 Comparison of measured and calculated shaft torque based on integration of wall forces and moment of momentum balance
Efficiency The efficiency of a pump has been defined in equation (2.2).3 Validation with experimental data
150% 140% 130%
T*/T*bep [%]
120% 110% 100% 90% 80% 70% 60% 50% 50% CFD_wall force integration CFD_moment of momentum balance Measurements
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.6. Shaft power is the product of torque and the impeller angular speed. Hydraulic power is the product of volume flow rate and the produced pressure head.20) as the ratio of the hydraulic power and the shaft power. and the torque based on integration of the wall forces. it is to be expected that the calculated pump efficiency will show some deviations compared to the experimental data.8 shows the calculated and the measured normalised efficiency of the pump. The pump efficiency is normalised with the efficiency at the design point.
Numerical analysis of a waterjet propulsion system
137
. The calculated efficiency is based on the head according to equation (6. Based on the graphs of head and torque.
The flow field variables at the monitoring points between the impeller and the stator blades should give a periodic solution with a frequency equal to the impeller blade passing frequency. Evaluation of convergence behaviour In a fully transient flow calculation it takes a number of impeller revolutions before the solution becomes periodic. The axial velocity has been normalised with the mean axial velocity.8 Comparison of measured and calculated pump efficiency. With this method the actual movement of the rotor with respect to the stator blades is taken into account.9 shows the axial velocity at the monitoring points for the design flow condition during the fifth revolution of the impeller. calculations have been made with the method for fully transient flow. Three points are located about half a diameter upstream of the impeller in the inlet pipe and three points are located in between the rotor and stator blades in the stationary part of the mesh.
138
. This enables a detailed analysis of the interaction forces between the rotor and the stator. Figure 6.Chapter 6. The locations of the three points are chosen arbitrarily. including the moving mesh option. The level of periodicity of the flow field is monitored at some monitoring points in the numerical domain. Numerical analysis of waterjet pump flow
120%
110%
100%
η/ηbep [%]
90%
80%
70% Measurements 60% CFD 50% 50%
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.
6.2
Transient flow calculations with moving mesh
Apart from calculations with the MFR method for quasisteady flow.3.
3 Validation with experimental data
1.2 and 3 are located upstream of the impeller.70
4. As a consequence.4
Normalised axial velocity[]
1. The average value and the amplitude depend on the location of the monitoring points.1
5
Point 4
1
1
Point 5 Point 6
2 3
0. The constant pressure boundary condition implies the pressure being steady at the nozzle exit surface. The pressure coefficient is defined as: p stat – p 0 C p = 2 2 1 .3
4
Point 1
1. due to the prescribed axial inflow velocity.9)
where p0 is the reference pressure at the nozzle exit plane.2
6
Point 2 Point 3
1.6. points 4.00
4.30
4.60
4.80
4.9
0.00
Number of revolutions [] Figure 6.20
4.ρΩ D 2 (6. The monitoring points downstream of the impeller show the expected periodic behaviour of the pressure.9 Normalised axial velocity at monitoring points during fifth impeller revolution for design condition. On the
Numerical analysis of a waterjet propulsion system 139
.10
4.90
5. The fluctuating pressure coefficients at the monitoring points is shown in figure 6. The signal shows that during the fifth revolution of the impeller the signal is periodic with frequency equal to the blade passing frequency (BPF).5 and 6 are located in between the impeller and stator. The monitoring points upstream of the impeller show fluctuations also containing higher frequencies.8 4. Points 1. Pressure fluctuations at the inlet side are related to the choice of the boundary conditions at both the inlet and outlet boundary.50
4. Ω the shaft speed and D the diameter of the inlet.40
4. Downstream of the impeller a periodic solution is found for all three monitoring points.10.
For the monitoring points upstream of the impeller constant values can be observed. ρ the density. any fluctuation in the overall pressure head is experienced at the inlet.
40
4. i.
other hand.00
4. Evaluation of the periodic behaviour of the solution is based on a Fouriertransformation of the fluctuations. Figures 6. The peaks at the impellerblade passing frequency and its higher harmonics can be recognised easily.10
4.12 show the Fourier transforms for both the axial velocity and the pressure at the design flow rate condition for the monitoring points. Points 1. In the calculations the mass flow rate is a constant prescribed value due to the prescription of a fixed velocity at the inlet side.50
4.
140
. points 4. in an actual pump.60
4.11 and 6.05
3
2
0 4.10 Static pressure coefficient at monitoring points during fifth impeller revolution for design condition.2 and 3 are located upstream of the impeller. which may limit the high frequency pressure fluctuations.Chapter 6.25
Pressure coefficient Cp []
5
0. Results of the Fourier analyses for other flow rates are presented in Appendix B. small fluctuations in mass flow will be present.20
4.e.00
Number of revolutions [] Figure 6. in between the impeller and the stator.2
4 6
0.70
4. Numerical analysis of waterjet pump flow
0.15
Point 1 Point 2 Point 3 Point 4
0. which are located downstream of the impeller.80
4.90
5.1
Point 5 Point 6
1
0.30
4. The Fourier transformations show a clear harmonic solution downstream of the impeller for both the axial velocity as well as the static pressure.5 and 6 are located in between the impeller and stator.
Frequency is normalised by the shaft frequency
x 10 5
3
Pressure point 4 4.12 Fouriertransform of fluctuating static pressure coefficient at monitoring points at location in between the impeller and the stator. Frequency is normalised by the shaft frequency.035 0.5 2
1.03 0.025 0.01
0.5 Pressure point 5 Pressure point 6
Pressure coefficient Cp []
4 3.5 1 0.02 0.015 0.6.3 Validation with experimental data
0.045 Axial velocity point 5 Axial velocity point 6 0.
Numerical analysis of a waterjet propulsion system
141
.005 0 0 3 6 9 12 15 18 21 24 27 30
Normalised frequency [] Figure 6.11 Fouriertransform of fluctuating normalised axial velocity at monitoring points at location in between the impeller and stator.5 3 2.5
0 0 3 6 9 12 15 18 21 24 27 30
Normalised frequency [] Figure 6.05 Axial velocity point 4 0.04
Normalised axial velocity []
0.
This phenomenon is also observed in calculations based on a potential flow method by Van Esch [9]. The efficiency of the pump is shown in figure 6. Both methods show the tendency of a slight increase of the head for the transient flow calculations.2) for both the quasisteady and the transient flow calculations. Numerical analysis of waterjet pump flow
Performance diagrams Transient flow calculations have been made for a number of flow rates. Both numerical methods as well as the experimental data are shown in this figure. The pump performance parameters.
200% 180% 160%
H/Hbep [%]
140% 120% 100% 80% 60% Measurements 40% CFD_quasi_steady 20% 0% 50% CFD_transient
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.Chapter 6.16 for the quasisteady and transient flow calculations. is shown in figure 6.13 Comparison of results of quasisteady and transient flow calculations of mass averaged head based on equation (6. A similar trend is seen for the torque as for the head. It is observed that the transient flow calculations predict a small increase of efficiency compared to the quasisteady flow results. Figure 6. where transient flow results show slightly higher values. according to equation (6.13 shows the mass averaged head curve based on equation (6.15. like head and torque are averaged over a number of timesteps.4). The experimental data is plotted as reference. This efficiency is based on the mass averaged head.The results for the head curves based on the area averaged quantities.2)
142
. The comparison for the impeller torque is shown in figure 6. Agreement with the experimental data is still acceptable for both the quasisteady and the transient flow calculations.14.
6.
Numerical analysis of a waterjet propulsion system
143
.15 Comparison of results of quasisteady and transient flow calculations of torque.3 Validation with experimental data
200% 180% 160%
H/Hbep [%]
140% 120% 100% 80% 60% Measurements 40% CFD_quasi_steady 20% 0% 50% CFD_transient
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.14 Comparison of results of quasisteady and transient flow calculations of area averaged head based on equation (6.4)
150% 140% 130%
T*/T*bep [%]
120% 110% 100% 90% 80% Measurements 70% CFD_quasi_steady 60% 50% 50% CFD_transient
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.
2 . These fluctuations in pressure distribution result in a fluctuating force on the impeller.4 3 .16 Comparison of results of quasisteady and transient calculations flow calculations of efficiency
6.5 . The counter rotating force is derived from a model which is based on the assumption of a variation in the pressure field when a rotating impeller blade passes a stationary guide vane in the stator bowl.Chapter 6.3
Rotorstator interaction forces
The fully transient flow calculations with the moving mesh can be used to evaluate the radial force acting on the impeller due to rotorstator interaction. Such a matrix can be used to determine the sequence in which stator blades encounter passing rotor blades. The resulting matrix is shown in table 6. For this pump the angles between all the possible combinations of rotor and stator blade pairs are determined and inserted in a 6x7 matrix. The resulting interaction force has thus a counterrotating direction compared to the impeller rotation and can possibly lead to backward whirling. Disturbances in the symmetry of the blade pressure distribution will result in a radial force.1. The currently analysed configuration has six rotor and seven stator blades. the sequence of stator blades with an impeller blade passing will be: 7 . Given the number of rotor and stator blades.6 . an estimation of the behaviour of the interaction force can be made. Realizing that angular distances decrease during rotation of the impeller. According to Brennen [10] and Dubas [11] this should give a radial force counter rotating at the blade passing frequency.
144
. Numerical analysis of waterjet pump flow
120%
110%
100%
η/ηbep [%]
90%
80%
70% Measurements 60% CFD_quasi_steady CFD_transient 50% 50%
60%
70%
80%
90%
100%
110%
120%
Q/Qbep [%] Figure 6.3.1.
7 265.9 222.6 188.7 85. at 60% of the design flow a significant reduction of the magnitude of the radial force is observed.4 171.6 128.0 60. This is in accordance with the expectations.9 342.3 Validation with experimental data
Table 6. The amount of scatter of the force signal at each of these five points in subsequent cycles gives an indication of the higher order fluctuations in the solution. which represent the timeaveraged magnitude of the force.1 17.7 145.6 8.1 317.1 Stator 7 308.6 248.17 shows the force map for different flowrates.0 300. given a situation with uniform axisymmetrical inflow and outflow.4 111. where the time averaged magnitude Cfr1 of the rotor
Numerical analysis of a waterjet propulsion system
145
.0 240.0 120. which is equivalent to five times the blade passing frequency. This effect can be observed more clearly in figure 6.3 334.1 137.9 282. The timeaveraged magnitude of the radial force Cfr1 for each flow rate is plotted as a circle in the force map.7 325. output is generated with a frequency of 960 Hz. This means that the timeaverage of both the horizontal component Cfx as well as the vertical component Cfy of the interaction force diminish.1 77. is located at the impeller axis.6.0 180.3 94.0
Figure 6.6
Rotor1 Rotor2 Rotor3 Rotor4 Rotor5 Rotor6
0.7 Stator 6 257. This quantity increases with decreasing flow rate.18. The forces show a periodic behaviour for all flow rates.9 162.3 274. The centre of the concentric circles.1 197.7 25. The average of the blade interaction force has a minimum value at the design condition.3 214.9 42.4 Stator 3 102.3 Stator 5 205. denoted as 100% flow rate.4 351.3 34.11)
During the calculations.10)
The two force components are combined to get the radial force component Cfr1: Cf r1 = Cf x + Cfy
2 2
(6.1
Angular distance between rotor and stator blades for pump with 6 rotor blades and 7 stator blades Stator 1 Stator 2 51. However.4 291.4 231.9 Stator 4 154. The forces are made nondimensional to obtain the force coefficients Cfx and Cfy defined as: Fx Cf x = 2 4 ρΩ D Fy Cf y = 2 4 ρΩ D (6.6 68.
The standard deviation shows a minimum near the design point.25
0.2
Radial force .04
0.6
0.4
0. which is in agreement with the expectations.05
0.70% flow rate
0.80% flow rate Radial force .17 Plot of rotorstator interaction force for different flow rates: instantaneous values (markers) and timeaverage of magnitude of force coefficient (dotted curves)
Time averaged radial force coefficient Cfr1 [
0.4 Timeaveraged force 0.03
.025
0.90% flow rate
Cfy []
Radial force .
0.35 Standard deviation
0.100% flow rate
0
Radial force .005
0 50%
60%
70%
80%
90%
100%
110%
0 120%
Q/Qbep [%] Time averaged rotorstator interaction force coefficient Cfr1 and standard deviation as function of the flow rate
Figure 6.4
Radial force .Chapter 6.6 0.3
0.2
0
0.2
0.100% flow rate
0.80% flow rate Time average .01
0.4
0.15
0.70% flow rate
0.035
0.60% flow rate Time average .60% flow rate Radial force .90% flow rate Time average .6
Cfx [] Figure 6. Numerical analysis of waterjet pump flow
stator interaction force and the corresponding standard deviation are plotted as function of the flow rate.110% flow rate Time average .015
0.110% flow rate
0.18
146
Standard deviation []
0.6
0.2
0.2
Time average .1
0.4
Time average .02
0.
3 show the normalised pump performance for four different inflow conditions.6.
6.4 Influence of nonuniform axial inflow
Model scale measurements on the same waterjet pump have confirmed the presence of multiple local minima and maxima of the radial force [12]. The values in the tables have been normalised with the result from the corresponding calculation with a uniform inflow distribution and design flow rate of 100%. This force originates from the interaction between the impeller trailing edge and the statorblade leading edge. The results of the method based on the massaveraged total pressure are shown in table 6. since the velocity distribution is not axisymmetric. an additional source for a radial force is introduced. In the remainder of this section the effects of the inflow on the radial forces will be analysed. according to equation (3. The nonuniform velocity distributions are imposed as boundary condition in the numerical analysis with the aid of a Fourier series approximation of the velocity distribution. The flow rate for all calculations is equal to the design flow rate. Tables 6. It can be observed that the variations in both head and torque are limited to less than two percent for all inflow conditions. Results of the transient flow calculations for uniform inflow show a radial force due to rotorstator interaction. These results have not been published yet. First the pump performance for the nonuniform inflow conditions will be reviewed. the measured distributions are shown in given in figures 5.13a to 5. For the calculations for nonuniform inflow velocity distributions. Waterjets operate in a strongly nonuniform inflow distribution as described in chapter 3.2).
6. With a nonuniform inflow velocity distribution. The efficiency shows a small decrease for increasing nonuniformity. the differences between the results of these two methods may increase even more.4.2 and 6. The effect of a nonuniform velocity distribution on the pump performance and the impeller forces will be investigated in more detail in this section.4
Influence of nonuniform axial inflow
The results discussed in the previous section are obtained for uniform inflow velocity distributions. Four distributions are used in the analysis.13d.
Numerical analysis of a waterjet propulsion system
147
.1
Pump performance for nonuniform inflow
The calculations with uniform inflow velocity already showed an interesting difference between the two methods for the determination of the pump head based on either massaveraged or areaaveraged quantities. All calculations performed with the nonuniform inflow distribution are carried out with the transient moving mesh option of the CFD method.2.
From equation (6.2)) for various inflow distributions.3
Normalised pump performance based on area averaged head (equation (6. Numerical analysis of waterjet pump flow
On the other hand.8
IVR [] 1.6 101.87 2.6 101.4 η/η_uniform [%] 99.0 103.3 105. The tongue introduces a clear asymmetry
148
. who claim an increase of 5% due to nonuniform inflow.7 103. are in disagreement with the general expectation.0 100.7 101.4) it is clear that the averaging of a nonuniform inflow velocity distribution will underestimate the suction head. the results as shown in table 6.0 T*/T*_uniform [%] 101.3. Therefore. A clear increase of the pump head and consequently the efficiency is found with increasing nonuniformity.4 η/η_uniform [%] 101.7 101.68 1.7 99.5 101. Hma/Hma_uniform [%] 101. The actual pump performance shows only a small decay due to the nonuniform inflow velocity distribution. Although an efficiency increase has been reported by Kooiker et al.0 101. this is (at least partly) a result of the way in which the pump head is measured. related to IVR.1 100. [13].9 102.5 101. Table 6.3 T*/T*_uniform [%] 101.4.2 101. it is concluded that the method of areaaveraged head determination is invalid for nonuniform velocity distributions.5 104.8
IVR [] 1.03 2. which results in a higher estimate of the produced head by the pump.03 2. Haa/Haa_uniform [%] 103.68 1.19
6.2
Background of radial forces acting on the impeller
Steady fluidinduced radial forces can be generated by a nonuniform pressure distribution at the impeller periphery or by an imbalance in the blade torque.Chapter 6.4)) for various inflow distributions. related to IVR. The first phenomenon is known to occur in centrifugal pumps with a spiral volute casing (see [14][16]).2 Normalised pump performance based on mass averaged head (equation (6.6 99.19 Table 6.0 98. This deviation can be neglected for practical nonuniform inflow distributions.87 2.
This will influence the local flow angles at the leading edge of the impeller blade. Even for the case of maximum level of nonuniformity. however. This may result in a net radial force.
6.19
Concept of radial force due to unbalanced blade torque
in the geometry.
Numerical analysis of a waterjet propulsion system
149
. Figure 6. Imbalance in blade torque may give a more significant contribution to the radial forces acting on the impeller. As a consequence.19.3
Flow rate fluctuations in the impeller channel
The CFD results are used to determine the local flow rate Qbb through the impeller channel between two consecutive blades. It appears that the nonuniform inflow velocity distribution is smoothed in the first part of the pump. The nonuniform inflow velocity may create a nonsymmetrical pressure distribution on the hub surface.4 Influence of nonuniform axial inflow
net radial force
Figure 6.20 shows the normalized local flow rate as a function of the rotor position for the four different inflow conditions. This source for a radial force can therefore be excluded. Waterjet installations with a stator bowl do not suffer from such type of asymmetry. the maximum deviation is less than 7% of the average. This will result in a torque contribution that is different for each blade.19. The volume flow rate is determined at a crosssectional plane at midchord of the impeller blades. The difference in torque contributions between two opposite blades will result in a net radial force on the impeller axis.4.2. The variations obtained from the CFD results are much smaller than the predictions based on the analytical approach used in section 3. Nonuniformity of the inflow velocity distribution will result during a revolution in variations of both the flow angle at the leading edge and the flow rate through an impeller channel. the blade loading and the blade torque will be influenced. This eliminates the possibility to derive a simple analytical model to estimate the radial and tangential forces on the impeller. for IVR=2. The concept of this imbalance is shown in figure 6.e. i.6.
3. as shown in subsection 6.22 shows a graph of the horizontal and vertical component of the force acting on the impeller for different IVR. The forces acting on the impeller can be approximated by a mean component Cfr0 and harmonic components Cfr1 and Cfr2: Cf r = Cf r0 + Cf r1 ⋅ ωt + Cf r2 ⋅ 2ωt (6.3. In this section the forces are evaluated that are caused by the nonuniformity of the inflow.Chapter 6. Numerical analysis of waterjet pump flow
110% 108% 106% 104% 102% 100% 98% 96% IVR = 1.20 Normalised channel flow rate Qbb as a function of the rotor position for various inflow distributions.12)
This is shown in a sketch in figure 6.4. each with a typical nonuniform inflow velocity distribution.03 IVR = 2.87 IVR = 2. For a uniform inflow distribution a periodic rotorstator interaction force is found.19
Qbb/(Q/N) [%]
Angle θ [degrees] Figure 6. which are related to the IVR
6.68 94% 92% 90% 0 45 90 135 180 225 270 315 360 IVR = 1.
150
. Figure 6. at the design flow rate.4
Radial forces for nonuniform inflow
The nonuniform inflow velocity distribution has an effect on the radial forces acting on the impeller.21 for the mean component Cfr0 and the first harmonic Cfr1.
68 2 IVR = 1. The direction of the mean force seems to be constant for all inflow conditions.22. This seems to be an indication of higher order harmonics (i.19 1
0
1 1 0 1 2 3 4 5 6 7 8 x 103
Horizontal force coefficient Cfx [] Figure 6.22 can be seen that the mean radial component Cfr0 shows a strong relation with the level of nonuniformity. The condition with the most severe nonuniform velocity distribution (IVR = 2. This direction may be related to the flow rate through the pump.
x 10 5
3
Vertical force coefficient Cfy []
4
3 Uniform inflow IVR = 1. which are related to IVR.03 IVR = 2.e. at design flow rate
Numerical analysis of a waterjet propulsion system
151
.4 Influence of nonuniform axial inflow
y Cfr1 Cfr0 x Figure 6. Cfr2 and Cfr3) in the system at these inflow conditions.22 Horizontal and vertical component of timedependent impeller force coefficient for different inflow velocity distributions.
ω
From the results as presented in figure 6.21 Sketch of radial forces acting on the impeller due to nonuniform inflow. which is equal for all presented conditions in figure 6.87 IVR = 2.19) shows quite some scatter of the time dependent force.6. which is in accordance with expectations.
19
1
0
1 1 0 1 2 3 4 5 6 7 8 x 10
3
Horizontal force coefficient Cfx [] Figure 6.87 IVR = 2.13)
The direction of the radial force shows a small increase with increasing nonuniformity. the direction of the steady force has changed for this flow rate. On the other hand. the differences due to variation in flow rate are significant. The direction angle α is calculated according to: Cf y α = atan  Cf x (6.5 degrees. Numerical analysis of waterjet pump flow
Figure 6. at 80% of the design flow rate
Analysis of mean radial force Cfr0 The results of the calculations for different flow rates and levels of nonuniformities show that both parameters influence the mean radial force Cfr0. Moreover.
x 10 5
3
Vertical force coefficient Cfy []
4 Uniform inflow IVR = 1.25. At an IVR of 2. A quantitative assessment is made in table 6.23 Horizontal and vertical component of timedependent impeller force coefficient for different inflow velocity distributions.03 a deviation of 4. The magnitude of both the mean force Cfr0 and the first harmonic Cfr1 are different compared to the results for 100% flow rate.23 shows the plot of the forces for 80% of the design flow rate. which are related to IVR. where the time averaged magnitude and the direction of the nondimensional radial force are listed. Results at the design capacity show a variation of less than 1.Chapter 6.4.9) degrees is found.6 (= 30.
152
.68 3 IVR = 1. The occurrence of increasing mean force Cfr0 with increasing nonuniformity is confirmed.5 .03 2 IVR = 2.
At 80% of design flow rate. become more important for these conditions.4 Influence of nonuniform axial inflow
The variation in direction seems to be related to the flow rate.550 6. The condition with the highest level of nonuniformity at design flow rate shows a chaotic behaviour of the unsteady forces. Figure 6.423 3.24 for 80% of design flow rate. i.03 2. The direction of the radial force will rotate accordingly. The results of the calculations with uniform inflow indicate a rotating radial force with a mean magnitude.569 5.478 2.68 1. This means that the higher order harmonics of the radial force.e. This is confirmed by the Fourier transformation of the fluctuating component of the radial force.819 4. when the level of nonuniformity is increased.24 shows that the shape of the locus of the force components becomes more and more elliptical.5 31.7 30.4
Flow rate 100% 100% 100% 100% 80% 80% 80% 80%
Analysis of unsteady forces The unsteady radial force can be examined in more detail. The resultant of the vertical and horizontal component is plotted in figure 6. This may be due to an increase of the residence time of the fluid in the impeller.3 25. Timeaveraged magnitude Cfr0 and direction α of radial force coefficient IVR 1.68 1.26. The conditions at 80% of the design flow rate and at the design flow rate and moderate levels of nonuniformity show two clear peaks at one and two times the blade passing
Numerical analysis of a waterjet propulsion system
153
.9 29.87 2. Cfr2 and Cfr3.2 28. if the mean force components Cfx0 and Cfy0 are subtracted from the results.87 2.24 for the design flow rate and in figure 6. the behaviour of the fluctuating forces is more regular.9 26. in which the impeller will rotate over a larger angle.9 25.3
Table 6.368 2. as shown in figure 6. This is a result of higher order fluctuations of the forces.19 Cfr0*1000 3.6.19 1.03 2.826 3.977 α 24.
87 IVR = 2.3
0.4
0.6
0.6 x 10
3
Horizontal force CfxCfx0 [] Figure 6.4
0.1
0.2
0.3
0.5
0. The conditions at design flow rate and an IVR above 2 show numerous additional peaks at other frequencies.1
0
0.25 Unsteady impeller force components for different inflow velocity distributions at 80% of design flow rate.19
0
0.4
0.1
IVR = 1.2
Uniform inflow Curvefit IVR = 1.2
0.3
0.2
Uniform inflow Curvefit IVR = 1.5
0.4
3
0.2
0.03
0
IVR = 2. Numerical analysis of waterjet pump flow
frequency.4
3
0.19
0.3
0.6
0.2
0.3
Vertical force CfyCfy0 []
0.
154
.4 0.3
Vertical force CfyCfy0 []
0.1
0
0.
x 10 0.
x 10 0.Chapter 6. This is represented in the chaotic behaviour of the radial force.5
0.24 Unsteady impeller force components for different inflow velocity distributions at design flow rate.4
0.3
0.1
0.4 0.1
IVR = 1.3
0.1
0.6
x 10
3
Horizontal force CfxCfx0 [] Figure 6.2
0.2
0.03 IVR = 2.5
0.68
0.68
0.1
0.87 IVR = 2.
09 0. are determined for an impeller with a centred axis. which are presented in this chapter.04 0. however.5 Nomenclature
0.06 0. The offcentred motion of the rotating impeller is called whirling [10]. the impeller will move away from its centred position due to the radial forces. The influence of the whirling motion on the occurrence of socalled reaction forces has to be investigated further. In practice.5
Concluding remark
The radial forces.26 Fourier transformation of the unsteady radial impeller force component (Cfr .
6.07 0.Cfr0) for different inflow velocity distributions and flow rates.05 0.6.5
A Cf Cp D F Fs g
Nomenclature
area force coefficient pressure coefficient diameter force surface shear force gravitational acceleration m2 m N N m/s2
155
Numerical analysis of a waterjet propulsion system
. This whirling motion of the impeller has not been taken into account in the currently presented calculations.1 0. This type of radial forces are also denoted as excitation forces.01 0 0 2 4 6 8 10 12 14
FFT100168 FFT100187 FFT100203 FFT100219 FFT080168 FFT080203
Frequency/shaft frequency [] Figure 6.
6.4.02 0.03 0.08
Magnitude []
0.
pp. Göhringer. 373382.. Lakshminarayana. A. Vol 114. T. 1992 [2] Nilson.. ‘Investigation of tip clearance phenomena in an axial compressor cascade using Euler and
156
.6
References
[1] Goto. 2000 [3] Aschenbrenner.H. A. Davidson. focusing on tip clearance flow’..F. ‘Study of internal flows in a mixedflow pump impeller at various tip clearances using threedimensional viscous flow computations’. Numerical analysis of waterjet pump flow
H pstat ptot Q r T* T vn
head static pressure total pressure flow rate radial distance torque coefficient torque normal velocity
m Pa Pa m3/s m Nm m/s
Greek symbols
α ρ Ω
Subscripts in n out r0 r1. 2000 [4] Kunz.. r2 t x y
radial force direction density shaft speed
deg kg/m3 rad/s
inlet plane normal direction outlet plane mean component of radial force coefficient harmonics of rotorstator interaction force coefficient tangential direction horizontal direction vertical direction
6. L. Moser. Basson.. W. H..... ‘Numerical and experimental flow analysis in a Kaplan turbine’. R. ‘A numerical comparison of four operating conditions in a Kaplan water turbine. Proceedings 20th IAHR symposium. Journal of Turbomachinery. Proceedings 20th IAHR symposium. Charlotte.Chapter 6. Charlotte. B. M.
pp. K.P.. Denus. University of Twente. M. pp. Italy. Baché. IngenieurArchiv. 54. 1964 [15] Badie....W. Oxford University Press. ’An assessment of pump impeller performance predictions by 3DNavier Stokes calculations’. proceedings FAST 2003 conference. A. [13] Kooiker. W.. 2003 [14] Stepanoff. Van Terwisga. Verbeek R.. University of Twente. ASME Fluids Engineering Division summer meeting. 1994 [11] Dubas.. 40. J. & Malalasekera. 1995 [6] Fox. T. ‘A finite element pertubation method for computing fluidinduced forces on a centrifugal impeller rotating in a volute casing’. C. “Centrifugal and axial flow pumps . personal communication. ‘Über die Erregung infolge der Periodizität von Turbomachinen’.. Vol 115. B..J. K. T. session A1 pp 5762. 413426. 269294.B. van. PhD thesis. R. & Van Essen. Journal of Turbomachinery. John Wiley. 1993 [5] Versteeg. J. H. R. 'Performance and cavitation analysis of a waterjet system on a cavitation tunnel'. ‘An introduction to Computational Fluid Dynamics’. 1997 [10] Brennen. A. vol.F. Favre. 1997 [8] Cugal. G.P. & McDonald. J.M. P. International journal for numerical methods in engineering.. PhD thesis. 1984 [12] Esch.T. 453467. ’Performance prediction from shutoff to runout flows for a complete stage of a boiler feed pump using computational fluid dynamics’.N. Longman Scientific & Technical. 'Hydrodynamics of pumps'.Theory. John Wiley & Sons.. Essex. ASME Fluids Engineering Division summer meeting.M.. pp. 1997
Numerical analysis of a waterjet propulsion system
157
. New York.. Ischia. Van Terwisga. 2nd edition. 1985 [7] Gülich. 1993 [16] Jonker..K. Vol..’Simulation of threedimensional unsteady flow in hydraulic pumps’..’Introduction to fluid mechanics’.G. ‘Analysis of unsteady potential flows in centrifugal pumps’.. 1997 [9] Van Esch.6 References
NavierStokes procedures. Third Edition. design and application”. B.. M..6.E.
Numerical analysis of waterjet pump flow
158
.Chapter 6.
An analysis is made of the overall performance indicators. Results are presented of the numerical analysis of such a propulsion system on full scale. The sensitivity of the flow rate
Numerical analysis of a waterjet propulsion system 159
. Care should be taken that the dimension of the nozzle exit area is equal to the exit area of the actual fullscale installation.1 for the location of these parts). At this interface a coupling method is used to create a fully connected domain. which are tuned to the performance of the propulsion systems in the currently sailing fleet. based on arbitrary matching of the cells.Chapter 7 Analysis of a complete waterjet installation
In the preceding two chapters the waterjet inlet and the mixedflow pump were analysed separately. which are required for the rotation of the impeller. thrust and power. In this chapter the complete waterjet installation is considered.1
Generation of the numerical model
The numerical domain of the complete waterjet installation is a combination of the two separate grids of the waterjet inlet and the mixedflow pump. like volume flow rate. A detailed analysis of the streamtube will reveal some new insights into the forces acting on the installation in both vertical and axial direction.
7. It should be noted that this interface between the models of the inlet and the pump is different from the two sliding interfaces. Comparisons are made with performance prediction and selection software of Wärtsilä Propulsion Netherlands (WPNLselect). This software is partially based on semiempirical relations. The interface between the two grids is located at the crosssection downstream of the inlet bend (see figure 5.
Analysis of a complete waterjet installation
nozzle exit
pump
inlet side Figure 7. Further finetuning is made with additional empirical data. The variation in volume flow rate will be much smaller. To a large extent. the actual flow rate through the waterjet becomes a part of the solution. The boundary conditions and parameters of the numerical method are identical to the ones used for the calculations of the separate parts. whereas it was prescribed by the boundary conditions in the calculations for the isolated inlet and pump. this program is based on the theory as discussed in chapter 2.1 Mesh of the complete waterjet installation.2 million. as indicated by the results presented in figure 2.Chapter 7. The accuracy of the method is continuously increased by adding the measured performance of newly installed waterjet installations.1 shows a plot of the final mesh of the full scale waterjet installation. with a boundary layer velocity profile. As a consequence. In this way a relatively large range of IVR conditions is covered. At the other sides on the domain beneath the hull and at the nozzle exit plane constant pressure boundary conditions are applied. The calculated volume flow rate through the waterjet is compared with the results of WPNLselect.2
Evaluation of volume flow rate
The calculations are made with a constant rotational speed of the pump and a varying ship speed. Since the currently sailing fleet performs according to the
160
.8 on page 36. At the inlet side of the domain a prescribed velocity distribution is applied.
7.9 on page 37. Figure 7. number of cells is about 1. The loss coefficients in this method are empirical values taken from experimental data.
as function of the nozzle size has been illustrated in figure 2.
49
Deviation [%] < 1% < 1% < 1% < 1% < 1%
Differences between the results of WPNLselect and the CFD analyses are within 1%.
7.52) on page 33.
Numercal analysis of a waterjet propulsion system 161
.
7. It is believed that the prediction of thrust in WPNLselect is quite accurate. or to select only the solid wall cells. Table 7.93 13. It is shown in equation (2.31 13.3 13. Agreement is satisfactory and it appears that the numerical method is suitable for further use to investigate the flow phenomena in the waterjet installation. With the selection of all wall cells of the numerical domain. The net thrust will be reduced due to the contribution of the drag of the hull.1 13. The thrust can be determined from the CFD results by a summation of the forces acting on the solid walls or by the simplified momentum balance method given by equation (2.19 13. that are part of the streamtube.08 13.4 13.3.3
Evaluation of waterjet thrust
The second quantity that can be compared with the results of WPNLselect is the thrust of the complete installation. and consequently the prediction of the volume flow rate will be accurate as well.1
Integration of solid wall forces
The procedure for the selection of the cells at the solid walls requires some attention.7. This trend has been captured well by the numerical method.7 on page 31. a relatively small error in the prediction of the volume flow rate can lead to a large error in the thrust of the installation. a large part of the hull structure is also taken into account.6
3
QCFD [m3/s] 12.0 13.52) that the thrust is related to the square of the volume flow rate. According to the expectations. The two options have been shown in figure 2. It is possible to either select all solid wall cells of the numerical domain. Consequently.3 Evaluation of waterjet thrust
predictions. only a rather limited difference in volume flow rate is found between the low and high ship speed condition.1 vship [kn] 31 35 39 43 47 Comparison of volume flow rate through waterjet based on prediction software and CFD calculations Qselect [m /s] 13. it is assumed that the accuracy of the performance prediction software is adequate.
The additional drag of the hull structure can be compensated for.1)
where Twj. when all solid wall cells of the numerical domain are taken into account in the force evaluation. the hull and the propeller can be split into two subsystems. The wall cells near the streamtube surface will be partly included in the streamtube.all is the integrated force of the pressure and wall friction on all solid wall cells and Dhull the drag force of the equivalent rectangular hull area.455 C f = 2.all – D hull (7.2)
with Dhull the actual friction drag of the equivalent hull. The process of wall cell selection is avoided.58 log ( Re l ) where the friction coefficient Cf is defined as: D hull C f = 1 2 . will be measured at the location of the inlet duct opening. This formula is used. without leaving a hole in the hull structure. because it is also used to extrapolate model scale bare hull resistance data to full scale predictions.
162
.Chapter 7. given the dimensions of the numerical domain. The method using drag compensation is also preferable from a ship builder’s point of view. Calculation of the drag of the hull boundary layer is based on flat plate boundary layer theory for high Reynolds numbers. A waterjet propelled vessel can not be split into a bare hull and a propulsion unit. vship the undisturbed velocity. It is concluded that the thrust of the installation has to be computed using all solid wall cells and that frictional resistance of the original hull area has to be compensated for. This is due to a subtle difference between a vessel with waterjet propulsion and a vessel with conventional propeller propulsion. due to the development of the boundary layer.ρv ship A hull 2 (7. however.3) (7. can not be done with a welldefined procedure. additional hull drag. The total waterjet thrust is then: T wj = T wj . The results show a significant influence of the choice whether these cells are included or excluded from the selection. In a bare hull resistance test. The friction coefficient is given by [1]: 0. and Ahull the area of the equivalent hull. The additional drag of the boundary layer will change once the waterjet installation with the inlet duct is installed. Analysis of a complete waterjet installation
Selection of the solid wall cells of the streamtube only. which is length times width of the used mesh. For such a propeller ship.
At higher speeds the momentum balance gives a lower thrust than the force integration and the prediction software.
7.5.3. which is the design speed for the investigated installation. The method of integration of the solid wall forces shows a very good agreement with the prediction software for the ship speeds up to 39 knots.52).
7. The method based on the momentum balance gives a good agreement up to ship speeds of 35 knots. The results have been normalised with the design thrust at a ship speed of 39 knots. The mass averaged inflow velocity is determined using a concentration scalar as described in detail in section 5.
Numercal analysis of a waterjet propulsion system 163
.7. both results are presented in the graph. The introduced error can be compensated with the socalled thrust deduction factor tj [2]. At higher speeds a small difference is found between the results of the force integration method and the results of the prediction software.3.3. which is applied to the streamtube. However.2 shows the thrust of the fullscale installation based on WPNLselect and on the CFD calculations. This phenomenon may be attributed to numerical inaccuracies. it can also be an indication that the simplification of the momentum balance is the cause for this. The deviations between the results of the two numerical methods are quite significant at higher ship speeds.3 at page 30. as described above.2
Momentum balance
The thrust based on the momentum balance. In this equation the contribution of the pressure distribution on the streamtube below the hull and the aft part of the hull is neglected. It is known.3
Results
Figure 7.4)
where Lwetted is the wetted length of the hull upstream of the inlet. as discussed in subsection 2. that the low pressure region along the cutwater will have a positive contribution to the thrust. The actual drag of the hull area of the numerical domain depends on the dimensions of the area A and the wetted length of the hull upstream of the waterjet inlet.3 Evaluation of waterjet thrust
The Reynolds number for the flat plate is defined as: ρv ship L wetted Re l = µ (7. In order to show the differences between results of the two methods that use the results of the CFD calculations. can be calculated with the simplified equation (2. This may explain the deviations found between the results of the force integration method based on considering the pressure and shear stress on all solid wall cells and the ones of the momentum balance approach.
7. Over the complete range of calculated conditions. Analysis of a complete waterjet installation
120%
Prediction software
115%
Integration of solid wall forces Momentum balance
Thrust/Thrust39 knots []
110% 105% 100% 95% 90% 85% 80% 30 32 34 36 38 40 42 44 46 48
Ship speed [knots] Figure 7.2 Comparison of thrust of waterjet installation based on WPNLselect with numerical predictions based on force integration of solid wall forces (Twj) and momentum balance applied to the streamtube (equation (2. The torque is obtained from the summation of the torque of the solid wall cells of the impeller.52)).5% is acceptable. however. a small overprediction of the power can be observed. Results are normalised with the required power of the full scale installation at the design speed computed with the prediction software. The deviation is more or less constant for all conditions.Chapter 7.4
Evaluation of required power
Evaluation of the required power is equivalent to the evaluation of the required torque. The increase in flow rate is only 4.
164
. Figure 7.5% for an increase of ship speed from 31 to 47 knots. since all calculations have been made at the same rotational speed. The deviation of about 2. due to the very small variation in flow rate.3 shows the power based on the calculations and the one obtained from the prediction software. The resulting torque is multiplied by the angular velocity of the impeller to obtain the required power.
0% 92. This result of the numerical analysis is not in agreement with the results of Van Terwisga [2].5% 95. Applying equation (2.0% 97.5% 90.all = –
∫
A tube
pz ⋅ dA + L wj . who claims that there is no net lift contribution. which can be up to 5% of the displacement of the vessel.hull
(7. Svensson [3] determined a lift force from pressure measurements in the intake and on the hull.5
Analysis of vertical force on waterjet structure
In the preceding section.35) for the vertical direction yields: L wj .0% 30 32 34 36 38 40 42 44 46 48
Ship speed [knots] Figure 7.0% 102.0%
Prediction software
107. since it is assumed
Numercal analysis of a waterjet propulsion system
165
. the prediction of the waterjet thrust is discussed. The vertical components of the momentum fluxes do not contribute.5)
where Lwj is a vertical force on either the waterjet or the hull. it is found that there can be a net force in the vertical direction as well.5%
CFD result
Power/Power39 knots []
105. This vertical force component is in fact a lift force acting on the waterjet structure. While calculating the thrust of the waterjet installation.5 Analysis of vertical force on waterjet structure
110. On the other hand. based on the integration of the forces acting on the solid walls.5% 100.7.3 Comparison of required power of waterjet installation obtained from prediction software and the one obtained from CFD calculations
7.
15) as: L wj .22 on page 151. The numerical results presented in this subsection are for a waterjet installation for very high speed vessels. For high speed applications (>60 knots) the IVR will be about 2 to 2. D and ϕ as well with equation (2. A more detailed analysis of the occurring lift force coefficient. Analysis of a complete waterjet installation
that both the velocity distributions at the inlet and nozzle exit are aligned with the horizontal direction. Equation (7.Chapter 7.19.08 to 0.
166
. Up to an IVR of about 1. Q the volume flow and vpump the averaged axial velocity upstream of the impeller.10. Inlet lift coefficient The actual lift force acting on the inlet structure is presented as a nondimensional coefficient. The lift coefficient is plotted as a function of IVR in figure 7. the vertical force is about 2. according to figure 6.2.4.5 the lift coefficient is indeed rather small. The most important difference from hydrodynamic point of view between a conventional waterjet installation and a highspeed application is found in the geometry of the inlet duct [4]. reveals that the largest lift is found at high IVR conditions. This will result in a lift coefficient of about 0. At an IVR of 2.85x103.081. It is concluded that the vertical force in the inlet is significantly larger (almost factor 30) than the vertical component of the impeller force due to nonuniform inflow velocity distributions. based on figure 7. ρ the density. and therefore high ship speed.⋅ 2 4 2 ρΩ D 4ϕ (7.all π C L = .4.6) can be expressed in specific pump parameters Ω. a direct comparison between the lift force coefficient on the inlet and the vertical force coefficient on the impeller.all is the calculated vertical force. The vertical force coefficient in the inlet for that condition is 0.all C L = ρQv pump (7.7)
The second factor on the right hand side is of order 1 for common waterjet mixedflow pumps. as a result of nonuniformity can be made.6)
where Lwj. according to: L wj . Thus.
(2.
Numercal analysis of a waterjet propulsion system
167
. the wake fraction and the geometry of the waterjet installation.2
0. (2. with aid of equations (2.52).2).5 Analysis of vertical force on waterjet structure
0. – ( 1 – w ) ⋅ IVR D nozzle (7.12) and (2.5 1 1.05
0
0. The thrust of a waterjet can be rewritten from equation (2.4 is presented as liftthrust ratio in figure 7. – ( 1 – w ) ⋅ IVR D nozzle (7.05 0 0.5
IVR [] Figure 7.2).11).15
CL []
0.4 Lift coefficient CL as a function of IVR
Inlet liftthrust ratio The lift force can also be compared to the thrust of the waterjet installation.5 2 2.1
0.13) to: D inlet 2 T = ρQv pump .5.9)
This shows that the liftthrust ratio is related to the lift coefficient. The liftthrust ratio becomes: CL L .= T D inlet 2 . defined in equation (2.7. The data for the lift coefficient as shown in figure 7.8)
where Dinlet and Dnozzle are two geometric parameter of the installation and w is the wake fraction.25
Lift force coefficient
0.
7.5 Liftthrust ratio L/T as a function of IVR Since the thrust decreases with increasing IVR. As a result.5
IVR [] Figure 7. which is in accordance with the simplified thrust equation (2. It is expected that any possible contribution to the vertical force will be a negative lift for high IVR. the liftthrust ratio shows a steep increase at high values of IVR: a significant lift force of about 1020% of the thrust is found. It is known from CFD analyses that the pressure in this region is very low for high IVR conditions (see for example figure 5.5 2 2. which acts on the streamtube surface.Chapter 7. which can lead to a reduced or even negative trim angle of the vessel.1
Pressure distribution on streamtube surface Evaluation of momentum balance in vertical direction
The CFD analyses reveal the presence of a significant lift force in a waterjet installation at the higher IVR conditions. Analysis of a complete waterjet installation
25%
Liftthrust ratio
20%
15%
L/T []
10%
5%
0%
5% 0 0.5 1 1. the lift can be the result of a net force acting on the hull surface or of a resultant force of the pressure distribution. the net lift force is equal to zero for all conditions.6. The contribution of the vertical forces acting on the hull surface are limited to a region near the cutwater. Based on equation (7. These two terms are neglected in the standard momentum theory.6).6 7. It is concluded that the lift force on the complete
168
. The lift force in the inlet creates a moment on the hull structure at high speeds.52).5). This amount of lift can not be neglected in general.
given the average pressure on each triangle and the normal direction. The resulting lift force of the streamtube pressure distribution is shown in figure 7.7.2
Calculation of vertical force on streamtube
The streamtube shape is determined for various operating conditions of the highspeed waterjet installation. The force of the integrated pressure distribution clearly depends on IVR. as discussed in the preceding section. A method to integrate the pressure distribution on the streamtube surface is required to confirm this. In order to determine the net force. acting on the hull structure. based on the Delaunay triangularisation method. because the effect of partially included solid wall cells can not be taken into account.6. At an IVR of about 1.8 the net lift force on the streamtube exceeds the total lift force on the inlet. which is in agreement with expectations as well. Accurate determination of the lift force. An example of the triangular mesh of the streamtube surface is shown in figure 7. is not possible with the applied method. At values of an IVR above 1. It is not possible to determine the contribution of the solid wall cells near the streamtube interface accurately. The force has been made nondimensional according to equation (7.6 Pressure distribution on streamtube surface
waterjet structure. At these IVR conditions. The net lift force can be determined.
Numercal analysis of a waterjet propulsion system
169
. the net force on the hull (near the cutwater) will be negative. For this numerical integration. The pressure distribution also depends on IVR. an integration of the pressure over the surface has to be made. The triangular mesh represents the streamtube surface and the rectangular cells represent the solid wall cells of the waterjet inlet.6.7.
7. This confirms the hypothesis of the contribution of negative lift of the wall cells near the cut water at higher values of IVR.6). The shape of the streamtube depends on the IVR at which the waterjet operates.0 the net force on the cutwater region will be positive. should originate from a pressure distribution along the streamtube surface. the surface of the streamtube is subdivided into triangles.
2
Inlet lift coefficient
CL []
0.05 0.
0.15 0 0.7
IVR [] Lift force coefficient of streamtube pressure distribution as function of IVR compared to inlet lift coefficient
170
.05 0 0.35 0.3 0.Chapter 7.1 0. Rectangular cells represent solid wall cells of waterjet inlet in blue.5 2 2.5
Figure 7.1 0.15 0.5 1 1.6
Example of triangular surface mesh of streamtube surface (in green).25 0. Analysis of a complete waterjet installation
Figure 7.4
Streamtube lift coefficient
0.
6. the actual geometry of the inlet geometry might have an important influence on the magnitude of the lift force as well.3
Concluding remark
A first attempt to quantify the forces.7
A Cf CL D D IVR L L p Q T v w z
Nomenclature
area friction coefficient lift coefficient diameter drag force inlet velocity ratio length of flat plate lift force static pressure flow rate thrust velocity wake fraction vertical direction m2 m N m N Pa m3/s N m/s 
Greek symbols µ ρ dynamic viscosity density flow coefficient angular velocity kg/ms kg/m3 rad/s
ϕ
Ω Subscripts inlet nozzle pump wj
inlet plane nozzle exit plane pump entrance plane waterjet
Numercal analysis of a waterjet propulsion system
171
. Further research of this is recommended before general conclusions can be drawn. which result from the pressure distribution along the streamtube. This force depends on the value of the IVR of the waterjet installation.7. has shown that there can be a significant force in vertical direction.
7.7 Nomenclature
7. However.
1996 [3] Svensson. 1989 [4] Bulten. N. H. R. van.. Ischia. & Verbeek. AIAA conference. Arlington. R. ‘Experience with the KaMeWa waterjet propulsion system’.’Waterjet hull interaction’. 1968 [2] Terwisga. Analysis of a complete waterjet installation
7. 2003
172
. Italy.’Design of optimal inlet duct geometry based on operational profile’.C. Paper No 891440CP.Chapter 7.. session A2. New York. PhD thesis.. Delft University.8
References
[1] Schlichting. McGrawHill.J. T. pp 3540. ‘Boundary layer theory’. Proceedings FAST2003 conference Vol I.
8.Chapter 8 Concluding remarks
8. With aid of the numerical results.1
Theory of thrust prediction for waterjet systems
The theory of thrust prediction is based on the method of momentum balance with a control volume involving the streamtube. for which the usefullness is generally accepted. It is shown that this approach is not completely correct. Partly because the streamtube approach does not take into account the complete waterjet geometry and partly due to the neglect of the contribution of the pressure on the streamtube surface. it has been possible to evaluate some general used assumptions in waterjet propulsion theory as well.1. Results have been obtained for both the flow through the inlet as well the flow through the mixedflow pump. The basic theory which is used to describe the flow in the system has been evaluated as well. Typical examples are the investigation of the streamtube shape and the time dependent forces acting on the impeller. The thrust of an open propeller is almost entirely transferred to the
Numerical analysis of a waterjet propulsion system
173
. resulting in a reassessment of the widely applied methods. The method of the momentum balance is derived from the theory for open propellers. Direct application of this theory to waterjet propulsion systems is not allowed because of an open propeller is an external flow machine and a waterjet is an internal flow machine.1
Conclusions
The numerical analysis of a waterjet propulsion system has led to an increase of the knowledge of the flow phenomena occurring in the system.
the thrust can for a certain part also be transferred to the ship through the pressure and friction on the solid surface of the inlet and the pump housing. it appears that the effect of the shortcomings of the applied theory are in general within a few percent. It is found that the velocity distribution is a result of accumulating vorticity in the flow. Typical examples are: • static pressure distribution inside the inlet • cavitation inception at cutwater • velocity distribution in the impeller plane • margin against flow separation in the inlet • shape of the inlet streamtube • lift force on the inlet geometry
174
.3
Waterjet inlet flow characteristics
The calculations of the waterjet inlet flow revealed that most characteristics are strongly related to the inlet velocity ratio (IVR=vship/vpump). Because of the dominance of viscous flow effects only a numerical method based on the NavierStokes equations is appropriate.
8. whilst keeping the overall calculation times and required hardware within acceptable limits. This is reflected in an overprediction of drag of profile sections. the used twoequation turbulence models do have a general shortcoming in the prediction of the pressure at a stagnation point. In practice. The vorticity is generated in the hull boundary layer and increased in the inlet. One of the important topics of the thesis was to investigate the origin and the consequences of the nonuniform inflow velocity distribution to the pump of the waterjet installation.2
Numerical aspects
For waterjets with a flush inlet it is unavoidable to get a nonuniform velocity distribution into the pump at the design point.Chapter 8.1. this is compensated for by most waterjet manufacturers using the socalled thrustdeduction factor. Currently available commercial Reynoldsaveraged NavierStokes (RANS) methods are well suited for the numerical analyses of both the flow in the waterjet inlet and the waterjet pump flow.1. where the flow is retarded rapidly. This necessitates the investigation of the effects on nonuniform inflow to the pump. However. For a waterjet installation.
8. It is demonstrated that standard twoequation turbulence models in combination with wall functions give results with acceptable accuracy. However. For open propellers similar phenomena play a role in the creation of the so called wake field behind a ship. Concluding remarks
ship through the shaft.
an additional contribution to the mean radial force is found. However.
8. In this respect these type of propulsion units can be regarded as miniwaterjet installation as well. the behaviour of the vessel may be influenced. The development of the flow in streamwise direction can be characterised as a diffuser flow. The pressure distribution on the solid part of the streamtube surface creates a net vertical force.8. The equivalent diffuser angle depends on IVR.2 Recommendations
Due to variations in design ship speed and power density of the installations.1
Recommendations Research topics for marine propulsion systems
The general applied theory for thrust prediction of waterjets has some shortcomings. This results in a change of the behaviour of the inplane forces. a dedicated inlet design is recommended for each specific ship. This lift force can be up to 20% of the thrust of the system. which means for example optimal cavitation inception margins and avoidance of flow separation in the inlet. Similar shortcomings might be present in the theory for the prediction of thrust for ducted propellers and thrusters.2.
Analysis of mixedflow pump in a waterjet propulsion system
175
. The magnitude and direction of this additional radial force depend on the amount of nonuniformity and the flow rate through the pump. specifically at high IVR. torque and efficiency. The consequences of the net vertical force occurring in the inlet have to be investigated in more detail. but which rotates about the impeller axis with the blade passing frequency. This is due to rotorstator interaction. In case of nonuniformity of the inflow. Calculations are based on the quasisteady multiple frames of reference method and the fully transient moving mesh method. For optimum performance of the inlet. the actual design IVR can vary per vessel. The shape of the streamtube upstream of the inlet can be approximated with a semielliptical shape.4
Waterjet mixedflow pump analyses
The numerical method is used to analyse the flow through the rotating pump impeller as well.1.2 8. Calculations with uniform inflow show a radial force that is almost constant in magnitude. the pressure distribution on the rotating impeller blades is strongly related to the inflow velocity distributions. The performance of the pump is not significantly influenced by variation of the inflow velocity distribution. For sufficiently large forces.
8. Differences between the two approaches has shown to be small for performance indicators like head.
The use of panel methods in the analysis of propellers should be evaluated.
176
.Chapter 8. Increase of the use of RANS methods in propeller and hull design may have an interesting consequence. the calculations presented in this thesis have been made within acceptable wallclock turnaround times. Application of a RANS method for such configuration eliminates the requirement of Kuttaconditions. i. certainly at offdesign conditions. Compared to a RANS method the required wallclock time for an analysis is small. centrifugal pumps and even complete ship hulls. single linux PCs. can be rather poor. It is thus recommended to start building experience with RANS methods for marine propulsion systems. Concluding remarks
Improvement of the thrust prediction method might give an improved insight in the thrust deduction factor. for which an additional Kuttacondition is required in a panel method at the nozzle trailing edge. but the obtained accuracy. which are currently in use to scale model scale experimental results to fullscale. given the fact that the performance of computers increases continuously. A similar study can be made for ducted propellers.e.2. the fullscale RANS calculation eliminate the use of Reynolds scaling laws.
8.2
Application of RANS methods
With the currently available hardware. This will even get better. Investigations should be made to determine which level of accuracy can be obtained with a RANS method.
In the first half of the bend the influence of viscous forces on the flow will be relative small and therefore the flow can indeed be regarded as irrotational. If a potential flow method is to be used. then an irrotational velocity field is required. In a viscous flow the high momentum fluid will move outwards in the second part of the bend. In a potential flow the high speed velocity region will vanish within a length of about 1 pipe diameter. after a bend occurs over a length of about 50 pipe diameters. before the velocity distribution is uniform again. This phenomenon can be attributed to
Numerical analysis of a waterjet propulsion system 177
. Velocity gradients occur by definition in a nonuniform velocity distribution. Ito [1] measured the pressure drop downstream of a 90 degree bend. where high velocity is found at the inner corner and low velocity at the outer corner. This is in accordance with the length which is required to obtain a fully developed velocity distribution without swirl again. It is shown that the velocity field is strongly nonuniform at normal waterjet operating conditions. Then it will take 30 to 50 pipe diameters.Appendix A
Stability of nonuniform velocity distribution
In chapter 3 the velocity distribution just upstream of the impeller is discussed. in addition to the one that occurs in a straight pipe. viscous flow become apparent. in the second part of the bend the differences between irrotational flow and the actual. It is shown that the pressure drop. For a velocity field free of vorticity. However. A typical example of a nonuniform velocity distribution in a potential flow is the flow through a 90 degree bend. velocity gradients in one direction are compensated by gradients in the other two directions. This type of velocity distribution has to be reproduced with the numerical method in order to obtain a correct analysis of the complete waterjet installation.
as defined in equation (3. The results of the calculations with a RANS method are presented in figure A. as discussed in section 3. The length of the pipe was set to 2 pipe diameters.4. vorticity stretching will not occur. analysing the stability of the nonuniform velocity field in a potential flow method and a RANS method. These velocity gradients are also present in the flow in a diffuser. As a consequence the flow is rotational. This vorticity is transported and redistributed in the flow. The viscosity causes a slow decay of the initial vorticity. For all conditions a very strong decay of the nonuniformity is observed. In case of a circular pipe geometry. Several runs have been made to ensure mesh independent results. It is to be expected that the axial component of the velocity gradient in the waterjet inlet contributes to a higher vorticity. reduction of the vorticity leads to smaller velocity gradients and thus a more uniform flow. the velocity distributions have been shown in figures 5. The boundary layer velocity profile has large velocity gradients in the direction normal to the wall. what leads to smaller velocity gradients and thus more uniform flow.13d. In a potential flow method velocity components in radial and tangential direction will be introduced to create the velocity gradients (in these directions).Appendix A Stability of nonuniform velocity distribution
the development and decay of vorticity in the flow. Figure A.
178
. Four different IVR conditions in the range from 1.2.1 shows the nonuniformity parameter ζ. For an accurate description of the viscous flow through a 90 degree bend vorticity can not be neglected. In a viscous flow method only velocities in the axial direction will result. It is questionable whether the typical nonuniform waterjet velocity distribution can be represented in an irrotational velocity field then. necessary to ensure irrotational flow. but that is not the object of this analysis.68 to 2. After a pipe length of about 1 diameter the flow field is almost completely uniform. These velocity components are constrained by the wall boundary conditions.13a to 5.1). Here a very small decay of the nonuniformity can be seen. until it is dissipated into heat.1
Test case with nonuniform pipe flow
For this test the nonuniform axial velocity distribution will be prescribed at the inlet of a circular pipe.
A. which will imply constrains on the admissible inlet velocity distribution. as function of the nondimensional pipe length for the potential flow method. On the other hand. This will not be sufficient to get complete uniform flow in the RANS calculation. Transformation into heat is thus the only remaining option.19 have been analysed. In this type of flow velocity gradients contribute to the vorticity. This velocity field is a result of boundary layer ingestion and retardation of the flow. This hypothesis can be verified.
5 2 2.68 0 0 0.25
0.87 IVR = 1.2
Development of nonuniformity ζ in flow with vorticity for circular tube
Numerical analysis of a waterjet propulsion system
179
.5
Figure A.05 IVR = 1.19 IVR = 2.15
0.03 0.3
0.5 1 1. Use of a potential flow method for the analysis of a waterjet installation is therefore not possible.25
0.
0.3
0.1
Development of nonuniformity ζ in potential flow for circular tube
0.2
0.05 IVR = 1.19 IVR = 2.1 IVR = 2.03 0.68 0 0 0.5
Figure A.1 IVR = 2.1 Test case with nonuniform pipe flow
It can be concluded that there are significant differences between both methods.15
0.87 IVR = 1.5 2 2.A.2
0.5 1 1.
Appendix A Stability of nonuniform velocity distribution
A.. ASME Journal Basis Engineering. p. 82.131. H.2
References
[1] Ito. ‘Pressure losses in smooth pipe bends’. Vol. 1960
180
.
7 to B. The solution is monitored at six positions in the domain during the solution process.12 as well. The Fourier analyses for the results of calculations with uniform inflow are presented in figures B.Appendix B
Fourier analyses of transient flow calculations
Results of the Fourier analyses of the axial velocity and pressure obtained for the transient calculations are presented in this appendix. Three points are located upstream of the impeller and three in between the rotor and the stator. The results of the pressure level upstream of the pump show small fluctuations with a frequency of about 13 times the blade passing frequency.11 to B. The Fourier analyses of the solution of these last three points can give a good indication whether the solution has become periodic. These pressure results are not used for the evaluation of periodicity of the flow field.6. The results of the Fourier analyses for design flow rate Q are presented in figures 6. The frequencies are normalised with the shaft frequency in all diagrams.1 to B.10 for design flow rate and in figures B. The results of the calculations with nonuniform velocity distributions are presented in figures B.14 for 80% of the design flow rate. Fourier analyses of these results do not provide additional information.11 and 6. Review off all results confirms the periodic behaviour of both the axial velocity and the pressure at the three arbitrarily chosen monitoring points. The results of the axial velocity at the monitoring points upstream of the pump are governed by the inlet boundary condition.
Numerical analysis of a waterjet propulsion system
181
.
5 4 3.05 Normalised axial velocity [] 0.02 0.5 2 1.045 0.01 0.5 1 0.5 4 3.5 2 1.04 0.05 Normalised axial velocity [] 0.005 0 0 5 Dimensionless pressure [] 4.04 0.5 3 2.01 0.035 0.03 0.03 0.035 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
Figure B.5 3 2.1
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for 60% of design flow rate
0.02 0.2
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for 70% of design flow rate
182
.045 0.025 0.Appendix B Fourier analyses of transient flow calculations
0.005 0 0 5 Dimensionless pressure [] 4.025 0.015 0.015 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
Figure B.5 1 0.
01 0.05 Normalised axial velocity [] 0.0.035 0.04 0.3
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for 80% of design flow rate
0.5 Dimensionless pressure [] 4 3.015 0.015 0.005 0 0 5 4.025 0.03 0.5 4 3.02 0.4
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for 90% of design flow rate
183
Numerical analysis of a waterjet propulsion system
.02 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.045 0.03 0.5 3 2.005 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.025 0.5 3 2.5 2 1.05 Normalised axial velocity [] 0.01 0.04 0.5 1 0.045 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30
Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
Pressure point 4 Pressure point 5 Pressure point 6
Figure B.035 0.5 1 0.5 2 1.
025 0.5 2 1.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.04 0.5 1 0.5 3 2.04 0.035 0.5 3 2.045 0.5 4 3.005 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.Appendix B Fourier analyses of transient flow calculations
0.5 2 1.6
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for 110% of design flow rate
184
.03 0.005 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.025 0.5 4 3.015 0.03 0.02 0.5
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller with uniform inflow for the design flow rate
0.01 0.5 1 0.02 0.015 0.035 0.05 Normalised axial velocity [] 0.05 Normalised axial velocity [] 0.01 0.045 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.
5 4 3.05 0.02 0.03 0.06 0.5 1 0.06 0.08 0.7
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=1.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.05 0.01 0 0 5 Dimensionless pressure [] 4.07 0.09 0.5 4 3.08 0.02 0.1 Normalised axial velocity [] 0.8
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=1.5 1 0.5 2 1.04 0.03 0.87 at design flow rate
185
Numerical analysis of a waterjet propulsion system
.0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
Figure B.68 at design flow rate
0.5 3 2.07 0.5 3 2.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.1 Normalised axial velocity [] 0.09 0.04 0.5 2 1.
06 0.1 Normalised axial velocity [] 0.04 0.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.08 0.09 0.04 0.5 2 1.5 1 0.07 0.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.05 0.19 at design flow rate
186
.03 0.03 0.1 Normalised axial velocity [] 0.07 0.02 0.Appendix B Fourier analyses of transient flow calculations
0.5 4 3.06 0.5 4 3.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.5 3 2.5 2 1.03 at design flow rate
0.5 3 2.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.02 0.5 1 0.05 0.09 0.9
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=2.08 0.10
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=2.
5 4 3.87 at 80% of design flow rate
187
Numerical analysis of a waterjet propulsion system
.06 0.1 Normalised axial velocity [] 0.5 3 2.03 0.11
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=1.0.5 4 3.5 1 0.02 0.08 0.07 0.09 0.02 0.5 1 0.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.05 0.05 0.68 at 80% of design flow rate
0.1 Normalised axial velocity [] 0.12
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=1.07 0.04 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.09 0.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.08 0.5 2 1.5 3 2.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.5 2 1.03 0.04 0.06 0.
07 0.08 0.02 0.5 1 0.09 0.02 0.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.05 0.5 2 1.19 at 80% of design flow rate
188
.04 0.5 2 1.Appendix B Fourier analyses of transient flow calculations
0.5 3 2.1 Normalised axial velocity [] 0.1 Normalised axial velocity [] 0.09 0.5 4 3.06 0.5 3 2.5 1 0.08 0.06 0.03 0.13
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=2.03 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.07 0.5 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Pressure point 4 Pressure point 5 Pressure point 6
Figure B.5 4 3.01 0 0 3 6 9 12 15 18 Normalised frequency [] 21 24 27 30 Axial velocity point 4 Axial velocity point 5 Axial velocity point 6
5 Dimensionless pressure [] 4.05 0.04 0.03 at 80% of design flow rate
0.14
Fourier transforms of axial velocity and pressure at monitoring points downstream of the impeller for IVR=2.
the water is ingested from below the the hull of the ship. Finally. commercial fastferries reach velocities of 50 knots. Four contributing factors are identified for the development of a nonuniform velocity distribution just upstream of the pump. The prediction of the thrust of a propeller is based on the momentum balance of a streamtube control volume. where a boundary layer with a nonuniform velocity distribution is present. a pump and a nozzle. which is about 90 km/h. which results in an increase of the nonuniformity. The inflow to the waterjet pump is nonuniform. As a first cause. Nowadays. The development of waterjet propulsion systems has made significant progress over the last few decades. For manoeuvring and reversing purposes an additional steering device can be integrated into the installation. This thrust is then transferred though the shaft of the propeller to the hull of the ship. Even at normal operating conditions. for a waterjet propulsion system. the inflow passes the bend in the inlet and the protruding
Numerical analysis of a waterjet propulsion system 189
.Summary
A waterjet propulsion system is used to propel ships. the water is subsequently retarded in the inlet. In contrast. forces are transferred to the hull not only through the shaft but also through the solid surface of the installation. The cause and the effects of this nonuniform inflow have been investigated. A standard waterjet installation can be divided into an inlet. The theory to describe waterjet propulsion systems is derived from open propeller theory. This results in a blade loading that varies during an impeller revolution. A critical review learns that some assumptions made for open propellers are not valid for waterjets. using a pump which produces a high speed jet.
Due to the high level of nonuniformity of the inflow. It is concluded that the nonuniformity is the result of the accumulated vorticity in the flow. Selection of the numerical method is based on the capability to capture typical flow phenomena in a waterjet installation: high Reynolds number. The numerical models of both the waterjet inlet and the mixedflow pump are validated with available experimental data. An estimation of the influence of this error on the prediction of thrust and torque shows that the actual deviations are acceptable. The Reynoldsstresses are obtained using the twoequation kε turbulence model. This enables the visualisation of the streamtube and the calculation of the mass averaged inflow velocity. The flow phenomena in a waterjet inlet are characterised by the inlet velocity ratio (IVR). and incompressible flow in a partially rotating frame of reference. Differences between predicted head and power in both methods are small. as well as the possibility to take into account the flow phenomena in the tip clearance region between the rotating blades and the stationary housing. The influence of the nonuniform velocity distribution to the pump is investigated as well. the ability to generate and transport vorticity in the flow is an important requirement. It is found that the magnitude of the radial interaction force depends on the flow rate though the pump. Due to this vorticity. The CFD calculations of the mixedflow pump are validated with experimental data for the pump head and the shaft power. A Reynolds averaged NavierStokes (RANS) method is chosen to perform all numerical analyses. This turbulence model is known to produce an error near a stagnation point. In this way the wake fraction of the waterjet installation is determined accurately. The deviation in pump performance is limited to a few
190
. Agreement between the CFD results and the experimental data is good for all calculated conditions. which is the ratio of the ship speed and the pump speed. and the typical velocity distribution is more or less independent of the actual design of the inlet.Summary
shaft which add to the increase in nonuniformity. It is shown that the actual shape of the streamtube depends on IVR. a stable velocity distribution is found. The fully transient moving mesh calculations with a uniform inflow velocity distribution provide the unsteady excitation forces on the impeller due to rotorstator interaction. The investigations are based on numerical analyses of the flow through the complete waterjet installation. Results of calculations of the waterjet inlet flow are compared with measurements of static pressure along the inlet and with the total pressure and velocity distribution at the impeller plane. The shape and location of the streamtube of the ingested water is determined with aid of a concentration scalar. timedependency. The calculations are performed with a quasisteady multiple frame of reference (MFR) method and a fully transient moving mesh method.
Two methods to determine the thrust are used: (i) the integration of the axial force component on the solid wall and (ii) the application of a simplified version of the integral momentum balance equation. has some shortcomings. The influence on radial forces is far greater. however. The origin of this mean force is an unbalanced torque on the impeller blades.
Numerical analysis of a waterjet propulsion system
191
. the magnitude and direction of which are related to the flow rate and the level of nonuniformity. Good agreement is found for the prediction of flow rate. The numerical results confirm the hypothesis that the simplified method to describe waterjet installations is not correct. An additional mean component of the radial force is found. Results of the calculations of the complete unit are compared with the results of the standard waterjet performance prediction and selection software of Wärtsilä Propulsion Netherlands BV. The deviation increases for higher ship speeds. The latter method is generally applied by ship building companies. due to a variation of the angle of attack during a revolution. A clear deviation between the two methods is found for higher ship speeds. thrust and torque of the installation. Both validated numerical models of the inlet and the pump are combined to form the complete waterjet installation. This can be partly attributed to the neglect of the influence of the hull in the vicinity of the waterjet inlet and partly to the neglect of the contributions of the pressure distribution acting on the streamtube. Analysis of the net force in vertical direction reveals a significant lift force at high speeds. It is concluded that the method based on the momentum balance for the streamtube control volume.percent for the calculated conditions.
Summary
192
.
de krachten die werken op een waterjet voortstuwingssysteem worden behalve via de as ook via de vaste wanden van de installatie naar het schip overgebracht. Dit heeft tot gevolg dat de belasting van de waaierbladen varieert gedurende een omwenteling.Samenvatting
Een waterjet voortstuwingssysteem wordt gebruikt om schepen voort te stuwen met behulp van een pomp. Allereerst wordt het water afgezogen uit het gebied onder het schip. De instroming naar de waterjet is nietuniform. De voorspelling van de stuwkracht van een schroef is gebaseerd op de impulsbalans van een stroombuis controle volume. Echter. Tegenwoordig worden met commerciële fastferries snelheden van rond de 50 knopen bereikt. De ontwikkeling van waterjet voortstuwingssystemen heeft de laatste decennia een enorme ontwikkeling doorgemaakt. De oorzaak en de gevolgen van deze nietuniforme instroming zijn onderzocht. Een stuurdeel kan geïntegreerd worden in de installatie om manouvreren mogelijk te maken. De theorie om het voortstuwingssysteem te beschrijven is afgeleid van de theorie voor open schroeven. Deze stuwkracht wordt door de as van de schroef op het schip overgebracht. Daar bevindt zich een
Numerical analysis of a waterjet propulsion system 193
. Een standaard waterjet voortstuwingssysteem is opgebouwd uit een inlaat. Een kritische analyse laat zien dat sommige aannames die gemaakt zijn voor open schroeven niet geldig zijn voor waterjets. die een waterstraal met hoge snelheid produceert. een pomp en een nozzle. Dat komt overeen met ongeveer 90 km/uur. Er zijn vier factoren geïdentificeerd die bijdragen aan de ontwikkeling van de nietuniformiteit van de instroming vlak voor de pomp.
Samenvatting
grenslaag met een nietuniforme snelheidsverdeling. in een systeem dat gedeeltelijk roteert. Vervolgens ondergaat deze instroming een vertraging in de inlaat van het systeem. nietsamendrukbare stroming met een hoog Reynolds getal. Het is aangetoond dat de vorm van de stroombuis afhangt van de waarde van IVR. wat resulteert in een toename van de nietuniformiteit. Resultaten van berekeningen van de stroming door de waterjet inlaat zijn vergeleken met metingen van de statische druk langs de inlaat en met de totale druk en de snelheid in het waaiervlak. Uit een schatting van de invloed van deze fout op de voorspelling van stuwkracht en koppel blijkt dat de afwijkingen acceptabel zijn. Er kan worden geconcludeerd dat de nietuniformiteit het resultaat is van een accumulatie van vorticiteit in de instroming. Door deze vorticiteit wordt een stabiele snelheidsverdeling gevormd en daardoor is de typische snelheidsverdeling nagenoeg onafhankelijk van de vorm van de inlaat. Deze kan worden gekarakteriseerd als een tijdsafhankelijke. Op deze manier kan het volgstroom getal (of wakefractie) van de waterjet installatie nauwkeurig worden bepaald. waardoor de nietuniformiteit verder wordt vergroot. De berekeningen zijn uitgevoerd met een quasistationaire multiple frame of reference (MFR) methode en met een volledig tijdsafhankelijke roterende
194
. Als laatste passeert de instroming een bocht in de inlaat en de as van de pomp. Het is bekend dat dit model een fout introduceert nabij een stagnatie punt. De CFD berekeningen van de mixedflow pomp zijn gevalideerd met experimenteel bepaalde waarden van opvoerhoogte en asvermogen. Dit geeft de mogelijkheid om de stroombuis te visualiseren en om de massagemiddelde instroomsnelheid te bepalen. De overeenkomst tussen de CFD resultaten en de experimentele data is goed voor alle berekende condities. Vanwege de hoge mate van nietuniformiteit zijn verder de productie en het transport van vorticiteit belangrijke aspecten. De eisen die worden gesteld aan de numerieke methode worden bepaald door de aard van de stroming door een waterjet installatie. De vorm en de ligging van de stroombuis waardoor het water naar de inlaat van de waterjet stroomt. evenals de mogelijkheid om de stroming in de nauwe spleet tussen de roterende waaier en het stationaire pomphuis goed te kunnen modelleren. De Reynoldsspanningen worden bepaald aan de hand van het kε turbulentie model. wat de verhouding geeft tussen de scheepssnelheid en de gemiddelde axiale snelheid vlak voor de pomp. Een Reynoldsgemiddelde NavierStokes (RANS) methode is gekozen voor alle numerieke analyses. De numerieke modellen van zowel de waterjet inlaat als ook de mixedflow pomp zijn gevalideerd met beschikbare experimentele data. is bepaald met behulp van een concentratiescalar. De stromingsfenomenen in een waterjet inlaat worden bepaald door de inlet velocity ratio (IVR). Het onderzoek is gebaseerd op een numerieke analyse van de stroming door de complete waterjet installatie.
Voor wat betreft de berekende waarden van opvoerhoogte en vermogen zijn de verschillen tussen beide methodes klein. Goede overeenkomsten zijn gevonden voor de voorspelling van het debiet door de pomp. De volledig tijdsafhankelijke roterende mesh berekeningen met uniforme instroming geven de instationaire excitatiekrachten op de waaier als gevolg van rotorstator interactie. De numerieke resultaten bevestigen de hypothese dat de vereenvoudigde methode voor de beschrijving van de waterjet installatie niet correct is. De afwijking in de pompprestaties is beperkt tot een paar procenten voor de berekende condities. tekortkomingen heeft. De grootte van de radiale interactiekracht blijkt afhankelijk van het debiet door de pomp.
Numerical analysis of a waterjet propulsion system
195
. Voor hoge scheepssnelheden is een duidelijke afwijking gevonden tussen de twee methoden. De twee gevalideerde numerieke modellen van de inlaat en de pomp zijn gecombineerd om een complete waterjet installatie te vormen. Resultaten van de berekeningen van de complete unit zijn vergeleken met het standaard waterjet performance predictie programma van Wärtsilä Propulsion Netherlands BV. de stuwkracht en het koppel van de installatie. als gevolg van de variatie in de aanstroomhoek van het blad tijdens een omwenteling. Een additionele tijdsgemiddelde component van de radiale kracht is aanwezig met een grootte en richting die afhangt van het debiet en de mate van nietuniformiteit. De invloed van de nietuniforme snelheidverdeling vlak voor de pomp is ook onderzocht. De laatste methode wordt algemeen toegepast door producenten van schepen. Analyse van de netto kracht in verticale richting laat zien dat er een significante liftkracht aanwezig is bij hoge scheepssnelheden. Dit kan deels worden toegeschreven aan de verwaarlozing van de invloed van de scheepswand in de buurt van de waterjet inlaat en deels aan de verwaarlozing van de bijdragen van de drukverdeling die werken op de stroombuis. Hieruit volgt de conclusie dat de vereenvoudigde methode. De invloed op de radiale krachten is echter veel groter. gebaseerd op de impuls balans voor het stroombuis controle volume.mesh methode. Twee methoden voor de bepaling van de stuwkracht zijn gebruikt: (i) integratie van de axiale component van de kracht op alle vaste wanden en (ii) de toepassing van een vereenvoudigde versie van de integrale impuls balans. De afwijking wordt groter naarmate de scheepssnelheid toeneemt. De oorzaak van deze tijdsgemiddelde kracht is een onbalans in het koppel van de verschillende waaier bladen.
Samenvatting
196
.
Rosmalen. De combinatie van het onderzoek met mijn werkzaamheden bij Wärtsilä Propulsion Netherlands BV was niet mogelijk geweest zonder alle studiedagen. die waarschijnlijk de meest waardevolle bijdragen aan dit proefschrift hebben opgeleverd. Dit advies heb ik opgevolgd en dit proefschrift is nu het resultaat van de combinatie van 9 jaar werken bij Wärtsilä Propulsion Netherlands BV en het promotie onderzoek aan de Technische Universiteit Eindhoven. Bert Brouwers het advies om ‘eerst maar eens aan het werk te gaan en dan later een keer een promotie onderzoek te komen doen’.Dankwoord
Tijdens mijn afstuderen kreeg ik van prof. Bij het zoeken naar verbeteringen van de waterjet installaties heb ik vele gesprekken en discussies gehad met Rob Verbeek. dr. In ieder geval hebben we de term ‘voortschrijdend inzicht’ bij het management geïntroduceerd. Hiervoor wil ik Bram Kruyt. Harry Hoeijmakers en mijn directe begeleider. Norbert Bulten. Bert Brouwers wil ik ook mijn tweede promotor prof. Bart van Esch.
Numerical analysis of a waterjet propulsion system
197
. iets waar vooral Bart mee te maken had. bedanken voor hun bijdragen. Zelfs in tijden van drukte was het mogelijk om tijd aan het onderzoek te besteden. Hanno Schoonman en Do Ligtelijn bedanken. september 2006. Het interessante van de combinatie van werken en promoveren is waarschijnlijk de directe feedback van de varende waterjet installaties. Naast prof. Ik ben me ervan bewust dat het begeleiden van een externe promovendus niet altijd even gemakkelijk was.
Dankwoord
198
.
Project for IHC Parts & Services to optimise dredger pumps Wärtsilä Propulsion Netherlands (formerly Lips propellers) in Drunen. 1997 . Switzerland Researcher at University of Twente. The Netherlands.Curriculum Vitae
29 april 1973 Born in Winterswijk.2002 Research engineer waterjets 2002 . Winterswijk Study Mechanical Engineering at the University of Twente. The Netherlands Specialisation: fluid dynamics and thermal engineering Graduation thesis on numerical analysis of a centrifugal pump impeller with two different RANS methods 1996 1997 Traineeship at Sulzer Innotec in Winterthur. from 2006 at Propulsor Technology
1997
Team leader CFD at Propulsor Technology department
20012006
PhD study at Technische Universiteit Eindhoven. The Netherlands
Numerical analysis of a waterjet propulsion system
199
.2006 CFD specialist department. the Netherlands. 19851991 19911996 VWO at RSG Hamaland.
Curriculum Vitae
200
.