UNIVERSITA’ DEGLI STUDI DI PISA

Master Thesis in Space Engineering

DEVELOPMENT OF A MODELING
FRAMEWORK OF THE FEEDING SYSTEM FOR
THE CHARACTERIZATION OF POGO
OSCILLATIONS

Candidate

Mario Amoroso

Supervisor

Dott. Ing. Angelo Pasini

Academic Year 2016-2017

Abstract
This thesis has been carried out within the framework of the MIT-UNIPI Project funded by MISTI Global
Seed Funds and entitled “Dynamic Characterization of POGO Instabilities in Cavitating Turbopumps”,
which provides a collaboration between the Massachusetts Institute of Technology and the University of
Pisa, aiming to jointly develop a novel theoretical foundation capable of characterizing the dynamics of
POGO oscillations and devising new design guidelines.
The first section of this thesis deals with a literature review of the past POGO experiences of NASA
human spaceflights and a collection of some procedures used to obtain a prediction of the dynamic
performances of space rocket turbopumps.
In the second part, a modeling framework defined in the time-domain has been developed to characterize
the steady-state and dynamic behaviour of each component of a typical feeding system for liquid rocket
engines. A typical water loop for experimental characterization of liquid rocket turbopumps has been
modeled according to the modeling framework in order to understand the best way to perform forced
experiments for the characterization of the transfer matrix of cavitating turbopumps necessary for
understanding the POGO instability phenomena that affect rocket launchers. The best results in terms of
capability of generating mass flow rate and pressure oscillations at the inlet of the inducer, have been
obtained by means of a device that produces a volume oscillation located downstream of the pump.

1
Contents
1 Detailed Review of the POGO Instability ...........................................................................7
1.1 Introduction ................................................................................................................................... 7
1.2 POGO Instabilities Episodes in NASA Human Spaceflight Vehicles .......................................... 8
1.2.1 Gemini – Titan II Experience................................................................................................ 8
1.2.2 Apollo – Saturn V Experience ............................................................................................ 10
1.2.3 1970 POGO State of the Art ............................................................................................... 15
1.2.4 Space Shuttle Experience .................................................................................................... 17
1.2.5 Recent Challenges ............................................................................................................... 18
2 Characterization of the Dynamic Transfer Matrix of Space Rocket Turbopumps ............21
2.1 Introduction ................................................................................................................................. 21
2.2 The Dynamic Transfer Matrix of a Cavitating Pump ................................................................. 21
2.3 Analytical and Experimental Turbopump Matrix Characterization ............................................ 22
3 Mathematical Model of the System ...................................................................................40
3.1 Introduction ................................................................................................................................. 40
3.2 Development of the Dynamic Equations for the Subsystems ..................................................... 41
3.2.1 Incompressible Duct: Straight (ID-S), Elbow (ID-E) and Tapered (ID-T) ......................... 42
3.2.2 Compressible Duct Straight (CD-S).................................................................................... 42
3.2.3 Silent Throttle Valve (STV) – Exciter ................................................................................ 43
3.2.4 Volume Oscillator Valve (VOV) – Exciter ......................................................................... 44
3.2.5 Tank (T) – Exciter ............................................................................................................... 45
3.2.6 Pump (P) ............................................................................................................................. 46
4 System Design Tools .........................................................................................................48
4.1 Introduction ................................................................................................................................. 48
4.2 Incompressible VS Compressible Solution ................................................................................. 48
4.2.1 Downstream Mass Flow Rate and Pressure Signal Comparison ........................................ 48
4.2.2 Conclusions and Results ..................................................................................................... 73
4.2.3 Duct Transfer Matrix Comparison ...................................................................................... 73
4.2.4 Conclusions and Results ..................................................................................................... 77
4.3 Hydraulic Loop System Design, Semi-Compressible Approach ................................................ 77
4.3.1 Conclusions and Results ..................................................................................................... 84
5 Conclusions and Future Developments .............................................................................85
6 Appendixes ........................................................................................................................86
6.1 Appendix A, Mathematical Model Equations ............................................................................. 86
6.1.1 Introduction ......................................................................................................................... 86

2
 6.1.2 Subsystems Mathematical Model ....................................................................................... 86
6.1.3 Incompressible Duct (ID).................................................................................................... 87
6.1.4 Compressible Duct Straight (CD-S).................................................................................... 93
6.1.5 Silent Throttle Valve (STV) .............................................................................................. 100
6.1.6 Volume Oscillator Valve (VOV) – Exciter ....................................................................... 104
6.1.7 Tank (T) – Exciter ............................................................................................................. 105
6.1.8 Pump (P) ........................................................................................................................... 109
6.2 Appendix B, Matlab Code ........................................................................................................ 113
6.2.1 Introduction ....................................................................................................................... 113
6.2.2 Steady-State System Parameters ....................................................................................... 113
6.2.3 Dynamic System Parameters ............................................................................................ 115
6.3 Appendix C, Simulink Modeling of the Hydraulic Loop ......................................................... 118
6.3.1 Introduction ....................................................................................................................... 118
6.3.2 Simulink Environment ...................................................................................................... 118
6.4 Appendix D, Simulink Modeling for the Comparison of the Incompressible Solution with the
Compressible Solution .......................................................................................................................... 127
6.4.1 Downstream Mass Flow Rate and Pressure Signal Comparison Circuit .......................... 127
6.4.2 Duct Transfer Matrix Comparison Circuit ........................................................................ 128
6.5 Appendix E, Table of Figures ................................................................................................... 130
6.6 Appendix F, List of Tables ....................................................................................................... 136

3
Introduction
The subsequent work aims to:

1. Present a review of the POGO instability episodes occurred in the past and collect the analytical
and experimental procedures exploited worldwide to study the behavior of the pump
withstanding unsteady conditions and to prevent the occurring of unstable phenomena;
2. Develop a mathematical model able to predict the steady and unsteady-state of the subsystems
operating within the hydraulic context of the pump;
3. Develop some tools useful to drive the design of an experimental apparatus, pointing out whether
to exploit the simplified incompressible assumption of the working fluid and, in case, provide a
solution able to take into account the compressibility effects.

4
Nomenclature
A Cross-section area, m2.
a , a0 Unperturbed acoustic velocity, m/s.
a Unsteady acoustic velocity, m/s.
CT Tank compliance, m4s2kg-1.
DH Hydraulic diameter, m.
e Unit vector.
f Frequency of perturbation, Hz.
fu Maximum frequency of perturbation of interest, Hz.
g Steady gravity acceleration, m/s2.
g Unsteady gravity acceleration, m/s2.
h Tank height, m.
H ij Transfer matrix component; ith row, jth column.

I Impedance, m-1s-1; Inertance, m-1s-1.

i Complex unit.
K Complex constant, kgm-2s-1.
k Complex constant, m-1.
kloss Loss coefficient.

kloss Steady loss coefficient.

kloss Unsteady loss coefficient.

L Duct Length, m.
M Mass flow rate coefficient.
m Mass flow rate, kg/s.
m Steady mass flow rate, kg/s.
m Unsteady mass flow rate, kg/s.
m̂ Unsteady mass flow rate, complex amplitude, kg/s.
P Pressure coefficient.
p Static pressure, Pa.
p Steady static pressure, Pa.
p Unsteady static pressure, Pa.

5
p̂ Unsteady static pressure, complex amplitude, Pa.
q Steady independent variable.
q Unsteady independent variable.
R Resistance, m-1s-1.
r Vertical tank oscillation amplitude, m.
rT Pump tip radius, m.

r Steady inertial acceleration, m/s2.

r Unsteady inertial acceleration, m/s2.
t Time, s.
V Volume, m3.
w0 Unperturbed axial velocity, m/s.
 Specific heat ratio.
 Density, kg/m3.
 , 0 Steady density, kg/m3.
 Cavitation number.
 Flow coefficient.
 Pump rotational speed, m/s.
 Radiant frequency of perturbation, rad/s.
 Head coefficient.

6
1 Detailed Review of the POGO Instability

1.1 Introduction
Propulsion stages can suffer from a dynamic coupling of the combustion process with structure and feed
system dynamics, called POGO. As the name suggests, this causes rapid positive and negative
accelerations along the thrust axis, as showed in Figure 1.1. POGO-type instabilities can result in severe
vibration, interference with the guidance systems, and possible destruction of the stage.
This instability is named after the children’s stick toy.

Figure 1.1 Typical occurrence of POGO vibration (NASA, 1970)

The so-called “POGO” instability consists of coupled vehicle structure/propulsion system oscillations, as
schematically showed in Figure 1.2, resulting in a severe longitudinal unstable phenomenon.

Figure 1.2 Block diagram of POGO feedback process (NASA, 1970)

An overview of more than 45 years of NASA human spaceflight experience is presented with respect to
the thrust axis vibration response of liquid fueled rockets known as POGO, inspired by the work of
Larsen, 2008 and NASA, 1970.

7
1.2 POGO Instabilities Episodes in NASA Human Spaceflight Vehicles
1.2.1 Gemini – Titan II Experience

The NASA history begins with the Gemini Program and adaptation of the USAF Titan II ballistic missile
as a spacecraft launch vehicle. It continues with the pogo experienced on several Apollo-Saturn flights in
both the first and second stages of flight. The defining moment for NASA’s subsequent treatment of pogo
occurred with the near failure of the second stage on the ascent of the Apollo 13 mission. Since that time
NASA has had a strict “no POGO” philosophy that was applied to the development of the Space Shuttle.
These efforts lead to the first vehicle designed to be POGO-free from the beginning and the first
development of an engine with an integral pogo suppression system.
NASA first identified POGO as a threat to spaceflight vehicles and their crews in the early 1960’s during
the tests of the Titan II launching vector, for the Gemini-Titan II program, showed in Figure 1.3.

Figure 1.3 Gemini-Titan

The USAF began test flights with the Titan II ballistic missile on March 16, 1962. Ninety seconds into the
first-stage flight the missile began a longitudinal vibration going from 10-13 Hertz for roughly 30
seconds, reaching a maximum amplitude of 2.5 g’s at about 11 Hertz.

8
Even for a military payload this design environment was excessive (the USAF considered 1.0 g as a
tolerable design load for the structure of the Titan II and it’s payload), especially considering manned
flights.
NASA required that the vibrations be kept below 0.25 g’s.
The first thought was that the POGO vibration might be caused by pressure oscillations in the propellant
feedlines and after a mathematical modeling evaluation, a vertical surge-suppressor pipe was added to
each of the oxidizer feedlines.
Test results show a worsening of the vibrational condition, with a maximum value of 5 g’s reached. A
partial explanation was given by sequent investigations: the accumulator on the oxidizer lines was
prompted to the closeness between the frequency of vibration of the structure and the oxidizer one, while
the fuel natural frequency was retained to be well above the structural ones; theoretically, then, having
lowered the oxidizer frequency enough w.r.t. the structure and with a fuel response that shouldn’t have to
be activated, the coupling action between propellant system and structure should have been weakened.
This didn’t happen because of the presence of cavitating bubbles at the inlet of both the oxidizer and the
fuel pumps in such a way that the fuel frequency were lowered resulting close to the structural one.
Before the installation of the surge-suppressor, the two oscillating response of fuel and oxidizer competed
through phasing, giving rise to a soft instability via thrust chamber pressure perturbation, then coupled
with the structure, while with the presence of the accumulator in the oxidizer feedlines, the competing
action of the oxidizer lines, missing, left free way to the action of the fuel lines perturbation which,
coupling with the structure, gave rise to a worse instability.
The addition of accumulators in each engine’s fuel line was shown to be essential to eliminate POGO on
the Titan II.
The next Titan II flew on December 19, 1962 with no standpipes, but increased fuel-tank pressure and
aluminum oxidizer feedlines instead of steel. Surprisingly, the POGO amplitude was lessened but no
reason for the effect was readily apparent. POGO on the tenth flight on January 10, 1963, was recorded at
a new low of 0.6 g at the spacecraft interface. But the NASA requirement for the Titan II remained as
0.25 g at most due to the larger role astronauts were to play in piloting Gemini compared to Mercury.
In a subsequent review with the commanding USAF general, the Titan II contractors argued POGO could
be solved by increased fuel-tank pressure, and a combination of standpipes in the oxidizer lines and
mechanical accumulators in the fuel lines.
The 17th test flight on May 13, 1963, reached a new low amplitude record for Titan II POGO of 0.35g.
Titan II launched again on September 23, 1963, and suffered a guidance malfunction unrelated to the
Gemini booster configuration. Pogo on this launch was reached plus or minus 0.75g.
An October meeting of the USAF management considered whether to follow through with plans to fly
Missile N-25 with oxidizer standpipes and fuel-side piston accumulators. Engine tests begun in August
had confirmed the fuel line resonance as the cause of the Missile N-11 failure and demonstrated that fuel
accumulators would solve the problem. The extensive testing was used to generate test-verified equations
describing the dynamics of structure, the propellant feed systems and the engines. Pump tests showed that
as inlet pressures were reduced toward cavitation, the pump started acting as an amplifier and large
oscillations resulted in the thrust chamber pressure. Aerospace and Space Technology Laboratories
argued strongly for the planned flight and won the crucial decision to fly as planned.
With both fuel and oxidizer suppressors installed, flight N-25 launched on November 1, 1963, recording
the lowest vibration levels ever on Titan II of only 0.11 g's, well below the 0.25 g required by NASA as
the upper limit for pilot safety.
The first and second unmanned Gemini launches (April 8, 1964 and January 19, 1965) didn’t show
significant levels of POGO, proving Gemini's spacecraft and launch.
The first Gemini crew (March 23, 1965), also didn’t notice any remarkable level of vibration, while the
Gemini V crew of Gordon Cooper and Pete Conrad reported experiencing POGO during
launch at 126 seconds, although the booster systems engineer didn’t notice any evidence on telemetry.
Unable to read the panel gauges to the desired degree of accuracy and finding speech difficult, the pilot
estimated the magnitude at 0.5 g.

9
Post-flight data analysis showed POGO onset after 92 seconds, lasting for 46 seconds, with maximum
amplitude of 0.38g at the spacecraft-launch vehicle interface, as showed in the graph of Figure 1.4.

Figure 1.4 Comparison of Gemini-Titan POGO levels (NASA, 1965)

The subsequent mission of the Gemini program showed low values of vibrations.
In retrospect to the Gemini-Titan experience, it was recognized that the longitudinal oscillations
experienced on previous Mercury flights on the Redstone and Atlas launch vehicles were also POGO,
with the astronauts withstanding about 0.45 g.

1.2.2 Apollo – Saturn V Experience

In parallel to the Gemini Program, NASA was developing the Saturn rockets to take the Apollo spacecraft
to the moon. The Saturn I vehicles all flew with no occurrence of POGO. No sign of vibration were
recognized on the first Saturn V launched on Nov. 9, 1967, showed in Figure 1.5, carrying the unmanned
Apollo 4 spacecraft.

Figure 1.5 Saturn V First Stage S-IC

10
Apollo 6, the second unmanned Saturn V launched on April 4, 1968, experienced pogo at 5 Hz between
105-140 seconds during first stage boost with 0.60 g maximum acceleration at the command module and
0.33 g at the aft of the vehicle.
More subsequent analysis of the Apollo 4 and 6 flights demonstrated a previously unappreciated
sensitivity of POGO to what may have been thought to be inconsequential changes to the Apollo
spacecraft. In fact, Figure 1.6 shows that, though the Apollo 6 mission presented a rocket structure
heavier w.r.t. the Saturn V staged for the Apollo 4 of just 45 kg’s and therefore, theoretically a slightly
less first natural frequency, the coupling response of the system resulted to be much more amplified for
the Apollo 6 as the oxidizer frequency line crosses the structural one during the flight. The POGO
phenomenon, being so sensitive to small changes, confirms its unpredictable and therefore dangerous
nature.

Figure 1.6 Illustration of Sensitivity to Small Changes: Comparison of AS-501 and AS-502 (Ryan, Robert S., 1985)

The POGO problem for the Saturn V launcher was underestimated. Several tests were conducted to
decrease the susceptibility of the rocket to longitudinal vibrations, but the perturbation amplifying role of
the cavitation phenomenon taking place at the inlet of the pump, was neglected in favor of a deeper
investigation on how to modify the inlet line natural frequency in order to erase the coupling feedback
with the structure vibrations.
The solution proposed was to inject helium bubbles into the selected line to decrease enough its natural
frequency w.r.t. to the structural response, but the test results showed that the frequency variation of the
line could not be controlled under flight acceleration and tank pressure and therefore could not be
implemented.

11
The same concept, less extremist, was successfully suggested and applied by using helium gas from the
tank pressurization system as trapped gas in the oxidizer pre-valve to create an accumulator, as in Figure
1.7.

Figure 1.7 Saturn I-C POGO Mitigation (Von Braun, Wernher, 1975)

Apollo 8, the third Saturn V and first manned flight, successfully demonstrated the effectiveness of this
POGO mitigation for the first stage, while, unexpectedly, the second stage experienced POGO at about 50
seconds before engine cut-off. Data analysis revealed an 18 Hz vibration of the center engine of the five
engine J-2 cluster, due to the LOX tank oscillation. The magnitude at the crew cabin wasn’t significant,
but the local amplitude at the engine mount reached dangerous level for the supporting cross beam
structure showed in Figure 1.8.

Figure 1.8 Saturn V Second Stage S-II Close up, J-2 Engine Cluster (Doiron, Harold H., 2003)

12
A mitigation was attempted for the subsequent Apollo 9, increasing the LOX tank pressure and therefore
the bulk modulus of the oxidizer line, but another POGO vibration was detected with a maximum
amplitude of 12 g’s, as showed in Figure 1.9.

Figure 1.9 Apollo 8 and 9 POGO Episodes (Fenwick, J., 1992)

Since the structural vibrational load limit for the engine cluster structure was set at 15 g’s and thus there
were substantially no margin with respect to the load experienced, it was applied the operational
procedure to shut off about one minute earlier the central engine, keeping on burning a bit longer the
others four.
The most famous Apollo 11, experienced a small POGO vibration after 75 seconds into the second stage,
while four different POGO episodes occurred during the Apollo 12 second stage burn reaching a
maximum amplitude of 8 g’s.
During the Apollo 13 second stage burn (April 11, 1970), two episodes of POGO occurred on the center
J-2 engine as expected from previous missions, but the third occurrence diverged severely and
acceleration at the engine attachment reached an estimated 34 g’s before the engine’s combustion
chamber low-level pressure sensor commanded a shut down.

13
Figure 1.10 shows a comparison between the Apollo 13 and previous Apollo mission center engine thrust
pad acceleration of the second stage.

Figure 1.10 Comparison of Center Engine Thrust Pad Accelerations (Ryan, Robert S., 1985)

After the installation of a helium-bleed toroidal POGO suppressor for the oxidizer side of the J-2 engine
for the subsequent Apollo missions, no further significant POGO episode occurred.
Figure 1.11 shows a comparison of the vibration level reached during the flights of NASA space
programs until the Apollo one and the French Diamant B space vehicle.

Figure 1.11 POGO Experienced in Flight (Rubin S., 1970)

14
1.2.3 1970 POGO State of the Art

In October 1970 the previous experiences with POGO, were collected in a monograph entitled
“Prevention of Coupled Structure-Propulsion Instability (POGO)”, authored by Sheldon Rubin of the
Aerospace Corporation for NASA’s Langley Research Center.
State of the art, criteria and recommended practices for mathematical modeling, preflight tests, stability
analysis, corrective devices or modifications, and flight evaluation were provided, highlighting the
concept that the POGO instability had to be eliminated rather than managed as a peculiar dynamic load
condition, given the serious threat represented. In particular, a better understanding of the dynamic
behavior of the pump and the structure, was indicated as a key point.
The instability was addressed to a non-linear behaviour of the damping of the vehicle’s longitudinal
modes and the compliance and the dynamic gain components of the pump matrix.
Under the assumption of small perturbations, the linearization process led to a feasible analytical way to
approach the oscillation phenomena with mathematical models that subsequently should have been
validated.
Moreover, since the rate of change of the system properties is relatively slow, a series of system
parameters can be assumed constant for the stability analysis at successive time of flight.
At that time, one of the major problems to tackle, together with the pump characterization, which is
nowadays still only a partially solved component, was the modeling of the structural modal parameters. In
fact, the most reliable source of structural-damping data was a carefully executed modal test of the full-
scale vehicle.
The approach to the pump characteristics was experimental; Figure 1.12 shows the cavitation compliance
against the cavitation parameter response of the Titan I pumps.

Figure 1.12 Cavitation Compliance of Titan Stage 1 Pumps (NASA, 1970)

15
Moreover, the dynamic gain of pumps does not increase as rapidly with reduction of cavitation index as
steady-state characteristics would indicate. Figure 1.13 presents this behaviour with respect to oscillation
data.

Figure 1.13 Pump Gain for Titan Stage I Pumps (NASA, 1970)

Figure 1.14 shows a schematic view of some accumulators solutions thought for the launch vehicles of
the Gemini and Apollo program, close to the pump inlet, as recommended in the NASA monograph.

Figure 1.14 POGO Suppression Devices (NASA, 1970)

16
The installations denoted by a), b), c) and e) successfully flight. In particular, the f) solution, thought to
decrease the line acoustic velocity in such a way to provide the decoupling of the structure-propulsion
feedback, didn’t work due to the modification of several line natural frequencies in the structural range.
The idea behind these concepts is to lower the first resonance frequency of the line well under the first
structural longitudinal mode and to increase the second frequency of the line well above the first
structural one. The first and second line resonance frequencies, according to the article of Norquist
L.W.S. et al., 1969, are:
1
𝜔1∗ = [𝜌𝐿𝑆 (𝐶𝑎 + 𝐶𝑏 )]−2
1
−
2
𝜌(𝐿𝑎 + 𝐿𝑏 )
𝜔2∗ = [ ]
1 1
+
𝐶𝑎 𝐶𝑏

Where the subscripts a, b and s respectively stands for accumulator, pump and line. It’s clear why the
accumulators were designed to have a low internal inertance and a high compliance.

1.2.4 Space Shuttle Experience

The Space Shuttle Phase C/D development began in 1972.

Great attention was paid to POGO, and for the first time, a POGO suppressor device was designed and
installed as an integral component of the propulsion system rather than being added as a successive
remedy. Studies showed that the best location was at the inlet of the LOX high-pressure turbopump.
Figure 1.15 shows the actual Space Shuttle POGO Suppressor device.

Figure 1.15 Space Shuttle POGO Suppressor Device

It’s a spherical container charged with hot gaseous oxygen, acting as a Helmholtz resonator, to attenuate
LOX flow oscillations in the 5-50 Hz frequency band for a smooth flow rate of LOX into the high
pressure turbopump.
Since it was clear the unpredictable nature of the POGO phenomenon, a much more conservative
approach was used for the analysis of the possible instabilities, in fact, the safety margins to account for

17
an instability behave were increased and uncertainty factors were applied to the components of the
models exploited.
The flight test program included instrumented flights on STS-1 through STS-5. The pogo instruments
were basically pressure measurements installed in the liquid oxygen feedline and the SSMEs, and
accelerometers on the Orbiter thrust structure. Examination of flight data indicated a POGO-free vehicle
and the POGO flight measurement responses were as expected. The instrumentation verified that the
suppressors were fully charged and operating throughout the boost phase of each flight. The flight data
confirmed and validated the data developed in the ground test programs.
The POGO effort for the Space Shuttle program taught that prevention is much more effective than
providing an eventual remedy.

1.2.5 Recent Challenges

In the 2000’s NASA gave birth to the “Constellation Program”, which objectives, among the others, were
the fabrication of two launch vectors: the Ares I and Ares V.
We know that these objectives were never accomplished due to cost issues and subsequent cancellation of
the program in 2011 in favor of the SLS, but nevertheless it’s interesting to highlight the studies
performed for the design of a POGO suppressor device.
L.A. Swanson and T.V. Giel, 2009, presented a trade-off between a branch and two annular accumulator
concepts as schematically showed in Figure 1.16 and Figure 1.17:

Figure 1.16 LO2 Feed Line and Branch Accumulator (A.Swanson and T.Giel, 2009)

18
 Figure 1.17 Annular Accumulators (A.Swanson and T.Giel, 2009)

Defining the compliance and inertance of the devices as:

𝜌𝐿 𝑔𝑆 𝑉𝑜𝑙
𝐶𝐴 =
𝛾𝑔 𝑃𝑔 𝑔𝑐

(𝑡𝑤 + 𝐷𝑝 ) 𝑑𝑠
𝐼= +∫
(𝑔𝑆 𝐴𝑝 ) 𝑔𝑠 𝐴(𝑠)

Where:
ρL is the liquid propellant (LO2 ) density
gs is the gravitational acceleration at sea level
Vol is the helium charge gas volume
γg is the charge gas ratio of specific heats
Pg is the absolute pressure of the charge gas volume
gc is the conversion between mass and force
tw is the feed line wall thickness
Dp is the communication port diameter
Ap is the communication port area
𝑠 is the accumulator liquid flow path
A(s) is the liquid flow path area at 𝑠

Reminding that the two general guidelines of this kind of design are 1)High Compliance; 2)Low
Inertance, with a safety check on the liquid height inside the accumulator, i.e. a margin of safety which
ensure that the gas isn’t injected in the main feedline, Figure 1.18 and Figure 1.19 clarify why the shaped
annular accumulator concept was the one selected:

19
 Figure 1.18 Accumulator Compliance versus Liquid Level (A.Swanson and T.Giel, 2009)

Figure 1.19 Accumulator Inertance versus Liquid Level (A.Swanson and T.Giel, 2009)

A passive control system for the gas parameters during the transient time of flight, was selected to
complete the POGO mitigation design analysis.
The whole POGO experience made clear that every family of launchers thought to be fabricate, has to go
through a POGO suppressing design analysis, exploiting models of increasing precision adapted to the
global design steps leading to the final configuration, to obtain a POGO-free launch vector.

20
2 Characterization of the Dynamic Transfer Matrix of
Space Rocket Turbopumps

2.1 Introduction
The dynamic transfer matrix of a component can be seen as the mathematical description of its physics,
since, connecting the upstream values of fluctuating quantities to the downstream ones in a peculiar way,
it unveils the nature of the component. The first steps in the analytical and experimental characterization
of the dynamic matrix date back to the work of Brennen, Acosta and their collaborators in the 70s
(Brennen & Acosta, 1976; Brennen, 1978; Ng & Brennen, 1978). However, more recent works have
given important contributions by evaluating the previously obtained results through a careful analysis of
the successive experimental and numerical data (Otsuka et al., 1996; Rubin, 2004).
It is well known that many of the flow instabilities acting on axial inducers, including the above outlined
POGO oscillations, are significantly influenced by the dynamic matrix of the propulsion system
turbopumps (Kawata et al., 1988; Tsujimoto et al., 1993, 1998). This chapter will be consequently
devoted to a collection of the past and recent efforts for the characterization of the dynamic matrix of
cavitating/non cavitating pumps and its influence on the flow instabilities acting on the machine.

2.2 The Dynamic Transfer Matrix of a Cavitating Pump

Conventionally, the dynamics of hydraulic systems is treated in terms of “lumped parameter models”,
which assume that the distributed physical effects between two measuring stations can be represented by
lumped constants. This assumption is usually considered valid when the geometrical dimensions of the
system are significantly shorter than the acoustic wavelength at the considered frequency. As a direct
consequence of this assumption, the dynamic matrix of a generic system can be written as:

 p d   H11 H12   pu 
 =   
Q d   H 21 H 22  Q 
 u

where p and Q are, respectively, the pressure and flow rate oscillating components, and the subscripts u
and d denote, respectively, the flow conditions upstream and downstream of the considered system. As a
consequence of the well-known electrical analogy, the negative of the real part of H12 is usually denoted
as the system “resistance”, the negative of the imaginary part of H12 is referred as the system “inertance”,
the negative of the imaginary part of H21 is the system “compliance”, and the negative of the imaginary
part of H22 is the system “mass flow gain factor” (Bhattacharyya, 1994; Kawata et al., 1988).
If we assume unsteady, quasi 1-dimensional flow with small oscillations, pressure and flow rate can be
written in complex form as follows:

p(t )  p  p  eit
Q(t )  Q  Q  eit
where p and Q (usually real) are the pressure and flow rate steady values, p and Q (usually complex)
are the pressure and flow rate oscillating components, is the frequency of the oscillations.

21
Under the above assumptions, the dynamic matrix of a “passive” incompressible system (as simple duct
lines filled with water or another liquid), as well as that of a non cavitating pump, typically has the
following appearance:

 p d  1 -R - iωL   pu 
 =   
Q d  0 1  Q 
 u

or, in other words, only a resistance R and an inertance L are present.

On the other hand, for a cavitating pump, the compressibility of the cavitating region leads to a more
complicated form of the transfer matrix. Typical appearance of the matrix is as follows:

 p d  1-(S +iωX) -R - iωL   pu 

 =   
Qd   -iωC 1- iωM  Q 
 u

where S + iX is the pressure gain factor, R + iL is the pump impedance, C is the cavitation
compliance and M is the mass flow gain factor. The last two parameters (C and M) are generally functions
of the volume of the cavitating region inside the pump blades.

2.3 Analytical and Experimental Turbopump Matrix Characterization

C. Brennen and A. J. Acosta, 1975, presented a paper which shows an analytical approach to evaluate the
components of the transfer matrix related to the discharge mass flow vibration.
Linearizing the dynamics by confining attention to small oscillations about a particular steady operating
point, the problem is therefore to determine the transfer function [Z] for the cavitating turbomachine
where:

𝑝̃2 − 𝑝̃1 𝑍 𝑍12 𝑝̃1

{ } = [ 11 ]{ }
𝑚
̃2 − 𝑚 ̃1 𝑍21 𝑍22 𝑚̃1

Being the input and output quantities non-dimensional fluctuations of mass flow rate and pressure at the
inlet and discharge sections of the turbopump. Mass flow rate and pressure are made dimensionless
through the following expressions:
∗ ∗
𝑝̃1,2 𝑚
̃ 1,2
𝑝̃1,2 = ;𝑚
̃ 1,2 =
1 2 𝜌𝐴𝑖 𝑈𝑡
2 𝜌𝑈𝑡

Where 𝑈𝑡 is the tip speed, 𝐴𝑖 the inlet area, 𝜌 the fluid density and 𝜔 = 𝛺𝐻/𝑈𝑡 a reduced frequency
where 𝛺 is the actual frequency of the oscillations and H the distance between impeller blade tips.
Being the reduced frequency small for the treated application, an expansion of the matrix elements
through its power is performed:

𝑍21 = −𝑗𝜔𝐾𝐵 + 𝑂(𝜔2 )

𝑍22 = −𝑗𝜔𝑀𝐵 + 𝑂(𝜔2 )

22
The main concept behind this paper is to take into account the different conditions of pressure and
velocity of the flow and shape of the blades through the radial dimension. Employing a linearized
cavitating cascade theory, flow solutions are obtained in order to calculate the cavity area, function of the
local shape of the blade and therefore of the radius coordinate, of the local cavitation number and of the
operative conditions. This area, can be seen as a local cavity volume on unity of the radial coordinate and
it’s exploited to calculate for example the local Compliance and Mass Flow Gain using the basic relations
that follows:

𝑈𝑡 𝛥(𝑚̃ 1∗ − 𝑚̃ 2∗ ) 𝛺𝑈𝑡 𝜕𝑉
𝑍21 = ( ∗ ) =𝑗 ( )
2𝐴𝑖 𝛥𝑝̃1 𝜑=𝑐𝑜𝑛𝑠𝑡
2𝐴𝑖 𝜕𝑝1 𝜑=𝑐𝑜𝑛𝑠𝑡

̃ 1∗ − 𝑚
𝛥(𝑚 ̃ 2∗ ) 𝛺 𝜕𝑉
𝑍22 =( ∗ ) =𝑗 ( )
𝛥𝑚 ̃1 𝑝̃∗ =0
𝐴𝑖 𝜕𝑈𝑡 𝑝̃∗ =0
1 1

The global parameters are then calculated by integration through the radial coordinate from the hub to the
tip of the inducer and then the matrix components are obtained.

Sheung-Lip Ng, 1976, performed the dynamic characterization of the pump matrix, exploiting an
experimental approach.
Using the same notations and hypothesis of the previous paper, he exploited a closed-loop test facility
where the main components were a tank, an inducer and two fluctuating valves placed upstream and
downstream with respect to the pump as in Figure 2.1:

Figure 2.1 Schematic diagram of the mathematical pump loop (Ng, 1976)

23
The set-up allowed direct measurements of the pressure perturbations through transducers and mass
perturbation through a Laser Doppler Velocimeter system upstream and downstream of the pump.
The steady state mass flow rate was controlled by a silent valve, while the mass oscillations were
provided by two identical siren valves represented in Error! Reference source not found.:

Figure 2.2 Functional schematic of the fluctuator valve (Ng, 1976)

The siren valve design consists of two slotted concentric cylinders with the inside one made of bronze
rotating within the outside stationary one, which was made of stainless steel, to avoid bonding with the
bronze cylinder. A piece of sintered bronze cylinder was situated coaxially and next to the slotted
cylinders providing a bypass to the flow when the slots were momentarily closed completely. The varying
amplitude of fluctuation, proportional to the relative amount of area covered, was provided by a
cylindrical sleeve which slid axially to cover the slotted cylinders and the sintered one.
Since the matrix parameters are four and the perturbation equations for the mass and pressure are two, the
relative amplitude and phases of the fluctuating valves where changed in order to obtain at least two sets
of linearly independent test situation and therefore equations. The tests were conducted under different
operative conditions (steady mass flow rate, cavitation number and frequency of perturbation). Then the
data, through a square fit method, allowed the extrapolation of the pump matrix parameters.
The tests were conducted for several combinations of relative amplitude and phases of the valves,
obtaining several sets of independent condition. Though theoretically the extrapolated matrix components
should have been coincident, the unavoidable errors led to difference from one set to the other, but what is
to be highlighted is at least the consistence of the trend showed by the matrix component plotted versus
the perturbation frequency as reported by the author.

S.L. Ng and C. Brennen, 1978, presented further results along the path of Ng’s PhD work. In fact, dealing
with the same test set-up, three different set of matrix parameters plotted against the perturbation
frequency are obtained in order to take into account respectively of absent, extensive and moderate
cavitating conditions, as represented in Figure 2.3 and Figure 2.4:

24
Figure 2.3 Left: the [ZP] transfer function for Impeller IV in the virtual absence of cavitation. The real and imaginary
parts of the elements (solid and dashed lines, respectively) are plotted against both the actual and the nondimensional
frequencies; Right: the [ZP] transfer function for Impeller IV under conditions of extensive cavitation (Ng and Brennen,
1978)

Figure 2.4 The [ZP] transfer function for Impeller IV under conditions of moderate cavitation (Ng and Brennen, 1978)

25
During the Joint Symposium on Design and Operation of Fluid Machinery in 1978, C. Brennen presented
the same experimental results obtained with Ng, plotted on the same graph for different cavitation
numbers and relating them to the results of an analytical model predicting the pump matrix components,
as showed in Figure 2.5:

Figure 2.5 Polynomial curve fitting to experimental pump transfer matrices, [ZP], obtained for Impeller IV at 𝝋=0.070
and a rotational speed of 9000 rpm. The real and imaginary parts of the matrix elements are presented as functions of
frequency by solid and dash lines respectively. The letters A to E denote matrices taken at fives, progressively diminishing
cavitation numbers, 𝝈, as follows: (A) 0.508 (B) 0.114 (C) 0.046 (D) 0.040 (E) 0.023 (Ng and Brennen, 1978)

According to the model, the flow through the impeller is divided into four parts and dynamic relations for
each part are used to synthesize the dynamics of the pump: (i) the relations between the upstream inlet
fluctuations and those at entrance to a blade passage; (ii) the bubbly flow region within a blade passage;
(iii) the single phase liquid flow in the remainder of the blade passage following collapse of the cavities
and (iv) the relations between the fluctuations at the end of a blade passage and the downstream
conditions. Despite these complications, the overall transfer function for the pump resulted predominantly
determined by the response of the bubbly region.
The model, by C. Brennen, 1977, an article collected in the Journal of Fluid Mechanics, seems to catch
the trends of the dynamic matrix components as showed in Figure 2.6:

26
Figure 2.6 Theoretical pump transfer matrices, [ZP], obtained for Impeller IV at 𝝋=0.070 as functions of reduced
frequency 𝟂. The lettered curves are for different fractional lengths, 𝜺, of the bubbly region and correspond to decreasing
cavitation numbers, 𝝈: (A) 𝜺=0.2 (B) 𝜺=0.4 (C) 𝜺=0.6 (D) 𝜺=0.8. The curves are for one specific choice of the parameters K
and M (See Brennen 1978) (Ng and Brennen, 1978)

C. Brennen, C.Meissner, E.Y. Lo and G.S. Hoffman, 1982, together with the previous experimental
results, presented a similar test campaign, represented in Figure 2.7, using a bigger impeller (10.2 cm
versus the 7.6 cm impeller in the 1978 experiments) and measuring mass flow perturbations with
electromagnetic meters (EM), judged, in comparison with the LDVs, more accurate:

Figure 2.7 Polynomial curve fits to the 10.2 cm impeller transfer matrices at 𝝋=0.070, a rotational speed of 6000 rpm and
various cavitation numbers as follows: (A) 0.37 (C) 0.10 (D) 0.069 (G) 0.052 (H) 0.044. The real and imaginary parts of the
matrix elements are presented as functions of frequency by solid and dashed lines respectively. The quasistatic resistance
from the slope is indicated by the arrow (Brennen et al., 1982)

27
The trend, at decreasing cavitation parameter, for the 10.2 cm impeller, is quite similar to the one of the
7.6 cm one only for low values of the reduced frequency.
On the other end, the bubbly flow model applied to the polynomial curve fits method with respect to the
expressions built for the 10.2 cm impeller test results, shows that the curves of the matrix components,
except for the real part of the compressibility component, are in accordance with the ones of the
experimental results as can be noticed in Figure 2.8:

Figure 2.8 Transfer functions calculated from the complete bubbly flow model with 𝝋=0.07, 𝝲=9 deg, 𝝉=0.45, F=1.0, K=1.3
and M=0.8. Various cavitation numbers according to 𝜺=0.2/𝝈 are shown (Brennen et al., 1982)

A. Stirnemann, J. Eberl, U. Bolleter and S. Pace, 1987, described a semi-experimental approach to the
characterization problem under non-cavitating and slightly cavitating conditions.
The main aspect of their work was to measure directly only the pressure oscillations, by means of four
couple of quartz transducers (stations n°1,2,5 and 6), placed symmetrically with respect to the pump as
showed in Figure 2.9:

Figure 2.9 Schematic representation of the basic system and its nomenclature (A.Stirnemann et al., 1987)

28
Stations n°2 and n°5 are installed at a certain distance from the inlet and discharge sections of the pump
because of noise reasons.
Then the four pipes are analytically characterized through their geometry and acoustic characteristics: the
ones from station n°1 to 2, the one from n°1 to 3 and the symmetric w.r.t. the pump. The characterization
is made via dynamic matrix as usual.
Dealing with the pump subsystem, three special ratios are computed: the pressure ratios and the ratio of
the two mass flow rate (inlet and discharge) with one of the pressure (the discharge one in this case):
𝑝3 𝑞3 𝑞4
𝐻34 = ; 𝑦34 = ; 𝑦4 =
𝑝4 𝑝4 𝑝4

The next step is to express these three components as a function of three pressure ratios experimentally
measured (stations n°1 and 2, 2 and 5, 5 and 6), together with the components of the four matrixes
expressing the behavior of the four pipes previously stated:

𝐻12 (𝜂11 𝛾12 − 𝜂12 𝛾11 ) + 𝜂12 𝐻25 𝛿12

𝐻34 = ∗
𝐻65 (𝑘11 𝛿12 − 𝑘12 𝛿11 ) + 𝑘12 𝛾12

𝐻12 (𝜂11 𝛾11 − 𝜂21 𝛾12 ) − 𝜂11 𝐻25 𝛿12

𝑦34 = ∗
𝐻65 (𝑘11 𝛿12 − 𝑘12 𝛿11 ) + 𝑘12 𝛾12

𝐻65 (𝑘21 𝛿12 − 𝑘11 𝛿11 ) + 𝑘11

𝑦4 =
𝐻65 (𝑘11 𝛿12 − 𝑘12 𝛿11 ) + 𝑘12

Being:

𝑝1 𝑝2 𝑝6
𝐻12 = ; 𝐻25 = ; 𝐻65 =
𝑝2 𝑝5 𝑝5

These expressions are then incorporated in a system of linear equations to finally calculate the four
components of the dynamic matrix of the pump as showed:

(𝐻34 )1 1 (𝑦4 )1
∗ ∗ ∗ 𝛼11
∗ = ∗ ∗ [𝛼 ]
∗ ∗ ∗ 12
[(𝐻34 )𝑁 ] [ 1 (𝑦4 )𝑁 ]

(𝑦34 )1 1 (𝑦4 )1
∗ ∗ ∗ 𝛼21
∗ = ∗ ∗ [𝛼 ]
∗ ∗ ∗ 22
[(𝑦34 )𝑁 ] [ 1 (𝑦4 )𝑁 ]

N is the number of independent conditions experimented. N=2 is the necessary and sufficient value to
obtain the four matrix components, but to have a perspective of the uncertainty of the method and
therefore an estimation of the errors, it was increased.
To solve the linear system for each frequency value, a method extended to the complex numbers from R.I.
Jennrich, 1977, was exploited.

29
The independent condition of oscillating flow and pressure was achieved changing the impedance by
means of a series of five accumulators placed either on the suction or on the discharge line, while the
oscillation was provided by an Electro-Dynamic Exciter placed at a different line w.r.t. to the series of
accumulators as schematically represented in Figure 2.10:

Figure 2.10 Schematic representation of the experimental facility (A.Stirnemann et al., 1987)

The following Figure 2.11 shows the results for the Mass Gain component of the pump matrix plotted
against the perturbation frequency, under slightly cavitating condition (0.2). The dotted lines represent an
estimation of the standard deviance:

Figure 2.11 Mass Gain obtained from excitation on the suction side (A.Stirnemann et al., 1987)

30
Y. Kawata et al., 1988, presented an experimental way to obtain the matrix components.
A direct measurement of the mass flow rate and pressure oscillation were performed at two measuring
stations, 6.5 m before the pump inlet and 8.2 m after the pump discharge section, as represented in Figure
2.12:

Figure 2.12 Test Loop for measuring the dynamic behaviour of the prototype multi-stage pump (Kawata et al., 1988)

If two independent conditions lead to one result for the matrix component, three independent
experimental loops get to three comparable results (theoretically coincident), to catch an estimation of the
robustness of the method; the 70 m long pipe and the accumulator are activated through three valves to
change the characteristics of the hydraulic loop leading to different experimental conditions.
The last step of the test campaign is represented to the analytical adjustment of the values of the matrix
components represented in Figure 2.13, since the measuring station are distant enough from the sections
of interest to deal with a likely modifying effect:

Figure 2.13 Corrected transfer matrix of the prototype multi-stage pump (Kawata et al., 1988)

31
S. Rubin, 2004, dealt with the mathematical modeling and interpretation of the frequency response of a
pump at a fixed flow coefficient and different cavitating conditions. The pump characteristics data are
extrapolated from tests conducted on a scaled model of the Space Shuttle Main Engine low pressure
oxidizer turbopump using room temperature water.
Pressure transducers and electromagnetic flowmeters were exploited at two measuring stations upstream
and downstream of the pump location to acquire data, while a siren-valve and a throttle valve exciter were
respectively placed upstream and downstream to provide the mass flow perturbation; they were activated
individually to achieve the two sets of data needed. Since the distance between the measuring stations and
the inlet and discharge sections of the pump wasn’t negligible, the actual pump transfer matrix was
obtained after the analytical step accounting for the impedance of the pipelines connecting the
downstream measuring station to the inlet pump section and the pump discharge section to the upstream
measuring station.
The new interpretation of the data, emerges from the intuitively assumption that the cavitation bubbles,
having an inertia, respond to inlet mass flow and pressure perturbations with a delay. This phenomenon is
taken into account assuming a complex value of compliance and mass flow gain, through a phase lag.
So, being the cavitation volume:

𝜕𝑉𝑠 𝜕𝑉𝑠
𝑉= 𝑝1 + 𝑚̇̇
𝜕𝑝1 𝜕𝑚̇1 1

The non-dimensional continuity equation is expressed as:

𝑚1 − 𝑚2 = −𝜌𝑉 = 𝐶̅ 𝑝1 + 𝑀
̅ 𝑚̇1

Using the Laplace variable 𝑠 = 𝑗𝜔, the complex compliance and mass flow gain are given by:

𝜕𝑉𝑠 𝐶 𝜕𝑉𝑠 𝑀
−𝜌 = ; =
𝜕𝑝1 1 + 𝜏𝑐 𝑠 𝜕𝑚̇1 1 + 𝜏𝑀 𝑠

With the non-dimensional transfer equations of the pump:

𝑝2 = 𝑇11 𝑝1 + 𝑇12 𝑚̇1

𝑚̇2 = 𝑇21 𝑝1 + 𝑇22 𝑚̇1

The transfer matrix non-dimensional components can be expressed as:

−𝑗𝜔𝐶 −𝑗𝜔𝑀
𝑇21 (𝜔) = ; 𝑇22 (𝜔) = 1 −
1 + 𝑗𝜔𝜏𝑐 1 + 𝑗𝜔𝜏𝑀

Having plotted the transfer matrix components against the frequencies of perturbation of 4, 7, 14, 21, 28,
35 and 42 Hz, the frequency dependent compliance and mass flow gain values and respective time delays,
where extracted as:

1 𝑅𝑒[𝑇21 (𝜔)]
𝜏𝑐 (𝜔) =
𝜔 𝐼𝑚[𝑇21 (𝜔)]

1 (𝑅𝑒[𝑇22 (𝜔)] − 1)
𝜏𝑀 (𝜔) = −
𝜔 𝐼𝑚[𝑇22 (𝜔)]

32
 1
𝐶(𝜔) = − 𝐼𝑚[𝑇21 (𝜔)](1 + 𝜔2 𝜏𝑐2 )
𝜔
1 2
𝑀(𝜔) = − 𝐼𝑚[𝑇22 (𝜔)](1 + 𝜔2 𝜏𝑀 )
𝜔

The non-dimensional pressure equation is expressed by:

𝑝2 = 𝐺̅ 𝑝1 − (𝑅 + 𝑗𝜔𝐿)𝑚̇2

It’s worth to notice that the pump impedance is applied to the discharge mass flow perturbation. This
choice is dictated by two reason: 1)The impedance experimental value is weak dependent form the
cavitation number, in particular the resistance can be assumed constant while the inertance face only one
change of value at increasing cavitation conditions; 2)The pump gain results to be real.
The pump gain and impedance can be extracted from the transfer matrix results as:

𝑇12
𝑍 = 𝑅 + 𝑗𝜔𝐿 =
𝑇22

𝑇12 𝑇21
𝐺̅ = 𝑇11 +
𝑇22

S. Rubin obtained, except for the mass flow gain which data were inconsistent, simple functional forms of
the real component of the pump parameters that fit the data, as showed in the following Figure 2.14,
Figure 2.15 and Figure 2.16:

Figure 2.14 Left: Compliance Lag data and fits; Right: Compliance data and fits (Rubin S., 2004)

𝐶0
𝐶(𝜔) =
1 + 9𝐶0 𝜔 2
𝜏𝑐0
𝜏𝑐 (𝜔) =
1 + 5𝜏𝑐0 𝜔 2

33
 Figure 2.15 Left: Resistance data and fits; Right: Inertance data and fits (Rubin S., 2004)

𝑅0 (1 + 2.3𝛼𝑅 𝜔2 )
𝑅=
1 + 𝛼𝑅 𝜔 2

𝐿0 (1 + 0.5𝛼𝐿 𝜔2 )
𝐿=
1 + 𝛼𝐿 𝜔 2

Figure 2.16 Pump Gain data and fits (Rubin S., 2004)

𝐺 = 𝐺0 (1 + 𝛼𝐺 𝜔2 )

The reference value, subscripted with o, at the different cavitation numbers used for the test campaign, are
shown in Table 1:

34
 Table 1 Values of fit parameters and references to equations and figures. Dual values for flow gain indicate uncertainty
(Rubin S., 2004)

To maintain a second order form in the Laplace variable s, taking into account the previous frequency
relations expressing the pump parameters, the continuity and pressure equations can be expressed as a
system of equations through the exploitation of auxiliary variables which have no true physical meaning;
The continuity equation is expanded as:

𝑚𝑐 + 𝜏𝑐0 𝑠𝑚𝑐𝑐 + 𝐶0 𝑝1𝑐 = 0

𝑚𝑐 − (1 − 9𝜏𝑐0 𝑠 2 )𝑚𝑐𝑐 = 0

𝑝1 − (1 − 5𝐶0 𝑠 2 )𝑝1𝑐 = 0

(1 + 𝜏𝑀 𝑠)𝑚𝑀 + 𝑀𝑠𝑚1 = 0

𝑚𝑐 − 𝑚2 + 𝑚𝑐 + 𝑚𝑀 = 0

While the pressure equation becomes:

𝑝2 = 𝐺0 (1 − 𝛼𝐺 𝑠 2 )𝑝1 − 𝑅0 (1 − 35𝑠 2 )𝑚2𝑅 − 𝐿0 (1 − 36𝑠 2 )𝑚2𝐿

(1 − 15𝑠 2 )𝑚2𝑅 − 𝑠𝑚2 = 0

(1 − 72𝑠 2 )𝑚2𝐿 − 𝑠 2 𝑚2 = 0

This formulation is heavier, but can be useful for a stability analysis through eigenvalues method.

A. Cervone et al., 2009, presented an interesting way to compute the calculation of the matrix
components.
The main concept underneath this study is to analytically characterized the facility under the assumption
that the compressibility has to be addressed only to the pump (under cavitating condition) and the tank
that behaves as an ideal capacitor, while the pipelines are treated as impedance as usual; each of the four
components of the pump matrix are expressed as function of the operative condition (steady flow rate and

35
cavitation number) and geometry of the inducer. In particular, the cavitating volume is expressed by
means of a polynomial relationship through the difference between the actual cavitation number and the
so called “choked cavitation number” (breakdown value), suggested by the data of a 10.2 cm impeller
from C. Brennen, 1994.
The test facility of the Osaka laboratory showed in Figure 2.17 clarifies the schematization underneath the
model concept:

Figure 2.17 Schematic of the experimental facility (Cervone et al., 2009)

The exciter highlighted above provides a known flow rate perturbation to the loop. Pressures and flow
rates at station n°1 and 2 are then calculated.
The whole procedure is repeated on an independent system. As has been already noticed in some of the
previous papers, the linearly independent condition is achieved varying the impedance of a pipeline.
At this point two independent sets of data of pressure and flow rate at station n°1 and 2 are available and
the four components of the pump matrix can now be obtained with what the authors logically call
“backward calculation” as showed:

𝑝̂1𝑎 𝑄̂1𝑎 0 0 𝐻𝑀11 𝑝̂2𝑎

𝑝̂1𝑏 𝑄̂1𝑏 0 0 𝐻𝑀12 𝑝̂2𝑏
[𝑇][𝐻𝑀] = [ ]= ̂
0 0 𝑝̂1𝑎 ̂
𝑄1𝑎 𝐻𝑀21 𝑄 2𝑎
[ 0 0 𝑝̂1𝑏 ̂
𝑄1𝑏 ] 𝐻𝑀22 [ ̂
𝑄 2𝑏 ]

Obviously the HM matrix encloses, together with the pump components, the contribution of the two
pipelines that links the two measuring station to the pump.
If the two sets of data were linearly dependent, the determinant of the T matrix would be zero; that’s why
the authors suggest that an intuitively way to check the degree of independence of the two conditions
selected is to calculate the value of the determinant of the data matrix, judging the correctness of the
selection through the magnitude of its value.
The author state that the best way to achieve the two linearly independent condition is to change the
impedance of the line which is not subjected to direct perturbation, therefore the suction line in this case.

A. Cervone et al., 2010, applied the same main concept to the 2009 paper to the Alta Space facility, in
Ospedaletto, Pisa.
The assumptions of the model are the same of the previous works (1-D flux, small perturbations, quasi-
steady response of the system components, incompressible flow in the pipelines); the analytical

36
characterization of the facility is performed by means of non-dimensional variables. The main difference
with the test campaign carried out in Osaka is the more precise characterization of the tank, by means of
continuity and momentum equations, used as a generator of perturbation (on the suction and discharge
line contemporary), by means of a controlled mechanical vertical oscillation.
Eight linear equations are obtained in order to compute the mass and flow oscillations at each strategic
section (begin and end section of the suction and discharge pipelines) as showed in Figure 2.18:

Figure 2.18 Top View of the Cavitating Pump Rotordynamic Test Facility (Cervone et al., 2010)

The procedure is repeated to obtain the same results for a different operative condition and again an
analytical model of this kind shows its power into the ability to predict which way is the best to proceed
with a setup modification in order to obtain the second set of data necessary for an experimental
characterization of the pump transfer matrix.
Modification of the impedance of the suction line is confirmed to be the most effective way; Figure 2.19
shows how the hydraulic circuit is modified for the achievement of the independent operative condition:

Figure 2.19 Suggested modification to the original facility setup for obtaining the second linearly independent test
configuration (added pipe lines are coloured in red) (Cervone et al., 2010)

37
Then, dealing with the actual measurement of the oscillating variables, it is showed that the mass and
pressure perturbation measurement can be approached in an indirect way, through the direct measurement
of the pressure in two section of each pipeline, obtaining the mass flow oscillation analytically through
the knowledge of the impedance of the pipelines.
The necessity to by-pass the direct measurement of the mass is due to the quite inaccurateness of the
frequency response of the electromagnetic flowmeters.
The results shows that it works for frequency above a certain value for a small oscillation of the tank,
while an increase of the amplitude of its mechanical oscillation is needed to overcame the low frequency
constraint.

G. Pace, L. Torre, A. Pasini, D. Valentini and L. d’Agostino, 2013, presented the results of the
experimental procedure described in the previous paper.
The tests were carried out on a high-head three-bladed inducer in the Cavitating Pump Rotordynamic Test
Facility (CPRTF) at Alta, Pisa.
The two configurations exploited are showed in the following Figure 2.20:

Figure 2.20 The “short” (top) and “long” (bottom) configurations of the test loop used for the experimental
characterization of the dynamic transfer matrix of cavitating inducers (Pace G. et al., 2013)

Figure 2.21 shows a comparison between the experimental and analytical results, plotting the non-
dimensional inducer matrix components against the frequency of perturbation. The bold lines represent
the results of the model:

38
Figure 2.21 Dynamic matrix for DAPAMITOR3 inducer: experimental points are in squares and the points obtained by
using the model are in star (Pace G. et al., 2013)

It’s evident that the results are quite different; the author suggest that the experimental setup lacks in
accuracy because of the ill-conditioned nature of the measurements and due to an uncertainty in the
dynamic modeling of the pipeline interposed between the inducer and the downstream measurement
section, not considered and likely home of high-compliant gas bubbles trapped able to modify the
compliance of the flow.

K. Yamamoto, A. Müller, T. Ashida, K. Yonezawa, F. Avellan and Yoshinobu, 2015, applied the same
experimental concept to the characterization of a resistance (orifice) and a compliance (accumulator).
The measurements, in fact, are carried out with four pressure transducers placed downstream and
upstream w.r.t. the component to characterize. The independent conditions are achieved exploiting two
mass flow exciters, upstream (piston) and downstream (rotary valve) of the component of interest.
The results are in good agreement with the analytical calculation, validating in a certain way the concept
of by-passing the mass flow direct measurement.

39
3 Mathematical Model of the System
3.1 Introduction
To characterize the dynamic behavior of the pump, a good modeling of all the subsystems interacting
with it, is essential. In our case we want to develop a reduced-order model for the characterization of the
pressure and flow rate oscillations in a given experimental facility. The model is based on the following
initial assumptions:

 Unsteady, one-dimensional flow.

 Small perturbations of the steady state flow (linearized equations under unsteady condition).
 The response of all the components of the system is assumed quasi-steady

Even if it has been shown that the last assumption is not valid in real pumps under cavitating condition
(Rubin, 2004; Tsujimoto et al., 1996, 2008), it could represent a good starting point for the simplified
reduced order analysis presented in this thesis. Under the made assumptions, pressure and mass flow rate
can be written in complex form as follows:

p  t   p  p  t   p  Re  pˆ  e  it 

m  t   m  m  t   m  Re m 
ˆ  eit 
where p and m are the pressure and mass flow rate steady values, p and m are the pressure and flow
rate oscillating components,  is the frequency of the oscillations.
A system of equations characterizing the steady and unsteady (oscillating) condition of the working fluid
is obtained for each component of the hydraulic loop, exploiting the continuity and momentum equations.
The definition of the steady condition is essential to the definition of the unsteady regime condition. The
following subsystems can be identified:

 Incompressible Duct Straight (ID-S)

 Incompressible Duct Elbow (ID-E)
 Incompressible Duct Tapered (ID-T)
 Compressible Duct Straight (CD-S)
 Silent Throttle Valve (STV)
 Volume Oscillator Valve (VOV)
 Tank (T)
 Pump (P)

The exciter will be represented by three different devices:

 Silent Throttle Valve (pressure drop oscillation)

 Volume Oscillator Valve (volume oscillation)
 Tank (vertical oscillation of the tank).

40
 Figure 3.1 Schematic view of the hydraulic loop system (Torre et al., 2011)

Figure 3.1 shows an example of hydraulic loop that can be represented by the mathematical model
developed in the following sections.

3.2 Development of the Dynamic Equations for the Subsystems

As a general approach, the steady and unsteady pressure and mass flow rate downstream (subscript d) of
each component are expressed as functions of their independent variables (respectively q n and q n )
q q q q
through suitable coefficients (such as Pname n
and Pname for the pressure or M name
n
and M name
n
for the mass
n

flow rate) as reported in the following expressions:

 pdname   Pname
qn
qn  pdname   Pname
qn
qn
 n  n
 name  name
md   M name qn md   M name qn
qn qn

 n  n

In particular, the model includes also the effects of body forces (such as the gravitational force expressed
through the gravitational acceleration g) and inertia related to the rigid body acceleration of the
component r . In the one-dimensional approximation, the work done by these forces can be obtained by
integrating the force along the streamline that connect the inlet and the outlet of each components. The
following expressions report the corresponding coefficient related to gravitational and inertial forces in
both steady and unsteady conditions (both g and r are supposed positive while g and r are oscillating
around zero):
 P g   xd e  dx  P g   xd e  dx
 name xu g  name xu g
 xd  xd
 Pname
r
    er  dx  Pname
r
    er  dx
 xu  xu

41
3.2.1 Incompressible Duct: Straight (ID-S), Elbow (ID-E) and Tapered (ID-T)
A duct can be considered incompressible if its length is so small that the compressibility effects can be
neglected. Further consideration in this regard will be provided in the next sections. The application of the
continuity and momentum equations for this component in steady state regime yields to the following
coefficients for the expression of the steady state pressure and mass flow rate downstream of the
incompressible duct element:

 PIDpu  1

 mdu kloss mdu 1 A2 xu A2 xd  mdu
 PID   2  A2 x  2  A2 xu
 u


 M ID  1
mu

where mdu  m 2 m is the averaged mass flow rate of the upstream and downstream sections. In unsteady
d u

condition, the oscillations of the downstream pressure and mass flow rate can be obtained through the
sum of the following contributions:

 PIDpu  1

 P mdu   kloss 2mdu  1 A xu A2 xd  mdu
2 2

 ID  A xu  A xu
 xd
 PID    Adx
mdu

 xu  x

 M mu  1
 ID

Whenever the assumptions of incompressible duct are satisfied, this model can be applied either to
straight duct (ID-S) or Elbow (ID-E) or tapered duct (ID-T) or even ducts with more complex geometry.
To fully characterize a 1-D incompressible duct the needed parameters are:

 ; m du ; A xu ; A xd ; kloss ; 
xd
dx
xu A x
; PIDg ; PIDr ; PIDg ; PIDr 
3.2.2 Compressible Duct Straight (CD-S)
In steady state condition, the compressible duct behaves as an incompressible duct:

 PCD
pu
S  1

 mdu kloss mdu
 PCD  S   2  A2 x
 u

 mu
 CD  S
M  1

While the unsteady behaviour of a compressible duct substantially defers from the incompressible case
according to the following coefficients:

42
 P pu    K1eik1L  K2 eik2L 
 CD  S   K2  K1   K2  K1  
 mu
 PCD  S    K2  K1    K2  K1    A xu
K1 K 2 eik1L K1 K 2 eik2 L 1


 P a    K1 K2 eik1L  K1K2 eik2L 
 CD  S   K2  K1   K2  K1  

 M CD
pu  eik1L eik2 L 
 S   A xu    K 2  K1    K 2  K1  

 M mu   K2 eik1L  K1eik2L 
 CD  S   K2  K1   K2  K1  
 a
 M CD  S   A xu    K22  K1    K12  K1   1
K eik1L K eik2 L

  

where the constants K1 , K 2 , k1 and k 2 are defined as follows:

K 
 1  0 a02 k1
  w0 k1  k 2



w 
w0  2 i  f 0 
DH 
 2
   i f
ik  
w0 

DH 
 0  k1 , k2

  a0  w0  a
0  w0 
2 2 2 2
0 a02 k2
K2 
   w0 k2

where 0 , w0 , a0 , L are, respectively, the unperturbed density of the fluid, the unperturbed axial velocity,
the speed of sound and the length of the duct. The independent variable a can be obtained from the
unsteady gravitational and inertial forces g and ( r ) according to the following equation:

da
dt
w

 f 0 a  g  r  ex
DH


To fully characterize a 1-D straight compressible duct the needed parameters are:

 ; m du
a
; DH ; L; f ; a0 ; PCD a
 S ; M CD  S 
since:

 DH2
A xd  A xu  4
;   0 ; w0  mdu
 A xu

3.2.3 Silent Throttle Valve (STV) – Exciter

The Silent Throttle Valve is a device used in the water loop to regulate the load of the pump without
introducing cavitation inside the rig. In steady state condition, the relevant parameters for the continuity
and momentum equations are:

 PSTV
pu
1

 mdu kloss mdu 1 A2 xu A2 xd  mdu
 PSTV   2  A2 x  2  A2 xu
 u


 M STV  1
mu

The unsteady behaviour of the Silent Throttle Valve can be obtained according to the following
coefficients:

43
 PSTV
pu
1

 P mdu   kloss 2mdu  1 A xu A2 xd  mdu
2 2

 STV  A xu  A xu

 mdu xd
 PSTV   x A x
dx

 u

 P kloss   mdu m2 du
 STV 2  A xu

 mu
 M STV  1

where kloss is the pressure loss coefficient in steady state regime and kloss is the unsteady pressure loss
coefficient used to excite the flow oscillations inside the rig during the forced experiments. To fully
characterize a Silent Throttle Valve the needed parameters are:

 ; m du ; A xu ; A xd ; kloss ; kloss ; 
xd
dx
xu A x
g
; PSTV r
; PSTV g
; PSTV r
; PSTV 
3.2.4 Volume Oscillator Valve (VOV) – Exciter
The Volume Oscillator Valve is a variable volume device capable of changing its cross-section area in
such a way to introduce volume flow rate oscillation. In steady state condition, the relevant parameters of
the Volume Oscillator Valve for the continuity and momentum equations are:

 PVOV
pu
1

 mdu kloss mdu 1 A2 xu A2 xd  mdu
 PVOV   2  A2 x  2  A2 xu
 u


 M VOV  1
mu

The unsteady behaviour of the Volume Oscillator Valve can be obtained according to the following
coefficients:

 PVOV
pu
1

 PVOV  2u
mu m
 A xu

 md md
 PVOV    A2 xd
 kloss mdu
 PVOVdu    A2 x
m

 u

 P mdu   xd dx
 VOV xu Ax
 mu
 M VOV  1
 M mOV  1
 VOV

where mOV   Vt is the mass flow rate oscillation produced by this device to excite the flow oscillations
OV

inside the rig during the forced experiments. To fully characterize a Volume Oscillator Valve the needed
parameters are:

 ; m du ; A xu ; A xd ; kloss ; 
xd
dx
xu A x
g
; PVOV r
; PVOV g
; PVOV r
; PVOV 
44
3.2.5 Tank (T) – Exciter
Following the direction of the flow in the water loop, the tank presents an upstream inlet section and a
downstream outlet section. The tank has either an air-bag on the top of the tank or an equivalent free-
surface air volume in which the pressure of the air ( pG ) can be controlled. In presence of the gravitational
force, the pressure at the inlet and outlet sections differs from the pressure of the gas also for the
hydrostatic head. In steady state condition, the general expression for the downstream pressure and mass
flow rate can be obtained from the following expressions:

 PTANK
pu
1

 mdu kloss mdu 1 A2 xu A2 xd  mdu
 PTANK   2  A2 x  2  A2 xu
 u


 M TANK  1
mu

The unsteady behaviour of the tank that combines both the properties of a pressure accumulator and a
gravity tank is described through the following coefficients:

 PTANK
pu
1

 PTANK  2u
mu m
 A xu

 md
 PTANK    A2 xd
md


 PTANK   loss
mdu k mdu
 A2 xu

 mdu xd
 PTANK   x A x
dx

 u

 M TANK
pu
   CTeq

 M TANK  1
mu


 M TANK   TeqA2 u
mu C m

 u

 M TANK   CTeq hrin  FS

r 2




Under the assumption of imposing the external forced oscillation as a vertical vibration of the tank, the
resulting compliance of the tank becomes:

CTeq 
AT CT

 
 AT   CT g  r  2 hinuFS 
x t 

where AT is the cross-section area of the tank and the compliance of the air-bag is

CT   pGG   pGG
V V

obtained by supposing an isentropic transformation in the air-bag. The pressure oscillation at the upstream
section of the tank can be related to the unsteady acceleration of the tank through the following equations
expressed in the time domain

45
 pu  CCT pG  m2u mu   hrin FS r
 Teq  A xu
 dp m m 
 G  u C d
 dt T

Or in the frequency domain:

 1 
mˆ d  1 i CTeq 2 mu  mˆ u
 
  hrin  FS rˆ
A
pˆ u   
 xu
i CTeq

To fully characterize a vertical oscillating Tank the needed parameters are:

 ; m du ; A xu ; A xd ; kloss ; 
xd
dx
xu A x
; CT ; AT ; hrin  FS ; hinu FS
x t 
3.2.6 Pump (P)
The pump is usually characterized in terms of non-dimensional number such as the head coefficient, the
flow coefficient and the cavitation number defined as follows:

  pd 2 p2u
  rT
 2 u
1 m m
d
   rT3
 pd  pv
  1 
 2
2 2
rT

The unperturbed characteristic curve of the pump     ,   is typically obtained from experimental
data. In steady state condition, the momentum and continuity equations yield to the following
coefficients:

 PPUMP
pu
1
 
 PPUMP   rT
2 2

 mu
 M PUMP  1

The dynamic behaviour of the pump is obtained by the unsteady momentum and continuity equation in
terms of the following coefficients:

 PPUMP
pu
1
  
 PPUMP   rT
2 2

  
 PPUMP   rT
2 2

 mdu xd
 PPUMP   xu A '
dx


 M PUMP
mu
1
 m
 M PUMP  1
C

46
where mC   Vt is the mass flow rate oscillation associated to the evolution of the cavity volume around
C

the impeller in cavitating regime. To fully characterize a cavitating pump needed parameters are:

 ; m du ;  ; rT ; 
xd
dx
xu A '
; ;  ;  ; mC ; PPUMP
g r
; PPUMP g
; PPUMP r
; PPUMP 

47
4 System Design Tools
4.1 Introduction
This chapter aims to show some useful tools to achieve fundamental informations about the quality of the
mathematical model exploited to characterize the system and to judge the effectiveness of different way
to generate mass flow rate and pressure perturbations inside the hydraulic loop, by means of four
experimental set-ups.

4.2 Incompressible VS Compressible Solution

As we could have noticed in the previous studies about the characterization of the dynamic matrix of the
pump, and, more in general, for the characterization of an experimental facility, the assumption of
concentrating the compressibility of the working fluid to the pump subsystem is one of the pillars of the
exploited models.
An analytical solution for the perturbation of a compressible fluid moving into a straight duct has been
obtained, leading to the possibility to relate the behaviour of the oscillating pressure and mass flow rate
into two geometrical and physical identical ducts, exploiting the incompressible duct straight solution and
the compressible one.

4.2.1 Downstream Mass Flow Rate and Pressure Signal Comparison

This section shows the results about the oscillation of the pressure and mass flow rate at the end of a
straight duct for the two cases, varying the frequency of perturbation, the acoustic velocity and the duct
length at different phase shift between the upstream pressure and mass flow rate perturbation, whose
oscillation amplitude is set as follows:

𝑘𝑔
 Upstream mass flow rate perturbation amplitude = 1 𝑠
 Upstream pressure perturbation amplitude = 103 𝑃𝑎

The simulation will show the oscillating behaviour of the downstream variables considered.
We expect that a combination of length of the duct, low acoustic velocity and high frequency of
perturbation, will lead to substantial differences between the two solutions. Figure 4.1, Figure 4.2, Figure
4.3 and Figure 4.4 show the comparison of the results of the two mathematical models against the
frequency, with the phase between the upstream mass flow rate and pressure perturbation equal to zero.

Case 1)
 5 𝐻𝑧 < 𝑓 < 140 𝐻𝑧
𝑚
 𝑎 = 1400 𝑠
 𝐿 =1𝑚

48
 Figure 4.1 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.2 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

49
Figure 4.3 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation at
fixed acoustic velocity and duct length

Figure 4.4 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

50
Figure 4.5, Figure 4.6, Figure 4.7 and Figure 4.8 show the comparison of the results of the two
mathematical models against the frequency, with the phase between the upstream mass flow rate and
𝜋
pressure perturbation equal to
2

Figure 4.5 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.6 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

51
Figure 4.7 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation at
fixed acoustic velocity and duct length

Figure 4.8 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

52
Figure 4.9, Figure 4.10, Figure 4.11 and Figure 4.12 show the comparison of the results of the two
mathematical models against the frequency, with the phase between the upstream mass flow rate and
pressure perturbation equal to 𝜋

Figure 4.9 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.10 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

53
Figure 4.11 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.12 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

54
Figure 4.13, Figure 4.14, Figure 4.15 and Figure 4.16 show the comparison of the results of the two
mathematical models against the frequency, with the phase between the upstream mass flow rate and
3𝜋
pressure perturbation equal to
2

Figure 4.13 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.14 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

55
Figure 4.15 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.16 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

56
Case 2)
𝑚 𝑚
 50 𝑠 < 𝑎 < 1400 𝑠
 𝑓 = 5 𝐻𝑧
 𝐿 =1𝑚

Figure 4.17, Figure 4.18, Figure 4.19 and Figure 4.20 show the comparison of the results of the two
mathematical models against the acoustic velocity, with the phase between the upstream mass flow rate
and pressure perturbation equal to zero.

Figure 4.17 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.18 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

57
Figure 4.19 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.20 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

58
Figure 4.21, Figure 4.22, Figure 4.23 and Figure 4.24 show the comparison of the results of the two
mathematical models against the acoustic velocity, with the phase between the upstream mass flow rate
𝜋
and pressure perturbation equal to
2

Figure 4.21 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.22 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

59
Figure 4.23 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.24 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

60
Figure 4.25, Figure 4.26, Figure 4.27 and Figure 4.28 show the comparison of the results of the two
mathematical models against the acoustic velocity, with the phase between the upstream mass flow rate
and pressure perturbation equal to 𝜋

Figure 4.25 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.26 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

61
Figure 4.27 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.28 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

62
Figure 4.29, Figure 4.30, Figure 4.31 and Figure 4.32 show the comparison of the results of the two
mathematical models against the acoustic velocity, with the phase between the upstream mass flow rate
3𝜋
and pressure perturbation equal to
2

Figure 4.29 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.30 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

63
Figure 4.31 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.32 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

64
Case 3)
 1 𝑚 < 𝐿 < 28 𝑚
 𝑓 = 5 𝐻𝑧
𝑚
 𝑎 = 1400 𝑠

Figure 4.33, Figure 4.34, Figure 4.35 and Figure 4.36 show the comparison of the results of the two
mathematical models against the duct length, with the phase between the upstream mass flow rate and
pressure perturbation equal to zero.

Figure 4.33 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.34 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

65
Figure 4.35 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.36 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

66
Figure 4.37, Figure 4.38, Figure 4.39 and Figure 4.40 show the comparison of the results of the two
mathematical models against the duct length, with the phase between the upstream mass flow rate and
𝜋
pressure perturbation equal to
2

Figure 4.37 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.38 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

67
Figure 4.39 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.40 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

68
Figure 4.41, Figure 4.42, Figure 4.43 and Figure 4.44 show the comparison of the results of the two
mathematical models against the duct length, with the phase between the upstream mass flow rate and
pressure perturbation equal to 𝜋

Figure 4.41 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.42 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

69
Figure 4.43 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.44 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

70
Figure 4.45, Figure 4.46, Figure 4.47 and Figure 4.48 show the comparison of the results of the two
mathematical models against the duct length, with the phase between the upstream mass flow rate and
3𝜋
pressure perturbation equal to
2

Figure 4.45 Pressure amplitude comparison between the compressible and the incompressible solution vs frequency of
perturbation at fixed acoustic velocity and duct length

Figure 4.46 Mass flow rate amplitude comparison between the compressible and the incompressible solution vs frequency
of perturbation at fixed acoustic velocity and duct length

71
Figure 4.47 Phase shift between the compressible and the incompressible pressure solution vs frequency of perturbation
at fixed acoustic velocity and duct length

Figure 4.48 Phase shift between the compressible and the incompressible mass flow rate solution vs frequency of
perturbation at fixed acoustic velocity and duct length

72
4.2.2 Conclusions and Results
These results confirm the criteria which state that the use of the incompressible model must be restricted
to duct lengths that do not exceed one-tenth of wavelength over the frequency range of interest,
mathematically:

a
L
10 fu

where fu is the highest frequency of interest in Hz, and a is the effective acoustic speed (Oppenheim and
Rubin, 1993). In fact, the two solutions are in great accordance when the duct length is small w.r.t. the
maximum length provided by the criteria.
The phase shift between the two solutions is negligible, in particular the phase difference of the two mass
flow rate never reaches the degree, while the maximum phase shift for the pressures is higher, but doesn’t
overcome the value of 10 degrees. It can be noticed that for the limit cases i.e. when the length of the duct
is equal to one, the two pressure solutions remain in good agreement with a maximum percentage error of
about 10%, while the two mass flow rate solutions can present an error of 20% in the limit cases,
overcoming 30% when the acoustic velocity drops to its minimum, with respect to the incompressible
upstream/downstream value.
In conclusion, as expected, getting close to the limit condition imposed by the Rubin-Oppenheimer
criteria, the two solutions tends to diverge and, in particular, the two mass flow rate present a more
developed tendency to depart moving towards the duct length criteria boundary under the point of view of
the amplitude of oscillations, while for the phase shift it can be noticed the opposite behaviour which is
not so appreciable as the amplitude phenomenon.

4.2.3 Duct Transfer Matrix Comparison

This section, conceptually, presents the same comparison of the previous section, exploiting the more
classical matrix notation to compare the oscillating behaviour of a 3 meter long straight duct using the ID-
S and the CD-S solutions.
Being the transfer matrix of an incompressible duct (using the complex domain):

 pˆ d  1  I d   pˆ u 
ˆ   ˆ 
 md  0 1   mu 
where Id is the so called “Impedance” (complex) of the duct, next plots show that, at decreasing acoustic
velocity of the line (water used as working fluid), is not correct to exploit the incompressible approach for
an appreciable range of frequency of perturbation (and length of the duct, which is taken constant in this
case), to model the unsteady-state of the oscillating system, by comparing the matrix components of the
above ID-S representation, with a general one which take into account the compressibility effects of the
fluid, exploiting our wave solution.

73
Figure 4.49 Comparison between the real components of the principal diagonal of the matrix of a 3 meters long ID-S
(black line) with a geometrically identical CD-S at different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue); a=700
m/s (cyan); a=500 m/s (yellow). The asterisks represent the data points resulting from the simulation process, the lines are
the corresponding polynomial fitting curves.

Figure 4.50 Comparison between the imaginary components of the principal diagonal of the matrix of a 3 meters long ID-
S (black line) with a geometrically identical CD-S at different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue);
a=700 m/s (cyan); a=500 m/s (yellow). The asterisks represent the data points resulting from the simulation process, the
lines are the corresponding polynomial fitting curves.

Figure 4.49 and Figure 4.50 represent the results of the diagonal of the CD-S matrix, which is composed
from two equal components, for a range of frequency from 0.1 Hz to 100 Hz, as for all the others
simulations of this section. The imaginary contribution of the compressibility is negligible w.r.t. the real
one, which can lead to an appreciable error even for high acoustic velocity and low frequencies.

74
Figure 4.51 and Figure 4.52 show that the Resistance component tends to depart from the ID-S value at
lower frequencies w.r.t. the imaginary component (product between the frequency in rad/s and the
Inertance) and that, for the value of the steady-state mass flow rate selected for these simulations (the
same resulted from the evaluation of the steady-state condition of the hydraulic loops studied in the next
section), for an appreciable range of frequency of perturbation, the imaginary part of the impedance is
predominant.

Figure 4.51 Comparison between the Re[H12] of a 3 meters long ID-S (black line) with a geometrically identical CD-S at
different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue); a=700 m/s (cyan); a=500 m/s (yellow). The asterisks
represent the data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

Figure 4.52 Comparison between the Im[H12] of a 3 meters long ID-S (black line) with a geometrically identical CD-S at
different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue); a=700 m/s (cyan); a=500 m/s (yellow). The asterisks
represent the data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

75
Figure 4.53 Comparison between the Re[H21] of a 3 meters long ID-S (black line) with a geometrically identical CD-S at
different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue); a=700 m/s (cyan); a=500 m/s (yellow). The asterisks
represent the data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

Figure 4.54 Comparison between the Im[H21] of a 3 meters long ID-S (black line) with a geometrically identical CD-S at
different acoustic velocity: a=1400 m/s (red); a=1000 m/s (blue); a=700 m/s (cyan); a=500 m/s (yellow). The asterisks
represent the data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

Figure 4.53 and Figure 4.54 show that the imaginary part of H21 tends to depart from zero at lower
frequencies w.r.t. the real one, being also dominant for an appreciable range of frequency of perturbation.
The algorithm behind the previous results can be exploited to check, knowing the range of frequencies of
our interest, an estimation of the acoustic velocity and the geometrical characteristics of the facility,
which error lies behind the choice to use the incompressible model for each straight duct component and,
in case this wasn’t negligible, to replace the ID-S with a CD-S to obtain more accurate results.

76
4.2.4 Conclusions and Results
These results, again, confirm the Oppenheim and Rubin criteria, in fact, the CD-S matrix components
tends to defer from the ID-S ones for an appreciable range of frequency of perturbation, at decreasing
acoustic velocity.
The results show that the real component of the principal diagonal of the matrix and the impedance of the
CD-S present the higher tendency to depart from the ID-S ones for an appreciable range of frequency of
perturbation and acoustic velocity.
In conclusion, as expected, the incompressible model fails to achieve a good prediction of the oscillating
behaviour of a straight duct for an appreciable range of combination of duct length, frequency of
perturbation and acoustic velocity of interest.

4.3 Hydraulic Loop System Design, Semi-Compressible Approach

This section presents the results of four different experimental set-ups, discriminating the capability to
create a detectable regime oscillating condition, aiming to the experimental characterization of the pump
transfer matrix. The longitudinal acceleration of the components has been neglected to simplify the
qualitative assessment of the set-ups.
For each one of the four set-ups, has been developed a model on Matlab/Simulink, to calculate the steady-
state condition of the hydraulic loop and then to perform the unsteady-state analysis of the system,
measuring the oscillating condition at the inlet section of the pump to have an idea of the magnitudes in
play. After having selected the elements composing the loop and the order in which they are placed with
respect to each other, the hydraulic loop can be built by connecting them through their two input/output.
The STV, the VOV and the tank, exploited as generators of perturbation, can be commanded in
frequency, phase and amplitude to observe, after having properly characterized all the parameters of
interest of all the subsystems of the loop, the time history of the two variables considered on whichever
signal line.
The upper line represents the pressure, while the lower, the mass flow rate signal. It can be noticed that,
while there are as many pressure sensors as the elements, we need less mass flow rate sensors, since they
have to be placed only at the downstream section of the elements taking into account the compressibility
effects. The results have been obtained assuming a non-cavitating pump and replacing the suction line
ID-S model with the CD-S one due to its length, since it approached the maximum length given by the
Oppenheim and Rubin criteria, at increasing frequency of perturbation.
Figure 4.56 and Figure 4.57 show the results for the hydraulic loop which exploits the longitudinal
oscillation of the STV as source of perturbation. As it can be noticed by the equations of the model of the
STV, the mathematical source of oscillation is the oscillating loss coefficient of the valve which has been
varied as a fraction of the steady-state one. The mass flow rate has been divided by the steady-state mass
flow rate, while the pressure by the dynamic pressure at the tip radius of the inducer, to achieve the non-
dimensional condition. The results have been obtained for a range of frequency of perturbation from 0.1
Hz to 30 Hz.
The behaviour of the pressure and the mass flow rate of the STV set-up, with the frequency of
perturbation, is the opposite. We can notice that at low frequencies, the STV is able to create appreciable
mass flow rate oscillation, which decrease quite heavily at increasing frequency, while the system cannot
create appreciable pressure oscillation at very low frequencies and about beyond 5 Hz the pressure
amplitude is weakly affected by the frequency.

77
Case 1)
 STV; the hydraulic loop is represented in Figure 4.55.

Figure 4.55 Hydraulic loop logic by Simulink

Figure 4.56 Comparison between the pressure amplitudes of oscillation at different values of the STV unsteady-state loss
coefficient: K=Kv/100 (red); K=Kv/25 (blue); K=Kv/5 (cyan); K=Kv/2.5 (yellow) where Kv is the steady-state STV loss
coefficient. The asterisks represent the data points resulting from the simulation process, the lines are the corresponding
polynomial fitting curves.

78
Figure 4.57 Comparison between the mass flow rate amplitudes of oscillation at different values of the STV unsteady-state
loss coefficient: K=Kv/100 (red); K=Kv/25 (blue); K=Kv/5 (cyan); K=Kv/2.5 (yellow) where Kv is the steady-state STV
loss coefficient. The asterisks represent the data points resulting from the simulation process, the lines are the
corresponding polynomial fitting curves.

Case 2)
 TVO; the 1st level Simulink logic of the hydraulic loop is represented in Figure 4.55. The tank
subsystem would act as the exciter instead of the STV, as it will be showed in detail into the
corresponding appendixes.

Figure 4.58 and Figure 4.59 show the results for the hydraulic loop which exploits the vertical oscillation
of the tank as source of perturbation. As it can be noticed by the equations of the model of the tank with
vertical oscillation, the mathematical source of oscillation is the vertical acceleration of the tank which
has been related to the vertical displacement. The qualitative behaviour of the mass flow rate and the
pressure oscillations created is similar, while quantitative a huge problem seems to verify, since, not only
at low frequencies this set-up is not capable to create an appreciable mass flow rate oscillation, but when
this start becoming detectable at high frequency, the pressure amplitude grows too fast towards such high
values that the model assumption of small perturbations is violated.

79
Figure 4.58 Comparison between the pressure amplitudes of oscillation at different values of maximum vertical
displacement of the tank: r=0.2 mm (red); r=1 mm (blue); r=5 mm (cyan); r=1 cm (yellow). The asterisks represent the
data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

Figure 4.59 Comparison between the mass flow rate amplitudes of oscillation at different values of maximum vertical
displacement of the tank: r=0.2 mm (red); r=1 mm (blue); r=5 mm (cyan); r=1 cm (yellow). The asterisks represent the
data points resulting from the simulation process, the lines are the corresponding polynomial fitting curves.

80
Case 3)
 VOV placed downstream of the tank (far upstream of the pump) as showed in Figure 4.60.

Figure 4.60 Hydraulic loop logic by Simulink with the VOV (red) placed downstream of the tank (blue)

Figure 4.61 and Figure 4.62 show the results for the hydraulic loop which exploits the longitudinal
oscillation of the VOV, placed right after the tank, as source of perturbation. As it can be noticed by the
equations of the model of the VOV, the mathematical source of oscillation is the VOV perturbation mass
flow rate, related to the oscillating volume occupied by the valve. We can notice that the system fails to
create an appreciable mass flow rate oscillation, which, beyond 5 Hz is weakly dependent by the
frequency, while this set-up, except for a narrow range of very low frequencies, seems to be able to create
an appreciable pressure oscillation, which is highly frequency dependent. The VOV set-up was very slow
to achieve an oscillating regime condition at very low frequencies.

Figure 4.61 Comparison between the pressure amplitudes of oscillation at different values of the VOV mass flow rate
amplitude of oscillation (in percentage of the steady-state mass flow rate): mvov=0.25% (red); mvov=1% (blue); mvov=5%
(cyan); mvov=10% (yellow). The asterisks represent the data points resulting from the simulation process, the lines are the
corresponding polynomial fitting curves.

81
Figure 4.62 Comparison between the mass flow rate amplitudes of oscillation at different values of the VOV mass flow
rate amplitude of oscillation (in percentage of the steady-state mass flow rate): mvov=0.25% (red); mvov=1% (blue);
mvov=5% (cyan); mvov=10% (yellow). The asterisks represent the data points resulting from the simulation process, the
lines are the corresponding polynomial fitting curves.

Case 4)
 VOV placed downstream of the pump, as represented in Figure 4.63.

Figure 4.63 Hydraulic loop logic by Simulink with the VOV (red) placed downstream of the pump (orange)

Figure 4.64 and Figure 4.65 present the results for the hydraulic loop which exploits the longitudinal
oscillation of VOV, placed after the pump, as source of perturbation. We can notice that the system
succeeds to create an appreciable mass flow rate oscillation, which, beyond 5 Hz is very weakly
dependent by the frequency, while, dealing with pressure oscillation, this set-up shows difficulties in
creating an appreciable oscillating regime condition at very low frequencies, presenting a quite relevant
dependence from the frequency of perturbation. The regime condition at very low frequencies, again, is
achieved much more slowly w.r.t. the other solutions.

82
Figure 4.64 Comparison between the pressure amplitudes of oscillation at different values of the VOV mass flow rate
amplitude of oscillation (in percentage of the steady-state mass flow rate): mvov=0.25% (red); mvov=1% (blue);
mvov=5% (cyan); mvov=10% (yellow). The asterisks represent the data points resulting from the simulation process, the
lines are the corresponding polynomial fitting curves.

Figure 4.65 Comparison between the mass flow rate amplitudes of oscillation at different values of the VOV mass flow
rate amplitude of oscillation (in percentage of the steady-state mass flow rate): mvov=0.25% (red); mvov=1% (blue);
mvov=5% (cyan); mvov=10% (yellow). The asterisks represent the data points resulting from the simulation process, the
lines are the corresponding polynomial fitting curves.

83
4.3.1 Conclusions and Results
The previous results seem to point to the VOV set-up as the most suitable to create an appreciable
oscillating regime condition, by placing such a device in the correct position w.r.t. the other components
of the hydraulic loop. Exploiting an electrical analogy, as the current tends to flow more towards the low-
resistance path, the oscillation of mass flow rate tends to be directed to a high-compliance path and, as it
can be noticed from the first of the two VOV set-up, having the tank very close and placed to the opposite
flow direction w.r.t. the pump, can result in “stealing” the oscillation from it. Since our objective is to
create an appreciable oscillation regime that “wraps” the pump, the VOV must be placed in a smart way.

84
5 Conclusions and Future Developments
A modeling framework defined in the time-domain has been developed to characterize the steady state
and dynamic behavior of each component of a typical feeding system for liquid rocket engine. A typical
water loop for experimental characterization of liquid rocket turbopumps (composed by a tank, a suction
line, a pump and a discharge line equipped with a silent throttle valve) has been modeled according to the
modeling framework. In order to characterize the transfer matrix of cavitating pumps, the oscillations of
pressure and mass flow rate upstream and downstream of the pump must be measured. The modeling
framework has been employed to understand the best way to generate the pressure and flow oscillations in
a forced experiment in view of the evaluation of the transfer matrix. In particular, three different main set-
ups of the forcer have been considered:

 Oscillation of the load of the silent throttle valve;

 Vertical oscillation of the tank;
 Oscillation of the volume inside a suitable device.

The best results in terms of capability of generating mass flow rate and pressure oscillations at the inlet of
the inducer have been obtained by means of a Volume Oscillator Valve located downstream of the pump.
Moreover, in view of the experimental measurement of the mass flow rate oscillations upstream and
downstream of the pump the goodness of the approach that uses pressure transducers at the inlet and
outlet of a duct to obtain the indirect measurement of the flow rate oscillation has been assessed by
comparing the dynamic behavior of incompressible ducts and compressible ducts. The main outcome of
this analysis is that the incompressible assumptions and the corresponding modeling framework can
introduce significant errors in the estimation of the mass flow rate oscillations especially at high
frequency. Therefore, a compressible model that relies on a reliable measurement of the speed of sound in
the duct should be employed to enhance the quality of the measurement of the mass flow rate oscillations.
The procedure exploited throughout the simulation processes, can be used to predict the oscillating
behaviour of any hydraulic loop and therefore, every kind of experimental set-up which aims to perform a
forced experiment, under the same assumptions, can be assessed. In the end, the dynamic model created
in the Simulink environment, can be easily exploited to assess the goodness of the experimental set-ups
designed to obtain (at least) two sets of linearly independent data for the experimental characterization of
the pump matrix.

85
6 Appendixes
6.1 Appendix A, Mathematical Model Equations
6.1.1 Introduction
This section aim to show the analytical process that led to the equations characterizing the subsystems in
Chapter 3.

6.1.2 Subsystems Mathematical Model

The pressure and mass flow rate in a given point of the test facility can be written in complex form, as
functions of time, as follows:

p(t )  p  p  e it
Q(t )  Q  Q  e it
p(t )  p  p  e  it
m(t )  m  m  e  it
 (t )      e it
g x (t )  g x  g x  e it
 
p  Re p  e  it

m  Re  m  e   i t

  Re    e   i t

g  Re  g  e 
x x
 i t

where p and Q (usually real) are the pressure and flow rate steady values, p and Q (usually complex)
are the pressure and flow rate oscillating components, is the frequency of the oscillations.
As pointed out before, the relevant components for the dynamic matrix are the oscillating ones; the set of
equations used for the evaluation of the matrices of the components of the facility are presented in the
following sections.
Aiming to describe the behavior of the system, an unsteady characterization, dividing the steady and the
assumed perturbation condition, exploiting dimensional and non-dimensional pressures and mass flow
rates linked to the following relations:

p p
ˆ  
 rT2 2
 2 rT 2
Q m m
ˆ   
 rT 3
 rT 3
 rT 3

86
where  is the water density,  is the pump rotational speed and rT is the pump outlet tip radius, is
performed for each one of the subsystems.

6.1.3 Incompressible Duct (ID)

General Case:
 
p p p
uuu
ggg
mmm

Governing Equation:
 D  
 Dt  t  u       u


  Du   u   u  u  p   g  1  u uC P  p   g  1  u uC 4  p   g  1  f u u
 Dt t 2
f
A 2
f
DH 2 DH

 D
 Dt     u


  Du  p   g  1  f u u
 Dt 2 DH
u 1 uu
   u  u  p   g   f
t 2 DH

Along the streamline in the 1-D approximation:

xd u xd xd xd xd 1 uu
xu  t  dx  xu  u  u  dx   xu p  dx  xu  g  dx  xu 2  f DH  dx
xd u x 1 1 xd 1 L
xu

t
 dx   ud2   uu2   pd  pu    g x  dx   uu uu f  
2 2 xu 2  D eq
m xd dx 1 2 1 2 xd 1 L
   ud   uu   pd  pu    g x  dx   uu uu f  
t xu A x 2 2 xu 2  D eq
m xd dx 1 2 1 2 xd 1
t xu
  ud   uu   pd  pu    g x  dx   uu uu kloss
A x 2 2 xu 2
m xd dx 1 2 1 2 xd 1 m m
t xu
  ud   uu   pd  pu    g x  dx  
A x 2 2 xu 2  A xu  A xu
kloss

2 2
m xd dx 1  m  1  m  xd 1 m m
t xu A x 2   A xd
        pd  pu  x  g x dx   kloss
 2   A x  2  A xu  A xu
  
u
u

87
Under the assumption of incompressible flow:

m  m  m  mu  md
2 2
1  m  1  m  xd 1 m m
       p d  p u    g x dx   kloss
2   A xd  2   A x
  u


xu 2  A xu  A xu

Taking into account the perturbation component of the pressure and the mass flow rate:


 mm  xd dx 1  m  m  1  m  m 
2 2

t  xu
      
A x 2   A xd  2   A xu 

1 mm mm
     
xd
  p d  p d  p u  p u    g x  g x dx   kloss
xu 2  A xu  A xu


 mm  xd  2

dx 1 m  m  2mm 1 m  m  2mm
2
  2 2

     
t
   
2 2
A x 2  A xd 2  A xu
xu

m m  mm  2 m m  
   1
  
xd
  p d  p d  p u  p u    g x  g x dx   kloss
 
2
2 A
xu
 xu

Subtracting the steady-state solution from the total one:

 
 m dx 1 2mm   1 2mm  
    1 2mm
 
 

xd
  p d  pu    g x dx  
xd
    kloss
t
     
2 2
xu A x 2  A 2 A xu 2 A 2
 xd  xu  xu

Leading to the steady-state momentum and continuity equations:

 2
 m  1  m  1
2

p  p   1 m m xd
      kloss    g x dx
 2   A xd  2   A xu  2  A xu  A xu
d u xu


m d  mu
 xd
 
2 2
1   rT 2  1   rT 2  1  rT 2  rT 2 g x dx
 d  u        kloss 
xu

 2  A xd  2  A xu  2 A xu A xu  2 rT 2

 d   u

Time-domain and frequency-domain unsteady-state momentum and continuity equations:

88
 
 
 p  p    1  2m  1  2m  1    2m   
m   
 m xd dx xd
 
 d k     g x dx
 2
     
2 loss 
t
u 2 2
 A xd 2 A 2 A u A
x xu
   xu  xu  x


 m d  mu


   
xd
  r 2 2    xd dx x g x dx
2 2
 r 2   r 2 
 r
       T      T      T   k    T
xu A x  2 rT 2
u

 d   A xd   A x   A x    
u loss
 u   d 
t
  
  
 d u

 
   
 p  e  it  p  e it    1  2m  1  2m  1   2m  xd dx  
k  i  

 it
 d   mu  e 

 2
    
u 2 2 2 loss
 A xd 2  A xu 2  A xu  x 
xu A
  

   g x dx  e  it
xd

 xu

m d  e  it  mu  e it

   2 2 2 2

 2   2 
ˆ  e  it   ˆ  e  it     rT      rT      rT   k  i  rT xd dx  
xu A x  ˆ  e 
 it
 d u
  A xd   A x   A x  loss

   u   d 



xd
 g x dx  e  it
 
xu

  2 rT 2
ˆ  it ˆ  it
 d  e   u  e

The inputs needed to evaluate pressure and mass flow rate from the previous dimensional relations are:

; m

  xd dx 
 A xu ; A xd ; kloss ;  x 
  u A x 


 xd
 u  xd
 x g x dx ; x g x dx
u

While the corresponding known data for the non-dimensional relations are:

; rT ; 

  xd dx 
 A xu ; A xd ; kloss ;  x 
  u A x 


 xd
 u  xd
 x g x dx ; x g x dx
u


89
6.1.3.1 Incompressible Duct Straight (ID-S)

Steady-state momentum and continuity equations:

 1 m m xd
 p d  pu   kloss    g x dx
 2  A xu  A xu xu


m d  mu
 xd

 d  u 
1  rT   rT 
2 2

kloss  u 2 2
x
g x dx
 2 A xu A xu  rT

 d   u

Time-domain unsteady-state momentum and continuity equations:

 
 p  p   1   
2m 
 
m 
 m xd dx
 
xd

 k     g x dx
 2
   t
d u 2 loss
 A xu u A
x xu
   x


 m d  mu


   
xd
  r 2 2   r   xd dx x g x dx
        T   k    T
 t xu A x
 u 2 2
 d   A xd    rT
u loss

  
  
 d u

Frequency-domain unsteady-state momentum and continuity equations:

 
 p  p   1   
2m  xd dx  

 d k  i     mu    xd  xu  g x
 2
 
u 2 loss
  x 
xu A
  A  xu 

 m d  mu


xd

ˆ 
  r 2 2
ˆ     T   k  i  rT d dx   x
  g x dx
 xu A x 
ˆ  xu
 d   A xd   2 rT 2
u loss

  
ˆ  ˆ
 d u

90
6.1.3.2 Incompressible Duct Elbow (ID-E)

Steady-state momentum and continuity equations:

 1 m m xd
 p d  pu   kloss    g x dx
 2  A xu  A xu xu


m d  mu
 xd

 d  u 
1  rT   rT 
2 2

kloss  u 2 2
x
g x dx
 2 A xu A xu  rT

 d   u

Time-domain unsteady-state momentum and continuity equations:

 
 p  p   1   
2m 
 
m 
 m xd dx
 
xd

 k     g x dx
 2
   t
d u 2 loss
 A xu u A
x xu
   x


 m d  mu


   
xd
  r 2 2   r   xd dx x g x dx
        T   k    T
 t xu A x
 u 2 2
 d   A xd    rT
u loss

  
  
 d u

Frequency-domain unsteady-state momentum and continuity equations:

 
 p  p   1   
2m  xd dx  

 d k  i     mu    xd  xu  g x
 2
 
u 2 loss
  x 
xu A
  A  xu 

 m d  mu


xd

ˆ 
  r 2 2
ˆ     T   k  i  rT d dx   x
  g x dx
 xu A x 
ˆ  xu
 d   A xd   2 rT 2
u loss

  
ˆ ˆ
 d  u

91
6.1.3.3 Incompressible Duct Tapered (ID-T)

Steady-state momentum and continuity equations:

 2
 m  1  m  1
2

p  p   1 m m xd
      kloss    g x dx
 2   A xd  2   A xu  2  A xu  A xu
d u xu

m d  mu
 xd
 
2 2
    g x dx
 d  u  1   rT    1   rT    1  rT   rT  kloss  xu 2 2
2 2 2 2

 2  A xd  2  A xu  2 A xu A xu  rT

 d   u

Time-domain unsteady-state momentum and continuity equations:

 
 
 p  p    1  2m  1  2m  1    2m   
m 
 m xd dx   xd
 
 k     g x dx
 2
      t
d u 2 2 2 loss
 A xd 2 A 2 A u A
x xu
   xu  xu  x


 m d  mu


   
xd
  r 2 2    xd dx x g x dx
2 2
 r 2   r 2 
 r
       T      T      T   k    T
xu A x  2 rT 2
u

 d   A xd   A x   A x    
u loss
 u   d 
t
  
  
 d u

Frequency-domain unsteady-state momentum and continuity equations:

 
 
 p  p    1  2m  1  2m  1     
2m  xd dx  
kloss  i  

 d   mu 
 2
     
u 2
 2 A 2 2 A 2 
xu A
x 
  A xd  xu  xu 

   g x dx
xd

 xu

 m d  mu

   r 2 2  rT xd dx  ˆ
2 2
  rT 2   rT 2 
ˆ  ˆ    
 d u
T

  A xd 
   
 A x 
    
 A x 
  k loss  i
 xu A x   
   u   d 


x g x dx
xd

  u 2 2
  rT
ˆ ˆ
 d   u

92
6.1.4 Compressible Duct Straight (CD-S)

Governing Equations (Continuity and Momentum):

 D  
   u      u
 Dt t

  Du   u   u u  p   g  1  u uC P  p   g  1  u uC 4  p   g  1  f u u

 Dt t 2
f
A 2
f
DH 2 DH
 D
 Dt     u


  Du  p   g  1  f u u
 Dt 2 DH

Linearization:
  0   '
p  p0  p '
u  u0  u '
g  g0  g '

Continuity equation:
D  
  u      u
Dt t
  0   '
  u0  u '   0   '    0   '     u0  u ' 
t
  0 
  u0    0     0     u0 
t
   '
  u0     '     0     u '
t
 '
 u0  '   0   u '
t
D 
  u0 
Dt t
D '
  0  u '
Dt

93
Momentum equation:
  u0  u '  1  u0  u '  u0  u '
    u0  u '    u0  u '    p0  p '     g0  g '    f
t 2 DH
  u0  1  u0  u0
    u0    u0     p0     g0    f
t 2 DH
u ' 1 2  u0  u '
   u0 u '  p '  g '  f
t 2 DH
u ' 1 2  u0  u '
   u0 u '   a02  '  g '  f
t 2 DH
Du ' 1 2  u0  u '
  a02  '  g '  f
Dt 2 DH
Du '  u0  u '
0  a02  ' 0 g ' 0 f
Dt DH

Linearized Governing Equations (Continuity and Momentum):

 D '
   0   u '
 Dt D 
    u0 
  Du '  a 2  '  g '  f  u0  u ' Dt t


0
Dt
0 0 0
DH

Cross Differentiation:
  D ' 
      0  u '   0 u '
2

  Dt 

  D u '  a 2 D   '   D g '  f D   u0  u ' 
2  
 0
Dt 2 0
Dt
0
Dt
0
Dt  DH 


General wave equation:

D2u ' D D   u0  u ' 
0   0 a02  2 u '  0 g '  0 f  
Dt 2
Dt Dt  DH 

D2u ' D D   u0  u ' 

 a02  2 u ' g ' f  
Dt 2
Dt Dt  DH 

D2u ' D D   u0  u ' 

 a02  2 u ' g ' f  
Dt 2
Dt Dt  DH 
D2 u '   u '   u '   u'
2
u ' u ' 2u ' u '
   u0 u '   u0    u0 u '   2  u0   u0    u0  u0   2 u '  2  2u0   u0  2 u '
2

Dt 2
t  t   t   t t t t t

D2 u '  g '     u u'   u u' 

 a02  2 u '   u0 g '   f   0   u0   0  
Dt 2     t D 
t   H   DH  
D2 u '  g '    u  u '  u    u  u 
 a02  2 u '   u0 g '   f   0   u '  0   u0    0   u ' u0  0 u ' 
Dt 2
 t    DH  t t  DH    DH   DH 
 

94
General wave equation:
D2 u '  g '    u  u '  u    u  u 
 a022 u '   u0 g '   f  0   u '  0   u0    0   u ' u0  0 u ' 
Dt 2
 t    DH  t t  DH    DH   DH 
 

Wave equation, particular case assumptions:

  u0 
  0
t  DH 
 u 
  0   0
 DH 
 g '  0

Hence:
2 u ' u '  g '    u  u '  u    u  u0 
 2u0   u0 2 u '  a022 u '   u0 g '   f  0   u '  0   u0    0   u ' u0  u ' 
2

t 2
t  t    DH  t t  DH    DH   DH 
 
2u ' u ' g '   u  u ' u 
 2u0    u0  2 u '  a02  2 u '   u0  0 u ' 
2
f  0 
t t  D 
t   H  t
2
DH 
u  u0  u '  w  w  z   w  z , t 
g  g0  g '  g  g  g  t   g  ge
ˆ it
u0  u0  z   0 0 w0  z   0 0 w  z 

  
u '  u '  z   0 0 w '  z   0 0 Re  w  z, t    0 0 Re  wˆ  z  e it  
 DH  constant

 w0  constant

1-D wave equation:

2 w 2 w 2  w
2
2  w
2
g  w0 w w02 w 
 2 w  w  a   f   
t 2 t z z 2 z 2 t  DH t DH z 
0 0 0

Tentative solution:
w  z, t   wˆ  z  eit

95

 wˆ  z   eikz  a

 wˆ  z 
  ikeikz
  z
  2 wˆ  z 
  k 2 eikz
 z 2

 wˆ  z   Aeik1 z  Beik2 z  a

 wˆ  z 
  ik1 Aeik1 z  ik2 Beik2 z
  z
  2 wˆ  z 
  k12 Aeik1 z  k22 Beik2 z
 z 2

2w 2w 2w  2 w g  w w w02 w 

 2 w0  w02 2  a02 2  f 0  
t 2
t z z z t  DH t DH z 
wˆ  z   it  2 wˆ  z   it 2  w  z   i t
2
ˆ  w w2 wˆ  z  it 
 2 wˆ  z  e  it  2i w0 e  w02 e  a ˆ  it  f  i 0 wˆ  z  e it  0
e  i ge e 
z z z DH z
2 0 2
 DH 
 wˆ  z   2
ˆ
w  z   2
ˆ
w  z   w w 2
 wˆ  z  
 2 wˆ  z   2i w0  w02  a02  i gˆ  f  i 0 wˆ  z   0 
z z 2 z 2  DH DH z 

 a02  w02  z2    2i w0  f Dw0  z     2  if  Dw0  wˆ  z   i gˆ

 2 wˆ z  2
 wˆ z  
 H   H 

 w02   2 w0 
 2i w0  f     i f 
 wˆ  z 
2
 DH  wˆ   
z D H  i gˆ
   wˆ  z   2
z 2  0 0
a 2
 w 2
 z  0 0
a 2
 w 2
 0  w02 
a
 w   w 
w0  2i  f 0  i  i  f 0 
 2 wˆ  z   ˆ   i gˆ
  2 H    wˆ z 
D w z D
 H
 
z 2
 a0  w02  z  a02  w02  a2
0  w02 
 w   w 
w0  2i  f 0  i  i  f 0 
 2 wˆ  z   ˆ   i gˆ
  2 H    wˆ z 
D w z D
 H
 
z 2
 a0  w02  z  a02  w02  a2
0  w02 

96
  w   w 
w0  2i  f 0  i  i  f 0 
i gˆ
  2  ikeikz   DH  ikz
 e  a  2
DH
 k 2 eikz
 a0  w0 
2
 a0  w0 
2 2
 a0  w02 
 w   w   w 
w0  2i  f 0  i  i  f 0  i  i  f 0 
i gˆ
  2  ikeikz    eikz   a
DH DH DH
 k 2 eikz
 a0  w0 
2
 a0  w0 
2 2
 a0  w0 
2 2
a
2
0  w02 
gˆ
a
 w0 
f  i 
 DH 
 w   w 
w0  2i  f 0  i  i  f 0 
 k 2 eikz    2  ikeikz    eikz
DH DH
 a0  w0 
2
 a0  w0 
2 2

 w   2 w0 
w0  2i  f 0     i f 
k2   2  ik  
DH DH 
 0  k1 , k2
 a0  w0 
2
 a02  w02 
Final solution:
  w   2 w0 
 w0  2i  f 0     i f 
k , k  k 2   D H  ik   DH 
0
wˆ  z   Aeik1 z  Beik2 z
 1 2
a 
 a02  w02   a02  w02 
a  gˆ
  w0 
 f  i 
  DH 

1-D continuity equation:

D '
  0  u '
Dt
D w
  0
Dt z
pˆ  Ce  Deik2 z
ik1 z

pˆ
 ik1Ceik1 z  ik 2 Deik2 z
z
1  p p  w
  w0    0
a02  t z  z
1  pˆ  wˆ
2 
i pˆ  w0    0
a0  z  z
1  pˆ 
 i pˆ  w0
a02  z 

   0 ik1 Ae  ik2 Be
ik1 z ik2 z


97
1
a02
 
i  Ceik1 z  Deik2 z   w0  ik1Ceik1 z  ik2 Deik2 z    0  ik1 Aeik1 z  ik2 Beik2 z 

 0 a02  ik2 eik2 z  0 a02 k2

1
 i e ik2 z
 w ik e ik2 z
 D     ik e ik2 z
 B  D  B  D  B
a02
0 2 0 2
 ieik2 z  w0ik2 eik2 z    w0 k2
 0 a02  ik1eik1 z  0 a02 k1
1
 i e ik1 z
 w ik e ik1 z
 C     ik e ik1 z
 A  C  A  C  A
a02
0 1 0 1
 ieik1z  w0ik1eik1z    w0 k1
  a2k    a2k 
pˆ   0 0 1  Aeik1 z   0 0 2  Beik2 z
   w0 k1     w0 k2 

Boundary Conditions:

 w  z, t   wˆ  z  eit

 p  z , t   pˆ  z  e
i t


   z, t   1 ˆ  z  eit
 a02

 wˆ  z   Aeik1 z  Beik2 z  a

  0 a02 k1  ik1 z  0 a02 k2  ik2 z
 pˆ  z     Ae    Be
    w0 k1     w0 k2 

 wˆ  z  0   A  B  a  A  B  wˆ  z  0   a
 
  0 a0 k1 
2
 0 a0 k2    0 a02 k1 
2
 0 a02 k2 
 ˆ
p  z  0     A    B   A    B  pˆ  z  0 
    w0 k1     w0 k 2     w0 k1     w0 k 2 
 A   B  wˆ  z  0   a

  0 a02 k2   0 a02 k1    0 a02 k1 
        B  ˆ
p  z  0      wˆ  z  0   a 
    w0 k2     w0 k1      w0 k1 

    a2k  
  pˆ  z  0    0 0 1   wˆ  z  0   a  
     w0 k1   ˆ
A        w  z  0  a
    a 2
k    a 2
k  
  
0 0 2 0 0 1
     w0 k2     w0 k1   

   a2k 
 pˆ  z  0    0 0 1   wˆ  z  0   a 
B     w0 k1 
  0 a02 k2   0 a02 k1  
   
    w0 k2     w0 k1  

Leading to the steady-state momentum and continuity equations:

98
 1 m m xd
 p d  pu   kloss    g x dx
 2  A xu  A xu xu


m d  mu
 xd

 d  u 
1  rT 2  rT 2
kloss 
xu
g x dx
 2 A xu A xu  2 rT 2

 d   u

Frequency-domain unsteady-state momentum and continuity equations:

   0 a02 k1  ik1L   0 a02 k2  ik2 L
 d p  p  z  L  
  Ae    Be
    w0 k1     w0 k2 

m d   A xd w  z  L    A xd  Ae  Be  a 
ik1 L ik2 L


     
     
1  0 a0 k1
2
 Aeik1L   0 a0 k2
2
 Beik2 L 
 d   z  L  
0   rT       rT k    rT 3  
2 3
   
  1 
 

k2 
 
 
  A  xd   A  xd  
 A xd
3 
 d    z  L   Aeik1L  Beik2 L  a 
  rT

 p  z  0   pu

 pu  p  z  0  
 mu
w  z  0  
mu   A x w  z  0    A xu
  
u

    rT   u  p  z  0   p  z  0      rT   u
2 2

 
  rT  u   A xu w  z  0    rT 3 u
3

w  z  0   A
  xu

Time-domain unsteady-state momentum and continuity equations:

99
 p  Re
 d  p e
d
 i t


m d  Rem  ed
 i t


 p e
 p u  Re

u
 i t

mu  Re
 m  eu
 i t


 d  Re

  e d
 i t

 d  Re
   e d
 i t


 u  Re   e u
 i t


 u  Re
   eu
 i t

 p  Re
 d p d e
i p
d
 
 e  it  p d cos t   p
d
 p d 
cos t   p
d


m d  Re

m d e
im
d
  
 e  it  m d cos t   md  m d cos t   md  
 p  Re
 u p u e
i p
u
 
 e  it  p u cos t   p
u
 p u 
cos t   p
u


mu  Re
 m e u
im
u
  
 e  it  mu cos t   mu  mu cos t   mu  

 d  Re

 e d
i
d
   
 e  it   d cos t    d   d cos t    d 

 d  Re  e d
i
d
 e  it    cos  t  
d
d    cos t  
d
d 

 u  Re

 e u
i
u
 e  it    cos  t  
u
u    cos t  
u
u 

 u  Re e u
i
u
 e  it    cos  t  
u
u    cos t  
u
u 

6.1.5 Silent Throttle Valve (STV)

General Case:
 
p p p
uuu
ggg
mmm

Governing equations:

100
 D  
 Dt  t  u       u


  Du   u   u  u  p   g  1  u uC P  p   g  1  u uC 4  p   g  1  f u u
 Dt t 2
f
A 2
f
DH 2 DH

 D
 Dt     u


  Du  p   g  1  f u u
 Dt 2 DH
u 1 uu
   u  u  p   g   f
t 2 DH

Along the streamline in the 1-D approximation:

xd u xd xd xd xd 1 uu
xu  t  dx  xu  u  u  dx   xu p  dx  xu  g  dx  xu 2  f DH  dx
xd u x 1 1 xd 1 L
xu

t
 dx   ud2   uu2   pd  pu    g x  dx   uu uu f
2 2 xu 2
 
 D eq
xd  m  1 2 1 2 xd 1 L
xu
 
t   A x
   ud   uu   pd  pu  xu  g x  dx   uu uu f  
 2 2 2  D eq
m xd dx xd   1  1 1 xd 1 L

t u A x
x
 m
xu t
 
 A x 
dx   ud2   uu2   pd  pu    g x  dx   uu uu f  
2 2 xu 2  D eq
m xd dx xd   1  1 2 1 2 xd 1

t xu A x
 m
xu t

 A x
 dx   ud   uu   pd  pu  xu  g x  dx   uu uu kloss
 2 2 2
m xd dx xd   1  1 2 1 2 xd 1 m m

t xu A x
 m
xu t

 A x
 dx   ud   uu   pd  pu  xu  g x  dx  
 2 2 2  A xu  A xu
kloss

2 2
m xd dx xd   1  1  m  1  m  xd 1 m m

t xu A x
 m
xu t

 A x
 dx   
 2   A xd
  
 2   A x
   pd  pu  x  g x dx  
 2  A  A
kloss
  
u
u  xu  xu

Under the assumption of incompressible flow:


m  m  m  mu  md
kloss  k loss  k loss
A x  A x  A x
 1   1 
    0
t  A x  t  A x  A x 
 1   1   A x
    2  0
t  A x   A x  t

101
 2 2
1  m  1  m  xd 1 m m
       p d  p u    g x dx   k loss
2   A xd  2   A x
  u


xu 2  A xu
 A xu

Leading to the steady-state momentum and continuity equations:

 2
 m  1  m  1
2

p  p   1 m m xd
      kloss    g x dx
 2   A xd  2   A xu  2  A xu  A xu
d u xu

m d  mu
 xd
 
2 2
    g x dx
 d  u  1   rT    1   rT    1  rT   rT  kloss  xu 2 2
2 2 2 2

 2  A xd  2  A xu  2 A xu A xu  rT

 d   u

Taking into account the perturbation component of the pressure and the mass flow rate:


 mm  xd dx 1  m  m  1  m  m 
2 2

t  xu
      
A x 2   A xd  2   A xu 

1 mm mm
       
xd
  p d  p d  p u  p u    g x  g x dx   k loss  k loss
xu 2  A xu  A xu


 mm  xd  2

dx 1 m  m  2mm 1 m  m  2mm
2
  2 2

     
t
   
2 2
A x 2  A xd 2  A xu
xu

m m  mm  2 m m m m  mm  2 m m    
   1
 1
 
xd
  p d  p d  p u  p u    g x  g x dx   k   k loss
   
2 loss 2
2 A 2 A
xu
 xu  xu

Subtracting the steady-state solution form the total one:

 
 m dx 1 2mm   1 2mm  
    1 2mm
  1 mm   

xd
  p d  p u    g x dx  
xd
    k   k loss
       
loss
t xu A x 2  A 2
2 A 2 xu 2 A 2 2 A 2
 xd  xu  xu  xu

Leading to the time-domain unsteady-state momentum and continuity equations:

102
 
 
 p  p    1  2m  1  2m  1    2m  
m 
 m xd dx 1   mm  
t xu A x 2  A 2
 k   kloss 
 2
     
  
d u 2 loss
 2 A 2 2 A 2
  A  xd  xu  xu   xu


 
xd
    g x dx
 xu

 m d  mu

  2 2
2
2
2
  
       rT      rT      rT   k     rT   xd dx  1   rT     k 
2
2
2

 d   A xd   A x   A x    t xu A x 2  A xu 
u loss loss

   u   d 


 
xd


x g x dx
 u 2 2
  rT

 d   u

Frequency-domain unsteady-state momentum and continuity equations:

 
 
 p  e  it  p  e  it    1  2m  1  2m  1      2m  xd dx  
k  i  

 it
 d   mu  e 
 2
     
u 2 2 2 loss
 A xd 2 A 2 A  A x 
xu
   xu  xu 



1
 
 mm  xd
k  e  it    g x dx  e  it
 
2 loss
 2 A xu
 xu

m d  e  it  mu  e  it

   2 2  2 
2
 2 
2

ˆ  e  it   ˆ  e  it     rT      rT      rT   k  i  rT xd dx  
xu A x  ˆ  e 
 it
 d   A xd   A x   A x  
u loss

   u   d 



xd
 1   rT 2 
2
g x dx  e  it
   ˆ
    kloss  e   it xu

 2  A xu   2 rT 2
ˆ  it ˆ  e  it
 d  e   u

 
 
 p  p    1  2m  1  2m  1    2m    xd dx  

1 mm   xd

 k loss  i 
  m u   k loss    g x dx
 2
       u A x    
d u 2 2 2 2
 A xd 2 A 2 A 2 A
x xu
   xu  xu
  xu

 m d  mu

   r 2 2  rT 2 
2
 rT 2 
2
 r xd dx  ˆ
 d   u    
ˆ ˆ           kloss  i T x 
T

   A xd   A x 
 u 
 A x 
 d 
 u A x 
 


x g x dx
xd
2

 1r 2 
  T     kˆloss  u 2 2
 2  A xu   rT

ˆ  ˆ
 d u

103
6.1.6 Volume Oscillator Valve (VOV) – Exciter
Following the same procedure exploited for the characterization of the IDs, with the presence of a forced
mass flow rate contribution given by the exciting purpose of the device into the unsteady-state equations:
Steady-state momentum and continuity equations:
 2
 m  1  m  1
2
1
 p  p    m m xd
      kloss    g x dx
2   A xd  2   A xu  2  A xu  A xu
d u
 xu


md  mu

Time-domain unsteady-state momentum and continuity equations:


p  p  1   2mm d  1 2mmu 1  2mm    m xd dx 
 
xd
    k      g x dx

      t
d u 2 2 2 loss
2 A 2 A 2 A xu A
x
xu

  xd  xu  xu
 V
 m d  mu   D  mu  m D
 t
 V
 m  mu  m d  mu   1 D  mu  1 m D
 2 2 t 2

 
 1 2m   1 2m   1 2m    
1  mu xd dx 

 d
p  p  
 2
     k  xu A x  mu 
      
u 2 2 2 loss
   A xd 2  A xu 2  A xu m u t
  
  
  1  
2m 1 2m 1   1 1  2VD xd dx  VD
2 VD t 2 xu A x  t
      kloss   
  2 A    
2
2 A 2 2
   xd  xu
t


 x
 xu
d
 
  g x dx

Frequency-domain unsteady-state momentum and continuity equations:

 
 
 p  e it  p  e it    1  2m  1  2m  1    2m  
k  i  
 xd dx  

 it
 d   mu  e 
 2
     
u 2 2 2 loss
 A xd 2 A 2 A  A x 
xu
   xu  xu 





  
1 2m   1
 
2m 1   1  xd dx  
kloss  i  

 it
   mD  e 
    
2 2
 2  A xd 2  A xu 2 2  x  
xu A

  
 xd
    g x dx  e  it
xu
  it
md  e  mu  e  it  m D  e  it

104
6.1.7 Tank (T) – Exciter
It is assumed that in steady-state condition, the tank acts as an ideal incompressible duct, without any
contributions of inertial and viscous losses.
General momentum equation:
xd u xd xd xd xd xd 1 uu
xu  t  dx  xu  u  u  dx   xu p  dx  xu  g  dx  xu  r  dx  xu 2  f DH  dx

Taking into account the 1-D streamline connecting the free surface of the tank and its downstream duct
section which communicates with the hydraulic loop, and defining some useful parameters, we obtain:
xout u x 1 2 1 2 xout xout 1 L
xFS  t  dx  2  uout  2  uFS   pout  pFS  xFS  g x  dx  xFS  rx  dx  2  uout uout f  D eq
1 xout
g xFS
hgFS out  g x  dx

1 xout
hrFS out   rx  dx
r xFS
1 xout u
hFSux tout  
u x t FSx
 x  dx
t
xout u x 1 2 1 2 1 L
 xFS

t
 dx   uout
2
  u FS
2
  pout  pFS   ghg   rhr   uout uout f
2
 
 D eq
1 2 1 2 1 L u
pout  pFS   uFS   uout   ghgFS out   rhrFS out   uout uout f     x hFSu x tout
2 2 2  D eq t

The compliance of the tank is taken into account by means of the dynamics of the interface of its upper
part (free surface):
dV
mout  min   T
dt
dVT dz FS
  AT
dt dt
dz
u FS   FS
dt

Under the assumption of adiabatic transformation:

1 dpT 1 dVT dV dp
pT VT   constant    0  T  CT T
pT dt VT dt dt dt
VT
CT 
 pT
pT  pFS
dpFS AT dz FS

 m  min 
  out
dt CT dt CT 

105
Taking into account the 1-D streamline connecting the upstream duct section of the tank which
communicates with the hydraulic loop and its free surface, and defining some useful parameters, we
obtain:

xFS u x 1 2 1 2 xFS xFS 1 L

xin

t
 dx   uFS
2
  uin   pFS  pin    g x  dx    rx  dx   uin uin f
2 xin xin 2
 
 D eq
1 xFS
g xin
hgin  FS  g x  dx

1 xFS
hrin  FS   rx  dx
r xin
1 xFS u
hinu    dx
FS x
x t
u x t xin t
1 2 1 1 L u
pFS   uFS  pin   uin2   ghgin FS   rhrin  FS   uin uin f     x hinu FS
x t
2 2 2  D eq t

Having obtained the previous relations, we can now eliminate the free surface pressure contribution:
 1 2 1 2 1 L u
 pout  pFS  2  u FS  2  uout   ghg   rhrFS out   uout uout f     x hFSux tout
FS  out

 2  D eq t

 p  1  u 2  p  1  u 2   ghin  FS   rhin  FS  1  u u f  L    u x h in  FS
in  
 FS 2 FS in
2
in g r
2
in
 D eq t
u x t

  g  hgin  FS  hgFS out    r  hrin FS  hrFS out  
1 1 2
pout  pin   ui2n   uout
2 2
1 L 1 L
  uin uin f     uout uout f   
2  D eq 2  D eq
u x in  FS

t

hux t  hFSux tout 

Dealing again with the free surface of the tank and the connected compliance, we aim to relate the
remaining parameters related to the free surface, to the time derivative of the vertical displacement of the
free surface, which will be related to the frequency of perturbation.

dVT
mout  min  
dt
dVT dz
  AT FS
dt dt
dz C dp d 2u FS d 3 z FS
u FS   FS   T FS   
dt AT dt dt 2 dt 3
dz FS
mout  min   AT
dt

106
 dpFS
mout  min   CT
dt
CT dpFS dzFS dp A dz
  FS  T FS
AT dt dt dt CT dt
dpFS AT dz FS

dt CT dt
1 2 1 1 L u
pFS   uFS  pin   uin2   ghgin FS   rhrin FS   uin uin f     FS hinu FS
x t
2 2 2  D eq t
Taking into account the assumptions of negligible viscous losses and small perturbations:
1 L
 uin uin f    0
2  D eq
in  FS
dpFS dpin duin duFS dhgin FS dhrin FS dr in FS d 2uFS in FS duFS dhux t
   uin   uFS  g  r   hr  hux t  
dt dt dt dt dt dt dt dt 2 dt dt
in  FS in  FS 2
dpFS dpin du dhg dh dr d uFS in FS
   uin in   g  r r   hrin FS   hux t
dt dt dt dt dt dt dt 2
dpFS dpin du dhgin FS dhin FS dr d 3 z FS in FS
   uin in   g  r r   hrin FS   hux t
dt dt dt dt dt dt dt 3
dhgin FS dhrin FS dz FS CT dpFS
  
dt dt dt AT dt

Aiming to obtain, as usual, unsteady-state momentum and continuity equations that relate the upstream
and downstream sections pressure and mass flow rate of the component, to allow the implementation of
the tank into our dynamic model:
e g  e z  cos gz
er  e z  cos rz
in  FS
1 dhg 1 dhrin FS dz FS CT dpFS
  
cosgz dt cosrz dt dt AT dt
d 3 z FS
AT dz FS dpin du dz dz dr 3 dz FS in FS
   uin in   g cosgz FS   r cosrz FS   hrin FS   dt h
CT dt dt dt dt dt dt dz FS dt ux t
dt
   d 3 z FS  
A   3  in FS   dz dp du dr
 T    g cosgz  r cosrz   dt  hux t   FS  in   uin in   hrin  FS
 CT   dz FS    dt dt dt dt
   dt  
    
dz FS 1 dpin 1 du 1 dr
   uin in   hrin FS
dt    d 3 z FS    dt    d 3 z FS   dt    d 3 z FS    dt
A   3   A   3   A   3  
 T    g cosgz  r cosrz   dt  hinu FS
x t
  T    g cosgz  r cosrz   dt  hinu FS
x t
  T    g cosgz  r cosrz   dt  hinu FS
x t

 CT   dz FS    CT   dz FS    CT   dz FS  
   dt      dt      dt  
              
 AT CT dpin  AT CT duin  AT CT dr
mout  min    uin   hrin  FS
   d 3 z FS    dt    d 3 z FS   dt    d 3 z FS    dt
   3      3      3  
 AT   CT  g cosgz  r cosrz   dt  hinu FS
x t
  AT   CT  g cosgz  r cosrz   dt  hinu FS
x t
  AT   CT  g cosgz  r cosrz   dt  hinu FS
x t

   dz FS      dz FS      dz FS  
   dt      dt      dt  
              

107
 dzFS
mout  min   AT
dt

dpin du dr
mout  min   CTeq   CTeq  uin in   CTeq  hrin  FS
dt dt dt
AT CT
CTeq 
   d 3 z FS  
   3  
 AT   CT  g cosgz  r cosrz   dt  hinu FS
x t

   dzFS  
   dt  
    
AT CT
CTeq 
 
 AT   CT g cosgz  r cosrz   2 hinuFSt 
x  

Particular case:
 Ain and Aout placed at the same height;
 Vertical tank vibration (coordinate z);
 g  gz

Obtain:
cos gz  1
cos rz  1
AT CT
CTeq 
 
 AT   CT g  r   2 hinuFSt 
x  
Obtaining the unsteady-state momentum and continuity equations in time and frequency domain:
 dpin 1 dmin dr
mout  min   CTeq  CTeq 2 min   CTeq  hrin  FS
 dt Ain dt dt
 AT CT
CTeq 
  
 AT   CT g  r   2 hinuFSt   

x


 p  p   1 

  2m 1  2m 1  2m 
kloss  min 
 
 m xd dx
t xu A x
   
  2
      
out in 2 2 2

  A xout 2 A 2 A

   xin  xin


 
 1   2m 1  
2m 1   2m  xd dx  

 xu
 i t  i t  it
 out
p  e  p  e        k  i   min  e

     
in 2 2 2 loss
 2 A 2 A 2 A  A x 
   xout  xin  xin 
 p  p    R  i L  min
 out in

  1 
mˆ out  i CTeq pˆ in  1  iCTeq 2 min  mˆ in  i 2CTeq hrin  FS rˆ
  Ain 

108
6.1.8 Pump (P)

Under the 1-D approximation inside the channel (only fully guided velocity):
 =0
u'
x2 x x2
u '22 u '12 p2 p1 1 2 2 1 2 2 2 1 P
x  t dx 
2

2
  
  2
 r2 
2
 r1  
x1
g x dx  
x1
C f u '2 dx  0
2 A
1

u'
x x x
p2 p1 u '22 u '12 1 2 2 1 2 2 2 2 2
1 P
     r2   r1    dx   g x dx   C f u '2 dx
  2 2 2 2 x1
t x1 x1
2 A
u'
x x x
p2 u2 u2 p1 2 2 2
1 P
  2  1  r2v2  r1v1    dx   g x dx   C f u '2 dx
  2 2 x1
t x1 x1
2 A
pt 2  pt1 u'
x2 2 x 2 x
1 2P
  r2v2  r1v1    C f u ' dx   dx   g x dx
 x1
2 A x1
t x1

pt 2  pt1  r2v2  r1v1  2 u'

x x2 x2
1 u '2 P 1 1
 rT 2 2

 rT
2 2
 
x1
C f
2  rT A
2 2
dx  2 2
 rT x  t dx   2rT2 x g x dx
1 1

In steady-state conditions:
pt 2  pt1
x2 2 x
1 2P
  r2v2  r1v1    C f u ' dx   g x dx
 x1
2 A x1

pt 2  pt1  r2v2  r1v1  2

x x2
1 u '2 P 1
 rT 2 2

 rT
2 2
 
x1
C f
2  rT A
2 2
dx  2 2
 rT  g dx
x1
x

Transforming in non-dimensional form:

pt 2  pt1  r2 v2  r1v1  2
x x2
1 u '2 P 1
 
 rT 2 2

 rT
2 2
 
x1
C f
2  rT A
2 2
dx  2 2
 rT  g dx
x1
x

 
  ,    0 ,  0    0 ,  0   0    0 ,  0    0 
 
  0  
  0 
 
  ,    0 ,  0    0 ,  0     0 ,  0  
 

In unsteady-state conditions:
 pt 2  pt 2    pt1  pt1  

 r
2  r2  v2  v2    r1  r1  v1  v1   

u'
x2 x
2 2 x
1 2 P
 Cf  u  u  dx  x  t dx  x  g x  g x  dx
x1
2 A 1 1

109
 pt 2  pt1    pt 2  pt1  
 r2v2  r1v1    r2v2  r1v1    r2v2  r1v1    r2v2  r1v1 
 
u'
x2 x x x x

dx   C f  2uu  u  dx  
2 2 2 2
1 2P 1 2 P
 Cf u dx   g x dx   g x dx
x1
2 A x1
2 A x1
t x1 x1

 pt 2  pt1   x2
1 2P 2x

 r2v2  r1v1    C f u dx   g x dx
 x1
2 A x1

 pt 2  pt1   x2
u'
x x

 r2v2  r1v1    r2v2  r1v1    r2v2  r1v1    C f  2uu  u  dx   dx   g x dx

2 2
1 2 P

 x1
2 A x1
t x1

pt 2  pt1  r2v2  r1v1  2 u'

x x2 x2
1 u '2 P 1 1
   Cf dx  2 2 x  t dx   2rT2 x g x dx
 rT 2 2
 rT
2 2
x1
2  rT A
2 2
 rT 1 1

1  u2  u1   r2v2  r1v1  2 u'

2 2
p2  p1
x x2 x2
1 u '2 P 1 1
 rT
2 2
 
 rT
2 2
2

 rT
2 2
 
x1
C f
2  rT A
2 2
dx  2 2
 rT x  t dx   2 rT2  g dx
x1
x
1

p2  p1

 2 rT2
   
1  u2  u1   r2 v2  r1v1  2
2 2 x x2
1 u '2 P 1
  2 2
 rT 2

 rT
2 2
 
x1
C f
2  rT A
2 2
dx  2 2
 rT  g dx
x1
x

u'
x2 x2
1 1
  2 2
 rT x  t dx   2rT2  g dx
x1
x
1

Expansion in Taylor’s series:

 
  ,     0 ,  0    0 ,  0   0    0 ,  0    0 
 
  0  
  0 
 
  ,     0 ,  0  
 0 ,  0     0 ,  0  
 
 
   0 ,  0    0 ,  0     0 ,  0   
 

Summarizing:

110
   1  m 2 dx
x x2
1
  ,     0 ,  0    0 ,  0     0 ,  0      2 2  g dx
   rT  t x1 A '  rT
2 2 x
x1

   rT 3 
x2 x2
dx 1
  ,     0 ,  0    0 ,  0     0 ,  0    x A '   2 rT2  g dx
   2 rT2  t x
1 x1

    rT
x2 x2
dx 1
  ,     0 ,  0    0 ,  0     0 ,  0    x A '   2 rT2  g dx
  t  x
1 x1

     rT 2 dx 
x x
1 2
  ,     0 ,  0     0 ,  0    
 t   x1 A '   2 rT2 x1 x
 g dx
 

 0 ,  0   0 ,  0     rT 2 dx 
x
1 2
x

  ,   
  x A '   2 rT2 x x
    g dx
 
 
 t  1  1

 1 VC  1 VC
 d   u   r 3 t    u  2 r 3 t


T

T

    1 VC
2

  1   
 d u
 d
2 rT3 t
 ˆ      xd dx 
ˆ d ˆ u   2 u    i    rT   ˆ
       L  xu A
Average oscillations of the average mass flow rate calculation:
 ˆ    ˆ  

ˆ d ˆ u  2   
u  d  1  2
   
 u  
  nˆ
 ˆ   
     x dx       xd dx  dˆ
i    rT x  i    rT 
d
   
    L  A       L  xu A

u


Leading to the steady-state characteristic pressure rise equation:

 1  u2  u1   r2 v2  r1v1  2
2 2 x x
1 u '2 P 1 2
   2 2    Cf dx  2 2  g x dx
  rT 2  2 rT2 x1
2  2 rT2 A  rT x1

  ,     ,      ,         ,    
 0 0

0 0 0

0 0 0


  0  
    
 0

  
  ,    0 , 0    0 , 0    0 , 0 
  



111
Time-domain unsteady-state characteristic pressure rise and continuity equations:
      r 2 dx 
x
1 2
x

  ,    0 , 0    0 , 0    T    2 2  g x dx
    t   x1 A '   rT x1

 1 VC  1 VC
 d   u   r 3 t    u  2 r 3 t


T

T

    1 VC
2

  1   
 d u   d
2 rT3 t

Frequency-domain unsteady-state pressure rise and continuity equations:

  ˆ      xd dx 
ˆ d ˆ u  2 u   i    rT   ˆ
      
 
 L xu A

 VˆC ˆ
ˆ VˆC
ˆ ˆ
 d   u  i     i
2 rT3
u
  rT3  
 
ˆ  ˆ  i VˆC
ˆ 1 ˆ
  2
 d  
ˆ
u 

d
2 rT3

112
6.2 Appendix B, Matlab Code
6.2.1 Introduction
The subsequent section aim to unveil the code used to calculate all the parameters to proceed with the
unsteady analysis of a wide-world typology of hydraulic loop system exploited to perform experiments
concerning a pump.
Knowing the geometry properties of the different subsystems identified in 3.2, composing the desired
loop, the code is able to calculate the steady-state operative condition and the unsteady parameters
exploited for the small-perturbation dynamic analysis, which is provided by a parallel cooperation
between a Matlab live script and a Simulink model described in 6.3.
The code is linked to the model equations derived previously and therefore, coherently with the
mathematical characterization of the subsystems, it will be presented the steady and unsteady-state
parameters required by the program for each one of the system elements.

6.2.2 Steady-State System Parameters

The following subsections show the parameters and the functions needed to obtain the operative
condition.

6.2.2.1 Incompressible Duct Straight (ID-S)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Loss coefficient: k_sd

Upstream duct diameter: Du_sd
Unperturbed density of the working fluid: d_sd

6.2.2.2 Incompressible Duct Elbow (ID-E)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Loss coefficient: k_ed

Upstream duct diameter: Du_ed
Unperturbed density of the working fluid: d_ed

6.2.2.3 Incompressible Duct Tapered (ID-T)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Loss coefficient: k_td

Upstream duct diameter: Du_td
Downstream duct diameter: Dd_td
Unperturbed density of the working fluid: d_td

113
6.2.2.4 Silent Throttle Valve (STV)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Loss coefficient: k_v

Upstream duct diameter: Du_v
Downstream duct diameter: Dd_v
Unperturbed density of the working fluid: d_v

6.2.2.5 Volume Oscillator Valve (VOV)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Loss coefficient: k_vov

Upstream duct area: Au_vov
Downstream duct area: Ad_vov
Unperturbed density of the working fluid: d_vov

6.2.2.6 Pump (P)

Experimental and geometrical parameters to input for the calculation of the steady-state operative
condition:

Unperturbed density of the working fluid: d_p

Rotational speed velocity of the volute: rpm
Pump tip radius: rt_p
Flow coefficient vector (non-dimensional): phi
Pressure rise coefficient vector (non-dimensional): psi

Since we expect a parabolic behavior of the pump characteristic curve, the next script shows the
interpolation procedure exploited to obtain the dimensional pressure rise of the pump as a second order
polynomial of the mass flow rate:

c_i = polyfit(phi,psi,2)
a = c_i(1)/(d_p*pi^2*rt_p^4)
b = c_i(2)*(rpm*pi/30)/(pi*rt_p)
c = c_i(3)*d_p*(rpm*pi/30)^2*rt_p^2

Since in a closed loop system the pressure rise of the pump has to be equal to the pressure drop obtained
through the others components, and, having all quadratic relations through the mass flow rate, for all the
pressure variation across the system elements, the solution of the operative condition i.e. the steady-state
mass flow rate, is given by the positive value of the mass flow rate which provides the intersection
between the concave pump parabolic characteristic curve and the summation of the elements convex
parabolic curves which results in an equivalent convex parabolic curve.
The operative point is calculated by the so called ‘OperativeCondition” function; the script presents the
steady-state solution for an example of a hydraulic loop composed by a tank, three ID-Es, four ID-Ss, a
pump and a STV.

114
function m0 = =OperativeCondition(a,b,c,k_ed1,d_ed1,Au_ed1,k_ed2,d_ed2,Au_ed2,...
k_ed3,d_ed3,Au_ed3,k_sd1,Au_sd1,d_sd1,k_sd2,Au_sd2,d_sd2,...
k_sd3,Au_sd3,d_sd3,Au_v,Ad_v,k_v,d_v)

m = roots([-a+0.5*k_ed1/(d_ed1*Au_ed1^2)+0.5*k_ed2/(d_ed2*Au_ed2^2)+...
0.5*k_ed3/(d_ed3*Au_ed3^2)+0.5*k_sd1/(Au_sd1^2*d_sd1)+...
0.5*k_sd2/(Au_sd2^2*d_sd2)+0.5*k_sd3/(Au_sd3^2*d_sd3)+...
+k_v/(d_v*Au_v^2)+1/(d_v*Ad_v^2)-1/(d_v*Au_v^2) -b -c]);

if m(1) > 0
m0 = m(1);
else
m0 = m(2);
end
end

The design flow coefficient can be calculated from the steady-state mass flow rate with the following
expression:

phi_op = m0/(d_sd1*pi^2*rt_p^3*rpm/30)

With the characteristic curve gradient at the operative point parameter named: dpsi_dphi_op

6.2.3 Dynamic System Parameters

The following sections show the parameters and the functions needed to obtain the unsteady-state
parameters, that will be consequently exploited into the Simulink workspace to simulate a small
perturbation experiment.

6.2.3.1 Incompressible Duct Straight (ID-S)

Geometrical and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Duct length: L_sd

Inertance calculation: I_sd = L_sd/Au_sd
Resistance calculation: R_sd = k_sd*m0/(d_sd*Au_sd^2)

6.2.3.2 Incompressible Duct Elbow (ID-E)

Geometrical and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Duct length: L_ed

Duct area as a function of the 1-D space variable (linearly decreasing in the example):
A_ed = pi*(Du_ed^2)/4 - z*0.001
Inertance calculation: I_ed = vpa(int(1/A_ed,z,0,L_ed))
Resistance calculation: R_ed = k_ed*m0/(d_ed*Au_ed^2)

115
6.2.3.3 Incompressible Duct Tapered (ID-T)

Geometrical and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Duct length: L_td

Taper coefficient: tap_coeff
Duct area as a function of the 1-D space variable: A_td = Au_td - z*tap_coeff
Inertance calculation: I_td = vpa(int(1/A_td,z,0,L_td))
Resistance calculation: R_td = m0*(k_td/(d_td*Au_td^2)+1/(d_td*Ad_td^2)-1/(d_td*Au_td^2))

6.2.3.4 Compressible Duct Straight (CD-S)

Under steady-state conditions, the CD-S behaves as an ID-S, therefore, for the steady-state operative
point calculation, it can be replaced by a geometrical identical ID-S, for what concerns the steady-state
parameters. Under unsteady-state conditions, as previously stated, the situation is different and a
dedicated subsection to describe the unsteady-state parameters and functions of the CD-S has to be
considered.
Geometrical and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Unperturbed density of the working fluid: d_cd

Friction loss coefficient: fr
Duct length: L_cd
Duct section diameter: D_cd
Duct section area calculation: A_cd = 0.25*pi*D_cd^2
Frequency of perturbation: f
Acoustic velocity: a0
Steady-state axial velocity calculation: u0_cd = m0/(d_cd*A_cd)
Exponential coefficient calculation: k_cd = (k1_cd,k2_cd) =
k_function(a0,D_cd,fr,2*pi*f,u0_cd);

Where the k_function is represented by:

function k_cd = k_function(a0,D_cd,fr,2*pi*f,u0_cd)

k_cd = roots([1 (u0_cd*2*2*pi*f +sqrt(-1)*fr*u0_cd/D_cd)/(a0^2-u0_cd ^2)-...

((2*pi*f)^2+sqrt(-1)*2*pi*f*fr*(u0_cd/D_cd))/(a0^2-u0_cd ^2)]);

end

6.2.3.5 Silent Throttle Valve (STV) - Exciter

Experimental and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Experimentally characterized valve inertance: I_v

Resistance calculation: R_v = m0*(k_v/(d_v*Au_v^2)+1/(d_v*Ad_v^2)-1/(d_v*Au_v^2))
Forcing coefficient calculation: F_v = 0.5*m0^2/(d_v*Au_v^2)

116
6.2.3.6 Volume Oscillator Valve (VOV) - Exciter

Experimental and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition:

Experimentally characterized valve inertance: I_vov

Resistance calculation: R_vov =
m0*(k_vov/(d_vov*Au_vov^2)+1/(d_vov*Ad_vov^2)+1/(d_vov*Au_vov^2))
Forcing coefficient calculation: F_vov =
m0*(0.5*k_vov/(d_vov*Au_vov^2)+1/(d_vov*Ad_vov^2))+0.5*I_vov

6.2.3.7 Tank (T) - Exciter

Geometrical and steady-state derived parameters to input for the calculation of the unsteady-state
oscillating condition taking into account the exciting action:

Unperturbed density of the working fluid: d_t

Unperturbed gas volume: V_g_t
Heat specific ratio of the gas: hsr
Unperturbed gas/working fluid pressure: p_t
Upstream duct area: Au_t
Downstream duct area: Ad_t
Tank loss coefficient: k_t
Experimentally characterized tank inertance: I_t
Tank section area: A_t
Gravity acceleration: g
Tank water height: h_t
Tank water height component w.r.t. the gravity acceleration direction: h_t_r
Resistance calculation: R_t = m0*(k_t/(d_t*Au_t^2)+1/(d_t*Ad_t^2)-1/(d_t*Au_t^2))
Compliance calculation: C_t = V_g_t/(hsr*p_t)
Equivalent compliance calculation: C_t_eq = A_t*C_t/(A_t+d_t*C_t*(g-4*pi^2*f^2*h_t))

6.2.3.8 Pump (P)

To evaluate the momentum and continuity equation parameters and functions of the pump, a non-
cavitating configuration, as a first approximation of its dynamic behavior, have been exploited:

Resistance calculation: R_p = -dpsi_dphi_op*rpm/(30*rt_p)

Semi-Experimental Inertance calculation: I_p = 29/(pi*rt_p)

117
6.3 Appendix C, Simulink Modeling of the Hydraulic Loop
6.3.1 Introduction
This section will show how the parameters and functions developed in the previous section, have been
used into the dynamic modeling environment of Simulink, to simulate the oscillating behavior of the
hydraulic loop.

6.3.2 Simulink Environment

Simulink software models algorithms and physical systems using block diagrams which can be connected
by way of signal lines to establish mathematical relationships between system components. We designed
our models to be hierarchical by organizing groups of blocks into subsystems, enabling to build discrete
components that reflect our real-life system and simulate the interaction of those components. For each
one of the previous components, an appropriate blocks logic able to satisfy the continuity and momentum
equations has been built. Each subsystem is provided by two inputs and two outputs lines that
respectively represent the upstream pressure and mass flow rate signal and the downstream pressure and
mass flow rate one. In this way we can build our experimental facility by connecting the selected
components taking into account their relative positions and simulate the experiment.

6.3.2.1 Incompressible Duct Straight (ID-S)

A detailed view of the ID-S subsystem is showed in Figure 6.1:

Figure 6.1 Blocks logic of the ID-S

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks are
multipliers, the small rectangular one performs the time derivative and the big rectangular one is an adder.
We can notice how the perturbation of the mass flow rate goes straight to the output, given the

118
incompressibility of the duct, while the pressure one (output n°1) is obtained taking into account the
resistance and the inertance of the duct calculated in the workspace of Matlab.

6.3.2.2 Incompressible Duct Elbow (ID-E)

A detailed view of the ID-E subsystem is showed in Figure 6.2:

Figure 6.2 Blocks logic of the ID-E

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks are
multipliers, the small rectangular one performs the time derivative and the big rectangular one is an adder.
We can notice how the perturbation of the mass flow rate goes straight to the output, given the
incompressibility of the duct, while the pressure one (output n°1) is obtained taking into account the
resistance and the inertance of the duct calculated in the workspace of Matlab.

6.3.2.3 Incompressible Duct Tapered (ID-T)

A detailed view of the ID-T subsystem is showed in Figure 6.3:

:

119
 Figure 6.3 Blocks logic of the ID-T

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks are
multipliers, the small rectangular one performs the time derivative and the big rectangular one is an adder.
We can notice how the perturbation of the mass flow rate goes straight to the output, given the
incompressibility of the duct, while the pressure one (output n°1) is obtained taking into account the
resistance and the inertance of the duct calculated in the workspace of Matlab.

6.3.2.4 Compressible Duct Straight (CD-S)

A general view of the CD-S subsystem is showed in Figure 6.4:

Figure 6.4 Blocks logic of the CD-S

The compressible case, as can be noticed, is of a much more complex implementation than the
incompressible one, and deserves a deeper insight to be appreciated in Figure 6.5.

120
 Figure 6.5 Blocks logic of the ID-S, left subsystem

The previous scheme represents the two subsystems on the left of the global compressible duct system.
Since the wave solution in closed form has been obtained in the complex/frequency field and the input
signal is a real/time function, the input has to be manipulated in order to assume the appropriate complex
suit.
Mathematically we have an input signal of this type:

𝑅𝑒𝑎𝑙(|𝐴|𝑒 −𝑖𝜔𝑡 𝑒 𝑖𝜙 ) = |𝐴|cos(−𝜔𝑡 + 𝜙)

The Subsystem purpose is this:

𝑅𝑒𝑎𝑙(|𝐴|𝑒 −𝑖𝜔𝑡 𝑒 𝑖𝜙 )  |𝐴|𝑒 𝑖𝜙

The lower sub-subsystem, aim to extrapolate the magnitude |𝐴| of the input signal and, since it works for
complex quantity only, it manipulates the input signal adding it the same signal with the appropriate time
delay and multiplying it with the unity complex number, to obtain a complex quantity of the same
magnitude of the initial time dependent signal. Mathematically we have:

|𝐴|(𝑐𝑜𝑠(−𝜔𝑡 + 𝜙) + 𝑖𝑠𝑖𝑛(−𝜔𝑡 + 𝜙)) = |𝐴|𝑒 −𝑖𝜔𝑡 𝑒 𝑖𝜙

sin(−𝜔𝑡 + 𝜙) = 𝑐𝑜𝑠(−𝜔𝑡 − 𝜋/2 + 𝜙)

Solving for the time delay 𝑇0 :

𝜋 𝝅 𝟑𝝅 𝟑
𝑇0 > 0 : 𝑐𝑜𝑠(−𝜔𝑡 − + 𝜙) = 𝑐𝑜𝑠(−𝜔(𝑡 − 𝑇𝑜) + 𝜙)  𝑇0 = − => =
2 𝟐𝟂 𝟐𝟂 𝟒𝒇

The upper sub-subsystem, aim to extrapolate the phase of the input signal and again, since it works for
complex quantity only, it manipulates the input signal in the same way of the lower circuit, but ending as

121
the input of the Complex Phase Difference block, which, extrapolates the desired result comparing it with
the basic complex exponential generated from a sine wave generator as showed:
𝜋
𝑒 −𝑖𝜔𝑡 = cos(𝜔𝑡) − 𝑖 sin(𝜔𝑡) = 𝑠𝑖 𝑛 (𝜔𝑡 + ) − 𝑖 sin(𝜔𝑡)
2

The phase is then multiplied to the appropriate complex exponential and then to the previously
extrapolated magnitude |𝐴| to achieve the desired solution |𝐴|𝑒 𝑖𝜙 .

Figure 6.6 Blocks logic of the ID-S, central subsystem

The central subsystem represented in Figure 6.6 provides the complex solution of the wave equation for
pressure and velocity perturbation of the working fluid. Without dealing with too many programming
details, since each one of the blocks, with the exception of the big rectangular ones that represent sums or
products, performs a function created in the Matlab workspace coherently with the CD-S solution, it can
be highlighted that the upper part work on the complex pressure perturbation signal, while the lower part
manipulate the complex velocity perturbation signal.

122
 Figure 6.7 Blocks logic of the ID-S, right subsystem

The right subsystem presented in Figure 6.7 manipulates the complex downstream signal obtained from
the wave solution procedure, back-transforming it in an oscillating real signal by performing (pressure as
example, of course the same stands for the mass flow rate) the following operation:
𝑝̂𝑑 𝑒 −𝑖𝜔𝑡 and subsequently 𝑅𝑒𝑎𝑙(𝑝̂𝑑 𝑒 −𝑖𝜔𝑡 ).

6.3.2.5 Silent Throttle Valve (STV) - Exciter

A detailed view of the STV subsystem is showed in Figure 6.8:

Figure 6.8 Blocks logic of the STV

123
The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks act as
multipliers, the small rectangular one performs the time derivative and the big rectangular one is an adder.
The Silent Throttle Valve acts as a generator of perturbation, through the oscillation, at a given frequency,
of its loss coefficient. Starting from the upper part of the subsystem it can be noticed that the second line
is the only one of all the subsystems previously considered, to start with an imposed input. It represents
our 𝑘̃𝐿𝑂𝑆𝑆 met during the characterization of the unsteady-state behavior of the valve. When the STV will
act as a non-oscillating valve, i.e. when the other two exciters will be exploited, 𝑘̃𝐿𝑂𝑆𝑆 = 0, which in
Simulink language means that the generator will be set to the ground option.

6.3.2.6 Volume Oscillator Valve (VOV) - Exciter

A detailed view of the VOV subsystem is showed in Figure 6.9:

Figure 6.9 Blocks logic of the VOV

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks are
multipliers, the small left rectangular ones perform the time derivative and the right rectangular ones are
adders. We can notice how the perturbation of the downstream pressure is composed by four adding
components, with one of them taking into account the exciting contribution, as described by the
corresponding momentum equation (top adder), while the mass flow rate one (output n°2) is obtained
directly taking into account the contribution of the forcing action of the device when exploited as the
exciter of the system (down adder), embraced by a sinusoidal wave generator of mass flow rate.

124
6.3.2.7 Tank (T) - Exciter

A detailed view of the tank subsystem is showed in Figure 6.10:

Figure 6.10 Blocks logic of the tank

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
circuit embraces the unsteady-state momentum and continuity equations. The triangular blocks are
multipliers, the small left rectangular ones perform the time derivative and the big right rectangular ones
are adders. We can notice how the perturbation of the downstream pressure is composed by three adding
components as described by the corresponding momentum equation (top adder), while the mass flow rate
one (output n°2) is obtained taking into account the contribution of the forcing action of the device when
exploited as the exciter of the system (down adder), embraced by a sinusoidal wave generator of vertical
acceleration of the tank. When the tank won’t oscillate, i.e. when the other two exciters will be exploited,
𝑉𝑒𝑟𝑡𝑖𝑐𝑎𝑙 𝑎𝑐𝑐𝑒𝑙𝑒𝑟𝑎𝑡𝑖𝑜𝑛 = 0, which in Simulink language means that the generator will be set to the
ground option.

125
6.3.2.8 Pump (P)

A detailed view of the pump subsystem is showed in Figure 6.11. In our simulations has been used a non-
cavitating configuration, which can be represented for a 1st level configuration, as a particular
incompressible duct, with its own impedance (lower part), set in the Matlab code and dependent on the
steady-state characteristic curve of the inducer.

Figure 6.11 Blocks logic of the pump

The input n°1 represents an incoming oscillating pressure while the n°2 the mass flow rate one. The
triangular blocks are multipliers, the small rectangular one performs the time derivative and the big
rectangular one is an adder. The pump circuit embraces the unsteady-state momentum and continuity
equations transformed in a dimensional form.

126
6.4 Appendix D, Simulink Modeling for the Comparison of the
Incompressible Solution with the Compressible Solution
Chapter 4.2 shows the results and conclusions of the comparison between two geometrical identical ducts,
perturbed in the same way, exploiting the wave solution for the compressible case and the incompressible
solution for the incompressible one.
The logic behind the CD-S and the ID-S subsystems is showed in the previous appendix, this section aim
to show the interaction between them to achieve the desired results.

6.4.1 Downstream Mass Flow Rate and Pressure Signal Comparison Circuit
A 1st level view of the Simulink logic behind the results obtained in 4.2.1 is showed in Figure 6.12:

Figure 6.12 Circuit for the comparison between the amplitude and the phase of the downstream mass flow rate and
pressure signal of the ID-S and CD-S models

The two green blocks on the left embrace the already unveiled CD-S and ID-S logic; the four small blue
rectangles are sensors which show the time dependent behavior of the variable of interest clearly depicted
in the scheme; the two gray blocks perform the phase shift of the two input signals, which is visualized on
the screen in the rectangular display placed on the top left corner of the two right yellow rectangular
subsystems acting as sensors.
Figure 6.13 unveils the logic exploited by the Phase Shift block.

127
 Figure 6.13 Circuit for the calculation of the phase shift between the CD-S and ID-S signal

It can be noticed that the subsystem presents the same component of the left section of the Compressible
Duct scheme, concerning the extrapolation of the phase from a sinusoidal signal, while the right sector
performs the subtraction of the phase obtained from the two input signals.
The phase shift is calculated subtracting the phase of the Incompressible oscillating variable of interest
from the Compressible one.

6.4.2 Duct Transfer Matrix Comparison Circuit

A 1st level view of the Simulink logic behind the results obtained in 4.2.3 is showed in Figure 6.14:

Figure 6.14 Circuit for the comparison between the components of the dynamic transfer matrix of an ID-S and a CD-S

128
Exploiting the steady-state mass flow rate resulted from the first of the loop analysis showed in Error!
Reference source not found., the ID-S matrix is known, while for the CD-S, as we already know, we
need two sets of linearly independent data obtained from the two green blocks on the left embracing the
already unveiled CD-S logic, with linearly independent upstream pressure and mass flow rate
perturbation. The four grey rectangular blocks extract the complex amplitude of the corresponding signals
to obtain the well-known matrix equation of oscillating variables in complex notation. The right part of
the circuit represents the four matrix components (generally complex) of the CD-S, with the four big
yellow rectangular blocks that calculate the corresponding components and the eight other ones that show
the corresponding value of the real and imaginary parts. The yellow top-left subsystem performs the
following calculation to get the H11 component:
 pˆ u 
mˆ u1  pˆ d2  pˆ d1 2 
 pˆ u1 
pˆ d1  
pˆ u2

ˆ
mu2  mu1ˆ
pˆ u1
H11 
pˆ u1

The yellow top-right subsystem performs the following calculation to get the H12 component:

pˆ u2
pˆ d2  pˆ d1
pˆ u1
H12  pˆ u2
mˆ u2  mˆ u1
pˆ u1

The yellow down-left subsystem performs the following calculation to get the H21 component:

 pˆ u 
mˆ u1  mˆ d2  mˆ d1 2 
 pˆ u1 
mˆ d1  
pˆ u


mˆ u2  mˆ u1 2
pˆ u1
H 21 
pˆ u1

The yellow down-right subsystem performs the following calculation to get the H22 component:

pˆ u2
mˆ d2  mˆ d1
pˆ u1
H 22  pˆ u2
mˆ u2  mˆ u1
pˆ u1

129
6.5 Appendix E, Table of Figures
FIGURE 1.1 TYPICAL OCCURRENCE OF POGO VIBRATION (NASA, 1970) ................................................... 7
FIGURE 1.2 BLOCK DIAGRAM OF POGO FEEDBACK PROCESS (NASA, 1970) .............................................. 7
FIGURE 1.3 GEMINI-TITAN ............................................................................................................................ 8
FIGURE 1.4 COMPARISON OF GEMINI-TITAN POGO LEVELS (NASA, 1965) ............................................. 10
FIGURE 1.5 SATURN V FIRST STAGE S-IC .................................................................................................. 10
FIGURE 1.6 ILLUSTRATION OF SENSITIVITY TO SMALL CHANGES: COMPARISON OF AS-501 AND AS-502
(RYAN, ROBERT S., 1985) ................................................................................................................... 11
FIGURE 1.7 SATURN I-C POGO MITIGATION (VON BRAUN, WERNHER, 1975) ......................................... 12
FIGURE 1.8 SATURN V SECOND STAGE S-II CLOSE UP, J-2 ENGINE CLUSTER (DOIRON, HAROLD H., 2003)
............................................................................................................................................................ 12
FIGURE 1.9 APOLLO 8 AND 9 POGO EPISODES (FENWICK, J., 1992) .......................................................... 13
FIGURE 1.10 COMPARISON OF CENTER ENGINE THRUST PAD ACCELERATIONS (RYAN, ROBERT S., 1985)
............................................................................................................................................................ 14
FIGURE 1.11 POGO EXPERIENCED IN FLIGHT (RUBIN S., 1970) ................................................................ 14
FIGURE 1.12 CAVITATION COMPLIANCE OF TITAN STAGE 1 PUMPS (NASA, 1970) .................................. 15
FIGURE 1.13 PUMP GAIN FOR TITAN STAGE I PUMPS (NASA, 1970) ......................................................... 16
FIGURE 1.14 POGO SUPPRESSION DEVICES (NASA, 1970) ....................................................................... 16
FIGURE 1.15 SPACE SHUTTLE POGO SUPPRESSOR DEVICE ....................................................................... 17
FIGURE 1.16 LO2 FEED LINE AND BRANCH ACCUMULATOR (A.SWANSON AND T.GIEL, 2009) ............... 18
FIGURE 1.17 ANNULAR ACCUMULATORS (A.SWANSON AND T.GIEL, 2009) ............................................. 19
FIGURE 1.18 ACCUMULATOR COMPLIANCE VERSUS LIQUID LEVEL (A.SWANSON AND T.GIEL, 2009) .... 20
FIGURE 1.19 ACCUMULATOR INERTANCE VERSUS LIQUID LEVEL (A.SWANSON AND T.GIEL, 2009) ....... 20
FIGURE 2.1 SCHEMATIC DIAGRAM OF THE MATHEMATICAL PUMP LOOP (NG, 1976) ................................. 23
FIGURE 2.2 FUNCTIONAL SCHEMATIC OF THE FLUCTUATOR VALVE (NG, 1976) ........................................ 24
FIGURE 2.3 LEFT: THE [ZP] TRANSFER FUNCTION FOR IMPELLER IV IN THE VIRTUAL ABSENCE OF
CAVITATION. THE REAL AND IMAGINARY PARTS OF THE ELEMENTS (SOLID AND DASHED LINES,
RESPECTIVELY) ARE PLOTTED AGAINST BOTH THE ACTUAL AND THE NONDIMENSIONAL
FREQUENCIES; RIGHT: THE [ZP] TRANSFER FUNCTION FOR IMPELLER IV UNDER CONDITIONS OF
EXTENSIVE CAVITATION (NG AND BRENNEN, 1978)........................................................................... 25
FIGURE 2.4 THE [ZP] TRANSFER FUNCTION FOR IMPELLER IV UNDER CONDITIONS OF MODERATE
CAVITATION (NG AND BRENNEN, 1978) ............................................................................................. 25
FIGURE 2.5 POLYNOMIAL CURVE FITTING TO EXPERIMENTAL PUMP TRANSFER MATRICES, [ZP], OBTAINED
FOR IMPELLER IV AT 𝝋=0.070 AND A ROTATIONAL SPEED OF 9000 RPM. THE REAL AND IMAGINARY
PARTS OF THE MATRIX ELEMENTS ARE PRESENTED AS FUNCTIONS OF FREQUENCY BY SOLID AND
DASH LINES RESPECTIVELY. THE LETTERS A TO E DENOTE MATRICES TAKEN AT FIVES,
PROGRESSIVELY DIMINISHING CAVITATION NUMBERS, 𝝈, AS FOLLOWS: (A) 0.508 (B) 0.114 (C) 0.046
(D) 0.040 (E) 0.023 (NG AND BRENNEN, 1978) .................................................................................. 26
FIGURE 2.6 THEORETICAL PUMP TRANSFER MATRICES, [ZP], OBTAINED FOR IMPELLER IV AT 𝝋=0.070 AS
FUNCTIONS OF REDUCED FREQUENCY 𝟂. THE LETTERED CURVES ARE FOR DIFFERENT FRACTIONAL
LENGTHS, 𝜺, OF THE BUBBLY REGION AND CORRESPOND TO DECREASING CAVITATION NUMBERS, 𝝈:
(A) 𝜺=0.2 (B) 𝜺=0.4 (C) 𝜺=0.6 (D) 𝜺=0.8. THE CURVES ARE FOR ONE SPECIFIC CHOICE OF THE
PARAMETERS K AND M (SEE BRENNEN 1978) (NG AND BRENNEN, 1978) ......................................... 27
FIGURE 2.7 POLYNOMIAL CURVE FITS TO THE 10.2 CM IMPELLER TRANSFER MATRICES AT 𝝋=0.070, A
ROTATIONAL SPEED OF 6000 RPM AND VARIOUS CAVITATION NUMBERS AS FOLLOWS: (A) 0.37 (C)
0.10 (D) 0.069 (G) 0.052 (H) 0.044. THE REAL AND IMAGINARY PARTS OF THE MATRIX ELEMENTS
ARE PRESENTED AS FUNCTIONS OF FREQUENCY BY SOLID AND DASHED LINES RESPECTIVELY. THE
QUASISTATIC RESISTANCE FROM THE SLOPE IS INDICATED BY THE ARROW (BRENNEN ET AL., 1982)
............................................................................................................................................................ 27

130
FIGURE 2.8 TRANSFER FUNCTIONS CALCULATED FROM THE COMPLETE BUBBLY FLOW MODEL WITH
𝝋=0.07, 𝝲=9 DEG, 𝝉=0.45, F=1.0, K=1.3 AND M=0.8. VARIOUS CAVITATION NUMBERS ACCORDING
TO 𝜺=0.2/𝝈 ARE SHOWN (BRENNEN ET AL., 1982) .............................................................................. 28
FIGURE 2.9 SCHEMATIC REPRESENTATION OF THE BASIC SYSTEM AND ITS NOMENCLATURE
(A.STIRNEMANN ET AL., 1987) ........................................................................................................... 28
FIGURE 2.10 SCHEMATIC REPRESENTATION OF THE EXPERIMENTAL FACILITY (A.STIRNEMANN ET AL.,
1987) ................................................................................................................................................... 30
FIGURE 2.11 MASS GAIN OBTAINED FROM EXCITATION ON THE SUCTION SIDE (A.STIRNEMANN ET AL.,
1987) ................................................................................................................................................... 30
FIGURE 2.12 TEST LOOP FOR MEASURING THE DYNAMIC BEHAVIOUR OF THE PROTOTYPE MULTI-STAGE
PUMP (KAWATA ET AL., 1988) ............................................................................................................ 31
FIGURE 2.13 CORRECTED TRANSFER MATRIX OF THE PROTOTYPE MULTI-STAGE PUMP (KAWATA ET AL.,
1988) ................................................................................................................................................... 31
FIGURE 2.14 LEFT: COMPLIANCE LAG DATA AND FITS; RIGHT: COMPLIANCE DATA AND FITS (RUBIN S.,
2004) ................................................................................................................................................... 33
FIGURE 2.15 LEFT: RESISTANCE DATA AND FITS; RIGHT: INERTANCE DATA AND FITS (RUBIN S., 2004) .. 34
FIGURE 2.16 PUMP GAIN DATA AND FITS (RUBIN S., 2004) ........................................................................ 34
FIGURE 2.17 SCHEMATIC OF THE EXPERIMENTAL FACILITY (CERVONE ET AL., 2009)............................... 36
FIGURE 2.18 TOP VIEW OF THE CAVITATING PUMP ROTORDYNAMIC TEST FACILITY (CERVONE ET AL.,
2010) ................................................................................................................................................... 37
FIGURE 2.19 SUGGESTED MODIFICATION TO THE ORIGINAL FACILITY SETUP FOR OBTAINING THE SECOND
LINEARLY INDEPENDENT TEST CONFIGURATION (ADDED PIPE LINES ARE COLOURED IN RED)
(CERVONE ET AL., 2010) ..................................................................................................................... 37
FIGURE 2.20 THE “SHORT” (TOP) AND “LONG” (BOTTOM) CONFIGURATIONS OF THE TEST LOOP USED FOR
THE EXPERIMENTAL CHARACTERIZATION OF THE DYNAMIC TRANSFER MATRIX OF CAVITATING
INDUCERS (PACE G. ET AL., 2013)....................................................................................................... 38
FIGURE 2.21 DYNAMIC MATRIX FOR DAPAMITOR3 INDUCER: EXPERIMENTAL POINTS ARE IN SQUARES
AND THE POINTS OBTAINED BY USING THE MODEL ARE IN STAR (PACE G. ET AL., 2013)................... 39
FIGURE 3.1 SCHEMATIC VIEW OF THE HYDRAULIC LOOP SYSTEM (TORRE ET AL., (12)) ............................ 41
FIGURE 4.1 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 49
FIGURE 4.2 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 49
FIGURE 4.3 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 50
FIGURE 4.4 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 50
FIGURE 4.5 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 51
FIGURE 4.6 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 51
FIGURE 4.7 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 52
FIGURE 4.8 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 52
FIGURE 4.9 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 53

131
FIGURE 4.10 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 53
FIGURE 4.11 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 54
FIGURE 4.12 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 54
FIGURE 4.13 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 55
FIGURE 4.14 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 55
FIGURE 4.15 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 56
FIGURE 4.16 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 56
FIGURE 4.17 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 57
FIGURE 4.18 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 57
FIGURE 4.19 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 58
FIGURE 4.20 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 58
FIGURE 4.21 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 59
FIGURE 4.22 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 59
FIGURE 4.23 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 60
FIGURE 4.24 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 60
FIGURE 4.25 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 61
FIGURE 4.26 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 61
FIGURE 4.27 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 62
FIGURE 4.28 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 62
FIGURE 4.29 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 63

132
FIGURE 4.30 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 63
FIGURE 4.31 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 64
FIGURE 4.32 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 64
FIGURE 4.33 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 65
FIGURE 4.34 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 65
FIGURE 4.35 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 66
FIGURE 4.36 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 66
FIGURE 4.37 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 67
FIGURE 4.38 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 67
FIGURE 4.39 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 68
FIGURE 4.40 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 68
FIGURE 4.41 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 69
FIGURE 4.42 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 69
FIGURE 4.43 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 70
FIGURE 4.44 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 70
FIGURE 4.45 PRESSURE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 71
FIGURE 4.46 MASS FLOW RATE AMPLITUDE COMPARISON BETWEEN THE COMPRESSIBLE AND THE
INCOMPRESSIBLE SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND
DUCT LENGTH ...................................................................................................................................... 71
FIGURE 4.47 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE PRESSURE SOLUTION
VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH .................... 72
FIGURE 4.48 PHASE SHIFT BETWEEN THE COMPRESSIBLE AND THE INCOMPRESSIBLE MASS FLOW RATE
SOLUTION VS FREQUENCY OF PERTURBATION AT FIXED ACOUSTIC VELOCITY AND DUCT LENGTH ... 72
FIGURE 4.49 COMPARISON BETWEEN THE REAL COMPONENTS OF THE PRINCIPAL DIAGONAL OF THE
MATRIX OF A 3 METERS LONG ID-S (BLACK LINE) WITH A GEOMETRICALLY IDENTICAL CD-S AT
DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000 M/S (BLUE); A=700 M/S (CYAN); A=500

133
 M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION
PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ............................... 74
FIGURE 4.50 COMPARISON BETWEEN THE IMAGINARY COMPONENTS OF THE PRINCIPAL DIAGONAL OF THE
MATRIX OF A 3 METERS LONG ID-S (BLACK LINE) WITH A GEOMETRICALLY IDENTICAL CD-S AT
DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000 M/S (BLUE); A=700 M/S (CYAN); A=500
M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION
PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ............................... 74
FIGURE 4.51 COMPARISON BETWEEN THE RE[H12] OF A 3 METERS LONG ID-S (BLACK LINE) WITH A
GEOMETRICALLY IDENTICAL CD-S AT DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000
M/S (BLUE); A=700 M/S (CYAN); A=500 M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS
RESULTING FROM THE SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL
FITTING CURVES. ................................................................................................................................. 75
FIGURE 4.52 COMPARISON BETWEEN THE IM[H12] OF A 3 METERS LONG ID-S (BLACK LINE) WITH A
GEOMETRICALLY IDENTICAL CD-S AT DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000
M/S (BLUE); A=700 M/S (CYAN); A=500 M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS
RESULTING FROM THE SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL
FITTING CURVES. ................................................................................................................................. 75
FIGURE 4.53 COMPARISON BETWEEN THE RE[H21] OF A 3 METERS LONG ID-S (BLACK LINE) WITH A
GEOMETRICALLY IDENTICAL CD-S AT DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000
M/S (BLUE); A=700 M/S (CYAN); A=500 M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS
RESULTING FROM THE SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL
FITTING CURVES. ................................................................................................................................. 76
FIGURE 4.54 COMPARISON BETWEEN THE IM[H21] OF A 3 METERS LONG ID-S (BLACK LINE) WITH A
GEOMETRICALLY IDENTICAL CD-S AT DIFFERENT ACOUSTIC VELOCITY: A=1400 M/S (RED); A=1000
M/S (BLUE); A=700 M/S (CYAN); A=500 M/S (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS
RESULTING FROM THE SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL
FITTING CURVES. ................................................................................................................................. 76
FIGURE 4.55 HYDRAULIC LOOP LOGIC BY SIMULINK ................................................................................. 78
FIGURE 4.56 COMPARISON BETWEEN THE PRESSURE AMPLITUDES OF OSCILLATION AT DIFFERENT VALUES
OF THE STV UNSTEADY-STATE LOSS COEFFICIENT: K=KV/100 (RED); K=KV/25 (BLUE); K=KV/5
(CYAN); K=KV/2.5 (YELLOW) WHERE KV IS THE STEADY-STATE STV LOSS COEFFICIENT. THE
ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION PROCESS, THE LINES
ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ................................................................ 78
FIGURE 4.57 COMPARISON BETWEEN THE MASS FLOW RATE AMPLITUDES OF OSCILLATION AT DIFFERENT
VALUES OF THE STV UNSTEADY-STATE LOSS COEFFICIENT: K=KV/100 (RED); K=KV/25 (BLUE);
K=KV/5 (CYAN); K=KV/2.5 (YELLOW) WHERE KV IS THE STEADY-STATE STV LOSS COEFFICIENT.
THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION PROCESS, THE
LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ...................................................... 79
FIGURE 4.58 COMPARISON BETWEEN THE PRESSURE AMPLITUDES OF OSCILLATION AT DIFFERENT VALUES
OF MAXIMUM VERTICAL DISPLACEMENT OF THE TANK: R=0.2 MM (RED); R=1 MM (BLUE); R=5 MM
(CYAN); R=1 CM (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE
SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. .......... 80
FIGURE 4.59 COMPARISON BETWEEN THE MASS FLOW RATE AMPLITUDES OF OSCILLATION AT DIFFERENT
VALUES OF MAXIMUM VERTICAL DISPLACEMENT OF THE TANK: R=0.2 MM (RED); R=1 MM (BLUE);
R=5 MM (CYAN); R=1 CM (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM
THE SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ... 80
FIGURE 4.60 HYDRAULIC LOOP LOGIC BY SIMULINK WITH THE VOV (RED) PLACED DOWNSTREAM OF THE
TANK (BLUE) ....................................................................................................................................... 81
FIGURE 4.61 COMPARISON BETWEEN THE PRESSURE AMPLITUDES OF OSCILLATION AT DIFFERENT VALUES
OF THE VOV MASS FLOW RATE AMPLITUDE OF OSCILLATION (IN PERCENTAGE OF THE STEADY-
STATE MASS FLOW RATE): MVOV=0.25% (RED); MVOV=1% (BLUE); MVOV=5% (CYAN); MVOV=10%

134
 (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION
PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ............................... 81
FIGURE 4.62 COMPARISON BETWEEN THE MASS FLOW RATE AMPLITUDES OF OSCILLATION AT DIFFERENT
VALUES OF THE VOV MASS FLOW RATE AMPLITUDE OF OSCILLATION (IN PERCENTAGE OF THE
STEADY-STATE MASS FLOW RATE): MVOV=0.25% (RED); MVOV=1% (BLUE); MVOV=5% (CYAN);
MVOV=10% (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE
SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. .......... 82
FIGURE 4.63 HYDRAULIC LOOP LOGIC BY SIMULINK WITH THE VOV (RED) PLACED DOWNSTREAM OF THE
PUMP (ORANGE)................................................................................................................................... 82
FIGURE 4.64 COMPARISON BETWEEN THE PRESSURE AMPLITUDES OF OSCILLATION AT DIFFERENT VALUES
OF THE VOV MASS FLOW RATE AMPLITUDE OF OSCILLATION (IN PERCENTAGE OF THE STEADY-
STATE MASS FLOW RATE): MVOV=0.25% (RED); MVOV=1% (BLUE); MVOV=5% (CYAN); MVOV=10%
(YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE SIMULATION
PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. ............................... 83
FIGURE 4.65 COMPARISON BETWEEN THE MASS FLOW RATE AMPLITUDES OF OSCILLATION AT DIFFERENT
VALUES OF THE VOV MASS FLOW RATE AMPLITUDE OF OSCILLATION (IN PERCENTAGE OF THE
STEADY-STATE MASS FLOW RATE): MVOV=0.25% (RED); MVOV=1% (BLUE); MVOV=5% (CYAN);
MVOV=10% (YELLOW). THE ASTERISKS REPRESENT THE DATA POINTS RESULTING FROM THE
SIMULATION PROCESS, THE LINES ARE THE CORRESPONDING POLYNOMIAL FITTING CURVES. .......... 83
FIGURE 6.1 BLOCKS LOGIC OF THE ID-S ................................................................................................... 118
FIGURE 6.2 BLOCKS LOGIC OF THE ID-E ................................................................................................... 119
FIGURE 6.3 BLOCKS LOGIC OF THE ID-T ................................................................................................... 120
FIGURE 6.4 BLOCKS LOGIC OF THE CD-S ................................................................................................. 120
FIGURE 6.5 BLOCKS LOGIC OF THE ID-S, LEFT SUBSYSTEM ..................................................................... 121
FIGURE 6.6 BLOCKS LOGIC OF THE ID-S, CENTRAL SUBSYSTEM .............................................................. 122
FIGURE 6.7 BLOCKS LOGIC OF THE ID-S, RIGHT SUBSYSTEM ................................................................... 123
FIGURE 6.8 BLOCKS LOGIC OF THE STV ................................................................................................... 123
FIGURE 6.9 BLOCKS LOGIC OF THE VOV .................................................................................................. 124
FIGURE 6.10 BLOCKS LOGIC OF THE TANK................................................................................................ 125
FIGURE 6.11 BLOCKS LOGIC OF THE PUMP ................................................................................................ 126
FIGURE 6.12 CIRCUIT FOR THE COMPARISON BETWEEN THE AMPLITUDE AND THE PHASE OF THE
DOWNSTREAM MASS FLOW RATE AND PRESSURE SIGNAL OF THE ID-S AND CD-S MODELS ............ 127
FIGURE 6.13 CIRCUIT FOR THE CALCULATION OF THE PHASE SHIFT BETWEEN THE CD-S AND ID-S SIGNAL
.......................................................................................................................................................... 128
FIGURE 6.14 CIRCUIT FOR THE COMPARISON BETWEEN THE COMPONENTS OF THE DYNAMIC TRANSFER
MATRIX OF AN ID-S AND A CD-S ...................................................................................................... 128

135
6.6 Appendix F, List of Tables
TABLE 1 VALUES OF FIT PARAMETERS AND REFERENCES TO EQUATIONS AND FIGURES. DUAL VALUES FOR
FLOW GAIN INDICATE UNCERTAINTY (RUBIN S., 2004) ...................................................................... 35

136
Reference Documents
 Hartmann M. J., Soltis R. F., 1960, “Observation of Cavitation in a Low Hub-Tip Ratio Axial
Flow Pump”, Proc. Gas Turbine Power and Hydraulic Conf., ASME paper n. 60-HYD-14.
 NASA, 1965, “Gemini Program Mission Report: Gemini V” MSC-G-R-65-4, Manned Spacecraft
Center, Houston, Tx.
 Rosenmann W., 1965, Experimental Investigations of Hydrodynamically Induced Shaft Forces
with a Three Bladed Inducer, Proc. ASME Symp. on Cavitation in Fluid Machinery, pp. 172-195.
 Sack L.E. and Nottage H.B., 1965, System Oscillations Associated to Cavitating Inducers, ASME
J. Basic Eng., Vol. 87, pp. 917-924.
 Badowski H. R., 1969, “An Explanation for Instability in Cavitating Inducers”, ASME
Cavitation Forum, pp. 38-40.
 NASA SP-8055, 1970, “Prevention of Coupled Structure-Propulsion Instability”.
 Rubin, S., 1970, “Prevention of Coupled Structure-Propulsion Instability (Pogo) on the Space
Shuttle”, Space Transportation System Technology Symposium (Cleveland), Vol. II – Dynamics
and Aeroelasticity. pp. 249-262, NASA TM X-52876.
 Natanzon M.S. el al., 1974, Experimental Investigation of Cavitation Induced Oscillations of
Helical Inducers, Fluid Mech. Soviet Res., Vol. 3 No. 1, pp.38-45.
 Brennen C.E. and Acosta A.J., 1975, “The Dynamic Performance of Cavitating Turbopumps”,
California Institute of Technology, Pasadena, California, USA, 91125.
 Von Braun, Wernher, 1975, “Saturn the Giant: Building the boosters to send men beyond Earth”,
Chapter 3, Apollo Expeditions to the Moon. NASA SP-350, Edgar M. Cortright, Editor.
 Brennen C.E. and Acosta A.J., 1976, The Dynamic Transfer Function for a Cavitating Inducer,
ASME J. Fluids Eng., Vol. 98, pp. 182-191.
 Ng S.L., 1976, “Dynamic Response of Cavitating Turbomachines”, Report No. E183.1,
California Institute of Technology.
 Jennrich, R. I., “Statistical Methods for Digital Computers”, Edited by K. Enslin, A. Ralston, and
H. S. Wilf, Wiley, New York, 1977.
 Kamjio K., Shimura T., Watanabe M., 1977, “An Experimental Investigation of Cavitating
Inducer Instability”, ASME Paper n. 77-WA/FW-14.
 Brennen C. E., 1978, “Bubbly Flow Model for the Dynamic Characteristics of Cavitating
Pumps”, Journal of Fluid Mechanics, 89, part 2, pp. 223-240.
 Ng S.L. and Brennen C.E., 1978, Experiments on the Dynamic Behavior of Cavitating Pumps,
ASME J. Fluids Eng., Vol. 100, No. 2, pp. 166-176.
 Braisted D.M. and Brennen C.E., 1980, Auto-oscillation of Cavitating Inducers, Polyphase Flow
and Transport Technology, ed. R.A. Bajura, ASME Publ., New York, pp. 157-166.
 Brennen C.E., Meissner C., Lo E.Y. and Hoffman G.S., 1982, “Scale Effects in the Dynamic
Transfer Functions for Cavitating Inducers”, Journal of Fluid Engineering, Vol.104.
 Ryan, Robert S., 1985, “Problems Experienced and Envisioned for Dynamical Physical Systems”,
NASA TP-2508.
 Stirnemann A., Eberl J., Bolleter U., Pace S., 1987, “Experimental Determination of the Dynamic
Transfer Matrix for a Pump”, Journal of Fluid Engineering, 109(3), 218-225, Sep. 01.
 Kawata Y., Takata T., Yasuda O., Takeuchi T., 1988, “Measurement of the Transfer Matrix of a
Prototype Multi-Stage Centrifugal Pump”, paper no. C346/88, IMechE, pp. 137-142.
 Yamamoto K., 1991, “Instability in a Cavitating Centrifugal Pump”, JSME Int. J., Ser. II, 34, 9-
17.
 Fenwick, J., Spring 1992, “POGO” Threshold: Pratt & Whitney Rocketdyne’s Engineering
Journal of Power Technology.

137
 Oppenheim, B.W., and Rubin, S., 1993, “Advanced Pogo Stability Analysis for Liquid Rockets”,
Journal of Spacecraft and Rockets, Vol. 30, No. 3.
 Tsujjimoto Y., Kamijo K., Yoshida Y., 1993, “A Theoretical Analysis of Rotating Cavitation in
Inducers”, Journal of Fluids Engineering, Vol. 115.
 Bhattacharyya A., 1994, Internal Flows and Force Matrices in Axial Flow Inducers, Ph. D.
thesis, Division of Engineering and Applied Science, California Institute of Technology,
Pasadena, California, USA.
 Brennen C. E., 1994, Hydrodynamics of Pumps, Concepts ETI, Inc., P.O. Box 643, Norwich, VT,
USA 05055.
 Otsuka S., Tsujimoto Y., Kamijo K., Furuya O., 1996, “Frequency Dependence of Mass Flow
Gain Factor and Cavitation Compliance of Cavitating Inducers”, ASME Journal of Fluids
Engineering, 118, pp. 400-408.
 Hashimoto T., Yamada H., Funatsu S., Ishimoto J., Kamijo K. and Tsujimoto Y., 1997, Rotating
Cavitation in Three and Four-Bladed Inducers, AIAA Paper 97-3026, 33rd
AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Seattle, Washington, USA,
July 6-9.
 Hashimoto T., Yamada H., Watanabe M., Kamijo K. and Tsujimoto Y., 1997b, Experimental
Study on Rotating Cavitation of Rocket Propellant Pump Inducers, J. Propulsion Propulsion and
Power, Vol. 13, No. 4, pp. 488-494.
 Tsujimoto Y., Watanabe S., Horiguchi H., 1998, “Linear Analyses of Cavitation Instabilities of
Hydrofoils and Cascades”, Proc. of US-Japan Seminar: Abnormal Flow Phenomena in
Turbomachinery, Osaka, Japan, Nov. 1-6.
 Watanabe S., Yokota K., Tsujimoto Y., Kamijo K., 1999, “Three-Dimensional Linear Analysis of
Rotating Cavitation in Inducers Using an Annular Cascade Model”, ASME Journal of Fluids
Engineering, Vol. 121, pp. 866-871.
 Zoladz T., 2000, Observations on Rotating Cavitation and Cavitation Surge from the
Development of the Fastrac Engine Turbopump, 36th AIAA/ASME/SAE/ASEE Joint Propulsion
Conference, Huntsville, AL, USA, July 17-19.
 Tsujimoto Y., 2001, Simple Rules for Cavitation Instabilities in Turbomachinery, lecture from
CAV 2001, 4th International Symposium on Cavitation, Caltech, Pasadena, CA, USA, 20-23 June
20-23.
 Furukawa A., Ishizaka K., Watanabe S., 2002, “Experimental Study of Cavitation Induced
Oscillation in Two Bladed Inducers”, Space Launcher Liquid Propulsion: 4th International
Conference on Space Launcher Technology, Liege, Belgium.
 Doiron, Harold H., 2003, “Pogo Prevention: A Comprehensive Guide to Theory and Practice”,
InDyne, Inc. Report No. IDI-H3588D-03-01R1.
 Enomoto N., Kim J. H., Ishizaka K., Watanabe S., Furukawa A., 2003, “Suppression of
Cavitation Surge of a Helical Inducer Occurring in Partial Flow Conditions”, CAV2003, Int.
Symposium on Cavitation, Osaka, Japan.
 Horiguchi H., Semenov Y., Nakano M., Tsujimoto Y., 2003, “Linear Stability Analysis of the
Effects of Camber and Blade Thickness on Cavitation Instabilities in Inducers”, CAV2003, Int.
Symposium on Cavitation, Osaka, Japan.
 Imamura H., Kurokawa J., Matsui J., Kikuchi M., 2003, “Suppression of Cavitating Flow in
Inducer by J-Groove”, CAV2003, Int. Symposium on Cavitation, Osaka, Japan.
 Shimura T., Shimagaki M., Watanabe Y., Watanabe M., Hasegawa S., Takata S., 2003,
“Cavitation Induced Vibration of LE-7A Oxygen Turbopump”, CAV2003, Int. Symposium on
Cavitation, Osaka, Japan, November 1-4.
 Rubin S., 2004, “An Interpretation of Transfer Function Data for a Cavitating Pump”, Proc. of
40th AIAA / ASME / SAE / ASEE Joint Propulsion Conference, Fort Lauderdale, USA.

138
 Kim J. H., Ishizaki M., Ishizaka K., Watanabe S., Furukawa A., 2006, “Suppression of Cavitation
Surge of Pump Inducer by Inserting Ring-Shaped Inlet Plate”, CAV2006, 6th Int. Symposium on
Cavitation, Wageningen, The Netherlands.
 Kimura T., Yoshida Y., Hashimoto T., Shimagaki M., 2006, “Numerical Simulation for Unsteady
Cavitating Flow in a Turbopump Inducer”, CAV2006, 6th Int. Symposium on Cavitation,
Wageningen, The Netherlands.
 Shimiya N., Fujii A., Horiguchi H., Uchiumi M., Kurokawa J., Tsujimoto Y., 2006, “Suppression
of Cavitation Instabilities in an Inducer by J-Groove”, CAV2006, 6th Int. Symposium on
Cavitation, Wageningen, The Netherlands.
 Kang D., Cervone A., Yonezawa K., Horiguchi H., Kawata Y., Tsujimoto Y., 2007, “Effect of
Blade Geometry on Tip Leakage Vortex of Inducer”, The 9th Asian International Conference on
Fluid Machinery, Jeju, South Korea.
 Larsen C. E., 2008, “NASA Experience with Pogo in Human Spaceflight Vehicles”, NATO RTO
Symposium ATV-152 on Limit-Cycle Oscillations and Other Amplitude-Limited, Self-Excited
Vibrations, Norway.
 Ohta T., Kajishima T., 2008, “Transition of Different Unsteady Cavitating Flows in 2D Cascade
with Flat Blades”, Proceedings of ISROMAC-12 - the 12th International Symposium on
Transport Phenomena and Dynamics of Rotating Machinery, Honolulu, Hawaii, USA.
 Sargent S. R., Sorensen K. P., McGlynn R. D., 2008, “Measurement of Cavitation Induced Strain
on a 3 Bladed Inducer”, Proceedings of ISROMAC-12 - the 12th International Symposium on
Transport Phenomena and Dynamics of Rotating Machinery, Honolulu, Hawaii, USA.
 Cervone A., Tsujimoto Y., Kawata Y., 2009, “Evaluation of the Dynamic Transfer Matrix of
Cavitating Inducers by Means of a Simplified Lumped-Parameter Model”, ASME Journal of
Fluids Engineering, Vol. 131, Is. 4.
 Cervone A., Torre L., Pasini A., d’Agostino L., 2009b, “Cavitation and Flow Instabilities in a 3-
Bladed Axial Inducer Designed by Means of a Reduced Order Analytical Model”, CAV2009, 7th
Int. Symposium on Cavitation, Ann Arbor, Michigan, USA.
 Iga Y., Yoshida Y., 2009, “A Study of Propagating Speed of Rotating Cavitation Based on
Numerical Analysis”, FEDSM2009-78083, ASME 2009 Fluids Engineering Division Summer
Meeting, Vail, Colorado, USA.
 Swanson A., Giel V., 2009, “Dynamic Analysis of the Ares I POGO Accumulator”, AIAA Paper
2009-4950, 45th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, Denver,
Colorado, USA, August 2-5.
 Cervone A., Piccoli E. A., Torre L., Pasini A., d'Agostino L., 2010, “Design of a Test Setup for
the Characterization of the Dynamic Transfer Matrix of Cavitating Inducers”, 46th
AIAA/ASME/SAE/ASEE Joint Propulsion Conference, Nashville, Tennessee, USA.
 Torre L., Pace G., Miloro P., Pasini A., Cervone A., d'Agostino L., 2010, “Flow Instabilities on a
Three Bladed Axial Inducer at Variable Tip Clearance”, Proceedings of ISROMAC-13 - the 13th
International Symposium on Transport Phenomena and Dynamics of Rotating Machinery,
Honolulu, Hawaii, USA.
 L. Torre, A. Cervone, A. Pasini, L. d’Agostino, Nov. 2011, “Experimental Characterization of
Thermal Cavitation Effects on Space Rocket Axial Inducers”, ASME Journal of Fluids
Engineering, Vol. 133, Is. 11, (doi: 10.1115/1.4005257) (ISSN: 0098-2202).
 Pace G., Torre L., Pasini A., Valentini D., d'Agostino L., 2013, “Experimental Characterization
of the Dynamic Transfer Matrix of Cavitating Inducers”, 49th AIAA/ASME/SAE/ASEE Joint
Propulsion Conference & Exhibit, San Jose Convention Center, San Jose, California, USA, July
15-17.

139
 Yamamoto K., Müller A., Ashida T., Yonezawa K., Avellan F. and Tsujimoto Y., 2015,
“Experimental method for the evaluation of the dynamic transfer matrix using pressure
transducers”, Journal of Hydraulic Research.

140
Acknowledgements (Ringraziamenti)
I’d like to express my gratitude to prof. Zoltan Spakovszky for his support and his welcome during the
project meetings occurred in Cambridge, Boston.
Special thanks go to prof. Angelo Pasini for his constant support and patience throughout the period of
research.
Il ringraziamento più importante, però, non può che essere riservato ai miei genitori, Anna e Vincenzo,
mio zio Antonio e mia sorella Alessia, per il loro costante e incondizionato supporto, senza il quale nulla
di quanto fatto sarebbe stato possibile.

141