https://ntrs.nasa.gov/search.jsp?

R=19770003597 2018-03-28T01:11:03+00:00Z

(::AS %-C?-150036) DYNAMIC BESPONSE OF N77- I05_0

C_.VITATING TUR_CHACHINZS (California Inst.
of Tech.) 203 p HC AIO/gF A01 CSCL 131
Unclas
G3/37 000o5
 Contract NAS 8-29313

Dynamic Response of
Cavitating Turbomachine s

by

Sheung- Lip Ng

Report No. E 183.1

Division of Engineering and Applied Science

California Institute of Technclogy

August 1976
 -ii-

ACKNOWLEDGMEN TS

The present work is a report of research on the dynamics of

cavitating pumps carried out mainly with the support of the National

Aeronautics and Space Administration under contracts NAS 8-2-9313

and NAS 8-28046 from the George Marshall Space Flight Center,

Huntsville, Alabama, administered by Mr. Loren Gross. In addition

there was support from the Office of Naval Research under contract

N00014-76-C-0156, and additional grants from the Ford Foundation and

Transwestern Pipe. All this support is gratefully acknowledged.

The present report is based on a PhD Thesis (California Institute

of Technology 1976) having the same title. Many individuals contributed

their help and advice. I would like to acknowledge them here.

Firstly, I would like to express my most sincere thanks to my

advisor, Dr. A.J. Acosta, for giving me the opportunity to work on this

project and for his constant guidance, encouragement and patience.

I also wish to thank Dr. C. Brennen for his close and indis-

pensable assistance in the research. To Dr. R. Sabersky goes my deep-

est appreciation for his concern during my stay here.

For their technical consultations on the design of the facility,

the help of Professor D. Welch, Dr. H. Shapiro and Mr. R. Rigley are

acknowledged. I am grateful to Messrs. N. Keidel, E. Daly, V. Sodha

and D. Laird for their technical assistance in the construction and as-

sembly of the system.

Thanks go to Messrs. E. Adler, D. Braisted, C. Constantin,

D. Carlson, J. DelValle, P. Mason and R. Higley for their valuable

help in various phases of the experiment.

Special thanks are due Ms. B. Hawk, J. Sarkissian and M. Suggs

in preparing this manuscript.

Finally, heartfelt thanks go to my parents for their extreme

patience and understanding.

 - iii-

ABSTRACT

Stimulated by the pogo instability encountered in many liquid

propellant rockets, the dynamic behavior of cavitating inducers is the

subject of the present thesis. An experimental facility where the up-

stream and downstream flows of a cavitating inducer could be perturbed

was constructed and tested. The upstream and downstream pressure

and mass flow fluctuations were measured. Matrices representing the

transfer functions across the inducer pump were calculated from these

measurements and from the hydraulic system characteristics for two

impellers in various states of cavitation. The transfer matrices when

plotted against the perturbing frequency showed significant departure

from steady state or quasi-steady predictions especially at higher

frequencies.
 -iv-

TABLE OF CONTENTS

Page

AC KN OW LE D GMENTS ii

ABSTRACT iii

TABLE OF CONTENTS iv

NOM ENC LAT URE vi

LIST OF FIGURES viii

I. INTRODUCTION 1

1.1 Background 1

1.2 Pogo Instability 2

1.3 Cavitation in an Inducer Pump 5

1.4 Lumped Parameter Model of a Turbomachine 7

1.5 Experimental Work I1

1.6 Pump Transfer Matrices 14

1.7 Goals of the Research 19

If. SYSTEM DESIGN Z4

Z. i Design Requirements and Problems Z4

2.2 Preliminary Experimental Setup 27

Z. 3 Analytical Model 30

2.4 Final Design of the Dynamic Pump Test Facility 34

Z. 5 Instrumentation 46

III. E XPERIMENTA L PROCEDUR E 86

3. 1 Steady State Tests 86

3. z Fluctuating Experiments 87

3.3 Impelle r .V Experiments 9Z

 -V-

IV. DATA REDUCTION 95

4. I RAWDATA Program 95

4. 2 TRANSCO Program 98

4.3 PUMA Program 99

V, PERF ORMAN CE I01

5. i Steady State Performance of the Inducers I01

I02
5. 2 Fluctuating Quantities

5. 3 Auto Oscillation 104

5.4 Pump Speed Variation 105

5.5 Pressure Drop Across the Smoothing Chamber 105

5.6 Accuracy of the Results I06

5.7 Linearly Independent Data Sets 108

5.8 Linear Transfer Matrix Assumption Test II0

VI. RESULTS AND DISCUSSIONS lZ9

6. I System Effects 129

6.2 Fully-wetted Experiments 13Z

140
6.3 Cavitating Experiments

VII. SUMMARY AND CONCLUSION 166

REFERENCES 169

APPENDIX A. Laser Doppler Velocimeter 172

APPENDIX B. Calibration of Mass Flow 176

Measurement System

APPENDIX C. Flow Charts of Computer Programs 181

APPENDIX D. Special Least Square Fit of the 190

Z-Matrix Calculation
 -vi-

NOMENC LATURE

A, inlet area
1

C compliance C =--_
-jQP u

DL dimensionless inertance

DR dimensionless pump resistance

gpm gallon per minute

H blade spacing

I identity matrix

J j=i

KB dimensionles s compliance

KBI imaginary part of dimensionless compliance

Pd-Pu
L ine r tance L -
-J Ou

upstream section, pump and downstream section inert.lnce

Lu, Lp, Ld

M B nlass flow gain factor

MBI imaginary part of mass flow gain factor

NPSH -P
net positive suction head P-oo v

rpm revolution per minute

R pump resistance

_P pressure rise across the pump

P upstream steady pressure

-CO

P vapor pressure
V

I
 -vii-

fluctuating pressure

Q fluctuating volumetric flow rate

P dimensionless fluctuating pressure

m dimensionless fluctuating mass flow rate

U inlet axial velocity

a

tip velocity
Utip
Y Z + I matrix

Z total transfer matrix between m_asuring stations

ZD downstream section transfer matrix

ZP pump transfer matrix

ZU up._:tream section transfer matrix

Ap
head coefficient
u_
P tip

density of fluid
P -P
-QO V

cavitation number _p U 2
- tip
U
a
flo_v coefficient
Utip

W dimensionless frequency

fluctuating frequency

Subscripts

U upstream

d downstream
 -viii-

LIST OF FIGURES

Page
Fig. 1.1 Occurrence of pogo instability. 20

Fig. 1.2 Typical cavitation performance of pumps. 21

Fig. 1.3 Typical fully-wetted pump performance curve. 21

Fig. 1.4 Theoretical and experimental comparisons of 22

compliance and mass flow gain factor.

Fig. 1.5 Cavitation compliance as measured from perturbed 23

inlet engine test.

Fig. 2.1 Schematic of preliminary experimental setup. 58

Fig. 2.2 Functional schematic of the fluctuator valve. 59

Fig. 2.3 Schematic diagram of the mathematical pump loop. 60

Fig. 2.4 Results of the mathematical pump model. 61

Fig. 2.5 Schematic drawing of the dynamic pump test facility. 62

Fig. 2.6 Photograph of the dynamic pump test facility. 63

Fig. 2.7 Schematic drawing of Impeller III. 64

Fig. 2.8 Photograph of Impeller III. 65

Fig. 2.9 Photograph showing fatigue fracture of impeller III. 66

Fig. 2. 10 Schematic drawing of Impeller IV. 67

Fig. 2.11 Photograph of Impeller IV 68

Fig. 2.12 Schematic drawing of Impeller V. 69

Fig. 2.13 Photograph of Impeller V. 70

Fig. 2.14 Drawing of the pump volute. 71

Fig. 2.15 Assembly drawing of the bearing cartridge. 72

Fig. 2.16 Schematic drawing of the smoothing chamber. 73

Fig. 2.17 Schematic drawing of the silent valve. 74

Fig. 2.18 Electronic feedback system of the silent valve. 75

 -ix -

Fig. 2.19 Servo-hydraulic system for the silent valve. 76

Fig. 2.20 Cross section of the fluctuator valve. 77

Fig. 2.21 Locations for the cross member bracing of the foundation. 78

Fig. 2.22 Vibration level of the foundation in direction parallel 79

to the pump axis.

Fig. 2. Z3 Vibration level of the foundation in direction perpen- 80

dicular to the pump axis.

Fig. 2.24 Schematic drawing of the deaeration system. 81

Fig. 2.25 Flow diagram of the instrumentation. 82

Fig. 2.26 Frequency response transfer functions of the two 83

LDV systems.

Fig. 2.27 Calibration curve of the Potter turbine flowrneter inwater 84

Fig. 2.28 Combined filter and ac amplifier frequency response 85

transfer function.

Fig. 3.1 Typical calibration curve of a pressure transducer. 94

Fig. 5.1 Fully wetted performance of Impeller III and V. 112

Fig. 5.2 Fully wetted performance of Impeller IV and 0.6 scale 113
model _rovided by Rocketdyne.

Fig. 5.3 Cavitating performance of Impeller V at 9000 rpm 114

where %0 = 0. 065.

Fig. 5.4 Cavitating performance of Impeller V at 11500 rpm 115

where _0 = 0. 052, 0. 076 and 0.095.

Fig. 5.5 Cavitation performance of a 0.6 scale test model 116

provided by Rocketdyne.

Fig. 5.6 Fluctuating velocities produced by operating the 117

upstream fluctuator only.

Fig. 5.7 Fluctuating velocities produced by operating the 118

downstream fluctuator only.

Fig. 5.8 Fluctuating velocities produced by operating both 119

the fluctuators together.
 -X-

Fig. 5.9 Fluctuating pressures produced by operating the 120

upstream fluctuator only.

Fig. 5. 10 Fluctuating pressures produced by operating the IZl

downstream fluctuator only.

Fig. 5. II Fluctuating pressure produced by operating both 122

the fluctuators together.

Fig. 5. 12 Photograph of Impeller V during auto-oscillation. 123

Fig. 5. 13 Steady pressure drop across the smoothing chamber. 124

Fig. 5. 14 Oscilloscope photographs of the fluctuating quantities. 125

Fig. 5. 15 Z-matrix produced by different combinations of 126

fluctuator settings.

Fig. 5.16 Linearity assumption test showing the fluctuating 127

response of the measurements against increasing
amplitude of upstream mass flow fluctuation.

Fig. 5.17 Z-matrix from measuring the third harmonics from 128
the fluctuating signals.

Fig. 6. 1 Photographs of Inducer IV during experiment (1), a 150

fully wetted condition.

Fig. 6. Z Z-matrix of experiment (1)0 fully wetted conditions 151

for Impeller IV where o = 0. 51 _ = 0.25 and c0= 0.07.

Fig. 6.3 Photographs of Inducer V during experiment (Z), a 152

fully wetted condition.

Fig. 6.4 Z-matrix of experiment (Z), fully wetted conditions 153

for Impeller V where a = 0.8Z _ = 0.1Z andco = 0. 088.

Fig. 6. 5 Lumped parameter electrical analogy of the fully 154

wetted experiment.

Fig. 6.6 Diagram showing the values of the system charac- 155
teristics used for the ZP pump matrix reduction.

Fig. 6.7 ZP-matrix of experiment (I), the fully wetted condi- 156
tion for lmpeller IV where o=0.51 _ =0.25 and CO-0.07.

Fig. 6.8 ZP-matrix of experiment (2), the fully wetted condi- 157
tion for Impeller IV where cy--0.82 4=0. IZ and CO=0. 087.
 -xi-

Fig. 6.9 Diagram showing the change of the pump resistance, 158
slope of the curve, with increasing frequency.

Fig. 6. I0 Vector dfagram showing the real and imaginary parts 159
of the Sears' function.

Fig. 6. II Photographs of the cavitating Impeller IV during 160

experiment (3) where o=0. II _ =0.26 and q0=0.07.

Fig. 6. IZ ZP-matrix of experiment (3), one of the cavitating 161

experiments for Impeller IV where o = O. 11 _ = O. Z6
and %o = O. 07.

Fig. 6.13 Photographs of the cavitating Impeller IV during 16Z

experirr, ent (4) where o = 0. 046 _ = 0. Z5 and _0=0.07.

Fig. 6. 14 ZP-matrix of experiment (4), the second cavitating 163

experiment for Impeller IV where o = O. 046 $ = O. 25
and _0 = O. OV.

Fig. 6. 15 Photographs of the cavitating Impeller V during 164

experiment (5) where o = 0. lZ $ = 0. lZ andq) = 0. 087.

Fig. 6.16 ZP-matrix of experiment (5), the cavitating experi- 165

ment for Impeller V where o = O. 12 _ --0. 12 and
%0 = 0. 087.

Fig. A 1 Schematic for the LDV processor. 175

Fig. B 1 Steady state velocity profile across the upstream 178

laser doppler velocimeter viewing channel.

Fig. BZ Steady state velocity profile across the downstream 179

laser doppler velocimeter viewing channel.

Fig. B3 Turbulence levels as measured by the LDV systems. 180

 I. _TRODUCTION

1.1 Backsround

Formerly, the designers of many hydraulic systems were con-

cerned solely with the steady state performance of the system since

this was the only information available. In recent times, the added

complexity has resulted in stability problems which calls for the study

of the unsteady performance. This added complexity may be due to an

improved and more sophisticated component which will give a more

complicated performance as in the case of a turbomachine or may

simply be due to additional network system. The instability may be a

transient problem like the case of boiler feed system (see Liao and

Leung 1972, and Liao 1974) or may be like the periodic oscillation en-

countered in hydropower installations (see Jaeger 1963 and Streeter

and Wylie 1974). The oscillatory type of problem is the one to be

studied here.

Many hydraulic systems suffer a vibrational problem which may

be severe if a natural resonance frequency is excited. The resonance

frequency may come from a structure vibration mode or may be solely

hydraulic in nature or a combination of both. This may happen when a

small excitation disturbs the system to operate away from its steady

state condition. The response of the system may be transient and

decay gradually back to the steady state or the response may be a

cyclic oscillation growing in amplitude until it causes the shut down of

the system, Such vibrational problems are attracting new attention as

increasing complexity is built into hydraulic systems; the added

 -2-

complexity creates an increasing number of excitable resonance modes.

Of all the hydraulic components involved in this vibration, the most

complex and least understood is the role played by the turbomachine.

Whereas other components may be passive in nature, the turbomachine

is active in the sense that it is designed to impart energy to the fluid as

in a pump or extract energy from it as in a turbine. Because of this,

the turbomachine may play a dominant role in the dynamics of the

whole system.

The steady performance which relates flow rate and pressure

rise of a pump for a given speed is well understood. But if a small

fluctuation in flow rate is superimposed on the steady state flow, there

is little available it.formation about the effect of this on the fluctuating

pressure rise across the pump. This problem is further complicated

by the fact that some turbomachines operate with a two phase fluid.

Besides the analysis of the two individual fluid flows, the interaction

between them has to be considered. This is exemplified by the grow-

ing concern about the influence on the structural vibrational instability

of the cavitating sodium pump used in nuclear power stations (see

Gibbs and Oliver 1973). Another example occurs in the liquid propel-

lant pump of a rocket engine where high pump speed causes the propel-

lant to cavitate; this example motivates the present work and will be

discussed in detail in the following section.

1. Z Pogo Instability

Occasionally, during the launching of some liquid propellant

rockets and space vehicles, a large amplitude longitudinal vibration

_,_PRODUCIBILITY OF Tt_
RI(HNAL PAGE IS POOR
 -3-

may gradually occur (see NASA SP-8055). This is called the "POGO"

instability. Although in most of the cases this vibration may decay

after about I0 - 40 seconds, severe damage to the structure and instru-

ments can occur besides discomfort and even bodily injury to astronauts.

This may also happen in a multistage rocket such as in the second stage

of Saturn engine where the occurrence of pogo has caused the pre-

mature shut down of an engine. In some extreme cases of pogo,

the vibration can attain an acceleration level of 17 g's with frequency of

5 - 60 Hertz (see NASA SP-8055). Recent data show that pogo is a

very frequently encountered phenomenon and also is a very severe prob-

lem (see Fig. i. I). Analysis indicates that each of these frequencies

coincides with a longitudinal vibrational mode of the rocket structure.

The cause of pogo is a closed loop feedback coupling of the pro-

pulsion system and the structure. In short, a small longitudinal dis-

turbance in the propellant tank creates an unsteady flow rate in the

feed line leading to the turbopump. The turbopurnp experiences an up-

stream fluctuation of pressure and flow rate and in turn creates its own

characteristic downstream pressure and flow rate fluctuations. The

unsteady supply of propellants to the combustion chamber leads to an

unsteady propellant burning rate which in turn produces an unsteady

thrust. The loop is closed when this fluctuating thrust causes arnpli-

fication of the longitudinal vibration of the structure.

Presently, there is no sure way to completely eliminate the

pogo problem besides suppressing it by minor modifications to the pro-

pulsion feed system. These methods try to alter the resonance

 -4-

frequencies of the propellant feed system. For instance, if the reso-

uance frequencies of the propellant feed system and structure are

known in advance, some means can be found to change the frequency of

the feed system such that they are decoupled. This includes, for

example, the injection of gas into the feed line or the incorporation of

a tuned gas accamulator upstream of the pump (see Norquist, Marcus,

Ruscio 1969). Since the gas contained in a gas accumulator can easily

be adjusted to any designed volume and pressure this device is the main

way of pogo suppression employed in most existing liquid propellant

rockets.

There is a possibility that the Space Shuttle may be even more

prone to pogo instability than previous rockets. The propellant pump

system in this case employs two physically separated pumps for each

propellant (see P, othe 1974). Besides the normal propulsion system

and structural system interaction, there is the additional possibility of

a sympathetic oscillation between the two pumps. Furthermore, the

Space Shuttle has a multibody, non-axisymmetr_c structure with many

vibrational modes; hence the pogo analysis is more complex. Along

with the investigation on the use of the conventional passive tuned gas

accumulator (see Lock and Rubin 1974), the feasibility of using a more

elaborate active suppression system (see Farrel and Fenwick 1973)

has also been explored; this method attempts to control the pressure

fluctuation upstream of the pump using a very complicated electronic

hydraulic servo feedback system.

All the approaches just mentioned are corrective in

nature and consist of modifications to the existing system in the

absence of sufficient dynamic information. Detailed analysis of the

 -5-

pogo instability encounters difficulty because of the lack of enough in-

formation about the turbopump. The study of turbopumps under fluctu-

ating conditions is the theme of the present study.

1.3 Cavitation in an Inducer Pump

To save weight and volume carried by liquid propellant rockets

and to yield similar savings of material in hydraulic systems in general,

:he design of a turbopump within a system is optimized to operate at

the maximum attainable speed (see NASA-8052). This introduces a

:nigh velocity at the outer tip of the pump impeller which in turn creates

a low pressure region there as indicated by the Bernoulli equation. As

liquids cannot withstand too high tension, the liquid undergoes a phase

change into vapor phase (see Knapp, Daily and Hamrnitt 1970). Further-

more, any other gas previously dissolved in the liquid may also come

out of solution. This phenomenon is termed cavitation and it can lead

to excessive noise, erosion damage to the material surface and de-

gradation of the pump performance. Increasing the upstream supply

pressure may suppress this cavitation but this requires the additional

strengthening of the container tank and thus decreases the payload of

the rocket. The degradation of pump performance due to cavitation can

be minimized by careful design of a first stage axial flow inducer pump

(see Stripling and Acosta 196Z, Acosta 1958, King 1970 and NASA

SP-8052). This first stage inducer pump will be designed to operate

under cavitating conditions but give enough pressure rise to eliminate

cavitation in the subsequent stages which are often of a mixed flow or

centrifugal type.
 -6-

The development of cavitation on the axial flow inducer pump is

4escribed by a parameter, the cavitation number. The cavitation num-

ber is the ratio of the static pressure head above vapor pressure of the

fluid to the dynamic head (for definition see nomenclature). As the

cavitation number is gradually reduced either b 7 lowering the upstream

_oressure or by raising the rotative speed, cavitation starts at the tip

clearance space, enveloping the outer blade passages of the inducer;

this is referred to as tip clearance cavitation. If the flow coefficient

(for definition, see nomenclature) is low enough, the downstream pres-

sure will force the tip clearance cavitation to flow upstream forming

"back flow" cavitation. As the cavitation number is further reduced,

small pockets of cavitation bubbles first build up on the blade surface

starting at the leading edge, spreading towards the trailing edge until

they eventually cover the whole blade surface. This changes the angle

of attack between the blade and flow direction preventing the blade from

putting mechanical work into the fluid. At this point, the performance

of the pump begins to deteriorate. Finally, if the cavitation number is

further reduced, the cavitation grows in size and partially blocks the

flow passage which leads to complete performance breakdown.

The cavitation performance of a typical pump is illustrated in

Fig. 1.2 where the pump head rise is plotted against net positive suc-

tion head for a constant flow rate and constant rotational speed. The

net positive suction head (NPSH) is the difference between absolute

pressure upstream of the inducer and the vapor pressure of the fluid.

As the upstream pre'ssure is reduced between A and B on the graph,

_i_I_EODUCIBILITY OF THI_
ORI(_[I_AL PAGE IS POOR
 -7-

the head rise is unchanged. Actually the head rise increases a little

due to the reduced viscous drag when the cavitation covers the blade

surfaces. As the NPSH is further reduced between B and C, the

performance of the inducer breaks down and the head rise decreases

drastically.

Turbopump inducers usually operate at a stage of cavitation

just before breakdown with large amounts of cavitation. When flow

rate and pressure fluctuations are imposed upstream of the inducer,

the cavities collapse and grow, thus inducing downstream mass flow

and pressure fluctuations. During a fluctuating cycle, the extremes

of the operating point may move into the region of performance break-

down. Such complex phenomena can alone lead to the instability of a

hydraulic system.

1.4 Lumped Parameter Model of a Turbomachine

It is essential at this moment to discuss how these complex

phenomena can be represented. The basic interest is a way to repre-

sent the fluctuating or oscillating output when a hydraulic system is

subjected to a given oscillating input. In the present context we are

concerned with fluctuating quantities that are small departures from a

mean steady state. Conventionally the dynamics of hydraulic systems

have been treated in terms of lumped parameter models.

The lumped parameter model assumes that the distributed

physical effects between two measuring stations can be represented

by a lumped constant. This assumption is valid when the hydraulic

system is relatively simple and when the physical lengths of the

system are shorter than the acoustic wavelength at the frequencies of

 -8-

interest. For example, the inertial effect of fluid between the two ends

of a long pipe can be represented by an inertial constant known in hy-

draulics as the inertance obtained by dividing the force (which is the

pressure drop multiplied by the area) by the acceleration (which is

simply the time derivative of the velocity). Similarly a compliance

value can be assigned to a pipe with a flexible wall by dividing the

change of fluctuating volumetric flowrate by the fluctuating pressure.

To clarify this, it is useful to use an electrical analogy. H we replace

the pressure by a voltage and the mass flow rate by an electrical cur-

rent, then the electrical analogies of inertance and compliance terms

are an inductance term and a capacitance which is connected to ground.

Further, if the pressure and volumetric flow rates are sinusoidal

oscillations, the equations relating these to the inertance and compliance

are

Pd _ U = -j_LQ U (1. I)

d _u -j_CPu (I. Z)

where l_d' _u and Od' (_u are the downstream and upstream fluctua-

ting pressure and downstream and upstream fluctuating volumetric

flow rate, L and C are the inertance and corn,pliance and 0 is the

frequency of oscill_tion. These two examples are quoted because they

will be used to ana!yze some of the results in a later chapter.

Previous lumped parameter models of the dynamic of pumps

were based on the steady state performance. The upstream and down-

stream quantities -were connected by terms like "compliance",

"dynamic gain factor" and "pump resistance" (see Rubin 1966, Vaage
 -9-

Fidler and Zehnle 1972). The pump performance curve which plots

the steady pressure rise across a pump against mass flow rate for a

constant rotational speed is shown in Fig. 1.3. The "pump resistance"

during flow oscillation is simply assumed to be the negative local gra-

client of the steady pump performance curve, i.e., the slope of the

total pressure rise with mass flow rate. The cavitating performance

of a pump was discussed and Shown in Fig. 1.2. The "dynamic pump

gain factor _' is defi_ed to be the fluctuating change of total pressure

rise across the pump with the fluctuating change of upstream pressure.

The dynamic model assumes that this can be taken to be the slope of

the cavitation performance curve. The "compliance" in these analyses

measures the change of cavitation volume with the change of upstream

pressure and is a most important parameter for dynamic analysis. A

basic simplification was made that the compliance could be represented

by a large t'bubble _ located at the pump inlet in quasi-steady equilib-

rium and this simulation gives rise to a "bubble spring constant".

Values of the bubble spring constant and thus the compliance were based

on static firing tests of the engine as will be mentioned in Section 1.5.

Analytical calculations of compliance based on quasi-steady

blade cavitation flcw were made by Brennen and Acosta (1973). These

calculations assume that the instantaneous cavitation size is deter-

mined by the instantaneous upstream pressure and these are related

by their steady state calculation. The result of this calculation

yielded a value much lower than experiment (see next section). Calcu-

lation of compliance contribution from other types of cavitation also

 -10-

exist in the literature. These include the effect of back/low cavitation

by Ghahremani (1970) and bubble cavitation by Brennen (1973).

The characterization of the cavitating inducer pump can be

broadened by the introduction of one more quantity_ '_the mass flow

gain factor '_ which is defined as the difference in the fluctuating mass

flow rates for a change in the upstream flow rate. A quasi-steady cal-

culation to find the mass flow gain factor from blade cavitation flow was

done by Brennen and Acosta (1975). The change of upstream mass flow

rates gives rise to a change of angle of attack on the blade thus chang-

ing the cavitation volume and this change of cavitation volume was cal-

culated from steady state performance. This quantity which was as-

sumed zero in all previous analysis was found to give a contribution to

the dynamic characteristics comparable with that of the compliance

term and bridged the discrepancy between compliance calculation and

measurement (see Fig. 1.4).

However, all of these analyses are quasi-steady in nature thus

are only valid at low frequency. At higher frequencies all these quan-

tities may have a more complicated behavior and may be complex. A

more general but incomplete calculation on the dynamic response of a

cavitating cascade has been presented by Kim and Acosta (1974). All

these calculations need to be verified by a carefully controlled experi-

ment where all the important variables could be directly measured.

In order to understand, predict, and hopefully, to eliminate the

occurrence of the pogo instability, lumped parameter models of space

vehicle dynamics have been constructed (see Rubin 1966, Rubin,

 -11 o

Wagner and Payne 1973). Such analyses are based on transfer func-

tions for each component involved in pogo. These include: the struc-

ture, fuel tank, feed lines, pump and combustion chamber. The trans-

fer functions relate the input and output of each component when the in-

put is a sinusoidal oscillation of a certain frequency. The input and out-

put can be acceleration in the case of structure or the pressure in the

case of a feed line, etc. These transfer functions are connected into a

mathematical feedback loop and the stability criterion of the loop is

investigated. The transfer functions for the pumps consisted of terms

like the pump resistance, compliance and pump gain factor. However,

the results of these analysis were not completely satisfactory because

of insufficient information about the values of the pump resistance etc.

under dynamic cavitat/ng condie/ons.

Similar calculations to investigate the system oscillations as-

sociated with cavitating inducers in a hydraulic loop were performed by

Sack and Nottage (1965). In their study, the instantaneous size and

shape of the cavities were calculated from steady state analyses.

Fanelli (1972) studied a similar problem concerning the dynamic be-

havior of large hydraulic power turbines with no cavitation. In his cal-

culation, he used the oscillating single airfoil result with a modification

factor which was used to correlate the single airfoil and cascade steady

state performance. However, these calculations were incomplete and

lacked an exact description of the dynamic behavior of the fluid flow.

1.5 Experimental Work

Three main categories of experiments on the dynamic behavior

 -12-

of turbomachines exist in the literature. These experiments included

cavitating and fully wetted conditions. The first type of experiment (see

Natanzon, et al 1974 and Karnijyo andSuzuki 1973) consists of operating a

cavitating inducer pump in a closed hydraulic circuit. Cavitation num-

ber, flow conditions and physical dimensions of the circuit were varied

until a self-sustaining hydraulic oscillation occurred in the system. A/-

though there were fluctuating pressure measurements p no fluctuating

flow measurements were taken so that only a limited quantitative anal-

ysis of the experirr.ent could be undertaken. These experiments con-

firmed that there was a strong interaction between the cavitating in-

ducer pump and the rest of the hydraulic system. In the pogo problem

the interaction is provided by the combustion chamberp structure, etc.

The experiments a_so showed that the pump could be the source of

energy which maintained this kind of oscillation.

The second category of experiments attempted to measure the

compliance of a cavitating inducer pump by perturbed inlet engine tests.

These tests measured a perturbation pressure transfer function of the

feed system by measuring the perturbed inlet and outlet pressure of

the pump caused by pulsing the suction flow rate. The pulsation was

provided by a hydraulically activated piston operating at various fre-

quencies. This transfer function when plotted against, frequency gave

the first fundamental frequency of the inlet pump system. The dynamic

characteristics of the feed system were modeled like a spring and mass

system. The mass value was given by the knovrn inertia of the feed

line. The spring constant was derived from the observed fundamental
 -13-

_requency. This calculated spring constant or compliance consisted of

contributions from cavitation and expansion of the feed line and the

value was taken to be the compliance of the pump, (see Rocketdyne 1969,

Vaage, Fidler and Zehnle 1972). The results of such tests have been

used in the pogo analysis of rocket engines.

Past efforts to measure the compliance with this type of method

showed a wide data scatter for different engines (see Fig. I. 5). This

is perhaps due to the oversimplification of the whole process. The

representation of the whole complex cavitation effect by only a single

lumped parameter, the compliance, was not justified. Furthermore,

the tests could not differentiate between the individual contributions

from bubble compliance, the compliance of the structure or the cavita-

tion effect due solely to the pump; also no account was taken of the in-

fluence of the fluid flow conditions or the different forms of cavitation

that could occur in a pump. The perturbed inflow engine tests also

revealed one interesting fact; namely that the pump gain did not retain

its steady state value under dynamic conditions (see Rubin 1966 and

Wagner 1971).

The third kind of experiment actually tried to measure directly

the fluctuating Quantitiee Anderson et al (1971) (see also Stevans and Blade

1971) attempted to measure the response of a radial bladed pump under

sinusoidal upstream disturbance with no cavitation. In their experi-

ments, the fluctuating pressures and mass flow rates were measured

but only one quantity, the pump impedance, which was defined as the

difference in oscillating pressure rise over the oscillating exit flow rate,

was analyzed. The results showed a wide scatter. This is probably

 -14-

due to the lack of reliable oscillatory mass flow measurements and be-

cause only one quantity was used to describe the pump dynamic perform-

ance, namely the pump impedance. The pump impedance so derived

also contains some uncertainty since this method of calculation de-

pends on the impedance downstream of the pump.

Two mass flow systems were employed in Anderson's experi-

ment. One involved measuring the oscillatory pressure drop across a

multiholed orifice in the discharge flow. Since Lhe dynamic measure-

ment of an orifice is not a simple extension of its steady calibration

the dynamic response of an orifice was calculated theoretically and

this was used for the calibration. No actual dynamic calibration was

attempted. The principle of the upstream mass flow meter was to mea-

sure the acoustic v/ave velocity in the moving fl-_id. The measurement

was done in a long pipe upstream of the pump. Since the acoustic x,e-

locity depends on the air bubble content, any unknown amount of air in

the flow affects the measurement. This effect is more severe for

cavitating tests where there may be substantial quantities of gas en-

trained in the upstream section. The main problem encountered in the

measurement of d:rnamic flow systems hinges on the development of a

reliable dynamic rilass flow measuring system.

1.6 Pump Transfer Matrices

It was reasoned that the analytical representation of a cavitating

inducer pump executing small periodic excursions about its mean con-

dition could be represented by groups like dynamic pump gain factor,

compliance, pump resistance and mass flow gain factor. This

 -15-

naturally led to an even more complete and general representation, the

ZP pump transfer matrix (see Brennen and Acosta, 1975) defined as

Pd

-lr zp,l
d - mu
=

[ZPzl
ZPl2

ZPz2J
Pu
(1.3)

where Pu and Pd are the dimensionless upstream and downstream

fluctuating pressures and m u and rn d are the dimensionless upstream

and downstream mass flow rates. These quantities are non-dimension-

alized by the following equations

U

d (1.4)
Pu - 1 Ug
d _P tip

r_ - d = (l° 5)
AiUtip

"_ "_ are the upstream and downstream pressure and

where P_ and Q_

volumetric fluctuation, p is the density of the fluid, Utip is the tip

speed of the pump and Ai is the inlet flow area. Therefore, if small

cyclic perturbations of a certain frequency are imposed upstream of

the inducer, the downstream oscillatory state can be calculated from a

ZP-matrix if this was known. Since there may be phase differences

between all the four pressure and mass flow quantities, the elements

of the ZP-matrix are complex in general. This representation is only

valid for small fluctuating pressures and mass flows and assumes a

linear relation between these quantities. An even more complete rep-

resentation should include the effect of pump rotational speed also {see
 -16-

Rubin, Wagner and Payne 1973), but this effect is small in many ap-

p].ications as demonstrated by Brennen and Acosta (1975) based on quasi-

steady blade cavitation calculations.

It is necessary at this point to interpret the ZP-matrix in terms

of the lumped parameter analysis as mentioned in Section 1.4. Since

ZPll is the ratio of fluctuating pressure rise to fluctuating upstream

pressure, this term will be called the pump gain term. The real part

is simply the dynamic pump gain factor, as defined in Section 1.4; but

there is no simple parallel for the imaginary part. This will be re--

ferred to as the imaginary pump gain factor. The ZPIZ is a ratio of the

Auctuating head rise to the fluctuating upstream flow rate. This term

'_ill be referred to as the negative of a pump impedance. The negative

sign is introduced here such that it will be consistent with the pump

resistance defined earlier. The real part of ZPIz is therefore the

negative of a dimensionless pump resistance DR as follows

Re ZPIz = - DR . (1.6)

The relationship between this and the dimensional pump resistance, R

is given by
2A.
DR =R x (1.7)
•
The imaginary part of ZPIZ is analogous to an inertial term. A

dimensionless inertance DL is defined in terms of the imaginary

ZPlZ by the following equation

Im ZPIz = -_DL (1.8)

REPRODUCIBILITY OF THE
0RIfi(NAL PAGE IS POOR
 -17-

where w is a dimensionless frequency defined as

(i.9)
- Utip

In this equation H is the tip spacing of the inducer pump. The relation-

ship between the dimensionless inertance DL and dimensional inertance

L is given by 2A.

DL = L pU*pti " (I. i0)

3imilarly a dimensionless compliance HB can be defined in terms of

Z PZ 1 by

im ZP21 - -wK B (1, 11)

where KB is related to dimensional compliance C through

Ut2i P
KB=C× (l.IZ)
I

No simple analogy can be drawn for the real part of ZP21. Therefore,

this will be referred to simply as the real compliance term. Finally,

the imaginary part of ZP22 is related to the mass flow gain factor

MB through

Irn ZP22 = -_M B (1.13)

as introduced by Brennen and Acosta (1975). Again there is no simple

physical interpreta_on for the real part of ZP22 and this will be re-

ferred to as the real mass flow gain factor term.

All the dimensionless groups defined in the last paragraph

through the ZP-matrix are not direct analogies to the lumped

 -18-

parameter groups discussed in Section 1.4. For example, the pump

resistance term in ZPI2 from Eq. (1.3) will be an exact analogy to

the pump resistance as defined in Section 1.4 only if the pump gain

factor term that appears in ZP 11 is identically zero and vice versa.

A similar argument applies to the compliance and mass flow gain

factor terms.

Experimental evaluation of the components of the ZP-matrix

requires that all the fluctuating pressures and mass flows be mea-

sured. After all the fluctuating pressures and mass flows are measured,

the ZP-matrix can be obtained by inverting Eq. (1. 1). However, since

the matrix is of order two, two linearly independent sets of data are

needed. In other words, to solve for four complex unknown numbers,

we need four complex linearly independent equations. Each data set

gives two equations and thus two data sets are needed. In the present

experimental study, more than two linearly independent data sets were

obtained and a special least square fit procedure was used to extract

the ZP-matrix.

In the present experiment, an inducer pump was installed in a

closed hydraulic 1Gop and fluctuations were imposed upstream of the

pump. Linearly independent data sets could have been obtained by

changing the physical parameters such as the hydraulic length or re-

sistance elements of the system. An alternative way was to provide

pulsation downstream of the pump by a second fluctuator. By varying

the amplitudes of fluctuation and the relative phase between the two

fluctuators, more than two linearly independent data sets could be ob-

tained and this was the scheme used in the present study.
 -19 -

1.7 Goals of the Research

The object of the present research was to measure the dynamic

response of inducer pumps under different cavitating conditions, the

result being represented by ZP-matrices as described in Section 1.6.

The elements of the ZP-matrix are expressed as functions of frequency.

A test impeller was required along with means to vary the sys-

tem pressure for cavitation Conditions and some way to change the

steady flow condition must also be provided. Two specially designed

valves were needed to provide the disturbances to the mean flow at a

given frequency. Oscillatory pressure and mass flow rate measure-

ments were made upstream and downstream of the pump and from

these transfer matrices were calculated. The difficulty of measuring

the fluctuating mass flow rate encountered in previous experiments

was overcome with the advent of laser doppler velocimeter. Since the

measuring stations were not immediately upstream and downstream of

the pump, the matrix thus calculated included some dynamic effects

due to the intervening hydraulic components. These system effects

were subtracted out to give the required ZP-matrix.

In the following chapters, we will discuss how such a facility

and its associated instrumentation were designed, constructedp tested

and used. This will be followed by a description of how the data were

processed and system effects removed. Finally, the experimental mea-

surements will be presented and discussed in the light of what is known

about unsteady flows through pumps.

-ZO-
 -Zl -

PUMP /

Ap

P-_- Pv

Fig. 1.2 Typical cavitation perforn_ance of purnp_.

AP

RESISTANCE

1;'tg. 1.3 Typical luliy-wetted pump performance curve,

 ll) -22-
0.9

I I
0 |

_ 0.8 I
I

Z
o._
(.9

0
_J 0.6
LL

Or)

O.5 \ R,, "_,,

0.4 \
KNO ',\ /°/,o,,
(THEO• R Y) \ ,, L ROCKETDYNE
• COMPLIANCE, Ke

_ 0.3
_ D_ MODELB(EXPT)

=E o.a
0
(.>
n
_E 0.1-
=:)

0
m 0 I I I I I
n-"
0.01 0.02 0.03 0.04 0.05 0.06 0.07
i---
TIP CAVITATION NUMBER, crT
,FJ+g. 1.4 Theoretical and exgerimental comparisons of
compliance and mass flow Rain factor.
 -23-

0.40 |

I I I I I I

_M 0.20

0.10 - F-I OX
O.08
OX
LR91
I
' LR91 OX

H-I FUEL
Z
0 0.01
H-I OX
LR87 FUEL
<_ 0.008
J-20X
0.006
>
LRS70X

._UEL
L .I I _ I
0.06 0.10 0.10 0.18 0.22 0.26
CAVITATION NUMBER

Fig. 1.5 Cavitation comp]iance as measured frona perturbed

inlet ep.g_ine test. ...
 -24-

II. SYSTEM DESIGN

_. 1 Design RecLuirements and Problems

As mentioned in the introduction, the present study was direct-

ed toward representing the dynamics of cavitating inducers by a ZP-

matrix. In order to measure this ZP-matrix for a pump, the minimum

number of fluctuating quantities to be measured are upstream and down-

stream pressure and upstream and downstream mass flow rate. These

ire needed togethez with the steady state information to calculate the

required matrix. The two most important desigr_ considerations were

_he fluctuating mass flow measurements and some kind of fluctuators

to impose the fluctuation.

An important consideration was the amplitude of the fluctuating

mass flow rate. T,h_is magnitude is dictated by the linearity assumption

requirements that the mass flow rate fluctuation will be allowed to

reach a maximum of five or six percent of the mean flow rate. Of

course, a lower value is more desirable in fulfilling the linearity re-

strictions; but this would decrease the signal to noise ratio. The dis-

charge of the pump is a highly rotational and a very turbulent flow. It

was expected that the turbulence level could be reduced to a minimum

of approximately I 0 to 0.5 percent by passing _hrough turbulence re-

ducing devices. This would give a maximum signal to noise level of

about 5 to 10. Of the signal analyzers available in the market at that

time, the most suitable one being considered for purchase required a

minimum signal to noise ratio of one (see Section 2.5.8). Thus a sig-

nal to noise ratio of about 5 is adequate for this purpose.

 -25-

Three mass flow measuring systems were considered keeping in

mind that the main task was to measure a fluctuating mass flow rate.

These were, a small pilot tube connected to a strain gauge pressure

transducer, a hot film anemometer and a laser doppler velocimeter

{see Appendix A). After considerable evaluation, the laser doppler

velocimeter was chosen. The reason for not using the pilot tube was a

result of the large phase shift and pressure loss due to the compliance

of the connecting line. The hot film anemometer was abandoned due to

its need for frequent calibrations and the need for a fluctuating calibra-

tion, in any case, to retain phase reference. The laser doppler velo-

cimeter was regarded as the most suitable choice at the time because

it was non-intrusive to the flow and measured the absolute velocity.

However, the laser velocimeter was still not completely satisfactory

±

because it only measured the velocity at one point of the flow which

does not truly represent the mass flow rate. The true mass flow rate,

as calculated by the laser velocimeter reading, can be given by locat-

ing the measuring station to be at the end of a converging nozzle to give

a flat velocity profile. This also reduces the turbulence level that is

needed for a better data processing. The converging nozzle will be aug-

mented by a smoothing chamber designed to give a low turbulence level.

This proposed scheme was tested in the preliminary setup described in

Section Z. 2. Z.

We needed some means to generate the five to six percent mass

flow fluctuation at a given frequency as mentioned above. The ideal

velocity profile against time is a sine wave, but a periodic wave form

with a high fundamental frequency content was acceptable..As explained

previously, the signals were analyzed by a signal analyzer which

 -26-

required a minimum signal to noise ratio of one, any other harmonics

being treated as noise. In the design of the fluctuator, special care

was taken that no high velocities region existed in the fluctuator. High

velocity would create low pressure areas that would drive the dissolved

air out of the solution and influence the measurement.

Several designs were considered. These included, for example,

the periodic squeezing of a piece of rubber with longitudinal holes

through it by a servo-hydraulic system or the linear displacement be-

£ween two slotted throttling plates similar to the design by Anderson

etc. (1971). These were abandoned because they were too costly and

difficult to fabricate. Finally, a variation of Anderson's design using

two concentric slotted cylinders, one rotating inside a stationary one,

to provide the variable throttling was decided upon.

The first proposal was to use two sintered metal cylinders one

inside the other. The cylinders were to have longitudinal bands in

which the pores were blocked either by machining on the surface or

filling with epoxy resin. The rotation of one cylinder with respect to

the other would provide the necessary modulation. The selection of

sintered metal was made because the throttling effect would be wholly

viscous and high velocity regions would be minimized. This approach

was abandoned because great difficulties were encountered during at-

tempts to manufacture the bands on the sintered metal.

A final design based on the foregoing principle was used. This

design consisted of two slotted concentric cylinders with the inside one

made of bronze rotating within the outside stationary one, which was
 -27-

made of stainless steel. The different metals were used to avoid bind-

ing. A piece of sintered bronze cylinder was situated coaxially and next

to the slotted cylinders providing a bypass to the flow when the slots

_'ere momentarily closed completely. The varying amplitude of fluctua-

te.on was provided by a cylindrical sleeve which slid axially to cover

either the slotted cylinders and/or the sintered metal cylinder. The

_mplitude of fluctuations was proportional to the relative amount of area

,:overed by the sleeve. This sintered bronze throttling bypass was in-

tended to provide a viscous resistance and to minimize the air being

ctriven out of the solution. This arrangement was called the "siren

valve". The design does have some drawbacks because it produces a

_quare-wave kind of fluctuation and the abrupt cutoff and opening of the

flow path causes high velocity regions. An arrangement asdescribed

,_bove was made and tested in the preliminary set-up described in the

next section.

2.2 Preliminary Experimental Setup

To develop the laser doppler velocimeter and the "siren valve"

fluctuator, a small pump test loop was constructed. This system con-

sisted of a 7 hp centrifugal pump, the fluctuator, a silent valve for

£he steady state throttling (for details of this valve, see Section 2.4.6),

transparent windows for the laser to pass through and a one hundred

_allon storage tank. These were connected by a 4 in. diameter plastic

_ipe, The fluctuator shaft was driven by a one hp variable speed

motor. Figure 2.1 shows a schematic of the setup.

Keeping in mind the requirements put forth in Section 2.1, the

 -28-

components of this setup were tested and re-designed to give the opti-

mum performance. This included the detailed design of the siren

valve, the turbulence reducing components inside the smoothing cham-

ber and the shape and fabrication o£ the laser doppler velocimeter

windows. After many modifications, a satisfying design was arrived

at and adopted in the final pump loop design with very slight changes.

Thus only a brief description of the important lessons learned will be

given below.

2.2.1 Fluctuator Valve.

Figure 2.2 shows a functional schematic of this valve. The

siren valve arrangement which has a diameter of 6 in. , was housed

in a 12 inch diameter and one foot long chamber. The flow was

brought in through a side entrance. After it passed through either the

rotating slotted path or the sintered bypass path, the flow exited

through the center of the fluctuator chamber into the smoothing cham-

ber. Two lessons were learned from this setup.

(i). The original slotted cylinders were made too thin causing

these parts to bend slightly and this eventually led to jamming of the

valve. The final design used cylinders with a thickness of 1/4 in.

(ii) The open area on both slotted cylinders was made equal

in area to the solid surface. This design gave periodic spike fluctua-

tions instead of the square wave intended. The design was revised to

make the inner one three-quarter solid surface and one-quarter open

and the outer cylinder remained half open and half solid area. This

design subsequently gave a satisfactory square wave fluctuation.

 -29-

For a detailed design of this valve, see Section 2.4.7.

2.2.2 Laser Doppler Velocimeter System and Smoothing Chamber

De sign

Two transparent windows were needed to let the laser beam

pass and be focused in the channel. This channel had a circular cross

section of 4 in. internal diameter. It was calculated that with this

channel size, the flow velocity inside the channel would be between

2 to 10 ft/sec and this was a convenient velocity to be measured by the

laser doppler velocimeter both in this preliminary setup and eventual-

ly in the final design. Two acrylic windows were installed in a view-

ing section to provide the laser beam access to the flow. It was found

that the surfaces of the windows had to be very flat and free of scratch-

es to avoid multiple scattering of the laser beam. This multiple scat-

tering of the beam would increase the noise level measured by the

laser velocimeter by several folds.

To reduce the turbulence level of the flow, the flow from the

siren valve passed through a smoothing chamber and a convergent

nozzle before it entered the laser viewing channel. This smoothing

chamber contained perforated metal plates, honeycomb material and

screens installed perpendicular to the flow to aim for a turbulence

level of less than 1 percent. To give such a low turbulence level, the

water emerging from the siren valve was first slowed down to less

than0.Sft/secby the increased area of the 12 in. diameter chamber.

Inside the chamber, it first passed through two plates of perforated

metal to cut down _ne large scale eddiesw then passed through the
 -30-

honeycomb to reduce the lateral direction turbulence. The water fur-

ther went through a series of screens to reduce any leftover small

scale turbulence and finally it went through a distance of about 12 in.

of straight chamber for further damping before it was brought

to the 9 to 1 area converging nozzle. Different shapes, sizes and ar-

rangements were tested before the final decision was made on the con-

figuration. This was incorporated into the final design and will be

given in the Section 2.4.5.

2.3 Analytical Model

.A_ter it was established that we could furnish the necessary

mass flow measurement system and design of the fluctuator, we began

the design of the final pump facility. It was decided that the inducer

pump would be housed in a closed hydraulic loop. The essential parts

of this loop consisted of the inducer pump, two fluctuators (one up-

stream and one downstrearr_ and an isolation tank between the fluctua-

tors. The main purpose of this isolation tank was to dynamically un-

couple the two fluctuators. All these parts were connected by hydraulic

pipe s.

If the dynamic behavior of the inducer could be represented by

the ZP-matrix as mentioned in the introduction, the fluctuating quanti-

ties upstream and downstream of the pump as excited by the fluctua-

tors would depend on the dynamic behavior of the rest of the hydraulic

system. The dynamic response of the hydraulic loop depended mainly

on its physical dimension. In order to have some idea of the physical

REPRODUCIBILITY OF THI_
ORI(}IINAL PAGE IS POOR

l
 -31-

dimensions allowable, a mathematical model of the pump loop was

constructed using one-dimensional linear theory. The constraints of

this mathematical model were the requirements that

(a) The percentage mass flow fluctuation should be at least

one to two percent of the mean. This percentage was the lowest al-

lowable limit which could be measured by the laser doppler veloci-

meter.

(b) The amplitudes of the fluctuating quantities should be large

enough that a compliance value ranging from 0.01 to 0. 001 ft. 2 could

be measured accurately. Although we intend to measure the ZP-matrix,

there was no estimation on the values of this matrix thus a compliance

value taken from engine tests as described in introduction was used as

a design evaluatior_ parameter.

(c) The pressure fluctuations upstream of the inducer should

not cause the cavitation number to vary more than 30 percent. A low-

er value of the variation would be more desirable but this implied

smaller mass flow fluctuation and this opposed the intention (a).

The mathematical model had four main elements; these were,

the cavitating pump, two variable throttling valves and an "isolation

tank" between the two valves (see Fig. 2.3). These elements were

connected by hydraulic pipe lines. It was intended to calculate the

dynamic response of this mathematical pump loop when the throttling

valves were varied about their mean value at a given frequency. The

characteristics of the individual elements are as follows:

(i) The pu-xxp- One of the impellers intended to be installed

 -32-

in the final pump loop was a larger three in. diameter model of the

nine degree helical inducer described by Acosta (1958). The steady

performance of the mnathematical pump was taken to be that of this nine

degree inducer. The fluctuating performance of the pump was cora-

L

posed of two parts; a pump resistance which was taken to be the grad-

ient of its steady state performance as in the model by Rubin and an

assigned compliance value which related the fluctuating mass flow dif-

ference to the fluctuating upstream pressure. During the calculation,

different compliance values were used covering the range from 0.01 to

0.00001 ft. 2 No direct relationship between compliance values and

cavitation number was assumed. This is because the information from

engine tests (see Fig. I. S) shows wide data scatter and the mathe-

matical model in this case covered all the possioilities.

(ii) The tw,_ valves - The reasons for using two valves for the

required fluctuations were mentioned in the introduction and will not

be repeated here. These valves were assumed to have a steady resist-

ance which defined the steady operating point of the pump loop and a

fractional sinusoiaal variation about the mean to give the fluctuating

action at a selected frequency. The fluctuating amplitude of the valves

could be varied independently and they could be operated with an

arbitrary phase difference. The exact values used for these resist-

ances were actually determined after the steady operating states were

chosen.

(iii) The i_olation tank - This tank wa_- situated between the

two fluctuating valves. It was assumed that the tank contained a large
 °33-

volume of air which had a large compliance value thus isolated the two

£1uctuators. The absolute pressure of the air volume also defined the

datum pressure of the system and thus controlled the cavitation num-

ber of the pump.

(iv) Hydraulic pipes - The four components mentioned above

were connected by hydrau]ic pipes. The pressure drop across these

pipes was assumed to consist of frictional losses as well as inertial

effects. The lengths of the pipes were varied during each calculation

but the size of the pipe was taken to be constant and equal to four in.

diameter.

A computer program for this mathematical model was devised,

This computer program was given the steady state operating point of

the system with some value of the total pipe length. Then based on

the assumed dynamic characteristics of the components, the mass

flow fluctuations and pressure fluctuations at various stations in the

circuit were calculated for frequencies of excitation between 0 and 50

Hertz. Different lengths of pipe sections were used and the whole

process was repeated for different values of compliance. Only solu-

tions within the required amplitude of mass flow fluctuation and with-

in the limits of the variation of the cavitation number were subsequently

considered.

A typical result of the calculation is shown in Fig. Z. 4. It can

be seen that for a certain amplitude of mass flow fluctuation, the up-

stream pressure fluctuation which is equivalent to cavitation number

fluctuation increases as the pipe length increases. This is because in

order to create the same amplitude of mass flow fluctuation, more

 -34-

pressure will be needed for the increased inertia due to longer pipes.

To limit the cavitation number fluctuation, it is then best to minimize

the total length of the hydraulic pipes. This was the most important

guideline learned in this analytical pump loop model and will be applied

to the final design of the pump test facility.

2.4 Final Design of the Dynami c Pump Test Facility

The design of a pump test facility to measure the dynamic trans-

fer characteristics of a cavitating pump was based on the confidence

and experiences gained from the preliminary experimental setup and

mathematical model. For a schematic drawing of the facility see

Fig. 2.5, and Fig. 2.6 shows photographs of it. The focus of

the facility was an impeller pump driven by a dc electric motor through

a step up gear box. Before the flow entered the pump, it passed

through a mass flow measuring arrangement. This arrangement con-

sisted of a smoothing chamber to reduce the turbulence level, a con-

f.

vergent nozzle to give a flat velocity profile as well as accelerating

the velocity to a suitable velocity to be measured by the laser doppler

velocimeter. A similar mass flow measuring arrangement existed

downstream of the pump. Following the downstream laser doppler

velocimeter station was a "silent If throttling valve to control the steady

state operation of the hydraulic flow. The flow passed through a tur-

bine flowrneter for mean flow calibration before it entered the down-

stream fluctuator. The flow then passed through an isolation tankp an

upstream fluctuator then finally flowed into the upstream smoothing

 -35-

chamber. The two fluctuators were driven by a phase lock dc motor

through an 8 ft. long connecting shaft. The whole hydraulic flow cir-

cuit was located on top of an eight in. I-beam frame supported on

five legs bolted to the floor. The system was supplemented by

auxiliary systems namely a deaeration system, an air bleed system

anda heat exchanger to condition the water used in the system.

As for instrumentation, a reference signal from a signal analy-

zer commanded the fluctuator motor to rotate very precisely at the

selected frequency. Sensors like the pressure transducers, accelero-

meters and laser doppler velocimeters located upstream and down-

stream of the pump measured the respective fluctuating quantities.

The output voltage signals of these instruments consisted of mean

values which were taken during experiment and fluctuating values

which were amplified, low pass filtered and finally recorded in analog

instrument recorders to be processed later. The detailed description

of the components follows:

2.4.1 The Impel.le r_s_s

(i) Impeller Ill I - This was a three bladed nine degree helical

impeller with Z.985 in. outer diameter. The solidity which is defined

as the ratio of blade length to blade spacing was 1.75 and the hub ratio

which is a ratio of hub diameter to tip diameter was 0.4. The design

1
This impeller was fabricated by
Ivi. C. Bidwell Co.
71 No. San Gabriel Blvd.
Pasadena, California 91107
 -36-

was based on a similar impeller used by Acosta (1958). A drawing of

Impeller Ill is shown in Fig. 2. 7 and a photograph is included as

Fig. 2.8. This impeller was made of aluminum with the surface hard

anodized.

The tip clearance between the impeller tip and housing was

0.005 in. A radial clearance of 0.002 in. was used at first but this

close clearance caused scratches on the acrylic housing and it was de-

cided to increase the clearance. The impeller had a straight, sharp

leading edge thus causing a high stress concentration on the root of the

leading edge. After about one year of testing during which time the

impeller had gone through 6 million fluctuating cycles and about 70

million rotating cycles, it had a fatigue fracture at the blade root (see

Fig. 2. 9). A sirnUar impeller, Impeller V, w_Ls made to replace this

one.

(ii) Impeller IV 2- This is a one-quarter scale model of the

first stage low pressure oxidizer pump of the Space Shuttle main

engine. See Fig. 2. 10 for a drawing and Fig. ?. II for a photograph

of this impeller. The inducer consisted of a four blade helical section

of seven degree tip angle followed by a tandem row of twelve highly

loaded blades. The outer diameter was 2. 983 in. and had a radial

clearance of 0. 004 in. A scaled model of the stator stage was also

fabricated. The impeller was made of aluminum which made it

structurally as well as hydraulically similar to the prototype impeller.

2
Fabricated by Contrua Inc.
430 No. Berry St.
Brea, California 92621
 -37-

After this impeller had been operated under cavitating conditions for

about eight hours, there was some cavitation damage to the leading

edge at the tip where the surface was found to be pitted. This was re-

paired by filing off the uneven part and the blade surface was reano-

dized.

(ill) Impeller V 3 - This was designed and constructed after

the fatigue failure of Impeller III. This impeller had the same basic

design as Impeller Ill except for a thicker root and a swept back lead-

ing edge. It was made of 304 stainless steel and was thus stronger

structurally than Impeller Ill. Figure 2. 12 is a drawing and Fig. 2. 13

is a photograph of this impeller.

2.4.2 Acrylic impeller Housing

This was a 6 in. x 6 in. x 8 in. laminated block of acrylic with

a 3. 007 in. bore. Part of this bore was enlarged to house the stator

when using Impeller IV. The transparent acrylic made it possible to

view and photograph the cavitating impeller.

2.4.3 Pump Volute

The downstream of the acrylic housing was joined to a volute

to turn the flow through a right aJngle. The central part of this volute

contained the bearing cartridge which will be described in 2.4.4.

Guide vanes were installed at the volute exit to reduce the rotation of

the flow. Figure 7..14 is a diagram of the aluminum volute casting.

3
M.C. Bidwell Company
 -38-

2.4.4 Motor, Gear Box and Bearing Cartridge

Preliminary calculations based on estimated performance char-

acteristics of the impellers indicated that the power required at 50

percent efficiency would be about 20 hp. The calculation was based

on a top pump speed of 12000 rpm which was believed to give a suit-

ably low cavitation number. A 2.0hp open loop SCR variable speed

dc motor was purchased (Sabina Electric) 4. This motor had a top

speed of 1750 rpm.

The rotation speed was increased via a straight, parallel gear

5
box with a step up ratio of 6.86. This brought the maximum rota-

tion speed to 12000 rpm. Originally, a set of spur gears were used

which gave excessive chatter, vibration and noise. These were re-

placed by a set of helical gears which reduced the noise to a tolerable

level. The output shaft of the gear box was connected to the impeller

shaft via a one foot long intermediate light aluminum connecting shaft.

This shaft is very stiff in the torsional direction but soft in bending.

The impeller shaft which was made of stainless steel was

housed in a bearing cartridge. During fluctuating flow experiments,

any axial movement of the shaft would constitute additional motion be-

tween the impeller and fluid flow and this was to be avoided. The design

of the bearing system needed to be very stiff in the axial direction but

free to rotate. As an estimation of the shaft axial displacement effect

4
Sabina Electric and Engineering, 1190 Batavia St., Orange, Ca.

Vi-Star Gear Co., Inc., 7312 Jefferson Str__et, Paraxnotmt, Ca.

 -39-

we may note that a typical axial fluid velocity of I0 ft/sec with a two

percent mass flow fluctuation at 40 Hertz will have a 5 percent error

if the axial displacement of the shaft is as little as 0. 0005 in. Thus,

a maximum value of 0. 0005 in. axial displacement was allowed in the

design of the stiffness of the bearings.

The bearing design consisted of a set of two class A7 preci-

sion angular contact bearings back to back located near the middle of

the shaft to restrain axial motion. This duplex bearing set had no

free end play nor radial run out and had in addition a very high spring

constant in the axial direction. The driven end of the shaft was

supported by a deep groove precision ball bearing. The bearings

were air mist lubricated. A mechanical seal having a stationary car-

bon face rubbing against a 440 C seal ring was installed outboard of

the duplex bearings to seal the bearing cavity. Provision was made

for an additional slinger and drain port so that the air mist flow would

tend to purge any small water leakage into a drain pot. The impeller

shaft was connected to the light intermediate drive shaft via a steel

plate that had sixty teeth machined on its circumference. A magnetic

pickup provided a frequency output equal to the pump rotational speed

in rpm. These details are shown in the assembly drawing of Fig. Z.15.

Z. 4. 5 Smoothing Chambers and Convergent Nozzles

In order to minimize the turbulence le,,el of the flow to be

measured by the laser doppler velocimeter, the two laser beam

measuring ports were preceded by smoothing chamber and a nine to

one area ratio convergent nozzle. The general description was given
 -40-

in Section 2.2. I and the details will follow.

The smoothing chamber had a diameter of 12 in. and a length

of 24 in. It was made of aluminum with surface anodized. The flow

entering the chamber first met two perforated metal plates in series

to break up any large scale eddies. The Perforated metal plates were

i/4 in. thick with 30 percent opening area of hole size 3/16 in. This

was followed by a piece of aluminum honeycomb material 2 in. thick

and 1/8 in. cell size. The honeycomb section damped down the lateral

turbulence velocity. The flow finally passed through two mesh 22

fine wire screens into approximately 12 in. of straight chamber. The

fine wire screens broke up any left over small scale turbulence and

the straight chamber provided enough length for any small scale tur-

bulence to be reduced by internal viscous damping. The flow paseed

through a nine to one convergent nozzle then entered the laser veloci-

meter measuring _ort as mentioned above. A schematic of the

smoothing chamber is shown in Fig. 2. 16.

6
2.4. 6 Silent Valve

This provided the steady state throttling action. The important

part of the valve _3 a block of elastomer which is 8 in. long and 6 in.

in diameter containing about 200 longitudinal holes. This elastomer

block was placed inside a stainless stell cylindrical housing and was

squeezed axially between two matching Bteel end plates. A hydraulic

cylinder applied force at one end to reduce the hole size thus

6
Supplied throu_,h the courtesy of Mr. C. A. Gongwer.
 -41-

providing extra resistance to the flow. The hydraulic loss is pre-

dominantly viscous making the operation of the valve very quiet;

hence the name "silent valve" and a schematic is shown in Fig. 2.17.

The oil pressure to the hydraulic cylinder was supplied by a

hydraulic oil pump via a servo valve. The servo valve was controlled

by an electronic feedback system governed by the output of the turbine

flowmeter. The steady flow rate could thus be preset by adjusting a

reference voltage input to the feedback amplifier. This elaborate

feedback system was needed because readjustment of the fluctuator

for fluctuation amplitude was liable to change the steady state opera-

tion of the system. Schematic diagrams of the feedback system and

hydraulic pressure system are shown in Figs. 2.18 and 2.19.

2.4.7 Siren Valve

The general description and function of this valve were dis-

cussed in Sectidns 2.1 and 2. Z. Only the essential dimensions will be

given he re.

The bronze rotating armature and stainless steel cage were

about 5 in. in diameter and 6 in. in length with a thickness of 1/4 in.

The number of slots manufactured was nine. The sintered bronze

cylinder used was 5 in. in diameter, 6 in. in length and 1/4 in. thick

with bead sizes of about 0.03 in. it was found out later that the pres-

sure drop across the sintered bronze was too high. Fifteen holes of

3/8 in. diameter were then drilled in the sintered bronze piece to

reduce the resistance. The slotted cylinders and the sintered bronze
 -42-

pieces were situated at the center of a 12 in. long and 12 in. diameter

chamber.

The two fluctuators, one upstream and one downstream, were

of the same design but had opposite flow direction. The two fluctuators

were driven by a synchronized speed dc motor as will be discussed in

Section 2.4.8 and the connection between the dc motor and the fluctua-

tots was provided by three couplings and a 6 ft. long and l in. diameter

hollow shaft between the fluctuators. The actual right angle connection

between the siren armature and the shaft was by a nine to one reducing

worm gear and this was the reason for using nine slots in the siren

valve. Figure" 2.20 shows an assembly drawing of the valve.

2.4.8 Fluctuator Motor 7

The signal analysis b V the Bafco required that the fluctuating

measurements of velocity, pressure and acceleration be phase refer-

enced to the referei_ce output signal. The connection for this phase

reference was through the fluctuator motor that drives the siren valve

to produce the fluctuating flows.

The electric motor was a one horse power dc motor with SCR

dc power supply. The rotational speed was controlled by the reference

7
Electric Motor Sabina Electric and Engineering
I196 Batavia Street
Orange, California 92667

Feedback Electronic Design Shapiro Scientific Instruments

3500 East Coast Highway
Corona DeIMar, California 92625
 -43 -

signal from the Bafco and the maximttm speed of the motor was about

2500 rpm. To provide the phase reference locking, an optical encoder

was instaUed at the free end of the motor to provide a feedback loop for

the speed control. A two way dial indicated the speed of the motor

relative to the signal. Zero dial reading indicated that the motor

speed had achieved synchrony with the reference signal.

2.4.9 Isolation Tank

The chief function of the isolation tank was to prevent inter-

actions between the two fluctuators in returning the flow to the inlet

side. This was accomplished by a plastic bladder containing about

one cubic foot of a_r. Control of the pressure inside the bladder also

fixed the static datum pressure of the system. The tank was made of

galvanized steel with a side port to house the plastic bladder. It had a

large cross sectional area providing a low flow velocity region where

any air bubbles carried by the water would float to the top and be bled

away. A coiled copper tubing heat exchanger inside the tank was con-

nected to the house chilled water system to absorb the heat generated

from the pumping action to maintain a constant temperature of the

water,

2.4. I0 Fotmdation and Foundation Bracin_

The whole hy-draulic system was supported on an 8 in. wide

flange I beam welded framework supported by five columns formed of

six in. steel pipe. These columns had welded flanges which could be

bolted to the concrete floor. This foundation was originally bolted to

REPRODUCIBILITY OF THE
0RI(KNAL PA6E Ia POOR
 -44-

the floor firmly to minimize the vibration of the system. During the

experiment, noise and vibrations from the electric motor, gear box

and pulsations from the cavitating flow were transmitted through the

steel supports to the floor and subsequently transmitted to the whole

building. This noise and vibration disturbed the comfort of the com-

munity within the building. To isolate the system, the bolts holding

the foundation to the floor were loosened and rubber insulation pads

were placed between the support and the floor.

However, during this phase of the experiment, there was grow-

ing concern over the vibrational level of the frame as a sufficiently

large amplitude vibration of the hydraulic structure would produce a '

faulty reading of the LDV velocity reading. Accelerometer measure-

ments of the I bea_'xl framework were subsequently observed to be too

high (in some cases about one half the velocity signal) and therefore the

foundation was again bolted tight to the floor. Further accelerometer

readings of the structure at the LDV measuring stations revealed that

even in this case the foundation was not completely satisfactory. These

tests at various locations of the foundation found that there existed in-

ternal vibrational modes due to the bending of the I-beams and support-

Lug columns. This unwanted effect was greatly- reduced by adding three

diagonal braces between the I beam and the vertical columns constuc-

ted of six in. pipe. Figure 2.21 shows the locations of the cross

members and Figs. 2. 2Z and 2. 23 show the vibrational level in two

directions above ti_e LDV viewing windows before and after the bracing

of cross members,
 -45-

2.4. 11 Air Bleed System

At the highest point of each section in the circuit, connections

were made to a water accumulator of a vacuum system. The water

accumulator was simply a small tank connected to a vacuum pump and

a drain valve at the bottom. Its function was to trap any water that was

entrained in the vacuum line and prevent this water from entering the

vacuum pump.

2.4. 12 Water Treatment

Distilled water was used in the system. Because of the many

different metals used in the system, the water was treated chemically

to prevent corrosion by maintaining a chromate ion concentration of

about . 06 percent by weight and maintaining a minimum ph of 8. The

water was deaereated for about three days before each experiment and

cooled to about 70°F. during experiments.

2.4.13 Deaeration System

This was a system to recycle the air between a vacuum tank

and a storage tank such that the water could be deaeration. The water

was pumped into the vacuum tank and sprayed inside via a set of noz-

zles forming small droplets. The increased relative surface area and

small volume enable the dissolved air to diffuse through the water and

be removed by the vacuum pump. The water droplet itself fell to the

bottom of the tank and was pumped into the storage tank. The vacuum

tank was elevated about 5 ft. above the ground giving a better suction

performance to the transfer pump pumping water into the storage tank.

The capacity of the storage tank was about 300 gallons and had a layer

of small plastic plates floating on the water surface to inhibit air from
 -46-

dissolving back into the water. The vacuum was provided by a Nash

3 hp vacuum pump. Figure 2.24 shows a schematic of the deaeration

system.

2.5 Instrumentation

This section contains a detailed description of the instrumenta-

tion. The interrelationships between these instruments are shown in

the signal flow diagram of Fig. 2.25.

2.5. 1 Laser Doppler Velocimeter (LDV) and Mounting Systems

For a preliminary description of this instrurnent, see Section

2.2.1 and for the theory, see Appendix A. In the final version of the

LDV system adopted for the facility, a 50 cm focal length lens caused

the two laser beams to cross at the center of the measuring sections.

These sections had four in. diameter cross-section and followed the

nine to one converging nozzles. The laser beams passed through one

in. thick fiat lucite windows on each side of the sections. A collimating

stop consisting of a piece of cardboard with a small hole let only the

weak beam pass into the photornultiplier. The photomultiplier tube was

mounted about two feet away from this window. It was found that the

dirt existing in the water provided enough light reflecting particles for

the LDV to function properly.

Originally, the lasers and the photornultipliers were mounted on

a light aluminum channel attached to the main beam of the foundation. This

arrangement proved to be unacceptable due to the vibration of the LDV

mounting frame. The resonant frequency of this frame was found to be

only 25 Hertz which was well within the experimental frequency range of

excitation. This source of error was avoided by mounting the lasers

 -47-

and photomultipliers on a sturdy massive steel structure constructed of

Z × 6 in. channels welded into a box beam suspended from the ceiling.

The natural frequency of this suspension system was found to, be about

one half Hertz and during subsequent tests no velocity excitations of

this mounting framework approached one percent of the fluid velocity

perturbations measured by the LDV systems. Allflow velocity meas-

urements are then referred to this "inertia" frame.

As can be seen from AppendixA, the ratio between the measured

velocity and doppler frequency depends on the angle between the two

laser beams. This angle is very small and very difficult to measure

precisely. Coupled with the fact that the actualmeasurement required

was the vol_u'netric flow rate and not velocity at a single point, the tur-

bine flowmeter signal was used to calibrate the output of LDV in volu-

metric flow rate.

By operating the pump loop with a small fluctuation and observ-

ing the signals on the oscilloscope, it was found that the laser doppler

velocimeter could measure velocity fluctuations down to one percent of

the mean flow. But the signal as measured by the upstream LDV had a

different fluctuating amplitude and phase than the downstream one even

when the flow was fully wetted and great care was taken to exclude the

possibility of air bubbles in the system. There then appeared to be two

cluestions that needed answering; first, were the two velocimeters actu-

ally working identically and secondly, was the observed difference be-

tween the two signals due to structural compliance. In the present con-

text assurance as to the first question was needed for the program to

continue. The second is addressed in the subsequent chapters.

 -48-

In order to investigate possible differences in the operation of

the two LDV systems, the two LDV photomultipliers were located side-

by-side and fed by the same laser beam with the aid of a beam splitter.

One would expect two identical velocity signals. It was found that there

was a small phase shift between the two systems. Since the LDV pro-

cessor electronics were both designed with a high frequency response,

the only time shifting component could come from the two similar but

not identical frequency-to-voltage convertors. (One of them was actu-

ally purchased at a later date and turned out to be a slightly different

version; see Appendix A). It was believed to be necessary then to

measure the transfer function of the combined LDV processor systems

including the frequency-to-voltage convertor. This was accomplished by

replacing the photomultiplier input with a frequency-modu/ated carrier

frequency generator 8. In this test the frequency generator fed the LDV

processor a signal with a main carrier frequency of about 80- 150 k

Hertz which was the normal doppler frequency. This frequency was

changed by an amount of 2-5 percent at a frequency of 5-50 Hertz.

This latter modulating signal was provided b V the Bafco signal analyzer

reference signal. By this means it was possible to simulate the be-

havior of a real photomultiplier signal of a mean flow with 2-5 percent

fluctuation. The output of the entire LDV system was then analyzed

by the Bafco return signal analyzer from which the amplitude and phase

shift ret'ponse of each LDV system could be determined. These are

shown as a function of frequency in Fig. 2. Z6

8
Wavetek 164

REPRODUCIBILITY OF THI_
ORI_AL PAGE IS POOR
 -49-

One very important point that should be mentioned here is that

_he upstream LDV did not work when the leading edges of the impeller

blades are highly loaded. This highly loaded condition causes a tip

clearance backflow vortex to intrude into the upstream viewing windows

and this higher turbulent flow blurs the phase locking mechanism of the

L DV system. This experimental limitation happens when the flow co-

efficient is 0.07 for the Impeller V and 0.05 for Impeller IV.

2. 5.2 Pressure Transducers

Three _.ntran 9 miniature strain gauge pressure transducers

were purchased. These had a 1/8 in. diameter pressure sensitive sur-

face and could be flush mounted on the wall of the system. These

proved to be a complete failure due to the following reasons. The first

problem encountered was the thermal effect of the flow on the transducer.

It was found that thermal transients due to fluctuating fluid velocity also

changed the resistance of the sensing elements thus giving a faulty meas-

urement of fluctuating pressure. This effect which is not unsimilar to

the hot wire effect is amplified due to the small thermal mass of these

transducers. Further problems were encountered when mounting these

delicate transducers to the system. Although great care had been taken

during mounting, small releasing or tightening of the mountingmech-

anismchanged the ti'ansducer balance and thus required frequent re-

balancing of the transducers. Finally, all these miniature transducers

were completely damaged beyond repair due to this mounting difficulty

and were discarded.

9
Entran Devices Inc., 145 Paterson Ave., Little Falls, New Jersey
 -50-

Two Statham 10
0-100 psia pressure transducers were purchased

to replace the Entr_n transducers. These pressure transducers were

chosen to have a natural resonance frequency of 10,000 Hertz. This

value was many orders higher than the frequency range of interest in

the present experiment. In this design, the pressure sensitive dia-

phragm was recessed in a tube fitted with a 1/8 female pipe thread.

These transducers were connected to the smoothing chambers via short,

1/2 in. long, plastic tubings. The short plastic tube helped to isolate

the transducer fron_ the vibrations of the structure. The compliance

effect of the short plastic tubing was calculated to be small. Provisions

to bleed any air trapped in the pipe connection or in the transducer were

made. The signal conditioning systems used for these transducers

11
were home-made b,_lancing resistance bridges and Burr-Brown

36ZOL/16 operational amplifiers.

I

2.5.3 Accelerometers

From Section 2.4. 10 on foundations and Section 2. 5. 1 on LDV

it can be seen that it is necessary to monitor the vibration of the struc-

ture at the point of mass flow measurement. Two _4g Statham'acceler-

orr_ers were installed directly above the LDV viewing windows. The

accelerometerls balancing units were similar to the pressure trans-

ducers system and the amplifiers used were two Burr-Brown 3640

operational amplifiers. Velocities calculated from the readings of

10
Statharn Instruments Inc. , 2230 Statham Blvd. , Oxnard, Calif.
11
Burr-Brown Research Corp., International Airport Industrial Pk.,
Tuscon, Arizona
 -51-

*hese accelerometers were subtracted from the LDV readings to give

the velocity of flow relative to the structure. A typical value of the

structure vibration gave a typical velocity correction of less than 5

percent.

_. 5.4 Turbine Flowrneter

To calibrate the volumetric mass flow rate of the LDV, a tur-

bine flowrneter was installed in the circuit just upstream of the down-

_tream fluctuator. It should be noticed that this turbine flowmeter

-:ould not measure the fluctuating velocity because the response of the

=urbine was too slow to give any fluctuating reading.

12
The particular turbine flowmeter was originally used on a J2

_ngine and its calibration curve is shown in Fig. 2. 27. The output is a

'_eriodic voltage with frequency proportional to the flow rate. This fre-

luency signal was converted to adc voltage by a frequency to voltage con-

]3
vertor with a conversion ratio of . 1 volt/100 GPM.

2. 5. 5 Pump Speed Magnetic Pickup

To measure the pump speed, amagnetic pickup transducer was

situated i/8 in. from a sixty tooth steel wheel mounted on the pump

shaft. The number of voltage pulses per second is equal to the speed in

rpm and this voltag_ pulse is counted by a HP 530ZA frequency counter.

In order to measure the variation of speed during the experi-

ment, these voltage pulses were fed into a frequency to voltage con-

vertor. The output of this frequency to voltage convertor was a dc

]'2 Potter Aero Company

13 Teledyne Philbrick (Model 4702), Dedham, Massachusetts

 -52-

voltage output adjusted to l volt/1000 rpm and any variation of the

fluctuating frequency in the pump speed would appear as small ripples

of the dc output (see also Section 5.4).

Z. 5.6 Axial Motion Proximity Transducer

To measure the axial displacement of the pump shaft during

14
fluctuating experiments, a Karman proximity transducer was pur-

chased and located on the pump volute with its sensing surface facing

the flat surface of the geared wheel. Any axialmovement between the

volute and the shaft would appear as a voltage proportional to the dis-

placement which was due to the slight yielding of the precision bearing

cartridge. This voltage output was connected to the Bafco signal

analyzer to measure the fluctuating displacement. No readings were

observed during a typical two percent upstream mass flow rate experi-

ment using these instruments. Since the resolution for the displace-

ment transducer and analyzer combination was 0. 0005 in. , this meant

that the axial displacement of the shaft relative to the volute is less

than 0. 0005 in. which was within the design requirement of the bearing.

2. 5.7 Amplifier and Filter Bank

The original signals like LDV output and pressure transducer

output were usually in the form of a small ac signal riding on a large

d c voltage. These signals also had a high noise content with values

comparable to the ac signals. It was intended to record these signals

14
Karman Sciences Corporation, P.O. Box 7463, Colorado Springs,
Colorado.
 -53-

by instrumentation tape recorders (see Section Z. 5. 10) and analyze by

_he Bafco signal analyzer. This presented difficulties as the tape re-

cording accuracy of the small ac was limited by the dynamic range of

the recorders. The accompanying recording of the large dc voltage

_llowed only a low amplifying factor on the tape recorder. The Bafco

analyzer (see Section 2. 5.9) also encountered difficulties as this anal-

_rzer had to filter out the large dc component and reject the high noise.

To improve the situation, the dc component was rejected and the

_c signal was amplified to a higher level such that the recorder would

not be saturated bE the dc but instead record a h_gh level of ac signa!.

Since the dc values were not recorded, the values were displayed on an

integrating voltmeter (see Section 2. 5.9) and written on data sheets.

Furthermore, the noise was filtered by a low pass filter. The dc re-

jection, ac amplification and noise filtering were accomplished by a

bank of ac filters followed by a bank of low pass filters. Six ac ampli-

fiers made from. conventional 741 operational amplifiers and six two

stage, four pole butterworth filters with a cut-off frequency of about

80 Hertz were fabricated to condition the six signals. The six signals

were upstream and downstream velocities, upstream and downstream

pressures and upstream and downstream accelerations.

However, these amplifiers and filters had their own character-

istic dynamic resp,_nse thus contributing an amplitude and phase change

at different frequencies. To measure these amplitude and phase

changes, a given ac voltage signal was connected to the individual com-

bined amplifier and filter input and the ac output signal was measured

at various input frequencies. The ratio of input and output amplitude

 -54-

was calculated. These, together with ratio and the phase change, were

plotted on graph paper. A curve derived from the filter characteris-

tic formula was fitted to these data points and a typical curve is given

in Fig. 2. 28. The fitted curves of the combined low pass filter and

amplifier were used in the data reduction procedure to calculate the

original signals.

15
_..5.8 Bafco 910 Signal Analyzer

This instrument is the most important one in the equipment. It

gives out a precise sine wave as reference at a pre-set frequency to

drive the fluctuator motor. Any fluctuating signal to be analyzed is

connected to the input end. The Bafco measures the in-phase and

quadrature component of the fluctuating signal rejecting the noise pres-

ent in the signal. The in-phase and quadrature voltages are displaced

_n two taut band meters. For easier reading, rear panel outputs of

the in-phase and quadrature voltages were connected to two digital

panel meters and readings were taken from these.

The above description is for real time analysis of the signa I.

For recording playback, an additional tracking unit was needed. This

tracking signal accepts the playback reference signal and sets the Bafco

to exactly the same frequency so that there is phase reference between

the Bafco and the signal to be analyzed. _The signal to be analyzed is

connected to the normal input terminal of the Bafco.

Forty db of noise and harmonic rejection were obtained by using

this analyzer. The performance was guaranteed at a signal to noise

level of one to one with a one percent accuracy on the measured

15
Bafco Inc., 717 Mearns Road, Warminster, Pennsylvania
 -55-

in-phase and quadrature measurements. This value dominates the

accuracy of the final results.

2.5.9 Integrating Voltmeter

To measure the mean dc component, an ordinary dc voltmeter

was not suitable because most voltmeters measure an instantaneous

voltage or measure the mean within a very short time interval and this

would not give a true mean voltage because of the low frequency used

in this experiment.

16
A NLS integrating voltmeter was incorporated into the system

and all the mean dc voltages were measured using an integrating thne

of one second. The one second integrating time averaged out all the

small fluctuating ac components.

2. 5.10 Recorders

The real tirne data analysis by the Bafco analyzer was time con-

suming and the .data were subjected to error due to the slowly drifting of

the mean operating state and the accumulation of air between measuring
17
stations during cavitating tests. Two HP-3960 instrumentation time

tape recorders were provided to record the signals during the experi-

ment. The recordings were played back at leisure after the experiment.

Each of the tape recorders had 4 channels and the recordings

were accomplished by always recording the reference signal on the

16
Non-linear Systems, Inc. , Model No. 2900, Del Mar, California
17
Hewlett Packard Company, 690 East Middlefield Road,
Mountain View, California
 -56-

first channel and fluctuating signals on the second, third and fourth

channels. The six recorded signals were two LDV signals, two pres-

sure transducer signals and two accelerometer signals. These tape re-

corders were FM recorders with a cut-off frequency at about 300 Hertz

and the recorders were checked to insure that there was negligible

phase shift between the individual recorded data channels.

Since the recordings were simultaneous for all channels' the

playback analyzed the signals for the same recorded time. This method

_us only demanded a shorter time for experiment and reduced the time

exposure of the inducer to the cavitation damage. Futhermore, shorter

experimental time meant that the drifting of mean state and air accumu-

lation in the system was minimized.

During playback, the playback reference signal had occasional

high frequency spikes and these spikes disturbed the tracking mech-

anism of the Bafco tracking unit. A low pass filter with a cut-off fre-

quency of about 500 Hertz was installed between the tape recorder ref-

erence output and the tracking unit. Although this low pass filter intro-

duced a phase shift,this was the same for all the channels and hence was

not corrected for in the analysis since we were only interested in the

relative phase between all the fluctuating signals.

2. 5.11 Selection Switch and Panel

All the outputs of the instruments were connected to a central

panelwhere BNC connections were provided in the front. These

signals were connected to the filters and ac amplifiers. The outputs

of the amplifiers were brought back to the panel and connected to

REPRODUCIBILITY OF TH_
0R/CANAL PAGE l,qP00_
 -57-

_nother set of BNC connections. A selection switch was installed on

he panel such that any signal on the panel could be connected to the

3afco analyzer and dc integrating voltmeter by turning a selection dial.

_See the flow diagram, Fig. 2. 25).

_. 5.12 Manometers and Pressure Gauges

Pressure gauges and water over mercury manometers were

-onnected to the pressure transducer locations for two purposes. The

'irst purpose was to provide the static calibration of the pressure trans-

ducers. For this a precision Bourdon pressure gauge provided the

:alibration above a_,_ospheric pressure and the manometers provided

:alibration for preesures below an atmosphere. The second purpose

was to measure the static pressure during experiments. This may be

redundant but prove,:1 necessary to ensure that the pressure transducers

_ere functioning properly.

Z. 5.13 Stroboscope Synchronizer

A stroboscope synchronized drive was constructed to keep the

inducer blades stationary visually. This device takes the pulsed signal

from the pump speed magnetic pickup, divides it by sixty to give one

pulse per revolution and the divided pulse train is used to drive the

stroboscope. Provisions were made on the driver such that a phase

lag can be imparted to the driving signal and the inducer can be

viewed from another angle.

 -58-

L
00
_J

4_

';i

w -,
.._

Z w
..,-i
,...o

_J
W
e-,
n- o
O_

b_
0 ,,j

W
I
0
o0

I_-_PRODUCIB[LITY OF _ "-
0RI_._L PAGE IS _ 7F
 -59-

>

c_

.IJ

{J

.-¢2

%.,4

L)
°,,

,,-,-
,p,1

1"4

4
°,,_
 -60-

f-t

t3

..C:

b_

Z
0 U

I-
F
-I z
0,_ t3

N"

°_,=_

_>
 -61-

Z
o
I.-
<Z O
_D V

I-- L:20
O Z
:D6- O
m
_J
LL <_ L:I6
LIJ
11:: 5 ' cO
23
O9 L=I2
03 _J
W D_

_4- 38-
O
_J
Lt_ 0
.co
I_ 3 - 036-
03
Q. /

1_2- 04-
0
w
L_J (.9
r'h
I- 72-
W
_J o
Q. cr
w
<0- nO I
0 10
FREQ (Hz)

Fig. 2.4 Results of the mathematical pump loop showing amplitude

of upstream pressure fluctuation for the same upstream
mass flow fluctuation. The curves were calculated from
various total hydraulic pipe lengths in feet when the pipe
diameter is 4 in.
 " -62-

PLAN VIEW

Fig. Z, 5 Schematic drawing of the dynamic pump test facility.

REPRODUCIBILrrY OF TI_,_
ORI(IgNAL PAGE IS POOR
 -63-

L_;FRODUCIBILI'I.'Y OF Tt
ORIQiNAL PAGE IS P00E
 -64-

-" 2 995 in "-

0.86 in.

L
T,E°

Fig. 2.7 Schematic drawing of Impel]e, Ill,

-65 -

I-.(

,-4

L_
O
..C

0
G

co

°,.._
-66 -

r •

Q_

_q

t_J

.,-4
q_

°L

r_

_sJ
o

t.-

c_

ol-6
 -67-

2.985 in.

?
1.82 in.

,!

Fig. 2. 10 ScherrxatJc drawing of h-npell_r I'V.

-68-

,-q
_J

.c"
O4

t_
O
_J
O

0.

o,q

REPRoDUC_BLLIq_X OF THg
O1R,].(_qA L pAG_ IB pOOR
 -69-

2 985i "-

l ; FLOW
2.47 in.

1.163 in

T°E.

Fig. 2. 12 Schernatic di'awing of Impeller V.

-70-

,.-j

O
c.

c_

_,_ODUCIB]I, IT¥ OF _
 -71 -

0
\

O_
-"/2-

t13

,4
 -73 -

E
e_

1-/3

e_
0
0

°_,,1

|
-g.

?.

_4

°,,,,I
-?4-

g-.(

,4
N
o

p,-

i//1
-7.5-

U
 -76-

t
f

i
I-.
l.J

=D))))))))))),;)))))i)))_-,
-'/7-

$
4,,)

u
c)

0
-?8-

t_

¢;

_J

E
Q_
0

Q;

0
°_.¢

m
cJ
o

.,._
 -79-

I I I"' 1 t I ._
0.40 -
(.9
Z
m
LOCATION AND DIRECTION a i
0.55 - OF ACCELEROMETER J t -
Z i I
0
p-
0.50 - I -
! I
_aJ

iii 0.25 - CROSS-MEMBER I I_=

o BRACING I
(D I

0.20 -
CI_
I BEFORE BRAClNG--.__ -
n
I
i1 0.15 - • -
0 AFTER BRACING I
I
i11 I
r_ 0.10 -
H-

n-j0.05
I

I I I I
0o 5 I0 15 20 25 50 55
FREQ (Hz)

Fig. 2.22 Vibration levet of the fo,mdation Jn direction parallel to

the pump axis.
 -80-

I I ! I I I
m i

o 0.20

z 0.18

MEMBER BRACING -
0.16 _.,T-._..._ LOCATION AND
0.14 ""- "' DIRECTION OF -
ACCELEROMETER
,,-fi,
0.12 I! .

! _ BEFORE I_
<I: 0.I0 t '_. BRACING _-
I I aj,'l I
n 0.08
IV l I •
0.06
I fill/ i" /IQ
_ I •tl : l -,
IPuQO I I
o.o4
0.02

I I i I I
0o 5 .10 15 20 25 :50 55
FREQ (Hz)
Fig. Z.23 Vibration level of l he foundation in direction perpendicular
to the pump axis.

R_RODUCIBILITY OF THE
ORII_NAL PAGE IS POOF.
 -8! -

(/)
ILl

n,'N OZ
nO
¢r) Z

QJ

f
i

0
°,,._

b2
p_

IM

_jLIJ
_U.j
W_
oO <_(.9
_,_
rv-v
_ nr-nr" OOZ
ILd_ 0 I-- ,,_
nr-i.-n I_ ¢/') I--
ILl
-82-
 -83-

UPSTREAM LDV

E2
DOWNSTREAM LDV --'J

5 I0 15 20 25 50 55 40 45
FREQ (Hz)

_DOWNSTREAM LDV

PHASE
ANGLES

UPSTREAM LDV

Fig. 2.26 1;'rtquency re._ponse transfer functions of the two

LDV systems.
 -84-

2O I I I I I I I

18

16

14

I l I I I i
_00 _50 200 25o 3oo 35o
FREQ (Hz)
Fig. 2. 27 Calibralion curve of the Potter turbine f]o_m_eter

iP. w_ter.
 -85-

I I I I

4.0 A A
v

BOTH STAGES

3.0

2.0
J

I I I I i

-20

-40
_D
Z

LLI -60
U_
<[
I
D_
BOTH STAGES
-80 -

-I00 -

I
O IO 20 30 40 50 60

FREQ (Hz)
Fig. 2.28 Co1_.,bir_d filler and ac a_nplifier frequency response
transfer funct:ion.
 -86-

Ill. EXPERIMENTAL PROCEDURE

Two types of experiments were performed, the steady and the

unsteady experiments. The steady state experiment established the

steady fully wetted and cavitating performance which in turn could shed

some light on the unsteady performance of the pump. The unsteady ex-

periment measured the transfer matrix for the inducer pump. As will

be described in thi_ and the following chapters, this matrix was meas-

ured both for fully wetted and cavitating flow,

3. l Stead), State Tests

3. i. 1 Fully Wetted Performance

This experiment measured the performance of the inducer pump

under fully wetted flow. The result will be expressed as the dimension-

less head coefficient against flow coefficient. To avoid cavitation, the

system was subjected to a high static pressure by pressuring the air

bladder. The usual pressure used was about 40 psig as constrained by

the design strength of the system. However, to achieve completely the

fully wetted condition, the pump speed had to be kept under 6000 rpm.

During the experiment, a certain pump speed was selected. The

flow rate as measured by the turbine flowmeter was varied by activat-

ing the silent throttling valve. The pressure rise across the impeller

was read from a mercury water manometer connected across the

system. From the pump speed, flow rate, and pressure rise across

the pump, the head coefficient and flow coefficient could be obtained.

V
 -87-

3. 1. Z Cavitating Performance

The pump speed was kept at a constant value. The flow rate

was also kept constant by activating the silent valve servo feedback

system. The cavitating state was varied by adjusting the static pres-

sure of the air bladder. In addition to the measurement of the pres-

sure rise across the pump, the absolute pressure upstream of the pump

was measured by a mercury manometer relative to atmosphere. A

standard barometer provided the absolute reference. The head coeffi-

cient was plotted as a function of cavitation numOer for a constant flow

coefficient.

3. 2 Fluctuating Experiments

Before each fluctuating experiment, the following quantities had

to be decided in advance. These were:

(a) Flow c_efficient

(b) Cavitation number

(c) The values of fluctuating frequency. In this study, frequen-

cies used were 4, 7, 14, 21, 28, 35, 42 Hertz.

(d) Pump rotative speed in rpm.

3.2. 1 Preliminary Preparation

To begin th,_ experiment in cavitating flow, the water was deaer-

ated for a period of two to three davs. This reduced the air content to

a molar concentration of about three to four parts per million as meas-

ured by a dissolved oxygen meter. 1 After the water was brought up into

1
Delta Scientific, Model 1010, Dissolved Oxygen and Temperature
Meter
 -88-

the test facility from the storage and deaeration system, it was circu-

lated slowly around the loop for a few hours with the L-npeller running

at a slow speed. These few hours of slow turning allowed enough time

for the rotating seal to be wetted and for any air bubbles to float to the

top and to be bled away. The heat exchanger cooling system was turned

on during this period and cooled the water to a steady state tempera-

ture of a-bout 68"F. The low temperature of water increased the solu-

bility of air in water thereby reducing the release of dissolved air from

water in the system. Such elaborate air content treatment was needed

because it was suspected that the amount of air content played an im-

portant role in the inception of cavitation and in the development of

cavitation in general. This procedure was not necessary for fully

wetted experiments. The oil pu.rnp for the silent valve servo feedback

control was switched on ready to control the flow rate. The last prepa-

ration was to check that the weak laser beam entered the photomulti-

plier pin hole.

3.2. 2 Calibration

The calibration of the LDV is shown in Appendix B. To cali-

brate the pressure transducer, connections to the mercury manom-

eters were bled to be free of air bubbles. These manometers were

used for pressure transducer calibration as well as a supplementary

system for steady pressure measurement. The pressure transducers

were statically calibrated in place for each test by applying a pressure

or vacuum to the air bladder. The vacuum readings were taken from

the manometer and the high pressure readings were taken from a
 -89-

precision Bourdon gauge. Typical calibration curves are shown in

Fig. 3. 1. The accelerometers were calibrated by orienting them rela-

tive to gravity in order to register signals corresponding to + one g.

The tape recorders were calibrated internally, then a reference sine

wave with a given amplitude from the Bafco was recorded on all chan-

nels for calibration during playback. The recording speed was 15 in.

per second. A slower recording speedwouldsave the amount of tape

used but would give a very unpleasant noise during playback in the

voice channel which was used for identification. The reference sine

wave signal used t=, drive the fluctuators was fed into the first channels

of both recorders. The six fluctuating signals from the electronic low

pass filter bank were connected to the second, _hird and fourth channels

of the two recorders. Finally, the monitor outputs of the tape record-

ers were connected to the oscilloscope.

3. 2.3 Test Operations

(i) The required steady mass flow rate which controlled the

flow condition and the upstream pressure which controlled the cavita-

tion number were calculated for a given inducer rotating speed. The

main motor control was switched on and then adjusted to the desired

speed. The flow rate was dialed from the variable resistance of the

• silent valve hydraulic control feedback system. The pressure of the

air bladder was v_ried until the required upstream pressure was

attained. The phase lock frequency was adjusted on both the LDVIs

processors until the doppler signal was properly locked by the phase

lock loop. This concluded the steady state adjustments of the system.
 -90-

(ii) For the fluctuating adjustments, the following items had to

be decided in advance: the fluctuator combinations and the frequency of

fluctuation. The necessity of several fluctuator combination sets was

discussed in the introduction and system design. These fluctuator com-

binations usually consisted of three or four sets. They were: upstream

fluctuator only; downstream fluctuator only; both fluctuators operating

simultaneously with each one contributing about half of the required

amplitude and lastly both fluctuators operating simultaneously with a

phase shift from the last setting of IZ0 degrees.

(iii) To suit the amplitude requirement the Bafco output signal

was dialed to amid-range frequency of, say 20 Hertz; and the fluctua-

tor motor was then switched on. About two minutes was needed for the

fluctuator to reach its final speed. This was indicated on the phase

error meter when it reached a null condition. For the particular fluc-

tuator combination desired, the sleeve on the appropriate siren valves

was adjusted until the upstream mass flow fluctuation had reached two

percent amplitude level. This was determined by the percentage ratio

of the vector sum of the in phase and quadrature component to the dc

value of the mass flow. All these readings were taken from the up-

stream LDV. The. setting which gave this condition would be used for

all other frequencies with the same fluctuator combination. Although

the other frequencies produced different amplitudes, they were of the

same order of magnitude, proved to be within the linearity requirements

and acceptable for data analvsis.

(iv) The shut off valve from the oil pump to the silent valve was

closed so that the oil pressure in the silent valve would be maintained

REPRODUCIBILITY OF THE
n'p.laINA.L PAGE IS POOR
 -91-

at the desired level and the pump speed was reduced to about 2000 rpm

from the previously adjusted higher speed. This valve would then be

reopened when the pump speed was raised again. This procedure was

adopted for a faster silent valve response time because the oil pump did

not have to pump from zero to full pressure when the valve was reopened.

The whole maneuver attempted to minimize cavitation damage to the £m-

peller as well as wear and tear to the system.

(v) The fluctuating frequenc Y was then adjusted to the first re-

quired frequency. The pump speed was raised to the test speed after

the fluctuators had reached the selected Irecluenc ¥, The silent valve

_hut off valve was reopened. The bladder pressure was readjusted to

the desired value if needed. Air bleed valves were temporarily opened

_:o let any trapped air to be bled away.

(vi) The two instrumentation recorders were turned on to record

for two minutes at irecluencies below 14 Hertz or one minute for fre-

quencies above 14 Hertz as these were the times needed for the tracking

-chit to work proper!y during subsequent playback. During the recording

time, all the dc levels of the velocities and pressures, which were indi-

cated on an integrating voltmeter with a one second integrating time,

were taken together along with the rotational speed and turbine flow-

meter readings. Photographs of the impeller and oscilloscope traces

were taken at this time if needed. After the recording was finished,

the pump speed was slowed down to about Z000 rpm and the next fre-

quenc_r of fluctuation was dialed on the Bafco ouptut. After the fluctua-

tot had achieved the set frequency, the same procedure as described
 -92-

in this paragraph was repeated. This procedure was again repeated

_or all other frequencies.

(vii) Next, new fluctuator combination sets were chosen and im-

plemented as mentioned earlier and followed by frequency adjustments,

recordings, etc. After all the fluctuator combinations were completed

and recorded, the experiment itself was finished.

3.2.3 Data Analysis Procedure

To begin the data playback from the recorders, the reference

signal, which was recorded on channel one of the recorders, was con-

nected through a low-pass filter to the tracking unit as described in

Section 2. 5. 10. The recorded ac calibration signal was played and the

output calibration knob of the recorder was adjusted. This was re-

peated for all chan_els.

The recorded signals which were the amplified, filtered ac sig-

nal of the velocity, pressure and acceleration were played back one

channel at a time and their in-phase and quadrature voltage values rela-

tive to the reference on the tape were taken and written on tabulated

data tables. Four random voltage readings of each recording were

taken and their average value was computed and used. These ac values

together with the dc values taken during the experiment were punched

into computer cards to be processed by the computer.

3.3 Impeller V Experiments

The procedures mentioned in Section 3.2 were the final proce-

dures developed and used for Impeller IV. Prior to these experiments
 -93-

and during the development of the experimental methods, date were ob-

tained on another impeller, the Impeller V. During that phase of the

research, the data acquisition and reduction procedure were somewhat

less refined. Thus, for example, there were no electronic filters to

reduce the high frequency noise. Consequently a lower signal-to-noise

ratio had to be analyzed by the Bafco signal analyzer. Tape recorders

were not available at that time and all data processing had accomplished

and logged in real time. This was done in the shortest time possible in

order that the steaav state conditions such as pump speed and flow rate

did not drift away from the set value. ClearIythelaterdatawere better

when the fluctuating quantities were recorded simultaneously and there-

fore analyzed at th_ same instant in time. An equally serious problem

during the cavitating test is the possibility of small air bubbles accumu-

lating in the downs'-ream smoothing section despite frequent bleeding

during an experiment. Although it is believed no such bubbles were ac-

tually collected, le_Lgthv test times make this chance more likely. The

experiments for Impeller V were carried out during the development of

the entire system. Nevertheless the results for this impeller are of

interest _n their own right and will be presented in Chapter VI for a

qualitative comparison with results from Impeller IV. However, for all

the reasons menticned, the reliability and accuracy of these data are

lower than for the _mpeller IV experiments.

 -94-

PRES. TRRNSDUCER
U-.STRTItFIH
0 CRL ] BRRT I ON Ctt-3 T_,,17Ot/1
t:)
I I ! i I

o
t"M

0
Q

IQD

Vt'_
•-,,s

LIJ

f,f'} 0

(1..

I=,
I

C) --

C:) --

/
/
0
0
=1'
! I I I I I
-2.0 O.O 2.0 I.l,.O 6.0 O.O 10.0
VOLTRF.,-E( VOLTS)

Fig. 3. 1 Typical calibration curve of a pressure transducer.

 -95-

IV. DATA 'REDUCTION

This chapter describes how the transfer matrix across the p_np

•_as calculated from the fluctuating pressure, velocity and acceleration

.-measurements. The data reduction was done in three steps and these

were accomplished by three computer programs. The necessity of

three separate programs will be evident in the following description.

Flow charts are provided in Appendix C. The first program,

RAWDATA, calculated the required dimensionless groups using the volt-

age readings taken fror_ experiment. The second program, TRANSCO,

-.-alculateda total transfer matrix between measuring stations and the

).astprogram, PUMA, reduced the total transfer matrix to the transfer

_natrix across the pt_p alone.

After playing back the recorded signals from the tape recorder,

the following quantities were punched onto computer cards:

(a) calibration data for pressure transducers.

(b) acceleration calibrations.

(c) mean values of (i) rpm (pump speed in revolutions per rain. ).

(ii) gpm (flow rate in gallons per minute).

(d) the in-phase and quadrature voltage readings of upstream and

downstream accelerometer s.

(e) the d-c, in-phase and quadrature voltage readings of the up-

stream and downstream laser doppler velocimeters.

(f) the dc, in-phase and quadrature voltage readings of the up-

stream and downstream pressure transducers.

 -96-

The analysis was set up to process a batch of data at the same

mean rotational speed, flow rate and upstream static pressure. This

implied that the head coefficient, flow coefficient and cavitating number

were identical. The cards were arranged in groups of the same fluctua-

for combinations following the experimental procedure and each of these

groups contained data for a complete set of fluctuating frequencies.

4. 1 RAWDATA Program

This program established the conversion factors of the instru-

_aaents from voltages to the appropriate units. The pressure transducer

and accelerometer calibration curves were calculated in the subroutine

CALIB. The LDV velocinoeter to volumetric flow rate conversion fac-

;-ors were obtained from comparing the dc voltage readings of each laser

doppler velocimeter to the turbine flowrneter reading. All the above cal-

culations were done in the MAIN program.

During the recording of the signals, all six of the ac signals

passed through a bank of amplifiers and electronic filters. The effect

of the amplifiers and the filters was corrected in the CORR subroutine.

This subroutine also corrected for the transfer function of the LDV

processor.

Based on the corrected in-phase and quadrature voltage readings

and the conversion factors obtained, the following quantities were cal-

culated:

(a) The upstream and downstream fluctuating pressures. They

were expressed in feet of water for amplitude and degree for phase angle

with respect to the sinusoidal reference signal.

 -97-

(b) The upstream and downstream fluctuating mass flow rates.

The in-phase and quadrature voltages were converted to velocities in

feet per second in the four in. viewing sections. Since the LDV meas-

ured velocities in an inertial reference frame, the accelerometer read-

kngs were converted into velocities and subtracted from the LDV

measurements. The calculated results were velocities relative to the

structure. The amplitudes of the fluctuations in velocity, expressed as

percentages of mean flow rate, were also calculated.

These eight quantities, upstream and downstream fluctuating

mass flow rates and fluctuating pressures expressed in amplitude and

phase angles were plotted against the frequency of fluctuation. These

graphs were checked for smoothness of the data. Any data point which

did not appear to fi: into a smooth curve would be rechecked for errors

in recorder playback, averaging process and/or key punching. Head

coefficients, flow coefficients and cavitation numbers were also calcu-

lated based on the mean values of pressure, flow rate, and rotational

speed.

The RAWDATA program finally calculated the fluctuating pres-

sure, fluctuating flgw rate and frequency in dimensionless form (see

nomenclature). These computations were done in DIMGP subroutine.

The calculated values were punched on computer cards; these cards had

to be rearranged in groups of the same frequencies to be processed by

the next program. However, within each group there was a complete

set of different fluctuator combinations.

A subroutine COMP existed in the program to calculate the com-

pliance and pump resistance as defined by Rubin. These were useful in

 -98-

+.he beginning of the project to understand tfle measurements, especially

in the fully-wetted experiment. However, these calculated quantities

were not analyzed in the final stages of the project.

4. Z TRANSCO Program

The computer cards with the dimensionless pressure and mass

flow rate values were reorganized in groups of the same frequency of

fluctuation as mentioned in the last paragraph. The program treated

the in-phase component as the real and the quadrature component as the

imaginary part of a complex quantity. Then the TRANSCO program cal-

culated a Z-matrix for each individual frequency using a special least-

square fit process. The details of the least-square fit procedure are

outlined in Appendix D. The real and imaginary parts of the elements

of this Z-matrix were plotted as a function of frequency. Their values

were also punched on computer cards to be processed by the next pro-

gram. The TRANSCOprogram also had the capacity to calculate the

Z-matrix using only a partial number of fluctuator combinations as long

as the total number of combinationswas greater than two. This was

used in the fluctuator combination linear independency analysis as de-

scribed in Section 5.7.

The Z-matrix so obtained was not the transfer matrix across

the pump as will be discussed in Chapter VI; it actually represented the

transfer matrix between the measuring stations. The differences be-

tween this Z-matrix and the matrix across the puxnp ZP consisted of the

inertia of the water and compressibilities of water and structure situ-

ated between the measuring stations and the pump. A third program
 -99-

PUMA was thus written to subtract the side effects and obtain the trans-

fer matrix for the pump alone, defined as the ZP-matrix.

4.3 PUMA Program

This program reduced the Z-matrix obtained in the previous pro-

gram to the transfer matrix of the pump alone. The transfer matrix

between the measuring stations was assumed to be composed of three

parts; the first part was the effect of the hydraulic system between the

upstream measuring station and the beginning of the pump and this was

expressed as an upstream matrix ZU; the second part was the effect of

the pump itself and was expressed as the required pump matrix ZP and

lastly the downstream matrix ZD was the transfer matrix connecting the

end of the pump and the downstream measuring station. The contents of

the ZU and ZD matrix were derived from dynamic fully wetted experi-

ment and will be given in Chapter VI. The above mentioned ZU, ZD and

Z-matrices which were calculated from TRANSCO program were used to

calculate the required ZP pump matrix with the following scheme.

The three matrices ZU, ZD and Z were first transformed into

YU, YD and Y matrices defined as

Y =Z+l (4.1)

where I is the identity matrix. These Y-matrices had the normal trans-

"fer property that they satisfied the chain rule. Thus,

Y = YD. YP.YU (4.z)

The YP-matrixwas simply obtained from

-I -I
YP = YD • Y. YU (4.3)

REPRODUCIBILITY OF TIIE
0RIOglqAL PAGE IS POOR
 -I00-

The required pump matrix ZP was then obtained from

ZP = YP - I (4.4)

All these processes were carried out in the PUMA program

using specially written routines f6r complex value matrixmanipula-

tion. The real and Emaginary parts of the elements of the ZPmatrix

were finally plotted as a function of frequency.

 -lOl-

V. PERFORMANCE

5. 1 Steady State Performance of the Inducers

5. 1. 1 Full), Wetted Performance

(I) Impeller III and V - The graphs of head coefficient against

flow coefficient (for definition of these see nomenclature} for

these two impellers are shown in Fig. 5. 1. There is only a

slight performance difference between the two impellers within

experimental error although the Impeller V has swept back lead-

ing edges. The agreement with Acosta's (1958) two inch

impeller upon which the design of the present impeller was

based is excellent. The theoretical calculation by Acosta (1958)

based on a nine degree straight cascade is also presented for

comparison.

(II) Impeller IV - The graph of head coefficient against flow co-

efficient for Impeller IV, the quarter scale model of the low

pressure oxi4izer pump of the Shuttle main engine is shown in

Fig. 5. Z wit}', similar results for a 0.6 scale test impeller

carried outby Rocketdyne (1974). It may be seen that the two

curves are not the same. This is very surprising because these

impellers are made from the same master tooling and the

Reynolds number difference is insufficient to account for the

discrepancy. The differences in the designs of the volutes may

be responsib]e for this discrepancy.

 -i02-

5.1.2 Cavitating Per fo rman ce

(I) Impeller V- The head coefficient against cavitation number

curves for constant flow coefficient are shown in Figs. 5.3 and

5.4 for different pump speed. All the curves show that the head

coefficient decreases as the cavitation number is reduced. This

cavitation breakdown is more severe when the flow coefficient

is lower. It can be seen from the curve that the slope at

breakdown increases more rapidly for a lower flow coefficient.

(11) Impeller IV - A cavitating performance curve similar to

Fig. 5.3 etc. for Impeller IV was not obtained because after

the fatigue failure of Impeller Ill, we tried to reduce the operat-

ing time of Impeller IV to avoid material failure. Indeed the

blade thickness of this impeller is particularly thin. However,

a cavitating performance of the 0.6 scale test impeller as per-

formed by Rocketdyne (1974) is given in Fig. 5. 5 and the quarter

scale model of Impeller IV is expected to perform similarly

under cavitating conditions.

5.2 Fluctuating Quantities

Typical fluctuating quantities measured during fluctuating experi-

ments are discussed here.

5. 2. 1 Fluctuating Velocity

Three typical curves of zero to peak amplitude velocity fluctua-

tion expressed as a percentage of the mean flow are given in Figs. 5.6,

5.7 and 5.8. A11 these figures come from an experimental run with
 - 103-

LTnpeller V for a cavitation number of about 0. 1. Each figure corre-

sponds to operation under one fluctuator combination setting at various

frequencies. Figure 5.6 corresponds to exercising the upstream fluctua-

tor only; Fig. 5.7 corresponds to activating the downstream fluctuator

only; and lastly Fig. 5.8 shows the velocity amplitude when both the

fluctuators are operating. It can be seen that the amplitudes are all

:under four percent and mostly under three percent of the mean flow.

The phase angles are all measured relative to the reference signal

which drives the flactuator motor.

5. Z.Z Fluctuating Pressure

The fluctuating pressure amplitudes and phase angles for the

same experiment described above are given in Figs. 5.9, 5. I0 and

5. II. It can be seen that the zero to peak amplitude of pressure fluctu-

ation is normally below 10 ft. of water. All the figures from 5.6 to 5. 1 1

were actually used to calculate a ZP-matrix for experiment (3) in

Section 6. 1.

It is convenient at this point to clarify one very important point

concerning the difference between oscillating static pressure rise and

_scillating total pressure rise across the pump. Since the pressure

transducer connections to the smoothing chambers were through a side

_ort on the structure wall, the static fluctuating pressure rise is actu-

ally measured. However, from the unsteady ]Bernoulli equation this

could be shown to differ from the total fluctuating pressure rise by a

very small arnount. The difference is represented by the term

U u'e jwt and this is found to be small compared with p'e jwt in this
 -104-

kind of flowwhere U is the mean flow and u'ejwt and p'ejwt are the

fluctuating flow and fluctuating pressure respectively. Therefore, no

distinction will be made between them in the present experiment.

5.3 Auto Oscillation

During the cavitating operation of the Impeller V, a self-sustain-

ing oscillation of velocity and pressure sometimes occurred. This is

not unlike the pogo oscillation encounters in liquid propellant rockets.

The auto-oscillation may be an intermittent phenomenon of about two to

three seconds duration or it may be a continuous phenomena. A typical

photographic picture of the cavitation and the corresponding fluctuating

velocity and pressure is given in Fig. 5. IZ and the steady state condi-

tion when this occu_'s is shown in Fig. 5. 3. In this particular case, the

upstream fluctuating velocity reached an amplitude of 22 percent of the

mean and the corresponding downstream amplitude was Z7 percent. The

fluctuating pressure rise was 47 ft. of water which was almost equalto

the steady pump head rise of about 50 ft. and all these occurred at a

frequency of 21 Hertz.

The boundaries of operating conditions when this occurs are not

known precisely, but it usually occurred when there existed intensive

backflow cavitation (as shown in photographs taken during the test)and

the Cavitation number dropped below 0. I. However, the auto-oscilla-

tion usually died down when the cavitation number was sufficiently low

so that the impeller suffered a complete cavitation breakdown.

 -105-

5.4 Pump Speed Variation

The variation of the pump speed to the fluctuating flow on the

inducer is important because a change in the pump speed means a

change in the angle of attack and this introduces one more variable to

the problem. The mean speed of the electric motor varied less than

0.8 percent about the preset speed. This driftwas usually less than

0. 5 percent as measured by the frequency counter.

The magnetic pickup output was connected to a frequency to

voltage converter with an output proportional to the pump speed. The

voltage proportional to pump speed was displayed on the oscilloscope

and shown to have a definite fluctuation at low fluctuating frequency. A

typical value was 0. 1 percent rpm fluctuation at 5 Hertz when the up-

stream mass flow fluctuation was about 2 percent. At frequencies high-

er than 10 Hertz, the pump speed variation was too small to be meas-

ured and is thought to be less than 0.05 percent.

Frornthe analysis by Brennen and Acosta (1975), this pump

speed variation effect is negligible. This is because the percentage

pump speed variation is to be compared to the dimensionless group

Pu
as defined in their paper. This numerical group when
2
PUti p o
evaldated for this experiment had a typical value of 30 percent which is

much higher than the measured O. 1 percent.

5. 5 Pressure Drop Across the Smoothing Chamber

The pressure transducers were attached on the walls of the

smoothing chambers. "I_the parametric evaluation of the system as

 - 106-

will be discussed in Chapter VI, it is assumed that there was uniform

pressure within the smoothing chamber. Only a small pressure drop

was expected due to the effects of the perforated metals, honeycomb

and screens.

The steady pressure drop across the chamber was measured by

a carbon tetrachloride water manometer for flow rates within the exper-

imental range (Fig. 5. 13). It can be seen that this steady pressure drop

was very srna11. To verify that the fluctuating pressure drop was also

small, two pressure transducers were attached on the first half and the

.qecond half of the smoothing chamber to measure the pressure drop

across the perforated metals, etc. During a typical flow of 180 gpm

and an axial velocity, of about 0. 5 ft/sec inside the chamber and a 2

percent mean speed fluctuation, there was no measurable steady

nressure difference above 0. Z psi and no fluctuating pressure difference

above 0.2 psi amplitude. These values are the accuracy limit of the dc

/oltmeter and Bafc_ signal analyzer. This demonstrated that the fluctu-

ating pressure drop across the smoothing chamber was negligible and

that the pressure within the chamber could be assumed uniform.

5.6 Accuracy of the Result

To assess the accuracy of the measured elements of the trans-

fer matrix, three essentialpoints will be discussed here. These are:

the accuracies of the instruments themselves; the capability of the sig-

nal analyzer to analyze signals with a high noise content; and the sta-

bility of the data reduction procedure which is needed to know how sensi-

tive the calculated transfer matrix was when subjected to a small change

in one of the input quantities.

/
 -107-

The laser doppler velocimeter and pressure transducer systems,

(including the transducer itself and the associated balancing unit and

amplifier) had a typical linearity of 0.01 percent. This linearity, which

came from the frequency-to-voltage converter for the LDV and came

from the amplifier for the pressure system, is needed to determine

the accuracies of the measured fluctuating quantities. The noise level

of the pressure transducer systems was 0. 1 percent whereas the LDV

had an electronic noise of 0. Z to 0. 5 percent as determined by measur-

ing the noise at a no-flow condition. The tape recorders had a typical

accuracy of I percent. The total accuracy thus far is about 1.5

percent.

Typical oscilloscope pictures of the signal traces during a fluc-

tuating flow condition are shown in Fig. 5. 14. It can be seen that the

fundamental frequency was marred by higher harmonics aJnd flow

noise. This is more apparent at the lower frequency of fluctuation

where the excitable higher frequencies are still within the response

range of the hydraulic system and the measuring instruments. The

nearly square wave traces at the low frequency are due to the design

o£ the fluctuators as mentioned in Section Z. 4.7. However, these sig-

nals introduced random fluctuation on the readings from the Bafco

analyzer and this represented the noise rejection capability of the

analyzer. Typically this scatter was less than 6 percent of the read-

ings and the reading itself had an accuracy of I percent. Thus, the

values of the measurements to be processed by the data reduction pro-

cedure had amaximum error range of 9 percent of the value.

REPRODUCIBILITY OF TH_
oRIGINAL PAGE IS PO0_
 -I08-

As mentioned in Chapter IV, the data reduction procedure first

corrects the data for the effect of filters and amplifiers from their com-

bined dynamic calibration curves. These curves were measured to an

accuracy of I percent. To investigate the stability of the least square

fit procedure in calculating the transfer matrix, a test data sample was

supplied to the program. The stability of the procedure was determined

by perturbing the elements of the data sample and the procedure was

considered stable if the corresponding perturbed transfer matrix only

changed by a proportionately small amount. It was found that the pro-

cedure was stable. Thus, it was concluded that the overallaccuracy of

the calculated transfer matrix was _bout I0 percent.

5.7 Linearly Independent Data Sets

The TRANSCO program used data from a number of different

fluctuation settings to calculate the required Z-matrix. Since the pro-

gram required only a minimum of two sets, we could use a partial num-

ber of total combination sets to calculate the Z-matrix for experiments

which involved more than two fluctuator settings.

To demonstrate this, let us take an experiment where three

fluctuator combinations were used; these were:

(A) Upstream fluctuation only

(B) Downstream fluctuation only

(C) Upstream and downstream fluctuation together.

We can produce four different Z-matrices by using the following combi-

nations :
(_) A + B +.C (v) A + C

03) B+ C (6) A+B

 -109-

A sample of this is given in Fig. 5. 15. The data showed a wide

scatter for different choices of combinations. The first combination

(@) is considered the most meaningful one since it comprises all the sets

and thus is the total average. It was observed that although the scatter

is wide, the Z calculated from the combination (6), A + 15, is always

closest to the total average. That is, the best partial combination

comes from using the purely upstream fluctuator and purely downstream

fluctuator actions. This result is because purely upstream and purely

downstream fluctuations produce the two most linearly independent sets

of data. The result which is farthest apart from the total average reo

sult is the combination ¥. This ¥ combination consisted of set A, up-

stream fluctuator alone and set C, both fluctuator functioning. This is

because the pressures and mass flow rate response of the two fluctua-

tots functioning is dominated by the upstream flactuator action. This

means that the two fluctuators' combined action is a very close resem-

blance to the set A. This will be clear when comparing the pressure

and mass flow rate between Fig. 5.6 and Fig. 5.8 also between Fig. 5.9

and Fig. 5. II. Consequently results from the A and C sets are not as

linearly independent as the other combination thus the data show a

wide scatter.

This linear dependency is also demonstrated in the denominators

in the equations calculating the Z elements in the least square fit pro-

cedure (see Appendix D). The denominator is a measure of the degree

of dependence of d_tta sets. It willbe zero for linearly dependent sets.

In the examples given in Fig. 5. 15, it was found that the combinations

involving single fluctuators had the maximum denominator.

 -110-

A simple analogy to the above discussion is in the fitting of

points to a straight line. In most of the cases, the two farthest points

usually define the best straight line.

5. 8 Linear Transfer Matrix Assumption Test

One of the basic assumptions in the present study of the dynam-

ics of cavitating pumps was that all the velocities and pressure fluctua-

tions were small. They should be small enough that they are small de-

partures from the mean state and a linear transfer matrix may be de-

fined. This was investigated by carrying out two types of tests at me-

dium cavitation nu.,_nbe

r.

The first test was an "amplitude" test in which we tried to verify

that if the amplitude of one fluctuating quantity, was increased by a cer-

tain amount, a11 of the other associated fluctuating quantities would in-

crease proportionally. For the conditions chosen this appears to be the

case for a sample graph shown in Fig. 5. 16. Transfer matrices calcu-

lated from different amplitudes of fluctuation but with the same steady

flow and cavitating condition showed that the values of the elements were

the same for all araplitudes within experimental error.

The second test invoked directly the property of a linear system

that the transfer function of a linear system at a certain frequency has

the same value whether it is measured with a pure sine wave signal or

whether this frequency is imbedded in noise with other frequencies pres-

ent. The signals in the present experiment approximate a square wave

and therefore there was a high third harmonic content in addition to the

fundamental. It was possible in the present system to measure the

 -111-

transfer matrix of the third harmonic of the command signal using the

special frequency rrJultipliercapability of the Bafco signal analyzer.

The third harmonics of the four Hertz and seven Hertz fundamental

were chosen for measurement (frequencies of 12 and 21 Hertz respec-

tively). The elements of the transfer matrix for these values were cal-

culated and are compared with the transfer matrix measured from the

fundaxnental in Fig. 5. 17. The occasional wide departure probably

occurs because of the low signal to noise ratio due to the low level of

the third harmonic _ignal. The low third harmonic content is showzJ in

Fig. 5. 14 (these traces with a fundamental frequency of 7 Hertz were

used to anallrze the 21 Hertz behavior shown in Fig. 5. 17). These two

tests confirmed tha; the linear assumption was justified in this experi-

ment.
 -llZ.

.40 |' ]_ I I 1 I

0 ACOSTA (1958)

\
\
.30 -
IMPELLER \
\
\
\ THEORETICAL
9 ° IMPELLER
b- \ (ACOSTA)
W .20 \
0
0 \
a k
IMPELLER l'rr
<I \
W
"1- \
\
k
.10 -
\
o \
oN
\
oN
\
.00 I I I I i i\
.00 .02 .04 .06 .08 .10 .12

,,,LOVV
Fig. 5. l Fully wetted performance o.t Impel!er iII and _¢.
 -113-

I I I I

.4O m

SSME-LPOTP
.6 SCALE
(ROCKETDYNE)

.30 -

Ii
i,i
0 .20 B

0
C_
IMPELLER Iv
Iii
"1-

.10 -

EPL
.00 I I I # I
.00 .02 .04 .06 .08 .10

FLOW COEF U°
UTIP
Fig. 5.2 _'ully wetted 1)erformm_ce cf Irnpe)l_r IV and O. 6 scale
rnodel by Rocke_dyne.

REPRODUCIBILITY OF THE
ORIGINAL PAGE IS POOR
 -114-

I ! •I 1 I '! I I I I I I I I I I 1 I I I

0
o

o.n

O0 o.
0'_"
0--

Z
0 0

II I'--
o d
_J Z
.J
m

0
o. _.
!
o
I-- w

_.

=8-

a. E

q ._ _
?° O

e_

u_

o _ _ _ _. o _. o.o.O °i,d

__ _

-_ "330::)CIV3H
 -I15-

I I I I I I I I I I I I I I I I I
0
N
O0
-o ®
o. 8 a. (M
0

•33oo OV3H $
dV.
 -116-

I i I
1

0
f_.
o. 0 0
0 d d
I! I! II

d d d d d

.LN31013-1309 GV3H
d_7
 -117-

I I I

I I I I
"SEt "06 "Sh "0 "Sh- "06 ° "S£I- "Ogt-
 -118-

q
i I I J I i 5_

7 _q

-// - .C

/ C

C_
0

U.J _o
Icl.

,"-, CK !
L.J 4
,,0 ,='4
(_"r 0
• 0
0
I , ,! ,, L I I I
"091 "S£I "06 "Sh "o "5b- "06- "S£t- "09t-
0

s_$_p u! _I3u_ eS_Nd .*4

U
0 _

I I i

" 0

U..
o_ _,

ctr- i.--, el.

...l ,.--,
._Jr,._)

(,..).,-I i
(,f') tJ.I,
0
I I I t 0
O'Ot 0"8 0"9 O'h 0"[ 0"0
 -119-

- I I I

p-

I I I . iI _°
I

. / ,;lo_
_ ,t/-I_;_
ujn-
:iil g i)

=>
I_ _/o n_

I! _-_:- cJ

o')"r-
Z,'_/ 0

edo- I I l ! , ,, _i_-" l#i

"091 "SOl "_ "Sh "0 "SII- "1_- "SOl- "Otil- .
• ..i I_

saa_$ap u! aISu_ as_q(:l; U 0

0_
.==4

o.
l I I I 53

o=
U !

_ ,,' 0
i

oi,,J
.4

°u,i

o:
o:3

z>i
O
it-i--. ¸
..J ..-- c;
t
_JU
l--, E) I
(J-J ,4
(/1 LIJ O
I
[ I I I I 0

O'Ol 0"$ 0"9 O'tl 0"_ 0"0

 -120-

q
I ! I

(1)bJ
.el
- !'
0 J
/
/
4°

°_
_v
tJ
r"

I
n-w_ I=
D_ -
I
# ",7-,
/
Z> #

_--UJ 1 o
J
_J
_ J
o.
# 2 r.,,
/Z pl
I
(_.) 9"
i 4_
(_-r- U
(DO..
1 I I ! I I o.
o
"091 "5£I "06 "Sh "0 "S_- "06- "S£1- "061- Q

saa_ap u! ai3u_ a_yd

o.
I i I I
u

0
!

0t
I
I

m !
J
i i v

C3i
IJ.J _.,
,IJ 8 o_
n" f-i I _4
L,
o
o_
-_ .,,4

Z> ! i,i.

I--l.IJ /
CI:E: I
I
.-I:::::)
- I' O'
M
-J U"}
,-,0')
(JILl
U')n"
E:)_- 0
! ! I ! 0
QWg
0"0 0"9 O'h 0"_ 0"0

._s,_ .To _3 u! uoT:_n:pnI_ azussa_d 30 _pu_.Idtuv

 -lZl-

I I I
• Set "0$ "Sh "0 'S_-
_:1.0

see.z3ep uT _I_U_ as_qd

(/_ ,,IJ

I I I I

u_
LL.

,p
 -IZ2-

el.
I ! I I I

/ m

0 •
• f
/
/
/
' ti 0
.o
o_
,/
OObJ /

Ujn- 4a

n-b_ n_
0.. -
0

c5£n / o

;'-- l.:J
Q::: ..J
-.I £._ " _J
-JZ
,-,rr
t.J I4, f "o
[.r) -r o
0
E:::)0..
I I I I I
"091 "S£t "06 "Sh "0 "St- "06- "$£t- "06; o i)

saa_p u! _I_U_ _s_qcl

I I I I

B f;l

• o

f
J-,I
°v u'_

l,JJ
0 _
rr"
i,

Z>
;--UJ !
cI:rr I
...I:::)- " !
i.
..JU')
•-, tr) /
4
Or)O::
I I I I
o'6 0"9 o'h o't. 0"0
 -IZ3-

nS

n_

o_ _

REPRODUCIBILITY OF TH_
_t_TG_AL PAGE IS POOt_
 -124-

I I I I I

l I I I I_
40 80 t20 I 60 2 O0

GAL/MIN

Fig. 5. 13 Steady pressure drop across the smoothing chainber.

 -125-

Fluctuating frequency = 7 Hertz

Fluctuating frequency = 35 Hertz

Fig. 5. 14 Oscilloscope pictures of the fluctuating frequencies. The

order of the traces from top to bottom are: upstream LDV,
downstream LDV, upstream pressure, downstream pres-
sure, upstream acceleration and downstream acceleration.

RE?RODUCIBILiTY OY T'H_
ORiG'TNAL PAGE IS POOF.
 -lZ6-

Zll VS. FREQ.

8
'1 i I I'ILCp) • ]1_,,,[ ])'

s In, o B _

. ..,,-"{""_' _ i
o ,,,,,_

.- ...

s.
"L I l I L A I
20.0 ._?.0 qO.O iO.O 0.0 10.3 20.0 30,0 qO.D
0,0 iO.O
FREQUENCT_ ,flZ) Ft_EQUENCY(HZ)

Z22 VS. FREO.

Z2t VS, FREO.
°_'[ I I v'e.i_l..cl,) o. ug_r_]T
& I t I W.mx t'_ _,_

P. R

iI i
_ _--_--_
_,'_1:_
N6

1 | _ I I I !
0.0 I0o0 20.0 30,0
FF_EQUENCT(HZ)

Fig. 5.15 Z-matrix produced by" different cornbinations of fluctuetor

settiz,g s.
 -127-

4_
u
m'_
{::I ,_
G)
N
U
-r-
,"i ...-4
0'I

z _ ._
re) _d

l
l
l
l

0
o 0 0 •"_ 0 _
cJ
_3±VM 30 "13 N I 3_INSS3Wd
I
o 0 0 M
{M R

NV31AI 30 % "-13A _V3NJ.SNM0(] °,.¢

 -17.8-

Z I] VS. FREQ. Z 12 VS. FREQ.

I I | mt.'_.(4,) 5 ]mRr._l) I I I' "I_.I_.(@} & IN¢_I)

.°

i
o . .e-

.-. -" Th,rd Hsrmmenic

1 I | •
c_ o
-j m,.J

_,, ThlTd Hlrrr_mc

9:

I i _ . ! I I I I l
0.0 I 0.0 20.0 30.0 _0.0 50.0 0.0 1O.0 20.0 30.0 tlO.O 50.C

FREQUENCY (HZ) FREQUENCY (PIZ)

Z 21 VS. FREQ, Z 22 VS. FREQ.

o
1 i I I_..'_.(
(,:,t.INTO,,(1) p.j I I IIRL[@) b IM_F.WI)

c3

• .._ -* (_

-- I

I 1 • I i I r I
o.o io.o 20.0 _0.0 _0.0 50.0 0.0 _0.0 2o.0 30.0 qo.o _.0

FREQUENCY (HZ) FREQUENCY (HZ)

Fig. 5. 17 Z matrix from measuring the Lhird harmonics from the

Iluc.tuatihg signals,
 -129-

VI. RESULTS AND DISCUSSIONS

This chapter will start with a discussion of the effects on the

measurements of the hydraulic components between the two measuring

stations. Effects of the hydraulic components between the measuring

stations and pump have to be subtracted out in order to obtain the re-

quired pump matrix ZP. Following this the components of the ZP pump

matrices will be plotted as functions of frequency. The behavior of the

matrix components will be compared with values predicted using quasi-

static analysis based on the steady state pump characteristics.

A total of five experiments using two impellers, Impeller IV and

Impeller V0 are presented. The five experiments were conducted in the

following ways: one in a fully-wetted condition, two under cavitating

conditions with Impeller IV, one under fully wetted conditions with

Impeller V and one under cavitating conditions for the same impeller.

As discussed in Section 3.3, the results for Impeller V are less reli-

able than regults from Impeller IV.

6. 1 System Effects

First, it is essential to define what hydraulic components com-

prise the pump and to identify the location of the measuring stations.

For the definition of the pump, it is convenient to choose the plane con-

necting the leading edges of the inducer blades as the beginning of the

pump and to choose the flange at the exit of the volute as the end of the

pump. In general practice, avolute is usually considered as part of a

pump. Further, the dynamics of the volute would be very hard to

analyze and to express in simple analytic form. _ Thus the volute is

 -130-

included as a part of the pump.

The identification difficulties of the measuring stations arise

because the pressure measurements were taken from the smoothing

chambers whereas the mass flow measurements were taken from the

LDV viewing sections. The pressure and mass flow measuring sta-

tions were physically separated by a converging nozzle (see Fig. 2. 5).

It is convenient to refer the measui'ing stations to a single upstream

and downstream point. The main concern here is the compliance effect

of the converging nozzles. The converging nozzles had thick metallic

walls and contained a relatively small amount of water. Their esti-

mated combined compliance effect was calculated to be too small to

cause any rneasurable fluctuating mass flow rate difference across it.

This meant that the fluctuating mass flow rate at the end of the converg-

ing nozzle must be very close to that at the end of th e smoothing cham-

ber. Therefore the flanges joining the smoothing chambers and the

convergent nozzles are considered as the upstream and downstream

reference stations.

Having defined the measuring stations and the pump, we divide

the hydraulic system between the measuring stations into three sec-

tions. These are: the upstream section between the upstream measur-

ing station and the beginning of the pump; the pump itself and a down-

stream section between the end of the pump and the downstream measur-

ing station. The next step is to identify the dynamics of the upstream

and downstream sections using lumped parameter models.

To estimate the effects of the upstream and downstream section,

we have to look closely at the construction and contents of each section.

 -131-

The upstream section consists of a convergent nozzle and a straight

pipe section leading to the pump. The nozzle and the straight pipe

have thick aluminum walls and thus negligible compliance. The volume

of water contained was also too small to contribute to any appreciable

cornpli&nce effect. Thus no compliance value was assigned to this sec-

tion. However, the inertance was calculated to be appreciable and the

individual contributions were 0. 150 sec 2 /ft2 for the convergent nozzle

and 0. 411 sec2/ft 2 for the straight section leading to the pump.

The downstream section consisted of the divergent nozzle im-

mediately following the volute and the downstream smoothing chamber.

The smoothing chamber contained a large volume of water and had a

thin wall relative to its radius. From the known compressibility of

water and the known volume of water inside the smoothing chamber,

this compliance was estimated to be 0.01 in3/psl. Assuming small

oscillating expansion and using linear theory for the radial expansion of

a thin wall cylinder, the compliance of the downstream smoothing cham-

ber structure also was shown to be 0.01 in3/psi. Thus the total com-

pliance was about 0.02 in3/psi. As for the inertial effect, there is

no accurate way to calculate the inertance for the complex separated

flow in the divergent nozzle. The inertance of the convergent nozzle

(0. 156 sec2/ft 2) was assumed for this. Finally the water inside the

smoothing chamber contributed 0. 12 sec2/£t 2 to the inertance. During

the preceding discussion, we did not mention the viscous resistance

across any sectioll. The viscous stress on the walls of all the sections

was calculated, appeared not to have any appreciable effect on the

measurements, and thus was neglected. The pressure drop due to the
 -132-

screens etc. in the smoothing chamber was measured in Section 5. 5

and was also found to be negligible compared with the measured resist-

ance_ Therefore, no resistance terms were included in the upstream

or downstream section dynamics. To summarize all the foregoing

factors we expect an upstream inertance of 0.57 sec2/ft 2, adownstrearn

inertance of 0.28 secZ/ft Z and a downstream compliance of 0.02 in3/psi.

The sum of the inertance terms, 0.85 sec2/ft 2, is relatively small corn-

pared with the inertance of the pump as will be discussed in Section 6.2.

(The calculated compliance of the downstream smoothing chamber was

found to be smaller than values determined from experiment and this

value will be modified in Section 6.2. ) All these inertance and compli-

ance values were used to extract the ZP pump matrix from the total

transfer matrix Z as measured betweenthe stations (implemented in the

PUMA program) see Fig. 6.6 for an electrical circuit representation of

these components.

6.2 Fully Wetted Experiments

Two Jets dexperirnentalresults for fully wetted flow (one for

each impeller) will be presented here. The Z total matrices will be

discussed first, followed by discussion of the ZP pump matrices.

6. 2. 1 The Experiments

Experiment (1)-In this experiment Impeller IV was operated in

the fully wetted condition. Although this experiment was designed to

avoid any cavitation, a small thread of cavitation was present at the

leading edge as seem from the photographs taken during the experiment

(see Fig. 6. 1). The actual static pressure used was 40 psig. The

mean speed and discharge were 9000 rpm and 165 gpm
 -133-

respectively which correspond to a head coefficient of 0.25 and a flow co-

efficient of 0.07. At this pump speed, a fluctuating frequency of 50

Hertz leads to a dimensionless frequency of about 0. 52. The calculated

cavitation number for this experiment was 0. 51. Three sets of experi-

ments with different fluctuator combinations were performed. The to-

tsl transfer matrix Z for this experiment is given in Fig. 6. 2.

Experiment (2)-This is another fully wetted experiment similar

to experiment (I) except here Impeller V was used. Again, there was

the unavoidable thread of cavitation as seen in the photographs of

Fig. 6.3. This experiment was operated at mean conditions of 6000

rpm and 128 gpm corresponding to a flow coefficient of 0. 088 and a

head coefficient of 0. 12. Four sets of fluctuator combinations were

used and the calculated total Z matrix is given in Fig. 6.4.

6.2.2 Discussion of the Z-matrices

The most important feature observed from these Z-matrices is

the non-zero values of Zzl (see Fig. 6. Z and 6.4). This confirmed

that there was a hydraulic system compliance between the measuring

stations which was addressed in Section 6. I. Although the photographs

of the inducers (see Figs. 61. and 6.2) indicated that there were small

traces of cavitation at the leading edge of the vanes, a simple estima-

tion of the compliance effect of these small vaporous bubbles from its

observed volume showed that the effect was much too small to account

for the values of ZZI observed. This compliance value must come

from the downstream smoothing chamber. We could further assume

that the pump was in a truly fully wetted state and that its pump gain

factors compliance and mass flow gain factors discussed in the

 -134-

Introduction were all zero. Therefore when the system effects of the

upstream and downstream sections were subtracted from the fully

wetted total transfer matrix, the resultant pump matrix should only con-

tain a pump resistance and a pump inertance term (Zlz only). Together

with the inertance of the upstream section and the inertance and com-

pliance of the downstream section, an electrical analogy was deduced

from the above discussion and its circuit diagram is shown in Fig. 6.5.

The total transfer matrix Z for this analysis is determined to be

R- _u(Lu+.Lp+ L d)
(6. 1)

-uvZC(Lu+ Lp+ L d) + j_ CR
/
where R is the pump resistance, C is the compliance of the down-

stream smoothing chamber and its water content, w is the fluctuating

frequency and L u, L P and Ld are the upstream inertance, pump Lnert-

ance and downstream inertance respectively.

It can be seen that the forms of the curves in Figs. 6. Z and 6.4

follow the form of Eq. (6. l) very closely. Thus this equation will be

used to investigate the experimental results. The compliance and total

inertance values were measured from the gradient of the graphs from

the imaginary ZZI and imaginary ZI2 by drawing a best fitted straight

line through the origin. Their measured values are tabulated in

Table 6. l with the estimated value as calculated from Section 6. I. The

pump resistance wi/l be discussed in the next section (6. Z.3) after the

system effects have been subtracted out.

 o135o

TABLE 6. 1

Compliance Inertance

L +L +L d
u p
Dimensionless Dimensioned Dimensionless Dimensioned

K C DL
2
in3/psi sec /ft 2

Exp. (I) 0.2 0.032 44 3.0

Imp. IV

Exp. (2) 0.066 0.030 27 2.7

[rap. V

L u + Ld

_2_s
timation
from Sec. 6. 1 0.0Z 0. 85

The system compliance values O. 032 and 0. 030 in3/psi be-

tween the two experiments agree very well. However, these values are

much higher than the calculated value of 0.02 in3/psi from the com-

pressibility of water and expansion of structure. This is because the

estimate just mentioned was very rough. Further, the experimental

value is actually the total compliance between the two, measuring sta-

tions instead of just the smoothing chamber. Compliance contributions

from other structural components and the volume of water between the

stations but outside the smoothing chamber were hard to estimate due to

'he complex geometry of each part. In the data reduction procedure,

•II these effects were attributed to the smoothing chamber and the value

of 0. 032 in3/psi was assigned to be the compliance of the chamber.

 -136-

The experimental measured inertances of 3.0 and 2.7 sec2//t 2

,__re much higher than the sum of upstream and downstream inertance.

:he difference is obviously the inertance value of the pump. One possi-

ole explanation for the higher inertance value of Impeller IV as com-

pared to Impeller V is that the Impeller IV has a thick hub section after

":he blades thus causing a restricted area. This restricted area should

,ncrease the inertance.

_.2.3 Discussions of ZP pump matrix

From the Sections 6. 1 and 6. 2.2, inertances and compliance

calues were assigned to account for the upstreana and downstream sys-

".era effect such that pump transfer matrices ZP can be extracted from

"_he total matrix Z. Figure 6. 6 shows the assigned system values and

circuit diagram aria Section 4.3 shows how the PUMA program accom-

plishes this. The reduced ZP matrices for experiments (1) and (Z) are

illustrated in Figs. 6.7 and 6. 8.

The expected zero values of ZPll, ZP21 and ZPZZ (see intro-

duction for their significance) for fully wetted flows are clearly demon-

strated in the figures. The occasional slight drifts from zero at higher

frequencies in Fig. 6.7 may be caused by an overly simplistic model

for the system effezts. This might be improved by a more sophisticated

model like the addition of an extra inertance after the compliance in

Fig. 6.6. However, the wide departure from zero for ZPZ1 and ZPZ_ -

of Fig. 6. 8 is considered to be purely experimental error.

We now focus our attention to ZP12 the pump irnpedence term.

The imaginary part of ZPI2 represents the inertance of the pump. If a

 -137-

simple inertance were assumed for the pump, then the imaginary part

of ZP1z would be a straight line passing through zero and decreasing

with frequency. Both experimental results show that this is very

nearly true except that a two points linear extrapolation at lowfrequency

in Fig. 6.7 doesnot go through zero. However, the non-zero value is

small and probably due to experimental error.

The first feature of the pump resistance (negative real part of

ZPIz) to be investigated is the linear low frequency extrapolation. We

expected this extrapolated value to agree with the pump resistance value

from the steady state performance curve (see Introduction). The linear

extrapolated values were 10 and 5 for Impeller IV and Impeller V. The

pump resistance value as obtained from their steady state performance

value for the corresponding flow coefficients were 15 and 5. 3. There

is good agreement between the two values for Impeller V. The discrep-

ancy between 10 and 1 5 for Impeller IV may be due to a more compli-

cated function of pump resistance at low frequency. This is not

surprising since Impeller IV is of a more complicated design

than Impeller V and also it has the addition of a stator stage.

As seen fr0m Figs. 6.7 and 6.8 when the frequency increases, the

pump resistance of Impeller IV increases until at a dimensionless fre-

quency of 0. 3 it starts to decrease. The pump resistance of ImpellerV,

also begins to decrease at a dimensionless frequency of 0.3. This

could be interpreted as the slope of the pump performance curve first

decreasing slightly (or remaining relatively constant) and then increas-

ing towards a horizontal slope as indicated in Fig. 6.9. This feature is

 -138-

:nore pronounced during the cavitating tests as will be seen in Section

i>. 3. It is interesting to speculate upon the effect when the pump re-

sistance crosses zero and becomes negative. The corresponding be-

havior of the slope in Fig. 6. 9 would be to reach a positive value.

This means an increase in the flow rate will increase the head rise

across the pump. This, in turn, woul'd tend to increase the flow

through the hydraulic circuit and hence the pum_,. Such an unstable

situation may be responsible for the auto-oscillation phenomena de-

scribed in Section 5. 3. Up to the present time, no negative pump re-

sistance has been measured either under fully wetted or cavitating con-

ditions (see al'so Section 6. 3). However, we anticipate that negative

pump resistance values could appear for experiments with very low

cavitation numbers. It is not possible to measure the pump transfer

matrix during auto-oscillation (as described in Section 5. 3) under the

present experimental setup.

In general, cascade theory is used to analyze the performance

of axial flow turbomachines. In the present studyp it would be desirable

to compare the experimental result with the analysis of the dynamic re-

sponse of a cascade subjected to a time varying inlet velocity. No

suitable analysis could be found, however. Although calculation on the

unsteady response of axial flow turbomachines to an upstream disturb-

ance exists in the literature (see Henderson 1972), the calculation is

for the case of spatial variation in the incoming velocity field and not

of time variation as in the present case. This calculation is therefore

not directly applicable here.

In an eflort to get a qualitative understanding of this phenomenon,

 -139-

the results of the oscillating single airfoil analysis will be recalled

here (see e.g. Fung 1968). Furthermore, the single airfoil result is a

imiting case of the cascade calculation when the cascade blades are

:.rery far apart.

The single airfoil analysis we shall examine is the special case

_f a planar airfoil subjected to a sinusoidal time varying vertical gust

_vith constant horizontal velocity. One of the results of this calculation

_s the complex lift coefficient which is the fluctuating lift force normal-

%zed by the product of the amplitude of sinusoida! vertical velocity and

_ree stream velocity. This is called the Sears' function and is repro-

duced in Fig. 6. I0. This can be considered analogous to the pump

impedance -ZI2. Here we shall examine the Sears' function at low

dimensionless frequencies. This low frequency range implies that the

wavelength of the sinusoidal gust is much longer than the chord length

which is the case in the present experiment. We may liken the real

part of the Seat's function to the pump resistance. We see from Fig.

6. 10 that the real and imaginary part of the Sears' function decreases

with increasing frequency when the frequency is low and assumes a

more complex behavior at a higher frequency, in the present experi-

ment, the pump resistance in general decreases with increasing

frequency. However, the imaginary part of pump impedence (-ZIz) is

positive and increases With frequency. This means that the phases be-

tween the real and imaginary part between the two cases are different.

No explanation can be offered here for this surprising finding. Overall,

this example shows that important unsteady effects can occur in the

unsteady fully wetted flow of turbomachines.

R_'RODUCIBILITY OF THI_
• 0R_I[NAL PAGE IS P00_
 -140-

'7. 3 Cavitatin B Experiments

The pump transfer matrices for three cavitating experiments

will be presented and discussed here. Some of the results will be com-

._ared quantitatively with the quasi-steady blade cavitation calculation by

Brennen and Acosta {1975).

S.3. 1 The experiments

]Experiment (3) - This experiment was performed with Impeller

IV. The mean conditions were 9000 rpm and 165 gpm. These condi-

tions correspond to a flow coefficient of 0.07 and a head coefficient of

0. 25. The mean cavitation number was 0. 11 which resulted in only a

moderate amount of cavitation. It can be seen in the photographs of the

impeller during the experiment (Fig. 6. 11) that there were distinctive

patches of tip clearance cavitation on the blades but little blade cavi-

tation. At this cavitation number, the condition of cavitation is far

from the region of performance breakdown as can be seen by referring

to the cavitation performance curve of the 0.6 scale model obtained by

Rocketdyne 1975 (see Fig. 5. 5). Three sets of results with different

fluctuator combinations were obtained and the calculated ZP pump

matrix is shown in Fig. 6. 12.

Experiment (4) - This experiment was also performed using

Impeller IV and the mean conditions were the same as in experiment (3).

The only difference was that the cavitation number had a lower value of

0. 046 resulting in more extensive tip clearance cavitation as shown in

Fig. 6. 13. Unfortunately, the fully developed tip cavitation obscures

somewhat the observation of cavitation elsewhere in the impeller.

During the experiment, the flow upstream of the pump was observed to
 -141-

nave entrained a small number of minute bubbles. The amount was

_-nuch too small to cause interference with the LDV measurement, how-

ever, a certain amount of bubble compliance as discussed by Brennen

1973 could be expected. The cavitation performance curve from Fig.

5. 5 indicated that this condition should cause slight performance de-

terioration. Again, the Zl:'-matrix was calculated from the results of

three different fluctuator settings and is shown in Fig. 6. 14.

Experiment (5) -- This cavitathng experiment with Impeller V

was run under similar mean conditions to those of experiment (Z). The

conditions were a rotation speed of 6000 rpm and a flow rate of 1ZSgpm

giving a flow coefficient of 0. 09 and a head coefficient of 0. 12. The

cavitation number was 0. 12. The increased blade spacing of this im-

peller compared with Impeller IV allowed clearer observation of cavi-

tation (see Fig. 6. 15). It can be seen that there was a moderate

amount of tip cavitation though not extensive enough to obscure the ob-

servation of any other forms of cavitation. The pump matrix from four

fluctuator combina:ion experiments is shown in Fig. 6. 16. As men-

tioned in Section 3.3, these data are less reliable than those of Impeller IV

and they are presented here for a qualitative comparison with that other

impeller.

6.3.2 Discussions

All the elements of the pump matrices ZP (Figa6. 12, 6. 14 and

6. 16) were plotted as functions of fluctuating frequency. The calculated

dimensionless frequencies (see Eq. 1. 9) are also shown and denoted by

DFREQ. We will compare these experimental results with the quasi-

linear calculations and discuss the significance of the matrix elements

 -142-

for the dynamic performance of pumps.

(i) ZPll -- The measured ZP 11, the pump gain factor term,

will first be compared with the steady state expectation.

As mentioned in Section 6. 3. 1, the mean flow conditions of

these experiments indicated no severe performance deteri-

oration from which one might expect small values of the

pump gain factor. Tangents to the steady state perform-

ance curve (Fig. S. 5) at the mean flow conditions suggest

low frequency values of 0.02 for experiment (3) and 0.09

for experiment (4). It is readily seen from Figs. 6. 12 and

6. 14 that the experimental values of ZPll are much greater

than these values; indeed the values approach unity at the

higher frequencies. This is of major importance because

large values of ZP 11 imply significant deterioration of

pump cavitation performance under dynamic conditions.

Another result of experiment (4) at low frequency is the

negative value of the linear extrapolation to zero frequency

from the two lowest frequency data points. This negative value

indicates a negative gradient of the cavitating performance

curve. However, this may be due to a more complex func-

tion at the lower frequencies and more data points at this

region are needed to verify this.

One of the most surprising features of Figs. 6. 12

and 6. 14 is the existence of an imaginary part which is not

predicted by the quasi-steady analysis. This means that

 -143-

there is a phase shift between the Upstream pressure and

the pressure rise across the pump even in the absence of

mass flow fluctuations. The real and imaginary parts of

ZPl 1 for experiments (3) and (5) have the same general

behavior but the curves of experiment (4) where the cavita-

tion number was lower have quite a different shape. This

may be due to differences in the way different degrees and

forms of cavitating flow respond dynamically. Experiment

(4) might have traces of blade cavitation at the leading edge

besides the extensive amount of tip cavitation whereas ex-

periments (3) and (5) had only tip cavitation. Unfortunately

the extensive tip cavitation of experiment (4) prevented the

observation of any blade cavitation.

However, one cannot rule out the possibility that the

observed ZPII for experiment (4) was a nonlinear effect

due to the magnitude of the fluctuation of the cavitation num-

ber. This is an experimental limitation which is inherent in

the present facility. Since the experiment requires a mini-

mum velocity fluctuation so that it can be measured by the

LDV, a minimum pressure fluctuation is required. The am-

plitude of this pressure fluctuation may bring the instantane-

ous operating point into the breakdown region where high

pump gain factors could be expected. During experiment (4),

the amplitude of the upstream pressure fluctuation was al-

most constant over the whole frequency range but differed

for different fluctuator settings. At somefluctuator settings,

REPRODUCIBILITY OF THg
0RICANAL PAGE IS POOR
 -144-
i

i
I,
the rp.sulting fluctuation of the cavitation number was quite

large. For the three fluctuator settings used in experiment

(4), the cavitation number fluctuated between 0. 0Z and 0.09,

between 0.04 and 0.05 and between 0.02 and 0.08. The low-

est value here is well into the breakdown region (Fig. 5. 5).

If straight lines are drawn in Fig. 5.5 between the cavitation

number pairs given above, the resulting slopes of these

lines would lead to pump gain factors of 0.78, O. 91 and 0.ZZ

i

respectively. These values may partially explain the large

values of ZP II measured but do not explicitly account for

the appearance of an imaginary part.

In an attempt to shed further light on these eflects,

photographs of the cavitation wcre taken at various points in

a cycle during the dynamic experiments (Fig. 6. 13) and

these were compared with the photographs of cavitation

breakdown taken by P, ocketdyne (1974) during steady state

experiments On the 0.6 scale model. It appeared that the

cavitation during fluctuating flow was nc, t as extensive as in

the breakdown state. The Rocketdyne photographs showed

that the cavitation extended all the 1way through _he four

blade inducer and into the twelve blade tandem stage. On

the other hand the photographs taken at the point of maxi-

mum cavitation in the dynamic experiments showed that the

cavitatlon only extended thro.,gh the tour blade section but

not into the twelvc blade stagc. In comparing the two sets

of photographs, the cavitation form o[ the dynamic

 -145-

experiment would not be extensive enough to cause break-

down. This appears to suggest that the large ZPll meas-

ured was due to dynamics of the flow. The contradiction be-

tween the photographic demonstration and cavitation number

fluctuation cannot be resolved until further experiment can

be carried out with much lower fluctuation of cavitation

number. However, the above discussion does not apply to

the ZPII of experiments (3) and (5) where the cavitation

number fluctuation was less than 25 percent of the mean

cavitation number.

In conclusion it was observed that the measured val-

ues of ZPII were much higher than the steady state pump

gain at the mean cavitation number. This together with the

unexpected existence of an imaginary part, demonstrated

that the dynamic performance of turbomachines can be in-

fluenced significantly by cavitation in a manner which is not

predicted by quasisteady analysis.

(it) Zloi2-- From Figs. 6. 12 and 6. 14, the linear extrapolations

of the pump resistance (negative of the real part of ZPI2)

from the two lowest frequency data points have values of 12

and 10 for experiments (3) and (4). However, the pump re-

sistance as measured from the steady state performance

curve would have a value of 15 for both experiments. At

higher frequencies, the pump resistance decreases and the

rate of decrease is higher when the cavitation number is

lower. This is not surprising since cavitation reduces the

 -146-

contact surface area between the fluid and the impeller sur-

face. This reduces the viscous drag on the fluid and could

lead to reduced resistance. The implications of further re-

duction of pump resistance and the possibility of negative

values was discussed in 6.2. 3 and will not be repeated her_

The low frequency values of the imaginary part of

ZP12 behave like an "ordinary" inertance term in the sense

that it has a definite negative gradient and is proportional to

frequency and the linear extrapolation goes through zero.

At higher frequencies, the slope of the imaginary part of

ZPI2 increases implying a decrease in the effective inert-

ance. Furthermore the rate of increase is higher for ex-

periment (4) where the cavitation number was lower. This

may be due to the reduction in the mass of fluid inside the

pump in the presence of cavitation. However, this cannot

explain the positive gradient shown in experiment (4) which

indicates a negative effective inertance. The possibility

that the imaginary ZP 1Z may become pos itive at lower values of

cavitation number than those investigated may have conse-

quences for auto-oscillation similar to those of the real

part of ZP12 .

(iii) ZPzl -- Surprisingly, experiment (3) produced no measurable

cavitation compliance as shown in Fig. 6. 12 although pho-

tographs of the impeller (Fig. 6. 11) showed distinct patches

of cavitation. However, as seen in Fig. 6. 14 experiment•

(4) did produce measurable cavitation compliance. The

 -147-

appearance of a real part of ZPzI is unexpected from the

quasi-steady blade cavitation calculations (see B rermen and

Acosta 1975). In order to compare the experimental re-

sults with the quasi-steady calculation, an imaginary com-

pliance is introduced into Eq. (1.11) and we define

ZPzI = -j_(K B+jKBI) • (6.I)

The values of K B and KBI were measured from the gra-

dient of ZP21 of experiment (4) at low frequencies and are

listed in Table 6. Z together with the quasi-steady estimates.

TABLE 6. Z

Compliance Mass flow gain factor

K
BI K B MBI M B

Experiment (4) 0.06 0.3Z Z 0.6

_uasi- steady blade O. 01 0. 53

cavitation calculation

The experimentally measured compliance value K B,

is much higher than the quasi-steady blade cavitation com-

pliance. This may be explained by the fact that the experi-

mental cavitation form was mainly tip cavitation. Thus the

compliance value measured was most probably due to tip

cavitation and not blade cavitation.

One interesting qualitative comparison can be made

betw.een the compliance data for experiment (4) and the

 -148-

bubble compliance calculations by Brennen (1973). The

real compliance of experiment (4)(K B as defined in Eq. (6.1),

the negative gradient of the imaginary part of ZP21) starts

at a positive value at zero frequency. But as the frequency

increases K B begins to decrease and finally crosses zero

into negative values. The imaginary compliance KBI

(gradient of real ZP21 ) starts at a small negative value at

zero frequency and as frequency increases, KBI first de-

creases but then increases and crosses zero into positive

values. These trends are qualitatively similar to the cal-

culated results of Brennen. This seemed to further con-

firm that the compliance value measured was a kind of bub-

bly tip cavitation.

(iv) ZP22-- This term is certainly measured to be non-zero

disproving the zero value assumption prior to the calcula-

tions by Brennen and Acosta (1975}. Even more surprising

• is the existence of a real part with values higher than the

imaginary par t. In a manner similar to the compliance

term, we modify Eq. {1. 13) to include an imaginary mass

flow gain factor MBI:

ZP22 = -jt_(M B+jMBI) (6. 2)

The low frequency approximation of the real and imaginary

mass flow gain factors were measured from Fig. 6. 14 for

experiment (4) and are listed in Table 6.2. The measured

 -149-

M B value is very close to the quasi-steady calculatedvalu_

This may be merely a coincidence since as demonstrated

earlier the main form of cavitation in this experiment was

tip cavitation and not blade cavitation. However, this does

mean that the tip cavitation can give rise to a mass flow

gain factor term. Note that the slopes of both the real and

imaginary parts of ZPZ2 do not change greatly at the higher

frequencies.

The ZPz2 of experiment (3) has very small values.

When this was compared with the ZP22 of experiment (I),

the fully wetted case, they were found to be comparable.

Since the ZP22 of the fully wetted case was considered to be

an experimental error, the measured ZP22 of experiment

(3) was also considered to be within experimental error

range.
-150-
 -151-
Z 11 VS. F_EO.
ii
I I !1111 • ii11

_R
_o

8.
i

w
|

I I I l
Q.O tO,O 20.0 30.0 _10.0 SO.I
F_OUENCT IMZ )

Z _1V5. F_[O.
Z 22 VS. FREQ.
° ,=
II I I I mLl._ tl i_il
I 1 I llilTIl _ *.'l"l.l_fTT

i,-. _. {, ..,

o.4"
.,!_. "

•. I"lr" i ! L ! .
_o

l ?

l o

I I 1 l I'
q
! I I I
e.o m.o 2o.o ao.o _o.o so.o 0.0
FFi(OUENCY (.Z) _o,o lo.o m.o ,_.o _.o
FFIEOI._NCIf (t4Z)

Fig. 6.2 Z-rnatriz of experiment (1), fully wetted conditions for

Impeller IV where o=0.51 Iv=0.25 and el=0.07.
 -1 52-

....4

REPRODUCIBILrrY OF THE
_](_NAL PAGE IS POOR
 - 153-

Z]] YS. FREO.

B
I I J _.It! i ill.In
I RIJUI) • _I#
J

B
.a o

..o
..°. -
- J _: i.... _.... " ._ ,-,,..:

.
Q

*_. °...q

8[ a = = = = = _ =

o.o '_ if.o 9 _.o ,,e

FREr]UEN[..T
_b.p
CMZ )
o W.O ' o.o ! ¢ ' l'o! "'
tf'.o 4' . _.o "P L:o SO,O
FREQUENCY (llZ ')

Z2! VS. FREQ. Z22 vs. PREO.

_.nl._t # • imlw¢ll

i I I l

• • .... * .

.... , ' , I I I, I I, I I,
SO.O
FREgtJZI_T (HZ) FREO(JENCT(HZ)

Fig. 6.4 Z-matrix of experinnent (2), fully wetted conditions for

Impeller V whe._e o=0.82 _=0. 12 a_d $=0.088.
 -154-

l
0

u
F m
7 ,,-4

'I
I o_
I
I
I
u
Q;

I
I
I
L _J

'I
do,
-155-

O
°,.4

(a
o,'4
a
N
u

E
0
• °

o
13_
N

=E

°_

D
r,4

,,o

REPRODUCIBILITY OF THE
0RK_NAL PA6E IS POOR
 -156- -- -- -Linear
extrapolations

ZPII VS. FREQ. ZPL2 V5. FREO.

8 I I I REALt_,:t B. IPVT_41)

c_

0 i I I I
,, .... -I -.-+ ..... }_ .... I .... I ;
N

Re L

9 a_
'% "..

"_ ml
"T
@ \ Re
I
R
"7 I

0.0
,L ;'o L'o " I_0.o ,i
_o.o 0.0
k',o ,o'o01.0
FREQUENCY (.HZ) FREQUENCY (HZ)

ZP21 vs. FREQ. ZP22 vs. FREO.

f I I _!41.( q,) •
I I l _nl.({,)_

o o

_8 : : l T- I -. - j .... -_ :- ___i
,---_ I _ . I .._l'_rl I I

_,f,"
Re
o

i[ = ._ 3 "-"

11i
[ ;_ i.o _lo _l.o L j.O
o.o _ .o . _.o _o.o O.O I .0
]0
I
0
FBEOUENC _ ( l-lZ) FREQUENCY (HZ)

Fig. 6.7 ZP-matrix of e_:periment (I), the fully wetted condition for
Impeller IV where 0=0. 51 _=O. g5 and _= 0.07.
 -157-

ZPi! VS. FREQ. ZPi2 VS. FAEO,

8
i [ l | I l Illll.lll) II IIIIr.llll

S. Ji

8
\

n o°

_o _a l. 1 I I

o4t_ • .... _ ""* ...... • ....... i

o.o
•

{ .o
FI_QUENCT
,° ,_-+
o

(NZ)
.T-----i=.=
;! 0ii/'ij_li,
I .n _l._

FREQUENCY (HZ)
_.a '_ lib.o em,O

ZP21 vS. FREQ. ZP22 v5. FREO.

1 !ll_li'l#'l • t_

i-_i D
Nt i _.-:,-"i'l'-'-';-I .... .." i
""r" ..... *'[ " _'%_ I .

---....
o.

:..' ,I°.'-,L_-t_ '_o

• - - i

,0.
=.°; Jo: 1'oi _'ol '_. SO,O

FREQUENCY (HZ) FFIEOUENCT tHZ )

Fig. 6.8 ZP-matrix of experiment (2), the fully wetted condition for
irnpelier V whore o =0. 82 t_ =0. 12 and 6=0. 087.
 -158-

F-
Z
w
m

0
U_
LL
W
O
CD

W
"I-
\
i \

FLOW COEFFICIENT

Fig. 6.9 Diagra.m showing the change of the pump resistance,

slope of the curve, with increasing frequency.
 -159-

IMAGINARY

0.50 -

2.0 .,11-VA L UES OF k

2 e" _I_

-0.2_.5 I ,_/u I0 \ 0.25 __0.50 0.75 1.00

I ' I

3"41_/
._/_I_ "K_0.8 • REAL

-O.P_5_-- k = 0 -J
I

_'_g. 6.10 Vector diagram showing the real and imaginary parts of
Sears' function, cp(k), as a function of the reduced frequency
k when k is defined as k-cub . In this equation, tu is the
oscillating frequency, b iZt_e semLchord length of the airfoil
and U is the mean velocity.
 REPRODUCIBILITY OF THE
ORIGINAL PAGE IS POI'_I-:

-160-

3,.

X_

O
o
eq

a. o_

l II ._

_ U
 -161 - -- _ -Linear
extrapolations

ZP11 VS. FREO. ZPi2 vs. FREQ.

I I I I _ Iq[._(l') L l_t=r.,_])

.4

'il
.P .--...4--
°..-

l..-v" L I I I . ] I I I

"Re'_ IB k.
*. +. 4 i.. .S,.&_L.L

,-,J
e,,,_
f

T "e > '6,

Re

T
8

0.0
L
_.o _.o i, . ,o o _o .o. 0._
L
t_.o ,, k'oLo
_.o _.0

FREOuENCY (HZ) FREOUENCY (HZ)

ZP21 V5 c_:n ZP22 VS. FREO.

1 I I _._-4,1,) L i,mo.qT I I I
o

_o
= __!.._ .... t_ I L. __, , I I I [ •

|1
o.o _.o
}o L Lo •0 .0 "0.0 I _ } 0 .0 _o _'0.0 _o.o

Fig. 6. 12 ZP-matrix of experime:at (Z), one of the cavitatir,£

experiments for Impeller IV where o=0. 11 _2=0.26
and 6=0.07.
 -16Z-

I
E_
.,-4

w..w

•! II U

0 Ir I_o

.d

° ,,,,i

I
 Linear
-163-
extrapolations

ZPII VS. FREQ. ZP12 VS. FREQ.

I I l _q.,'gL.(#) & InP_X I) I

I I I P£RL(,I,) L II'_qO( I )

,. ...L.

/ ." "- "--.4--- " "-.

- ///
o
_ e
/
l I ., I I
_j,_ , , , _.. , _° . "--.+..-

-, o* - ° --e-
,t,o°"

"0 _ R
T \Y
\
-r

8 _-. = -_. : • e

0.0 i o I,o _.o ,

'tm o _-_0.0 0.0 L_,
.o L
_B.o
FREQUENCY (HZ)
I,
_b,o i,
_.o
_oi.o
FREQUENCY (MZ )

ZP2! vs. FREQ. ZP22 vS. FREQ.

__ L ' I I I, ' I i i
_8 O.o
_k -'"
0

o
? Re

-1

4 o.o i_ o l,
_.o '" l,
i
S_.o .o
;ooIo L L L'o1]
FREQUENCY (HZ) FBEgUENCY (HZ)

Fig. 6. 14 ZP-naatrix of experiment (4), the second cavitating

experiment for Irrpeller IV "_vhere ,7=0.046 _ ::0. Z5
and 4 = O. 07.
 REPRODUCIBILITY OF TH_
ORIGINAL PAGE IS POOR
-164.

u'3

,d
 -165-

ZPII VS. FREQ. PI2 vs. FAEQ.

s. i I | RULC}I | pl_tl)
$ | I ! lini¢(,) i llzr_ I )
e_

zi

• oO_.. o.°-t =.
°
t _

.... v"""- .... z" I , I'W •

i" I I I
_o
°,o.,_.

s.
m
i

•
0.0 @ I .0 'o
t,t
._.o _
I,
'_.o SO.O
jli!i i,i
Q.0 ,,e t .o 1_.o '_ _.o _"
I,I t,
'lb.o
FI_0LIENCY _.rlZ ) FP,[OUENCY (HZ)

ZP21 vs. FREQ. ZP22 vs. FREQ.

2
I U'Itl L i'wllr.lll ) | II[ILIII • II'Illy_I I i

_z [ [ ." ,.l-.| .... 4- .... L . I i ..P .... r'*

o.

! 'ili,,,t = II 1 3 :.0.0!,I0! 10! t,, i,

, . JIb,o ,e _.o
o.o" llr.o e, i_.o ,e 'JIb.o,e 'iro.o 50.0 SO .0

FI_OU[NC'r (MZ) IENCT (l'lZ)

FR(OL,

Fig. 6. 16 ZP-matrix of experiment (5), the cavitating experimezlt

for Impeller V where o=0. 12 _ =0. 12 and _=0.087.
 -166-

VII. SUMMARY AND CONCLUSION

The dynamic response of a cavitating inducer pump has been in-

restigated experimentally in the present research. The repeatability

_nd the agreement of experimental results when carefully controlled and

when the system elfect is removed demonstrate that it is possible to

,_easure the transfer function of a pump operating at constant rotational

_peed. The present experiments show that the dynamic response of

cavitating pumps has a very complex behavior in frequency which cannot

)e explained by extrapolations from steady-state performance nor

quasi-steady analysis. These findings appear in all of the eight coeffi-

Cients of the transfer function matrix. This is particularly so for the

pump gain, ZPll' the results for which were unexpected. There is

rather a large phase lead for this term and the increase in magnitude

with frequency of pump gain represents a deterioration o£ cavitation

performance under dynamic conditions.

The remaining coefficients of the pump transfer function all

show a behavior with frequency that cannot fully be accounted for by the

quasi-steady assumption. Some particularly significant features in this

respect are:

(i) The pump resistance appears to become smaller at the

higher frequencies for both impellers.

(2) There are both real and imaginary parts for the pump "com-

pliance" unlike the approximations that have been made to

date in the field. In addition, there are important changes

with frequency that cannot be accounted for with quasi-steady

 -167-

cascade theory.

(3) The tests on Impeller IV reveal that the compliance is

much larger than that deduced from quasi-steady cascade

theory. Observations reveal that the predominant form of

cavitation is bubbly tip clearance flow and not blade

c avitation.

(4) The mass flow gain factor (associated with the term ZPzz)

appears to be significant and should be taken into account

in pump dynamic representations. It too, is seen to have

real and imaginary parts. The quasi-steady blade cavi-

tation calculation of this effect predicts only an

imaginary part.

There are a few areas in which future experimental research

should be directed to improve the experimental results:

(i) There is a need for a better fluctuator. As can be seen

from the oscilloscope traces of the signals, they all ap-

pear to be in the form of square waves at the low fre-

quency end. This produces a low signal-to-noise ratio

for the basic frequency component and thus reduces the

accuracy of the result. An ideal fluctuator will produce

a precise sine wave which can be easily analyzed.

(ii) Tha laser doppler veloc_meter measures the velocity at

a point inside the flow channel. This is satisfactory but

not ideal. A true mass flow rate measuring system will

be needed to give a more true representation of the ZP-

matrices. Hopefully, the new system will not need the

 -168-

downstream smoothing chamber which introduces a large

system compliance.

(iii) The present data acquisition system of recording and

playback is very time consuming. The total time required

for one ZP-matrix consisting of experiment, playback and

signal analyzing, data averaging and computer processing

needs a minimum of thirty man-hours. A considerable

amount of time win be saved if the experiment can be re-

Corded digitally. The digital records can be processed

directly by the computer.

(iv) With +he improvements above it would be possible to carry

out the experiments at much lower levels of excitation,

0. 5 percent mass flow fluctuation or less.

(v) Finally, there is a need for a tip cavitation compliance

calculation similar to the bubble compliance calculated by

Brennen (1973) such that a quantitative comparison can be

made directly to the present experiments.

 -169-

REFERENCES

Acosta, A. J., 1958, "An experimental study of cavitating inducers",

Proc. of Second O.N.R. Symposium on Naval Hydrodynamics,
August 25-29 (ACR-38).

Acosta, A. J. , 1975, "Unsteady effects in flow rate measurement of

the entrance of a pipe", Technical Brief, California Institute of
Technology, Division of Engineering and Applied Science.

Anderson, D.A., Blade, R.J. and Stevens, W., 1971, "Response of a

radial-bladed centrifugal pump to sinusoidal disturbances for non-
cavitating flow", NASA, TN D-6536.

Brennen, C., 1973, "The dynamic behavior and compliance of a stream

of cavitating bubbles", J. Fluids Eng., Vol. 95, Series l, No. 4,
1973, pp. 533-542.

Brennen, C., and Acosta, A. 3., 1973, "Theoretical, quasi-static

analysis of cavitation compliance in turbopurnps", Journal of Space-
craft and Rockets, Vol. I0, No. 3, pp. 175-180.

Brennen, C. , and Acosta, A. J. , 1975, "The dynamic transfer function

for a cavitating inducer", ASIvIE, 75-WA/FE-16.

Fanelli, M. , 197Z, "Further considerations on the dynamic behavior

of hydraulic turbo-machinery", Water Power, pp, 208-2ZZ.

Farrel, E.G. and Fenwick, 3. R., 1973, "Pogo instabilities suppres-

sion evaluation", NASA Report CR-134500.

Fung, Y.C., 1968, "Aeroelasticity", Dover Press.

Ghahremani, F.G., 1970, "Turbopump cavitation compliance", Report

No. TOR-0059(6531-01)-2, Aerospace Corporation, E1 Segundo,
California.

Gibbs, K.P. and Oliver, A.G., 1973, "Special requirements for pumps
for sodium cooled fast reactors". Pumps for Nuclear Power Plant,
Convention sponsored by the Fluid Machinery and Nuclear Power
Plant Groups of the Institute of Mechanical Engineers, University of
Bath, April Z2-25, pp. 119-124.

Goldstein, R. J. , and Adrian, R. J. , 1971, "Measurement of fluid vel-

ocity gradients using Laser Doppler Techniques", Review of
Scientific Instruments, Vol. 42, No. 9.

Henderson, R.E. , 1972, "The unsteady response of an axial flow

turbomachine to an upstream disturbance", Cambridge University,
England.

Jaeger, C. , 1963, "The theory of resonance in hydro-power systems,

discussion of incidents and accidents occurring in pressure systems,
J. Basic Eng., Vol. 85, pp. 631-640.

5"6nsson, L. , 1974, "Laser velocity meter for water flow with low pa'r-
ticle concentration", Journal of Hydraulic Research, Vol. IZ, No. I.

REPRODUCIBILITY OF THE
ORII_I_IAL PA6E IS POOB.
 -170-

Kamijyo, K. and Suzuki, A., 1973, "An experimental investigation of

flat-plate helical inducers for rocket turbopumps", National and
Aerospace Laboratory of Japan, NAL TR-345.

Kim, J.H. and Acosta, A. J., 1974, "Unsteady flow in cavitating

turbopump", J. Fluids Eng., Vol. 96, pp. 25-28.

King, J.A. , 1970, "Low speed inducers for a rocket engine feed
system", NASA Report CR-72716, R-8272.

Knapp, R. T. , Daily, J.W. and Hammitt, F. G. , 1970, Cavitation,

McGraw-Hill, Inc., New York, N.Y.

Liao, G.S., 1974, "Protection of boiler feed pump against transient

suction decay", J. Eng. for Power, Vol. 96, pp. 247-255.

Liao, G.S. and Leung, P., 1972, "Analysis of feedwater pump suction
pressure decay under instant turbine load rejection", J. Eng. for
Power, Vol. 34, pp. 83-90.

Lock, M. and Rub[n, S. , 1974, "Passive suppression of POGO on the

Space Shuttle", NASA Report CR-13Z45Z.

NASA, 1970, "Prevention of coupled structure-propulsion instability",

NASA Report SP-8055.

NASA, 1971, "Liquid rocket engine turbopump inducers", NASA

Report SP-8052.

Natanzon, M.S., Bl'tsev, N.]., Bazhanov, V.V. and Leydervarger,

M. R. , 1974, ',Experimental investigation of cavitation-induced oscil-
lations of helical inducers", Fluid Mechanics, Soviet Research,
Vol. 3, No. I, pp. 38-45.

Norquist, L.W.S., Marcus, J.P. and Ruscio, D.A., 1969, "Develop-

ment of close-coupled accumulators for suppressing missile longi-
tudinal oscillations (POGO):, ALA_ 5th Propulsion Joint Specialist
Conference, AIAA Paper No. 69-547.

Rocketdyne Report, 1969, "Investigation of 17-Hz, closed-loop insta-

bility on S-II stage of Saturn V", Rocketdyne Division, Rockwell
International, Report No. R-7970.

Rocketdyne, 1974, "Test results of SSME low pressure oxidizer turbo-

pump model inducer", Internal letter No. R/H 4194-3074.

Rothe, K., 1974, "Turbopump configuration selection for the Space

Shuttle Main Engine", ASME 74-FE-23.

Rubin, S. , 1966, "Longitudinal instability of liquid rockets due to pro-

pulsion feedback (POGO):, J. Spacecraft and Rockets, Vol. 3, No. 8,
pp. 1188-1195.

Rubin, S., Wagner, R.G. and Payne, J.G., 1973, "POGO suppression
on space shuttle - early studies", NASA Report CR-2210.

Sack, L.E. and Nottage, H.B., 1965, "System oscillations associated

with cavitating inducers", J. Basic. Engr., Vol. 87, Series D, No.
4, pp. 917-925.
 -171-

Stevans, W. and Blade, R.J., 1971, "Experimental evaluation of a

pump test facility with controlled perturbations of inlet flow", NASA
Report No. TN D-6543.

Streeter, V.L., and Wylie, E.B., 1974, "Waterhammers and surge

control", Ann. Rev. Fluid Mech., Vol. 6, pp. 57-73.

Stripling, L.B. and Acosta, A. J. , 1902, "Cavitation in turbopumps -

Part I", J. Basic Engr., Vol. 84, No. 3, pp. 326-338.

Wagner, R. G. , 19710 "Tital LI engine transfer function test results",

Report No. TOR-0059 (G471)-9, Aerospace Corporation, El Segundo,
California.

Vaage, R.D., FicUer, L.E. and Zehrde, R.A., 1972, "Investigation of

characteristics of feed system instabilities _',Final Report MCR-Y2-
107, Martin Marietta Corporation, Denver, Colorado.
 -172-

APPENDIX A

LASER DOPPLER vELOCIMETER

The laser doppler velocimeter requires a transparent working

..quid with. some light scattering particles moving with the fluid. The

_rinciple is based on the doppler frequency shift due to the movement

of a reflecting surface. In this particular case, a laser beam is split

into two beams, a weak one and a strong beam. Particles in the fluid

-3catter the light in the strong beam. This scattered light is mixed

with the weak beam in a photomultiplier. The two beams, being of

_lifferent frequency, produce a voltage output from the photomultiplier

vcith frequency equal to the doppler beat frequency as given by

n
_Jd = ([ )u. (ks s -ki)

_here Vd is the beat frequency, n is the refractive index of the fluid,

k is the vacuum wavelength of the incident radiation, u is the particle

velocity, and k. and k are the unit vectors in the direction of the
"-1 _SS

incident and scattered light. Thus the doppler frequency is directly

proportional to the velocity of the flow. For a more detailed descrip-

tion see Jibnsson 1974, Goldstein and Adrian 1971.

In the present system which was conceived and designed by

Dr. S. Barker formerly at this institute, the beam of a 5row He-Ne

1
continuous laser i_ split into two parallel beams by a precision glass

block tilted at a slight angle to the beam direction. One of the beams

1
Spectra Physics No. lZO
 -173-

which will eventually shine directly into the photomultiplier Z" _s weak-

ened by a neutral density filter and is referred to as the weak beam

whereas the other one is referred to as the strong beam. The direction

of the measured velocity is in the plane of the two beams and perpen-

dicular totheir bisector. The velocity of the fluid lathe direction defined

above, V, isgivenby
_oVD
V=
2n sin(0/Z)

where Vs is the beat frequency, _o is the laser wavelength, n

is the refractive index of the fluid and 0 is the angle between

the two beams. The two beams-are focused to cross at the center of

the flow channel by a 50 cm (19.3 in.) converging lens. The two beams

intercept with a scattering volume of about 0.79 in. in length and

0. 006 in. width. The output of the photomultiplier is an ac voltage

having a frequency equal to the doppler frequency as given above.

The rest of the system consists of electronic processing of the

doppler signal to give a voltage directly proportional to the velocity.

The signal out of the photomultiplier is amplified first by a pre-amp

stage and finally by two more stages. The last amplifier stage has a

voltage limiter and shapes the signal to approximate a square wave.

3
This output is connected to a phase-lock loop where the output is a

more precise square wave. The improved wave form leads to a better

signal-to-noise ratio. The square wave is fed into a frequency divider

4
and hence into a frequency to voltage converter.

Z RCA No. 8645 photomultiplier

3 Signetics NE 565A
4 Anadex No. PI-408R
 - 174-

The frequency divider divides the frequency into a more convenient

frequency for the frequency to voltage converter which can only accept

up to 50k Hertz. The normal operating range of the doppler frequency

is of the order of 80K-200k Hertz. Finally, the output of the fre-

quency to voltage converter is a voltage proportional directly to the

measured velocity. These details are sketched in the flow diagram

of Fig. A1.
5
Two such systems were constructed to be used in this study.

They are essentially identical except for slight differences in the fre-

quency to voltage converters. The two converters were purchased at

6
different times and have different model numbers. The calibration

of these systems is given in Appendix B.

5
CIT Aeronautics electronics shop, Mr. V. Sodha
6
The latter Anadax model PI-608R
 -175-

o3
_zo
o
_- ,
G.-r. _

.:K
Z

rt-
W
,,z,
O >4 >ft. O

9 ,-,_5 _
JcL O u
83 4-
O

Pr"
O
:.D
0
z
I]'
uja-

' o
--
,,_ -r" IZ)
13
L_

,.4
a3 I

o3 I 7 _ Z ',z"
,,i > nn (.D
"i- 1.1.1! O
w(.) Z 0
O ! d
n,- I O
13_ d
x
I:K >O "_O I..,JO
LI_I

}
IO
_J
[1.
G_
t G..J
_on-
O LI_
,<
:3
!
n- 5 5°
_o 8,
° ,,,,_

LI_I
O3
D n
_J
 -176-

APPENDIX B

CALIBRATION OF MASS FLOW" MEASUREMENT SYSTEM

In order to truly represent the mass flow rate by the laser dop-

pler velocimeter measurements, it is desirable to minimize the bound-

ary layer in the section where the velocity is measured. The nine-to-

one converging nozzle just prior to the measuring station was designed

to do this. To assure that we have obtained a good flat velocity profile

_cross the channel, a small Pitot tube was used to traverse the two

laser viewing sections to measure the steady velocity profile. It was

found that the boundary layer thickness was typically about 10 percent

of the radius as shown in Figs. Bl and B2. The slope of the profile

was due to the reduced cross sectional area occupied by the Pitot tube.

The laminar boundary layer calculation gives a thickness of about 5 per-

cent of the radius az the viewing station. This calculation used the re-

sult of a flat plate laminar boundary layer starting from the end of the

convertent nozzle. The discrepancy between measured and calculated

values is due to the crude assumption made in the calculation.

There is an important issue of whether the oscillating velocity

profile has the same shape as the steady velocity profile since the shape

affects the basic assumption of mass flow measurement. This was re-

solved by Acosta (1975) who found that the thickness of oscillating

boundary layer had littleeffect on this type of mass flow measurement.

To evaluate the smoothing chambers, the turbulence level

against flow rate was measured. This was accomplished by comparing

the rms output of the LDV as measured by a HP 3400 true rms meter
 -177-

v¢ith the dc value. The resultant measurement is actually an inseparable

combination of the fluid flow turbulence and the electronic noise of the

LDV processor. The result is shown in Fig. B3 which demonstrated

:hat the turbulence level £s mostly under l percent of the mean flow.
 =178-

///////////////////////////////////
1.0

0.8 v

0.6

0.4

0.2 v

Y
R
0 I I I I
0.2 0.4 0.6 0.8
U
UMAX

-I.0_
Fig. B1 Steady state velocity profile across the upstream laser
doppler velocimeter viewing channel.

REPRODUCIBILrrY
OF TIlE
ORdeAL PAe_. m POOR
 -179-

Y
O.2 0.4 0.6 0.8 .0
U
-0.2 -
UMAx
W

-O.Z_ -

-0.6 -

Fig. B2 Steady state velocity profi!e acro_:s tile downstream laser

doppler vclocimeter viewing channel.
 -180- '

I I

-r'-

:>.

_Q

4.J
W
,.Q

Z Ca
I

>-
p_

0
_J
W

_ .
b_

¢¢1

I I 0
m
o _. 0
-- 0

39N39NSWN±
39V.[N3OW3d
 -181-

APPENDIX C

FLOW CHARTS OF COMPUTER PROGRAMS

The block flow charts of the three programs mentioned in

Chapter IV are given here. These are:

(I) RAWDATA

(Z) TRANSCO

(3) PUMA
 -18Zo

ko,_o
0 B

i:x,o

&)
=:_, = co

°_

n_

U _ 0 I

0
._J _u
LL.

u_
0 _o'_ _ 0
u _u
.J
rn 0

----1
i ..=i

I
m

4 9_
f_ x,)
,,-,-d I t'-
v_ U
" " '['- _ uF
_',_ 0
U
\ Z' 0 0

0
c
t-- i
c
,C 0 _
ig o
0
0
,.1 [-

6
&
0
0
©
 -184-

°,
m

0
L
D,.

II
Im

_-0 '
0 •
0"=
IIU

•_ 0
_u
®11
_z

....

o
g

<
rO
 -185-

TRANSCO BLOCK FLOW CHART

This program finds the transfer f_mct_on complex matrix Z,as determined by
pressure _,d mass flo_,' measurements made at the upstream and downstream measur-
ing station:, and provided to the program by data cards. One $1_Tis composed of
NUMWAY Z matricel at eac.h oE NFREQ Irequencies. The program computes and
processes one set entirely bL'fo:'e it starts on {he next set. NSETMX sets form one
"bunch" of sets. Eunches of s<.Ls are processed until all data cards have been used.

Data cards for one set are grouped by frequency. For each frequency, one card
identities the frequency, and NK (NK_ Z) cards follow, each providing a pressure and
mass flow at the upstream and at the dow'nstream measuring stations. Lf NK > Z, the
program may compute Z matrices using difierent combinations (specified in the a_ray
NWAY by the user _or each set) of the NK data cards."

[ S_',_'_T ]

NO

!
YES

0
i

READ DATA CA/_DS "

/'or this bunch of sets: - .
NSETMX = Number of sets in this b_nch.
NFREQ = Number of frequencies per set.
NK = Number of data cards per /requency
(not counting the Irequency data card)_
: (cards/setting) X setzinSs

i/
I
300 DO-LOOP IS = I. NSFITMY5_ I
!

l START OF ]00 DO-LOOP

Continued on next page

 -186-

READ DATA CARDS

for this set

-- Number of different ID first digits

,NUMWAY = Number of calcuAations of transfer functions at each
frequency, each calculation using a different combinatior,
NWAY of flow settings
NWAY (I, J) -- Combinations of flow settings identified by LD first
(for I = I. digit of data cards
NUMWAY
J=l,9)
N Plot = YES Lf and only iI plots of Z versus frequency desired
for this set.
N Terms = YES For printout of Z x upstream f/ow
N Punch = YES For punch of Z

I
]00 DO-LOOpr JJ = I, NFRE_

7
stream pressures and mass flows for]
[:Lead and write upstream and down-
each frequency. " J

I _ubr out_rle TKANZ

calculates the complex

h i CALL TRANZ IIL transfer function

a given frequency,
and combination
Z at
J J,
of data,

[ NW, according
de fir, it, ion:"
to the

DMOV Z = DMOD
D POU D POD

NO _

I Note: The Z matrices are

WKITE
Plot TERMS,
Z vs Freq, or | ZlZ (JJ. NW) = Complex
Punch Z, | number
1 in the
|Row I form"
Freq. Combination
• As requested for this set. | Column 2 Nu.rnber

GO TO A NO
 -187-

PUMA BLOCK FLOW CHART

T1hls program finds the pump transfer function complex matrix ZP, given ZU
and ZD matrices and • ZT matrix input on data cards. One "set" isc6mpleted when
NUMWAY ZP matrices at each ol NFREQ frequencies is calculated, printed, plotted,
and punched. The program repeats this process NSETMX times to complete one "bunch
of sets". Bunches of sets will be processed for however many dat_&re In the input deck.

I
NSETMX, READ NFREQ,
Data NUMWAY
Cards:

300 DO-L.OOP. JS=l. NSET_!(li,_e 3S)

Start of 300 DO-LOOP i .

l
INPUT:

specifying experimental conditions:

DATE, IDa, P.PM, GPM, HDRISE, HDCO, FLOC

CAVNO, FLUU. FLUD, FLUP

!
Calculate
All elements of ZU and ZD
matrices for e&ch frequency
according to/ormulas in the
program:

i
for ,II elements of ZT matrix
_or each frequency and each of
NUMW&Y combination numbers.

I
Continued on next page
 (
!
Starts of ZZO DO-LOOP (
(One loop
matrices
calculates HUMWAY
at one frequency. )
ZP
I
_ ' ii -'- CTOMATO form. the

-'" i ' ! ' ' _ L.-) variables representing

-- ' -' Lmatrix in its agrurnent.

SO DO- LOOP = l, NUMw^Y
I
Start o_ ZIO DO-LOOP
(One loop calculates ZP for one combination nurnber_
JL

Inverts a Z X 2 complex
matrix by an algorithm
using a 4 X4 real matrix --'
and the IBM MATINV
SUBROUTINE COMATI 1
z outine.

rFROMAT ¢onvertl the

J 2 x 2 complex ZP matrix
_ into 4 separate variables.
| each w_th frequency &rid
} &rid combination number
I_. subs c ript s.

Continued on next page

 -IB9 -

NO _ YES

Plots ZP versus frequency

for each combination number.
CALL ALLPLT using a separate graph for
each complex element of ZP.

B,_RODUCIBII.,141_ OF _IB
ORIGINAL PA_S m PO0_
 -190-

APPENDIX D

SPECIAL LEAST SQUARE FIT OF THE Z-MATRIX CALCULATION

The formula for the transfer matrix Z is

where all the elements are complex.

To solve for the Z elements for a given set of mean flow con-

ditions, a minimum of two linearly independent sets of measurementst

I

Pu'Pd'm U and m d are needed as mentioned in the introduction. How-

ever, in the present experiment more than two data sets were obtained

and a least square fit procedure was needed to find the best fit Z-

matrix. Denote these sets of experimental data by

i i i i
Pu' Pd' mu' md i = l,Z . . . I

where I is the total number of experimental sets. The procedure sub-

stituted these data into Eq. (1) and calculated the two residue values

for each experimental set assuming the Z elements were known. Next,

the summation of all the square of the amplitudes of the complex resi-

dues for each equation of Eq. (1) was calculated separately. The two

unknown elements in each equation were varied independently by differ-

entiating to minimize this residue summation. This resulted in two

linear equations which were then solved to give the two unknown Z

elements. This was repeated for the other equation to give the other

two elements.

The results of this calculation are:

 -191-

S5S2-$3S 6
ZII = - 1 (D3)
SIS5-$3_ 3

SIS6-_3S 2
ZI2 = (D4)
SIS5-$3_ 6

where
I I

Sl=_ piup: (D5) SZ =/, Pd p: (D6)

i=l i=l

1 I

S3 = _..J
m u mu (D7) s4= md Pu (DS)
i=l i=l

I I
,,,-,,,,,-,,m,,,--

= mu mu u
(DI0)
i=l i=l

and
$4S 5 -$7S 3
Z21 - (DI I)
S1S 5 -S3_ 3

SIS7-_3S 4
Z22 = - 1 (D12)
S1S5-S3g 3

where
i
(DI3)
i=!

These equations were used to least square fit the experimental data.