THE DESIGN OF THE

MIT BLOWDOWN TURBINE FACILITY

by

Alan H. Epstein
Gerald R. Guenette
Robert J.G. Norton

GTL Report No. 183 April 1985

This research was sponsored by the Office of Naval Research,

Mr. M. Keith Ellingsworth technical monitor, and by Rolls-Royce Inc.
 2

ACKNOWLEDGEMENTS

Mr. R.R. Moritz of Rolls-Royce Inc. has been instrumental in the

conception, design, and construction of this facility. Messrs. James Nash,

Joseph Marksteiner, N.R. Andrews, and V. Dubrowski, and Mrs. S. Shemer have

contributed much to the assembly of the rig. Mr. Yu-Zhang Cao has assisted

in the instrumentation development. The authors gratefully acknowledge

many useful discussions with Professors E.E. Covert and J.L. Kerrebrock,

Dr. K.W. Bushell, and Dr. A. Wood.

 3

SUMMARY

This report details the design, construction, and preliminary testing

of a short duration (0.4 sec) test facility capable of testing 0.5 meter

diameter, film cooled, high work aircraft turbine stages under conditions

which rigorously simulate actual engine conditions. The simulation capa-

bility of the facility extends up to 40 atm inlet pressure at 2500 0 K (4000 0 F)

turbine inlet temperatures. The facility is intended primarily for the

exploration of unsteady, three-dimensional fluid mechanics and heat transfer

in modern turbine stages.

 4

TABLE OF CONTENTS

Page

Acknowledgements 2

Summary 3

Table of Contents 4

List of Tables 6

List of Figures
1.0 Introduction

2.0 Facility Scaling 11

2.1 Facility Configuration 16

3.0 Detailed Facility Description 22

3.1 Main Valve 23

3.2 Test Section 26

3.3 Eddy Brake 29

3.4 Tanks and Auxiliary Systems 37

4.0 Data Acquisition System 39

4.1 Analogue Front End Characteristics 42

4.2 Channel Setup and Checkout 42

5.0 Flow Instrumentation 44

5.1 High Frequency Response Measurement 44

5.2 Low Frequency Response Measurement 46

5.3 Traversing Mechanisms 47

6.0 Heat Transfer Measurement 49

6.1 Methods of Determining the Surface Heat Flux 49

6.2 Direct Measurement of Heat Flux 51

6.3 Bonded Gauges 52

6.3.1 Low Frequencies 52

6.3.2 High Frequencies 56

6.3.2.1 One-Dimensional Heat Flux Analysis 57

6.3.3 Calibration of the Insulator Thermal Properties 58

6.3.3.1 Phase Shift Calibration 59

6.3.3.2 Step Response 61

6.3.4 Thin Film Temperature Sensors 62

6.3.4.1 Thin Film Sensitivity 62

 5

TABLE OF CONTENTS (Cont.)

Page

6.4 High Output Thin Film Gauges 66

6.4.1 Frequency Response 67
6.4.1.1 Response to a Step Change in Heat 68
Transfer Rate
6.4.1.2 Steady State Response to a Sinusoidal 70
Heat Transfer Rate
6.4.2 Sensor Design Considerations 72
7.0 Initial Experiments 77
8.0 Conclusions and Summary 79
Appendix A - Blowdown Dynamics 121
A.1 Blowdown of Supply Tank 121

A.2 Corrected Speed 123

A.2.1 Equation of Motion 123

A.2.2 Turbine Corrected Speed 124

A.2.3 Turbine Power 124

A.2.4 Eddy Brake Power 125

A.2.5 Corrected Speed Equation 126

A.3 Conditions for Constant Correct Speed 128

Appendix B - Eddy Current Brake Theory 131

B.1 Load Requirements 131

B.2 Eddy Current Brake As Blowdown Turbine Load 133

B.3 Eddy Brake Configuration 134

B.4 Excitation Magnets 137

B.5 The Loss Drum 139

B.6 Eddy Current Brake - Simplified Theory 140

B.7 Magnetic Reynolds Number - Influence of Induced Field 141

on Braking Force
B.8 Eddy Current Brake - Induction Motor Model 142

References 151
 6

LIST OF TABLES
Page

Table 2.1 Composition of Thermophysical Properties 14

Table 2.2 MIT Blowdown Turbine Scaling 15

Table 3.1 Comparison of Eddy Brake Drum Materials 32

Table 3.2 Eddy Current Brake Configuration Summary 33

Table 4.1 Data Acquisition System Specifications 43

Table 7.1 Comparison of Initial Test Results 78

 7

LIST OF FIGURES

Page
Figure 2.1 Candidate Facility Configurations 80
Figure 2.2 Typical Predicted Blowdown Turbine Mechanical 81

and Corrected Speed Variation

Figure 2.3 Typical Predicted Blowdown Inlet Pressure and 82

Turbine Pressure Ratio History

Figure 2.4 Predicted Variation in Turbine Torque Produced 83
and Eddy Brake Torque Absorbed
Figure 3.1 Blowdown Turbine Facility External View 84

Figure 3.2 Blowdown Facility Simulation Capability 85

Figure 3.3 Main Annular Plug Valve Scheme 86
Figure 3.4 Main Valve Construction Detail 87

Figure 3.5 Valve Actuation Sequence 88

Figure 3.6 Valve Analytical Model Constituents 89

Figure 3.7 Typical Valve Model Prediction 90

Figure 3.8 Comparison of Measured Valve Behavior With Model 91

(Measured Plots Shifted by 30 ms)
Figure 3.9 Turbine Facility Flow Path 92

Figure 3.10 Test Section Detail 93

Figure 3.11 Eddy Current Brake Torque vs. Rotational Speed 94

Figure 3.12 Candidate Eddy Brake Magnet Configurations 95

Figure 3.13 Original and Modified Brake Magnet Detail 96

Figure 3.14 Brake Magnet Current Switching Circuit 97

Figure 3.15 Magnet System Electrical Performance 98

Figure 3.16 Measured Brake Power Absorption 99

Figure 3.17 Facility Oil Heating System 100

Figure 3.18 Facility Gas Mixing System 101

Figure 3.19 Auxiliary Instrumentation Calibration Tunnel 102

Figure 4.1 Data Acquisition System Functional Layout 103

Figure 4.2 Data Acquisition/Analysis Computer Systems 104

Figure 5.1 Four Sensor Cylinder Probe Measures Total and Static 105
Pressures and Two Flow Angles. Frequency Response is
d.c. to 12 kHz.

Figure 5.2 Single Sensor Total Pressure Probe, 1.2 mm O.D. 106
 8

LIST OF FIGURES (Cont.)

Page
Figure 5.3 Aspirating Probe Measures Total Temperature and 107
Pressure. Frequency Response is d.c. to 20 kHz.
Figure 5.4 Stationary Frame, Radial-Circumferential Translator 108
Figure 5.5 Shaft Mounted Rotating Frame Radial Translator 109
Figure 6.1 Heat Flux Gauge Cross-Section 110
Figure 6.2 Effect of Adhesive Thickness on Bonded Gauge 111
Frequency Response
Figure 6.3 Phase Shift and Attenuation Variation With Heat 112
Flux Frequency for Bonded Gauge
Figure 6.4 Sensor Test Specimen Geometry 113
Figure 6.5 Wheatstone Bridge Nomenclature 114
Figure 6.6 Heat Flux Gauge Layout for NGV 115
Figure 6.7 Step Response Calculatino Nomenclature 116
Figure 6.8 Sinusoidal Response Calculation Nomenclature 116
Figure 6.9 Thin Film Sensor Sensitivity and Frequency Response 117
as a Function of Insulator Material and Thickness

Figure 6.10 Thin Film Sensor Frequency Repsonse Variation With 118
Insulator Thickness
Figure 7.1 Blowdown Turbine Facility Test Raw Tata 119
Figure 7.2 Typical Variation of Correct Speed and Weight Flow 120
Figure A. 1 Eddy Brake Excitation History Required for Constant 130
Corrected Speed Conditions

Figure B.1 Eddy Current Brake Geometry 152

Figure B.2 Eddy Brake Rotational Speed Characteristics 153
Figure B.3 Magnet Detail 154
Figure B.4 Linear Eddy Current Brake Configuration 155
Figure B.5 Induced Current Nomenclature 156
 9

1.0 INTRODUCTION

As the performance required of high pressure, cooled turbines con-

tinues to increase, there is an insistent need to explore many of the

details of turbine fluid physics which have hitherto been overlooked,

either because they seemed of second order, or were simply too difficult to

attack with the technical tools at hand. Three-dimensional effects on

turbine flowfield falls perhaps in the first category, while unsteady heat

transfer and rotational effects may belong in the latter. Both of these

problems could be studied at full scale in an engine, but that would be an

extraordinarily complex task. The engine environment is extremely harsh,

making instrumentation very difficult. The limited access to the turbine

flow path exacerbates the problem, as do the necessary safety and opera-

tional constraints which restrict parametric investigation. Also, of

course, an engine is an expensive device to operate. One solution to the

instrumentation access and parameter variation problem is a test facility

designed for the full turbine environment. This can be even more expensive

to build and operate, however, (and as difficult to engineer) as a large

development gas turbine engine. The one such facility at the NASA Lewis

Research Center will be, perhaps, the penultimate turbine test rig. It

requires a considerable fraction of the agency's turbine research resources

and thus is beyond the scope of university research.

There is also a need for lesser facilities which can simulate most of

the important high pressure turbine parameters but in a more benign environ-

ment and at a much lower cost. The study of turbine fluid physics is, in

particular, well suited to the use of transient facilities since most high

speed turbine aerodynamics and heat transfer phenomena occur on blade

passing time scales. Several such facilities are in use or under construc-
 10

tion. In the research community, this includes light isentropic piston

cascade tunnels, blowdown cascades, and a small shock tunnel driven turbine.

With the exception of the latter, these are stationary cascades designed

primarily either for airfoil heat transfer or aerodynamic studies. The

shock tunnel is a heat transfer tool since its very short test time and

small size preclude extensive aerodynamic instrumentation.

The goal of this project is to design, build and utilize a short

duration turbine test facility benefiting from the experience gained from

past turbine efforts and the MIT Blowdown Compressor Facility. The philo-

sophy directing this endeavor is summed up in the following guidelines:

1. The full fluid physics of the external flowfield and heat transfer must

be simulated, including temperature ratios, viscous effects (Reynolds

No.), rotational effects (Rossby No.), compressibility effects (y and

Mach No.), at the proper corrected speed and weight flow.

2. A full stage (nozzle guide vanes and rotor) must be tested with pro-

visions for a following blade row.

3. The turbine should be as large as possible to facilitate instrumentation.

4. The test time must be sufficiently long for aerodynamic surveys (several

tenths of a second).

5. Instrumentation access is a prime criterion.

6. The construction and operating cost must be of academic scale.

This document describes the Blowdown Turbine Facility, which has been

developed under the preceding criteria. It is designated to test 0.5 meter

diameter film cooled, high work turbines under rigorously scaled conditions.

An integral part of this work is the simultaneous development of the high

frequency response instrumentation (aerodynamic and heat transfer) and the

data acquisition system necessary to realize the facility potential.

 11

2.0 FACILITY SCALING

Proper scaling of the turbine operating environment requires an under-

standing of the basic physical forces influencing fluid flow and heat

transfer. In so far as we know, it is the ratios of forces and states

which control fluid physics. Proper scaling for a rotating flowfield with

heat transfer includes viscous forces (Reynolds No., Re=pUD/i), temperature

ratios, rotational effects (Rossby No., Ro=U/QD), thermal properties

(Prandtl No., Pr=Cp p/k), compressibility (y), and inflow geometry (corrected

speed and weight flow); where p is the fluid density; U, the fluid velo-

city; D, a characteristic length; p., the viscosity; Q, the rotational

speed; Cp, the specific heat at constant pressure; k, the thermal conduc-

tivity; and y, the ratio of specific heats. Turbulent intensity may also

have an important influence on the heat transfer. For a film cooled turbine,

the mass flux ratio, the blowing parameter (pcUc/pAU.) is of interest,

where the subscripts c and m refer to the coolant and the main flow.

We can write the scaling equations quite conventionally as follows,

prime denoting the scaled conditions.

Temperature Ratios:

T' T
9 - (2.1)
T' T
m m

T' T
c c (2.2)
T' T
m m

The subscripts g, m, and c refer to the main gas flow, the metal, and the

coolant gas respectively.

Compressibility:

yI = y (2.3)
 12

Reynolds Number:

Re' = p'U'D' I (2.4)

Re p U D p

'
For Re' = Re, Eq. (2.4) can be written as

-P' = p'- -m I' T1

- D
- (2.5)
P p m'y T D'

where P is the total pressure and m is the molecular weight. The

rotational effects may be scaled either as Rossby number or corrected

speed. Here we have chosen corrected speed.

N'D' ND (2.6)
y'T'/m' y T/m

for y' = y, Eq. (2.6) becomes

N' T' m D (2.7)

N T m' D'

Buoyancy effects due to rotation scale as the Rayleigh number, Ra

2
Ra = r Re rAT 2 8)
Ro

where r/D is geometry. Thus, the above equations imply that buoyancy

effects are properly modelled so long as

Pr' = Pr (2.9)

The key scaling is that of temperature ratio. It is primarily the

very high temperature of the full scale turbine which renders investiga-

tions both difficult and expensive. By reducing the absolute level of

temperature while maintaining the proper gas/metal temperature ratios (Eqs.

2.1 and 2.2), the fluid mechanics and heat transfer can be accurately

simulated (so far as we know) under relatively benign conditions. Note

that, by reducing the absolute gas temperature, the pressure required for
 13

Reynolds No. matching (Eq. 2.5) is reduced, thus reducing the facility

construction and operating costs. Also, the mechanical rotating speed of

the turbine is decreased, thus reducing rotational stresses and simplifying

the mechanical design and rotating system instrumentation. The decrease in

rotational speed also reduces the frequency response required of fixed

frame instrumentation for a given rotational frame resolution.

At low speeds, fluid flow is not strongly affected by variations of

the ratio of specific heats, y. For transonic and supersonic turbines,

however, y matching may not be simply ignored. At a given temperature, y

is a function primarily of gas composition. The y can be adjusted through

an appropriate combination of high and low y molecules, i.e., monatomic and

large molecular weight gases. Through judicious selection of the constitu-

ents, the molecular weight of the gas mixture can be increased as compared

to that of air for a given y. Maximizing the molecular weight has the

advantages of: a) decreasing the pressure level required at constant

Reynolds No. (thus reducing construction costs); and b) decreasing the

speed of sound of the gas (thus reducing the turbine rotational speed with

the benefits listed above). The heaviest monatomic gas whose price is

consistent with large volume consumption is Argon.

The properties of several argon-refrigerant mixtures at typical scaled

conditions are presented in Table 2.1. For the main flow, the 75% Argon-

25% Freon-12 mixture is the composition of choice because of Freon-12's low

cost ($2 per Kg) compared to the heavier mixtures. The film coolant flow

must be at low temperature (~200 0 K). This requires a saturation tempera-

ture lower than that of Freon-12; thus Freon-14 is selected for the coolant.

The scaling laws were applied to a typical modern turbine design with

the results illustrated in Table 2.2. Given the choice of gas species, the
 TABLE 2.1
COMPOSITION AND THERMOPHYSICAL PROPERTIES

YMIX = 1.28, T = 400 0 F, P = 14.7 psia

Refrigerant Argon - Refrigerant Mixtures

MMIX CPMIX RMIX YMIX MMIX KMIX Pr

Mass% Mole% MIX
BTU
BTU BTU BTU lbm T
lbm-mole) (lbm-OR lbm- 0 R (sec-ft) sec-ft-0 F

R-12 50.94 25.54 60.63 0.1500 3.276x10-2 1.279 1.785x10- 5 3.548x10- 6 0.755

R-13 49.58 27.32 57.58 0.1579 3.449x10- 2 1.279

R13B1 57.84 26.90 69.26 0.1313 2.867x10- 2 1.279 1.939x10- 5 3.515x10- 6 0.724

R-14 47.80 29.36 54.06 0.1682 3.674x10- 2 1.279 1.994x10- 5 4.222x10- 6 0.794

R-22 52.55 33.84 55.69 0.1633 3.566x10- 2 1.279 1.761x10- 5 3.648x10- 6 0.788

R-115 39.97 14.69 56.76 0.1602 3.498x10- 2 1.279 1.925x10- 5 3.935x10- 6 0.783

Air at 17800 28.97 0.3066 6.856x10- 2 1.289 3.892x10- 5 15.86x10- 6 0.752

14.696 psia
 15

TABLE 2.2
MIT BLOWDOWN TURBINE SCALING

Full Scale MIT Blowdown

Fluid Air Ar-R12

Ratio Specific Heats 1.27 1.27
0
Mean Metal Temperature, Tm 1118 K (1550 F) 0
295 0 K (72 0 F)
Metal/Gas Temperature Ratio, Tm/Tg 0.63 0.63
0 0
Inlet Total Temperature,
Tg
1780 K (2750 F) 478 0 K (4000 F)
Cooling Air Temperature 790 0 K (9600 F) 212 0 K (-770 F)

Airfoil Cooling Air Flow 12.5% 12.5%

True NGV Chord 8.0 cm 5.9 cm

Reynolds Number* 2.7 x 106 2.7 X 106

Inlet Total Pressure, atm (psia) 19.6 (289) 4.3 (64)
Outlet Total Pressure, atm (psia) 4.5 (66) 1.0 (14.7)
Outlet Total Temperature 1280K (1844 0 F) 343 0 K (160 0 F)
Prandtl Number 0.752 0.755
Rotor Speed, RPM 12,734 6,190
Mass Flow, kg/sec 49.00 16.55
Power, watts 24,880,000 1,078,000
Test Time Cont. 0.2 sec.

* Based on NGV chord and isentropic exit conditions.

 16

desired y sets the mixture ratios. The metal temperature selection sets

the inflow and coolant temperatures. Reynolds number similarity then deter-

mines the pressure. The Prandtl number is not a free variable. Fortuitously

in this case, the Prandtl number of the mixture is quite close to that for

high temperature air. The mechanical rotational speed is set to keep the

corrected speed constant. Note that the rotating system stresses have been

reduced by a factor of four and the required instrumentation bandwidth by a

factor of two. The mass flow is down from full scale by a factor of three,

while the power produced (and thus the size required of the power absorber)

is reduced by a factor of 24.

Using the scaled inlet conditions in Table 2.2, it can be seen -that,

for a typical high pressure turbine with a 4 to 1 pressure ratio, the

turbine outflow is at approximately standard pressure and temperature,

further simplifying the instrumentation task.

Clearly, the scaling has enormously reduced the resources required

for turbine testing while preserving the important fluid physics.

2.1 Facility Configuration

Many facility configurations were examined which could meet most, if

not all, of the scaling criteria enunciated in the preceding section. All

were based on transient testing techniques, both to reduce the construction

and operating costs and to exploit many of the transient heat transfer and

fluid dynamic testing techniques developed over the last 20 years. Since

the predominant source of unsteadiness in turbines is rotor-stator inter-

action, a test time on the order of 1,000 to 10,000 blade passings (0.1 to

1.0 seconds) should suffice for most studies.

Four configurations were studied in some detail, either because they

 17

seemed particularly simple, utilized equipment on hand, or built upon

previous experience.

The first configuration consisted of high pressure storage tanks (100

to 200 atmospheres) feeding a flow heater in line with the turbine (Fig.

2.la). A large flywheel could absorb the turbine power produced. The

advantage here is that, with the heater and fast acting regulators in the

flow path, constant turbine inlet pressure and temperature can be main-

tained. The principal disadvantage is that the high pressure gas storage

precludes the use of a heavy gas mixture (the heavy gases condense at the

high pressure).

The second configuration is a variation on the first in which the high

pressure gas storage is replaced by a room pressure expulsion bladder so

that a heavy gas mixture can be used. The bladder would be contained in a

pre-heated shell since little pressure drop would be available for an

inline flow heater. The scheme has the disadvantage of failing to maintain

full Reynolds number similarity. Also, the technology required for a

1OM 3 5000K expulsion bladder appears risky.

The third configuration considered was the original concept which

motivated this project. In this scheme, the turbine is directly coupled to

a compressor whose inlet is the turbine's outlet. Thus, the turbine and

compressor turn at the same mechanical speed and share the same flowpath.

Since a choked turbine produces power proportional to the tip Mach No.

squared and a centrifugal compressor does work as the square of the tip

Mach No., the turbine-compressor pair are matched over a considerable range

and operate at constant corrected speed without the need of a control

system. The corrected weight flow is constant so long as any orifice in

the flow path remains choked. If we now place the turbine-compressor test
 18

section between two tanks, the upstream supply tank filled with the gas

mixture at appropriate conditions and the downstream dump tank evacuated to

full vacuum, the facility is complete with the addition of a valve or

diaphragm between the supply tank and test section.

The design of such a blowdown tunnel was carried out in some detail.

A compressor was chosen in this design over a flywheel for two reasons.

The first is that the compressor can maintain a closer match in corrected

speed compared to the flywheel since, with the flywheel, the turbine mech-

anical speed must always increase (if only a small amount), while with the

compressor, the turbine speed can slow to match the drop in inlet tempera-

ture that is inherent to a blowdown from a fixed volume tank. Thus, the

flywheel may be appropriate for constant inlet condition schemes such as

the two discussed above but an energy absorber is really required for a

blowdown. The second reason is that the large power output of a transonic

turbine requires a very massive flywheel, complicating the mechanical

design.

The compressor design proved to be troublesome, however. The problem

is that the very high work output of a large transonic turbine is consider-

ably greater than any existing single centrifugal compressor stage can

absorb. Since the turbine and compressor are constrained to turn at the

same mechanical speed (excluding gearing), the only way to increase the

power absorption is to increase the compressor diameter. The power scales

with the square of the compressor diameter but the moment of inertia scales

as the fourth power of diameter. Thus, the compressor grows into an enorm-

ous flywheel whose aerodynamic characteristics are of secondary importance.

As the compressor diameter passed one meter in the design, it was clear

that a safe mechanical design of the rotating system would be very difficult.
 19

(Note that the aerodynamic design is not so difficult since the compressor

efficiency does not effect the power absorption, only the pressure rise.

The pressure rise does, however, determine the volume of the dump tank

required. The principal aerodynamic design consideration for the compressor

is that the weight flow vs. speed characteristic match that of the turbine.)

Various compressor designs were then attempted, including multiple

centrifugals, a supersonic axial compressor, and the adaptation of a multi-

stage aircraft engine high pressure compressor. These schemes did not seem

to offer much in reducing either the mechanical complexity or the technical

risk of the project.

Another configuration investigated replaced the compressor with an

eddy current brake. An eddy brake consists of a moving conductor (metal

disk or drum) in an imposed magnetic field. The motion induces secondary

currents (eddy currents) in the conductor which opposes the applied field,

thus generating a braking force. The power is dissipated as heat in the

conductor. The eddy brake is attractive because the braking torque is

proportional to the square of the rotational speed. Thus, an eddy brake

coupled to turbine will, as the compressor, match the turbine characteristic

without the need of a control system. The eddy brake offered the advantage

of a high power density design, simplifying packaging and rig layout.

Unlike a mechanical friction brake, the eddy brake has no moving parts

other than the conductor and thus relatively few mechanical and dynamic

design problems.

Due to the above potential advantages of the eddy current brake and

also due to the investigators' expertise in compressor design and ignorance

of eddy current brakes, the eddy brake configuration was chosen for the

blowdown turbine.
 20

At this point, a turbine stage was selected about which to size the

facility. Instrumentation considerations dictated as large a turbine as

possible. In fact, even the largest production high pressure turbines have

typical blade spans of only 6 cm. Large size means large mass flow,

however, and it is mass flow which drives the facility costs. Since a

turbine from a 50,000 lb. thrust engine was beyond the resources of this

program, a 0.5 meter (20 inch) diameter turbine size was selected. This

choice was consistent with the resources in hand, comparable in size to

that in the NASA High Pressure Facility, and has 2/3 the span of the largest

high pressure turbines.

Although the goal of the blowdown turbine project is the exploration

of phenomena not readily studied in other facilities, it is desirable to

verify early in the program that turbine performance in the blowdown tunnel

is the same as in more conventional facilities. Thus, an existing,

conventionally tested turbine was selected for the tunnel, both to provide

a baseline for performance comparisons and to reduce program costs. The

turbine selected is of Rolls-Royce Limited design and manufacture. It is a

high work, transonic design of the late 1970's. The inlet conditions are

those listed as full scale in Table 2.2. Additional design parameters of

interest include:

2
Turbine Loading, AH/U - 2.3

Total Pressure Ratio - 4.5

Specific Output, CPAT/T - 0.29 kJ/kg *K

Cooled and uncooled blading is available for both the rotor and the nozzle

guide vanes (NGV's).

A calculation of the blowdown tunnel performance was used to demon-

strate the overall concept and to facilitate tank sizing. A mathematical

 21

model of the turbine blowdown process with eddy brake can be found in

Appendix A. The results of the model are shown in the following figures.

Figure 2.2 illustrates that, although the rotational speed changes during

the test, the corrected speed is calculated as constant to better than 1%

over 200 ms, as are the pressure ratio and thus the corrected weight flow

(Fig. 2.3). The calculated torque balance is shown in Fig. 2.4.

The size of the downstream dump tank is set by the overall run time

desired at a particular turbine pressure ratio. The supply tank size is

set by the allowable change in temperature ratio during the test time

(<1%). The coolant flow is fed from a supply tank separate from that for

the main flow. Its size is set so that the volume ratio of the supply

tanks equals that of the flow rates. Thus, the blowing ratio across the

film cooling holes is maintained constant over the test time.

 22

3.0 DETAILED FACILITY DESCRIPTION

The MIT Blowdown Turbine Facility (Fig. 3.1) consists of a supply tank

separated by a large, fast-acting valve from the test section containing

the turbine stage which discharges into a vacuum dump tank. Prior to a

test, the entire facility is evacuated (0.1 torr) by a mechanical vacuum

pump. The supply tank and valve are then heated by oil circulating in a

jacket around the tank. After the desired temperature is reached, the

fast-acting valve is closed and the supply tank pressurized with the

appropriate gas mixture. The turbine (which is still in vacuum) is brought

to operating speed by a small electric motor. The test is initiated by

simultaneously opening the valve and energizing the eddy brake magnets.

The steady test time starts about 200 ms after the valve actuation begins

and lasts for 200-300 ms until the downstream throttle unchokes.

The facility has been sized to allow testing at twice design Reynolds

No. (2x(3x10 6 )) in argon-freon, or at design Re in air, i.e., the

facility design inlet total pressure is 10 atm in the laboratory which

corresponds to 40 atm at engine conditions. The maximum laboratory inlet

total temperature is 533 0 K (5000 F), thus facilitating carbon steel con-

struction and the use of elastomer seals. This corresponds to a scaled

engine temperature of 2300 0 K (3680 0 F) assuming a 1250 0 K (1790 0 F) metal

temperature or a 2500 0 K (4040 0 F) for a 1350 0 K (1970 0 F) metal temperature.

Thus, the facility operating envelope, shown in Fig. 3.2, includes, the full

range of present and proposed engine designs.

The following sections describe the facility in four parts: the main

valve, the test section, the eddy current brake, and the tanks and

auxiliary systems.
 23

3.1 Main Valve

The main valve separates the supply tank from the test section. The

valve must seal the 10 atm supply tank pressure against the test section

vacuum at temperatures up to 530 0 K (500 0 F), open fully in 50 ms, and

provide a disturbance-free inflow to the test section. These are challeng-

ing requirements and the valve design and manufacture developed into a

major engineering effort.

Several classes of valves were reviewed, including explosively or

mechanically ruptured diaphragms (as used on the MIT Blowdown Compressor

Tunnel), small diameter valves followed by expansions and flow straighteners,

and large diameter valves with integral flow contraction. The last type

was chosen because it seemed to offer both the minimum flow establishment

time and smoothest inflow profile, although at the expense of increased

size of the moving parts.

An annular plug design was selected, as conceptually illustrated in

Fig. 3.3. in the closed position, the u.9m diameter cone-shaped plug seals

against the inner and outer annuli of the flow path. To open, the cone

slides forward along its shaft to form a smooth annular contraction from

the supply tank into the test section. The valve flowpath is contoured to

provide a smooth and continuous area contraction to keep the boundary

layers thin.

Since the high temperature operating requirement precludes the use of

aluminum for the moving parts, the entire valve is constructed of mild

steel and, therefore, is fairly massive (100 Kg). This requires that large

forces be applied to accelerate and decelerate the slider in a 50 ms time

period. Conventional very high pressure hydraulic technology could be

used, but the high temperature operation and geometrical constraints con-
 24

siderably complicate the application. Thus, in the interests of mechanical

simplicity, a pneumatic pilot-operated design was selected in which the gas

pressure in the supply tank provides most of the force required for accel-

eration and deceleration.

The valve operation proceeds as follows (with reference to Fig. 3.4).

Initially, the slider is closed, the supply tank (A) is pressurized, and

both the downstream flowpath (B) and the damping chamber (C) in the slider

(D) are evacuated (the damping chamber is connected to the dump tank via

three 2.54 cm diameter lines in one of the shaft support struts (G)). To

start the opening sequence, the 10 cm diameter pneumatic pilot cylinder (H)

on the valve centerline is actuated from a small high pressure argon cylin-

der (I) by a solenoid valve (J). The pilot cylinder accelerates the slider

forward, breaking 0 ring seals (K) and (L). The downstream flowpath now

fills with gas since the nozzle guide vanes act as a downstream orifice.

The damping chamber is also filling but at a very low rate since the area

of the gap formed by the orifice plate (E) and the power cylinder (M) is

very small, 0.5 mm. Thus, a pressure difference is developed across the

conical slider surface, which provides the force necessary to accelerate

the slider. This is the acceleration period (Fig. 3.5). Once the slider

moves sufficiently far so that the power cylinder is no longer over the

orifice plate, the damping chamber rapidly fills to supply tank pressure

through the annular gap, reducing the net force on the slider to zero.

This is the coast period. The slider motion next carries the damping

cylinder (N) over the seal plate, closing down the gap from the damping

chamber to the supply tank. Further slider motion now compresses the gas

in the damping chamber, decelerating the slider smoothly to a stop. Excess-

ive bouncing of the slider is reduced by the small leakage through the seal
 25

plate-damping cylinder gap, by leakage through the vacuum lines in the

shaft, by non-adiabatic effects, and by residual pressure in the pilot

cylinder. Any residual mechanical energy in the system is absorbed by

springs (0) and a crane damper (P).

The valve was designed with the aid of a fairly detailed mathematical

model of the system's fluid mechanics and dynamics. The variables consid-

ered include slider mass, orifice areas and discharge coefficients, power

and damper cylinder lengths, friction, leakage, damper and spring dynamics,

and pilot pressures and volumes as shown in Fig. 3.6. All fluid flow is

assumed one-dimensional and adiabatic. The orifice areas and discharge

coefficients were first calibrated by monitoring the damping chamber

pressure while moving the slider with the pilot cylinder alone and then

fitting the computer model to the data. The predicted pressure, force,

velocity, and position behavior is shown in Fig. 3.7 for a full-scale

blowdown run at room temperature. A comparison of the calculated versus

measured performance (Fig. 3.8) shows good agreement (slightly more bounce

in the actual valve than predicted), thus confirming the fidelity of the

model and the proper operation of the valve. Note that the valve is full

open (full open is defined as the minimum position in which the smallest

valve area is equal to the largest downstream area) 50 ms after the main

seal breaks.

Several mechanical design features of the valve are worth noting. The

slider is carried on graphite rings (Q) which slide on the main shaft (R).

The shaft surface is carbonized to a hardness of Rockwell 55C to eliminate

wear. The main slider seals against a self-aligning (elastomer mounted)

seating ring (S), thus reducing the manufacturing tolerances required. To

insure controlled thermal growth and reduce facility setup time, the main
 26

frame, shaft, and slider are oil heated. The slider position is monitored

by a fiberoptic array in the shaft. The bifurcated fiberoptic bundles are

brought through a pressure/vacuum seal to externally mounted illuminators

and sensors. The chamber pressure is measured with a high temperature

pressure transducer mounted on the seal plate.

The valve dynamics are not independent of supply tank pressure and

gas composition. Damping, in particular, is sensitive to the operating

conditions. If operation considerably off the design point conditions is

required, the dynamics can be fine tuned with an external throttle valve

on the vacuum line from the damping chamber to the dump tank. Coarser

adjustment must be done by altering the lengths of the power and damping

cylinders.

The main valve has been tested at room temperature in air, carbon

dioxide, and an argon-Freon mixture, and at several supply tank pressures

in order to further verify that the computer model accurately simulates the

valve dynamics, as well as at design temperature and pressure (400*F at 65

psia). Performance has been completely satisfactory and very close to

prediction.

3.2 Test Section

The test section contains the inlet section, nozzle guide vanes,

rotating assembly (including the turbine rotor and eddy brake), the exhaust

channel, and the throttle. Figure 3.9 shows the rig flow path while the

test section is detailed in Fig. 3.10. MIT was responsible for the prelim-

inary design of the test section while Rolls-Royce Inc. was responsible for

the detailed design and construction.

The test section is designed to operate at room temperature. A sili-

 27

cone rubber thermal barrier is used at the upstream mating plane with the

main valve to restrict heat flow from the hot valve. Water circulation in

small manifolds on the inner and outer annulus walls helps maintain room

temperature conditions in the test section. Three struts support the

upstream inner annulus and carry the inner boundary layer bleed, the coolant

flow, and instrumentation wiring. These struts mate with the three struts

in the main valve. Immediately downstream of the struts are boundary layer

bleeds on the inner and outer annulus walls which are sized to remove up to

30% of the total mass flow in order to insure clean inflow to the turbine.

The bleeds are manifolded together and discharge to the dump tank. There

is sufficient space at this station to permit the later addition of turbu-

lence generation grids. The boundary layer bleed is followed by a rapid

flow path contraction to the nozzle guide vanes (NGV's), thus duplicating

the geometry in the Rolls-Royce Limited rig in which this stage was

originally tested.

The rotor flow exits into a constant area duct which in turn discharges

radially into the dump tank. A sleeve throttle on the annular discharge is

used to adjust the downstream flow and thus set the stage pressure ratio.

The rotating assembly is supported by two 100 mm diameter, high-speed

angular contact ball bearings with the rotor and eddy brake drum overhung

from either end. The bearings are operated in vacuum and are lightly

grease lubricated to avoid problems with oil contamination. The short run

times preclude the need for bearing cooling. Vibrometers are located on

each bearing mount to monitor the rotating system condition. The rotational

power is provided by a 10 KW d.c. drive motor. The turbine rotor hub

contains a mounting can for rotating system electronics. A 50 channel slip

ring assembly is mounted to this can through a flexible coupling. A 732

 28

count per revolution (12 per blade) optical encoder monitors the rotor

angular position and speed. The encoder, slip rings, and drive motor

operate in the laboratory environment (STP air) and are isolated by carbon

face seals from the test section environment (vacuum to 10 atm). The rotor

cooling air path is isolated from the main flow path by standard knife edge

seals.

The rotating system has been tested to 150% of design speed without

the rotor mounted and 115% as a complete unit. No vibratory problems have

been encountered. One per rev movement at the bearing housing as inferred

from accelerometer measurements is 5 pm (0.2 mil) during a full pressure

test.

The eddy brake consists of a thin (6 mm), 1/4 meter diameter Inconel

718 drum rotating in the magnetic field of 10 U-core magnets. The magnet

assembly can be mounted on load cells (as indicated in Fig. 3.9) which

would then act as the turbine torque meter. The construction of the torque

meter has been deferred, however, in favor of attempting to calibrate the

eddy brake torque vs. speed characteristics to the 0.5% level, thus obviating

the need for a separate torque meter. The eddy brake is described in more

detail in the following section.

What happens if the brake fails to actuate? Turbine rotor runaway is

a concern in all turbine rigs, not just the blowdown turbine facility. The

solution adopted in this case has been to design the maximum safe rotational

speed to be equal to the calculated peak runaway speed. This has been

estimated at 12,000 rpm (twice the normal operating speed) by assuming that

all the potential energy in the supply tank is converted to rotational

energy, a conservative over-estimate. The rotating system design was

greatly simplified by the fact that 12,000 rpm is the operating speed for
 29

which this particular rotor was designed for conventional rig testing at

Rolls-Royce Limited.

Instrumentation and optical access is provided by three 13 cm wide

windows spaced 1200 apart around the outer case. These windows extend from

upstream of the NGV's to 11 cm downstream of the rotor. The NGV's, whose

coolant originally entered from the tip, have been modified to a hub feed

configuration in order to provide unobstructed access for instrumentation

and optical studies.

The rotating system bearings and shaft have been sized to carry the

additional load of a rotating system probe-traversing mechanism mounted

immediately behind the rotor.

3.3 Eddy Brake

Although simple in concept, the eddy current brake proved to be a

challenging design problem. Basically, an eddy current brake consists of a

conductor moving through an applied magnetic field. This induces a second-

ary current in the conductor and thus an induced magnetic field which

generates a force in opposition to the conductor motion. This force (or

torque in a rotating geometry) provides the braking. The power absorbed

appears simply as the resistive dissipation of the induced current in the

moving conductor, i.e., as heat. When the induced magnetic field is small

compared to the applied magnetic field, the torque is linearly proportional

to the velocity past the magnet poles. As the induced field strength grows

relative to that of the applied field, the induced field begins to exclude

the applied field from the conductor. This reduces the incremental rise of

breaking force with speed (i.e., reduces the slope of force versus speed

curve) until the critical speed is reached, at which point the induced field
 30

strength equals that of the applied field and the braking force begins to

decrease with increasing speed.*

A simple model of the basic eddy current brake torque versus speed

characteristic is the induction motor model which closely approximates the

brake behavior up to the critical speed, wo'

T = kB2 (3.1)
0 1+(W/u )2
0

where T is the torque; k, a constant established by geometry and material

properties; B 0 , the applied magnetic field strength; w, the angular

velocity; and wo, the critical velocity at which the induced field

strength equals that of the applied field. This relation is plotted in

Fig. 3.11. At speeds low compared to wo, the torque versus speed rela-

tionship is linear. Clearly, we wish to design the brake so that the brake

critical speed, wo, is well above the turbine operating speed and would

prefer to be on the linear part of the curve.

The critical speed is a function of geometry and material properties,

S (2g/-r (3.2)
o p~aAr

where p is the permeativity of free space; a, the conductivity of the

conductor material; 2g the gap between adjacent poles of the magnet; T,

the pole pitch of the magnets; r, the radius from the center of rotation;

and A, the thickness of the conductor.

The geometric design parameter, k, can be written as

2
k = nA pa r M (3.3)
p eff

where Ap is the cross-sectional area of the magnetic pole face normal to

A detailed analysis of the eddy brake can be found in Appendix B.

 31

the direction of the applied field; n, the number of poles; and M, an

empirical field fringing parameter approximately 1.2 to 1.7 (M~Ag/Ap where

Ag is area of the applied field on the drum, per pole).

The simplest design is one in which the moving conductor serves as a

heat sink for the resistive dissipation. It is uncooled during the test

with its temperature rising as power is absorbed. The principal design

problem arises from the high rotational speed required by the turbine

compared to other applications more typical of eddy brakes. This forces

the conductor radius (r) and thickness (A) to be small (Eq. 3.2) in order

to keep the critical speed high, which in turn reduces the conductor volume

and thus the amount of mass available as a heat sink.

Two design solutions were adopted which enhanced the viability of the

eddy brake concept. The first was to use a drum rather than a disk config-

uration. With the drum geometry, the conductor moves at a uniform linear

velocity past the magnet poles, while for the disk, the velocity varies

with radius. The drum thus reduces the peak velocity for a given average

velocity and power absorption. Another substantial advantage is that the

analysis of the drum configuration is considerably simpler than that for

the disk with a concomitant reduction in design uncertainty.

The second design decision was to optimize the choice of drum material.

Commonly, stainless steels are used in eddy brakes because of their high

resistance, high strength at temperature, and ready availability. A com-

parison with aerospace nickel alloys in Table 3.1 shows that a substantial

advantage can be realized in both resistivity and strength at temperature

by using the more exotic materials. Inconel 718 was chosen for the drum

based on the availability of a suitable forging.

 32

TABLE 3.1
COMPARISON OF EDDY BRAKE DRUM MATERIALS

Resistivity (a) Yield Strength (0.2%)

Alloy 10 b. (Qe cm) 20 0 C 550 0 C (ksi)

314 S.S. 76.8 50 30

321 S.S. 71.8 30 22
347 S.S. 72.8 35 24
A-286 91.0 105 88
Inconel 625 129.0 71 59
Inconel 718 124.0 172 154

A cantilever drum design was adopted for reasons of simplicity. This

imposes a limit on the drum length to diameter ratio due to structural

dynamics stability considerations. (The drum is mechanically excited at

its rotational frequency times the number of magnet poles.) The drum

structural dynamics were analysed by Rolls-Royce Inc. and the 15 cm drum

length selected has proved free from dynamic problems. The calculated

temperature rise in the drum, ignoring conduction at the end, is 340 0 C

(600 0 F) at the design Reynolds number (3x10 6 ) for the ACE turbine

presently installed and would be twice that at the maximum facility inlet

pressure condition. The design geometry of the eddy current brake is

summarized in Table 3.2.

The eddy brake magnets are required to provide a 0.7 tesla magnetic

field across the 1.25 cm gap, rise from zero to full field strength in 50

ms, and provide a constant strength field for periods up to one second.

Brake critical speed considerations discussed above dictated 20 magnet

poles. Initially, an induction motor-like design consisting of 20 poles on

a common circular yoke (Fig. 3.12) was considered. This was rejected in

favor of a 10 U-core configuration for reasons of manufacturing cost and

uncertainty in the electro-dynamic design of the drum. The U-core

 33

TABLE 3.2
EDDY CURRENT BRAKE CONFIGURATION SUMARY

Nominal Operating Point----ACE Turbine design point at scaled test conditions

Power P = 1,078,000 watts
Speed N = 6190 RPM

Machine Configuration------Cylindrical electrically conducting "loss" drum

attached to main shaft, excited by circumferential
array of alternating d.c. electromagnets.

Loss Drum

Material Inconel 718

Physical properties
Magnetic permeability II = 4T x 10-7 H/m
Electrical conductivity cx = 0.801 x 106 (S-_M)~
Density p = 8.19 g/cc
Specific heat Cp
C, = 427 J/kg-K
Configuration
Mean radius r = 0.1619 m (1.00 in)
Thickness A = 6.35 x 10-3 m (0.25 in)
Axial length (min. active) w = 0.1524 m (6.0 in)

Excitation Magnets

Cores
Material Grade M-6 transformer stock and ingot iron
Saturation limit Approx. 20,000 Gauss
Number of poles n = 20
Pole width (circumferential) 2a = 0.0254 m (1.00 in)
Pole length (axial) 2b = 0.1524 m (6.00 in)
Pole pitch T = 0.0509 m (2.003 in)
Mechanical airgap 1 = 0.0127 m (0.50 in)
Coils
Turns N = 444 turns per coil #23AWG magnet wire
Resistance Rc = -12 ohms per coil
Inductance Lc =
-0.12 henrys per coil
Power dissipation Pc = 2700 watts per coil @ 15A
excitation

Machine Characteristics

Magnet time constant Tc =

0.010 sec
Drum effective conductivity Geff = 0.721 x 10- 6 (Q-m)~l
Drum axial resistance RD = 3.27 x 10-5 S
Induced current ieg =
175,000 amps @ 1 MW dissipation
(equivalent axial)
Drum heating AT = 290 0 C per 106 joules absorbed
 34

arrangement permits ready adjustment of the magnet gap, g, which could be

used to change the brake critical speed, wo, should the initial

calculation have proved in error. However, the manufacturing costs of the

U-core were not substantially less than the induction motor design would

have been; the brake dynamics closely followed prediction so that magnetic

gap adjustment was not required, and the U-core design had a flux leakage

problem not initially appreciated. Thus, in retrospect, the induction

motor design would now be the arrangement of choice.

The cores are laminated from 0.38 mm thick transformer steel (M6) in

order to reduce the rise time of the magnetic field. A magnetic return

path inside the eddy drum is laminated from the same material.

The magnet coils were initially designed with 444 turns of 20 AWG

copper wire on a 4.5 cm high core. The magnets in this configuration did

not behave as designed, however, and produced a field across the drum only

75% of that required for full scale facility operation. The measured

torque versus magnet current characteristic indicated that the magnet cores

were saturating at relatively low gap field strength. Re-analysis of the

magnet design revealed a serious field leakage problem between the arms of

the U-cores. Basically, the arms of each core and its neighbors form

additional gaps in parallel with the gap across the drum. Since these

magnetic gaps behave much as parallel resistors, the field through the drum

gap is considerably reduced, by a factor of three in this case. A rela-

tively simple solution was adopted which consisted of (a) the fabrication

of new coils wound from smaller wire (23 AWG) to reduce the coil length to

2 cm, (b) positioning the coils at the tips of the magnet arms, and (c)

inserting an ingot iron filler in the U-shaped space between the tops of

the coils and the top of the U-core. Figure 3.13 shows both the old and
 35

new configurations. The net effect of modification is to reduce the

leakage by 75%. Reduction of the coil, wire diameter increases the power

dissipated in the coils to 60 KW and therefore increases the coil

temperature rise to approximately 25 0 C in one second.

The magnet field must be switched on rapidly so that the eddy brake

turn on time is comparable to the 50 ms opening time of the main valve, and

must be turned off within one second to prevent overheating and coil burnup.

The power source for the magnets is a 250 hp d.c. motor generator set rated

at 250V, 600A continuous service. A simple switching circuit (Fig. 3.14)

is used to turn on the current. A water cooled 0.4 ohm ballast resistor is

used to establish the operating point on the d.c. generator load line. A

vacuum contactor rated at 50 kV, 150A continuous service is then used to

switch the magnet coils in parallel with the ballast resistor. Thus,

during the brake operating period, 800 amps at 200V d.c. (at design Reynolds

number) is drawn from the generator (which has sufficient inertia to

provide this level of current for one second). The magnets are de-energized

by switching off the shunt field excitation of the d.c. generator and

allowing the current to decay. This prevents the high voltage arcing

problems associated with the opening of a large inductive circuit. The

generator is protected by both fuses and a fast acting d.c. circuit breaker.

The magnet coils are also fused with a 2 to 4 second time constant. One-

kilojoule varistors are included to protect the switch and magnet coils.

Isolation amplifiers are used to protect the data acquisition system

channels monitoring the brake currents and voltages from high voltage

transients.

The ballast resistor consists of two 2-meter lengths of 0.476 cm (3/16

inch) diameter 0.04 cm (0.016 inch) wall thickness type 316 stainless steel
 36

tubing. Forty liters/min of water are pumped through the tubing at 12 atm

pressure by a 3 hp gear pump. At this flow rate and pressure, the tube

wall heat transfer coefficient, 20 kW/m 2 (750,000 BTU/ft 2 hr), is

sufficient for the resistor to absorb 200 KW. Interlock circuitry shuts

off the generator field should the resistor cooling flow be lost.

The performance of the power system is illustrated in Fig. 3.15, which

shows the total current drawn from the generator, the current through the

magnet coils, and the voltage across the magnet coils as a function of

time. The magnets are switched on at time zero and the generator field

switched off at 1.1 sec. The small droop in coil current (and thus magne-

tic field strenth) is due to the overloading of the generator. This is

advantageous because it extends the turbine constant corrected operating

speed period since the fraction of the brake power absorption attributable

to stored rotational energy (moment of inertia) decreases with the square

of the rotational speed of the turbine. This can be inferred from Fig. 2.2.

The performance of the eddy current brake was measured by energizing

the magnets while the turbine was rotating in vacuum. Thus, the brake is

absorbing only the stored rotational energy and the torque can be calcula-

ted from the change in rotational speed. Tests were performed with and

without the turbine disk-blade assembly mounted to the shaft in order to

provide two levels of stored energy (moments of inertia). The moment of

inertia was measured using a torsion spring assembly and the absolute

accuracy of the brake power absorption is constrained by the accuracy of

the moment of inertia measurement.

A measured power versus speed curve is shown in Fig. 3.16, from which

a critical speed of wo = 770 radians/sec can be inferred.

 37

3.4 Tanks and Auxiliary Systems

Although physically large, the dump and supply tanks are relatively

minor facility cost items. The supply tank volume is set by the require-

ment for a small inlet total temperature drop over the test time, while the

dump tank volume controls the time to unchoking of the turbine outlet

throttle and thus the run time at constant corrected flow. The volumes are

1OM 3 (364 ft3 ) and 16m 3 (570 ft 3 ) respectively. The supply tank is

double-walled with a 1.6 cm (5/8 inch) gap between walls for oil circula-

tion. The tank is trunnion-mounted on a wheeled carriage to eliminate

thermal growth problems. The tank is pressure rated at 10 atm.

The oil system is based on a commercial unit consisting of an 80 KW

electric heater, 212 1/m (55 gpm) pump, expansion tank, and controls, Fig.

3.17. The joints in the system are welded and connections to the supply

tank and valve are by way of flexible metal hose. The heat transfer fluid

is a commercial oil (Dowtherm G) rated to 288 0 C (550 0 F). The oil can be

valved to a pressurized water-cooled heat exchanger to cool down the

facility after a test, if required for disassembly.

The tanks are evacuated with a 255 m 3 /hr (150 cfm), mechanical

vacuum pump through appropriate valves and piping. The dump tank and test

section can be evacuated to 20 4m of Hg. A water-cooled heat exchanger in

the pump inlet line allows the tunnel to be pumped down immediately after a

test to facilitate rapid turnaround.

The argon and freon test gas is metered and mixed as the supply tank

is filled. A 15 KW electric heater provides the power to vaporize the

freon and warm the argon which is supplied from a tank farm located else-

where in the building, Fig. 3.18.

Power for the rotor drive is provided by a 12 KW motor generator

 38

welding set. A preset digital tachometer initiates the start of the test

as the commanded rotor speed is reached.

The auxiliary facility equipment is sized for a 2-3 hour tunnel setup

time. The actual cycle time is governed more by instrumentation respotting

and data review times than by mechanical considerations.

A 15 cm (6 inch) diameter enclosed jet tunnel has also been constructed

using a loop of the 10 cm (4 inch) vacuum system piping and ball valves,

Fig. 3.19. This forms a bypass circuit around the test section and is

designed to be used for instrument calibration. A choked orifice in line

with the heated supply tank establishes the mass flow while the dump tank

volume limits the run time.

Because the first build of the facility is with uncooled hardware,

both for reasons of simplicity and to furnish timely comparisons for the

CFD calculations, construction of the coolant injection system has been

deferred.
 39

4.0 DATA ACQUISITION SYSTEM

The specification of the data acquisition system (DAS) is principally

determined by the requirements for detailed examination of the time

resolved heat transfer and fluid flow in the high speed turbine. The 6 kHz

blade passing frequency and relatively thin NGV wakes (1/5 to 1/10 of the

blade spacing) dictates an analogue bandwidth of 30 to 100 KHz for each

channel. The need for simultaneous heat transfer and pressure measurement

along the airfoil surfaces requires a large number of channels, 25 to 100.

Thus, 5 to 20 million samples per second are required for the 0.1 to 0.4

second test time (considering sampling theorem restrictions), thus

producing 1 to 8 million total samples.

The total number of samples that need be taken during a test can be

reduced by noting that the data acquisition system need not run at the

highest data rate during all parts of the blowdown. Information during the

initial startup transient and during post-test flow sloshing is of use

principally for tracking transducer drifts, and thus can be taken at rela-

tively modest data rates. The variable data rate requirement can be met

with a programmable multispeed clock.

The commercially available data acquisition system which comes closest

to meeting these requirements is a CAMAC-based transient acquisition system

with local internal memory for each 4-channel cluster. The CAMAC system

met most of the blowdown tunnel system requirements but was weak in several

important areas. The first was the 10-bit resolution, but 8-bit accuracy

of the units is marginal for the heat transfer gauges, not because 1 part

in 256 is insufficient heat flux resolution, but because of the need to

track transducer signals over a wide pretest to post-test range. The

second weakness was the maximum of 32,000 samples of memory per channel.
 40

The third was the very high memory cost.

As an alternative to the commercial system, a custom designed DAS was

considered. This was attractive since the system could be tailored to this

application, and it was affordable because a large high-speed memory was

available at no increment cost to the experimental portion of the program.

This is the random access memory (RAM) required by the companion computa-

tional fluid dynamic (CFD) effort. The preliminary design of the system

was done at MIT while the final design and construction was performed by an

outside vendor. Not including the cost of the memory, the system is 40%

less expensive than the commercial system in its initial configuration and

would be 80% less in a fully expanded version.

The DAS is designed for 100 high speed, 12-bit channels with 200 Klz

per channel maximum sampling rate. Up to 96 lower speed channels, organized

into 8 groups of 16, are multiplexed onto 8 of the high speed channels.

The high speed channels are used for the heat transfer gauges and high

frequency pressure transducers which record the blade to blade flow, while

the low speed channels are used for overall facility monitoring, pitot

rakes, etc. The sampling rate is controlled by four programmable multispeed

clocks (20 Hz to 200 KHz) which are individually set with a sampling rate

and number of samples prior to a test. The data is stored in 32 M byte

solid state (RAM) memory for post-test readout to a host computer. The

present configuration consists of 45 high speed channels, four 16-channel

multiplexers, and a 64-bit wide digital input stream.

The data acquisition system is organized as illustrated in Fig. 4.1.

The analogue front end, digital sequencer, and control microprocessor is

located adjacent to the blowdown facility. The digitized data is trans-

mitted as parallel 32-bit words to the computer room 200 feet distant. A
 41

first in first out (FIFO) buffer smooths the data flow into the semicon-

ductor memory (to allow for memory refresh). The maximum digital data rate

is 8.4 M words/sec. or 34 ri bytes/sec. The memory is interfaced to a

Perkin-Elmer 3242 computer which, along with its attached array processor,

is dedicated to the companion CFD effort, and to a Digital Equipment Corp.

PDP 11/70 computer, which is used for the data reduction. The DAS is

locally controlled by a Rockwell AIM-65 (6502 pP) microcomputer, which is

used to preset the test conditions. The AIM-65 is connected to the

PDP-11/70 by an RS-232C serial data link, Fig. 4.2. The AIM-65 program

resides in read only memory (ROM) or can be downline loaded from the host.

The overall test sequence is controlled by the PDP-11/70, which

instructs the AIM-65 to set the sampling rates and burst lengths and then

requests the exclusive use of the 32 M byte memory from the P-E. The DAS

can then be armed and placed under the control of the digital sequencer

used to operate the tunnel. After a test is complete, the data is read

from the memory by the PDP-11/70 for storage on disk and later analysis.

A typical time sequence would run the high speed channels at 20

kHz/ch from the start of test (t = 0) to the start of the matched test time

(t = 250 ms), at 200 kHz during the test time (t = 250 ms to t = 550 ms),

at 20 kHz to the first tank "slosh" (t = 2 sec), and at 20 Hz for the next

60 seconds as the entire facility comes to equilibrium.

As the data is stored in the PDP-ll/70, it is organized into a form

compatible with the data-based management system developed for the Blowdown

Compressor Facility. The organization has been modified to account for the

much larger size of the data stream from the blowdown turbine.
 42

4.1 Analogue Front End Characteristics

The 12-bit high speed channels are organized five to a printed circuit

card. Each card also contains two 1-bit digital input channels and two

party bits. The card output is arranged into two 32-bit words, which are

sequentially read onto the 32-bit data bus (Motorola VMEBUS) for trans-

mission to the memory.

Each high speed channel consists of a band limited input pre-amplifier,

sample and hold amplifier, and 12-bit 200 kHz analogue to digital converter.

The input pre-amplifier span and offset can be changed by jumpers on the

P.C. card to 0.5, 5V, 0-1V, 0-10V. Channel specifications are listed in

Table 4.1.

Signals not requiring the full bandwidth of the high speed channel are

input to one of the eight 16-channel, low speed multiplexers (MUX). Each

of the 16 multiplexer channels are sequentially switched to a high speed

channel input. If all 16 channels of a multiplexer are not required, the

microprocessor may select a smaller number (1 to 16) during the pretest

setup. Each multiplexer can be individually set. The number of high speed

channel groups can be preset at the same time.

4.2 Channel Setup and Checkout

Effective operation of a short duration test facility requires that

each input signal be checked before each test. With up to 216 separate

channels, conventional plugging and unplugging of cables for connection to

oscilloscopes or voltmeters can be both tedious and likely to induce

errors. To facilitate pretest transducer checkout, the DAS was equipped

with a numerical display which can function as a digital voltmeter and an

analogue output (12-bit D/A converter) for connection to an oscilloscope.

 43

Thus, using the input keyboard, individual high speed and multiplexer

channels can be sequentially called up for pretest checkout without the

need to disturb the cabling.

The entire system can also be run by the host computer (PDP-11/70) as

a data logger, with the output of as many as 40 channels displayed on the

screen simultaneously, updated at a rate of about 3 Hz. This data can also

be stored for later analysis. This mode is commonly used for transducer

checkout and calibration.

TABLE 4.1
DATA ACQUISITION SYSTEM SPECIFICATIONS

Input Impedance 1 M2
Input Range 0.5V, 5V, 0-1V, 0-10V
Resolution 12 bits
Linearity 0.012%
Absolute Accuracy 0.1%
Aperture Uncertainty 100 ps
Max Sampling Rate 200 kHz/channel
Max No. of Channels:
High Speed 100
Low Speed 96
Max Number of Samples 20,000,000
 44

5.0 FLOW INSTRUMENTATION

Flow instrumentation for the Blowdown Turbine closely resembles that

used on the MIT Blowdown Compressor Facility. It consists of time resolved

instrumentation including: traversing probes to measure total and static

pressures and air angles, wall static transducers on the outer casing wall

and NGV and rotor blade surfaces, and a total temperature-total pressure

probe. Low frequency response pitot instrumentation can be used where the

flow is quasi-steady, as in the annulus upstream of the NCV's.

5.1 High Frequency Response Measurement

There are four traversing probes which are initially intended for

traversing behind the rotor or the NGV's with the rotor removed. The

four-way probe (Fig. 5.1) consists of four semiconductor strain gauge

pressure transducers mounted around a 3 mm diameter cylinder. The trans-

ducers are arranged such that the probe can simultaneously measure total

and static pressure and azimuthal and radial angles ( 200 ). Data reduction

is done with the aid of calibrations performed as a function of Mach

number. (These calibrations have only been carried out to Mach 0.9,

however, and will be extended for turbine operation.) The probe is water

cooled (or heated) to increase thermal stability. Frequency response is

determined by the cylinder size and rolls off above 12 kHz, which is

probably marginal for the rotor outflow. The four-way probe is the only

instrument small enough to traverse between the NGV's and the rotor,

although interference and blockage effects must be determined.

Total pressure can be measured independently using a cobra head impact

probe with a transducer mounted one diameter in from the conical probe

entrance (Fig. 5.2). This probe is smaller in diameter (2 mm) and thus has
 45

higher frequency response than the 4-way probe. It has also proved to have

better d.c. stability. Another version of this probe is under construction,

incorporating a smaller transducer so that the probe diameter is only 1.4 mm.

This class of transducer has poor stability, however, so that only a.c.

information will be available.

Rotor outflow total temperature will be surveyed using a dual wire

aspirating probe (Fig. 5.3), which can simultaneously measure total temp-

erature and pressure with a frequency response of d.c. to 20 kHz. This

probe is under development for a compressor loss study. The probe's d.c.

accuracy is 1%, while its resolution is 0.3%. We are considering attempting

to manufacture a 1/2 scale model of the probe (1.5 mm diameter) for use in

the blowdown turbine.

Wall static pressures will be measured on the casing above the rotor

and on the NGV and rotor blades. At present, four transducers are being

flush mounted on the NGV suction side toward the rear of the blade, at

locations previously instrumented at Rolls-Royce Limited. These will serve

both as a proof test and a measure of the unsteady rotor influence on the

NGV loading. One transducer is to be mounted on the rotor blade close to

the leading edge. This is primarily a test article (although a transducer

has been operated on a compressor blade at similar stress levels with

satisfactory results). The transducers on the blades must be permanently

mounted. Because of their high cost ($1000 per mounted transducer), a

cautious approach has been adopted.

Drift and stability are always an important consideration when using

high frequency response instrumentation. Development of semiconductor

strain gauge transducers over the last 15 years has brought the performance

of the best units to a level equal to that of typical conventional trans-

 46

ducers (0.04%/*C). Conventional scanning valve pressure recording systems

rely as much on continuous on-line calibrations as they do on transducer

stability for measurement absolute accuracy. In the blowdown turbine, a

similar scheme can be used for -traversing probes. The probes reside in a

cavity in the tunnel wall, along with a stable reference transducer, so

that a reference pressure is known immediately preceding and following the

traverse. Thus, it is only the drift during the short traverse time (15 to

200 ms) which can influence the measurements. Also, since the transducers

are in vacuum immediately preceding the test and at a stable pressure a

few seconds following the test, transducer sensitivity can be absolutely

established. The net absolute accuracy of the measurement is about 1% for

the more stable probes (total pressure). In general, the pressure trans-

ducers agree to within 0.5% after the test. The static pressure transducers

to be mounted on the blades cannot be dynamically calibrated by this tech-

nique and their absolute accuracy has yet to be established.

5.2 Low Frequency Response Measurement

The steady test time of the facility is long enough to permit the use

of carefully designed pneumatic instrumentation. This is to be used either

for comparison with similar measurements taken at Rolls-Royce Limited or as

a lower cost scheme for measurements in regions of essentially steady flow.

The comparison studies use conventional Kiel probes four chords down-

stream of the rotor exit, duplicating measurements taken at Rolls-Royce

Limited. These will then be compared with the high frequency total

pressure probe results.

Static taps will be used around the NGV chord to record "tube averaged"

airfoil static pressure distributions. The advantage of the pneumatic over

 47

the high frequency measurements is simply one of cost and time. A dense

array of pressure taps (20 per blade) can be installed, but all need not be

connected to transducers during each run since the external transducers can

be moved. Faulty transducers can be readily replaced. Also, the trans-

ducer is in a more benign environment than when surface mounted. The

magnitude of the errors induced by the "tube" averaging will be inferred

from comparison with the high frequency flush mounted transducers at

similar chordwise locations.

5.3 Traversing Mechanisms

Two traversers have been designed to permit survey of the rotor outflow

with the high frequency response probes. One is for use in the stationary

frame which mounts to an instrumentation access window, and the other is

designed to be mounted on the rotating assembly directly behind the rotor.

The stationary traverser is designed to make radial traverses at four

circumferential locations in 100 ms (see Fig. 5.4). The radial motion is

driven by a stepping motor-cam system. Radial position versus time is

linearized electronically in the motor drive circuitry. The circumferen-

tial travel, powered by a hydraulic/pneumatic blowdown actuator, is

designed to move the unit at constant velocity along a tangent to the test

section outer case. (A linear motor drive was considered for its finer

speed vs. position control, but rejected due to high cost.) The distance

covered is 7 cm, allowing the outflow from 2-1/2 NGV passages to be

surveyed. The radial and azimuthal probe positions are monitored by linear

differential variable transformer (LDVT) transducers.

The rotating traverser is still in design. A typical configuration is

illustrated in Fig. 5.5. The probe and a diametrically opposed dummy

 48

balance probe traverse a single radially outward scan powered by the centri-

fugal load. The probes are mounted on cross-connected hydraulic pistons to

insure balance and to control the scan velocity. The probe release is

activated through a swash plate mechanism from the stationary frame.

Construction of the rotating traverser has been deferred due to overall

facility workload.
 49

6.0 HEAT TRANSFER MEASUREMENT

One of the major aspects to be considered in the design of a high

pressure turbine is the heat transferred from the freestream gas to the

profile surface. Poor prediction of the heat load, resulting from a lack

of understanding, can result in reduced profile lives and the use of

excessive amounts of external cooling air. The objective of the MIT

blowdown turbine heat transfer measurements is to improve the understanding

of the effects in a turbine which lead to increased heat loads. This will

allow the design procedures to be improved, and hence reduce the heat load,

leading to longer turbine lives and increased efficiency. Various aspects

of the flow in a turbine which affect the surface heat load are tabulated

below:

ASPECT LOW FREQ. EFFECT? HIGH FREQ. EFFECT?

"Mean" flowfield
+

Secondary flow + ?
Rotation +?
Overtip leakage + ?
Transition +
+

Wake impingement
+

Periodic inlet flow changes -+

Film cooling +
+

The total heat load in a turbine therefore consists of both low and

high frequency effects. The technique used to measure the surface heat flux

in the MIT rig should therefore have sufficient bandwidth to observe both

the "mean" and varying heat flux.

6.1 Methods of Determining the Surface Heat Flux

Four methods were considered for the measurement of surface heat

fluxes. These were:

1) Placing temperature sensors on the profile surface and solving the 3-D
 50

transient profile heat conduction equation to determine the local heat

flux.

2) Using calorimeter gauges inserted into the profile surface.

3) Manufacturing profiles (or sections of profiles) from an insulator and

using 1-D semi-infinite theory to determine the surface heat flux.

4) Direct measurement of the surface heat flux from temperature difference

measurements across an insulator applied to the profile surfaces.

Of these four options, the last one was selected as the best. The

disadvantages of the other three were:

a) This requires a significant number of temperature sensors over the

whole of the profile to obtain sufficient resolution in inferring the

local heat flux. Also, this technique requires assumptions to be made

about the heat loss through the root and tip of the profile. These

assumptions may not significantly influence the "mean" section inferred

heat transfer rates, but may introduce errors at the root and tip

sections. This technique has been used by other investigators and

results show significant sensitivity of the inferred heat flux to small

errors in the measured surface temperature. Also, a 3-dimensional

inverse heat conduction code is required for the data analysis.

b) Calorimeter gauges can be used to determine the mean heat flux by

monitoring the temperature rise of the calorimeter slug. If high fre-

quency data is needed, the slug has to be made so that it is

semi-infinite at these frequencies, and a 1-D analysis is used. The

problem with this type of gauge is that the temperature rise of the

slug differs from that of the surrounding material, hence introducing a

temperature discontinuity. This would significantly affect the boundary

layer profile development at the slug and hence result in an inferred

 51

heat transfer rate, which is not the same as it would be if the slug

were not present.

c) This procedure removes thermal discontinuities along the blade surface,

but requires extensive manufacturing to produce either completely new

blades or to place inserts in the blade surface. The rotor blade

stress levels make this a challenging problem.

6.2 Direct Measurement of Heat Flux

This method was chosen to obtain the heat flux on the profile surfaces.

This uses two temperature sensors on either side of an insulating substrate

applied to the profile surface (as shown in Fig. 6.1). A measurement of

the temperature drop across the insulator enables the surface heat flux to

be calculated directly. The direct proportionality of heat flux and insul-

ator temperature drop is, however, only valid below a certain frequency,

this frequency reducing as the insulator thickness increases. An analysis

of the frequency range over which this assumption is valid will be detailed

later in this chapter.

With this type of gauge, the insulating substrate can be applied over

the whole of the profile surface, thereby eliminating temperature discon-

tinuities.

The temperature sensors used are thin film resistance thermometers,

and these will be discussed later. Two types of heat flux gauges are

currently being investigated. These both use the same insulator material,

(a polyimide, Kapton) but differ in the insulator thickness. One type is

referred to as a bonded gauge, where the insulator is a sheet of order 25

m (1 mil*) thick. Thin film temperature sensors are manufactured on this

*1 mil = 1/1000 inch.

 52

sheet before it is bonded to the profile surface using polyimide glue of

thickness 2.5 to 12.5 pm (0.1 to 0.5 mils).

The other is referred to as a high output gauge where the insulator is

of the order 0.5 pm thick. In this case, the lower surface thin film is

directly deposited on an electrical insulation coating, such as metal oxide,

on the airfoil surface. The profile is then dipped in polyimide and the

second thin film deposited over the insulator.

Since the bonded gauges are fabricated independently and then adhes-

ively bonded to the airfoils, they are considerably easier and less

expensive to fabricate than the high output gauges which must be sputtered

or deposited directly on the blade surfaces. The difference between these

two types of gauges is the frequency range over which the temperature

difference (Tj - T 2 ) can be used directly to infer the surface heat

transfer rate. The thicker bonded gauges will be shown to have direct

readout of heat flux up to 20 Hz, while the very thin gauges have direct

readout to 35 kHz. It will be shown in Section 6.3.2, however, that high

frequency heat transfer rates can be inferred from the bonded gauges. The

theory for the two types of gauges will now be discussed.

6.3 Bonded Gauges

6.3.1 Low Frequencies

As mentioned previously, the insulator for these gauges is of the

order 25 pm (1 mil) thick. At very low frequencies, the surface heat

transfer rate can be inferred directly from the measured temperature drop:

Q = - kdT/dx (6.1)

where:

is the surface heat transfer rate (per unit area)

 53

k is the conductivity of the insulator (Kapton)

T is the temperature at a depth x from the upper surface

of the insulator.

At low frequencies, this can be rearranged as:

Q = -k(T - T 2)/d (6.2)

where T, and T2 are the temperatures of the upper and lower surface of

the insulator respectively, and d is the thickness of the insulator.

The frequency range over which this assumption is valid has been

estimated using the analysis presented later in Section 6.4.1.2. In this

analysis, a sinusoidal heat transfer rate variation is applied on the top

surface of the insulator. The profile surface temperature is assumed

constant and the insulator is assumed bonded to the profile surface by an

adhesive of known thickness with the same thermal properties as the

insulator. The assumption of equal thermal properties may be reasonable

for the bonded gauges since the Kapton and adhesive used are both

polyimides.

This analysis can then be used to compare the inferred heat transfer

rate (obtained from the temperature drop across the insulator), to the

magnitude of the actual heat transfer rate.

The results for this analysis at low frequencies for a Kapton insul-

ator are presented in Fig. 6.2. These are two cases for 0 and 12.5 pm (0.5

mil) adhesive thickness. It is readily apparent from this figure that the

low end frequency response is dependent quite strongly on the thickness of

the adhesive. For the case with zero adhesive thickness, the assumed heat

transfer rate is greater than 90% of the actual value up to frequencies of

30 Hz. This frequency drops to 15 Hz when 12.5 pm adhesive is used.

 54

As stated previously, these predictions assume that the adhesive has

the same thermal properties as the insulator. If this is not true, then

the effect of the adhesive may be different. The low end frequency

response can therefore only really be obtained from calibration of the

gauges.

The temperature drop across the insulator can be estimated. At the

turbine design Reynolds number, typical heat transfer rates (for air)

around the blade surface are given by (at a 387 0 K gas temperature, with

the profile at 288 0 K):

Leading edge: O = 335 kW/m 2

Mean value: = 114 kW/m 2

Lowest value: 6 = 65 kW/m 2

The test gas used in the NIT turbine rig is argon-freon, and this has

different properties than air. This mixture has the same Prandtl number,

and so the expected heat transfer rates with argon/freon are (assuming the

same Nusselt number and gas to wall temperature difference):

Leading edge: C= 215 kW/m2

Mean value: C= 73 kW/m 2

Lowest value: C= 42 kW/m 2

The insulator has a thickness of 25 pm (1 mil), and an expected

thermal conductivity of 0.1442 W/m*K. If the adhesive is assumed to be

half the insulator thickness and to have the same thermal properties, this

results in temperature drops across the insulator and gauge of:

Insulator Gauge

0 0
Leading edge: Ti - T2 = 24 K 36 K

0
Mean value: T, - T2 = 11 K 170 K

Lowest value: Ti - T2 = 70 K 10.5 0 K

 55

assuming that the profile surface temperature and gas temperature are 288 0 K

and 387 0 K respectively. The temperature drop across the insulator at the

design Reynolds number therefore ranges from approximately 10 to 30 0 K.

The thin film temperature sensors used with this type of gauge,

therefore, do not need a very high temperature sensitivity to produce an

acceptable output signal.

Exact scaling of the real turbine heat flows requires that the ratio

between the gas and the upper surface of the insulator temperatures be the

same as the gas to wall temperature in the engine. The MIT rig is set up

so that the initial gas and profile temperatures satisfy this ratio. The

heat transferred from the gas to the insulator during the run results in a

temperature drop across the insulator, which reduces the effective gas to

insulator temperature ratio, and hence results in non-exact scaling of the

engine flow conditions. The most probable place where this reduced temper-

ature ratio would have the most effect on correct simulation of engine

Nusselt numbers is expected to be where the profile boundary layer is

transitional. The very thin gauges, however, have a significantly smaller

insulator temperature drop, and hence will have the correct gas to insula-

tor temperature ratio. It is proposed to compare the very thin and bonded

gauge outputs to see how accurately the bonded gauges simulate engine

conditions. It can also be seen from the estimated insulator temperature

differences that these are strongly dependent on the local heat transfer

rate. Since the heat transfer rate varies considerably around a profile,

the temperature of the upper surface of the insulator will also vary

considerably. The surface temperature of engine profiles is, however,

almost isothermal (except at the leading and trailing edges where the

temperature is slightly hotter to avoid transient thermal stresses). The

 56

bonded heat transfer gauges therefore will not measure an isothermal heat

transfer rate. If the data is to be used for comparisons with predictions,

however, the non-isothermal nature of the wall temperature would not be

important.

6.3.2 High Frequencies

If a high frequency signal is present at the top surface of the insul-

ator, the signal is attenuated through the insulator. For an insulator of

a given thickness, there is a certain frequency above which no surface heat

flux information reaches the lower surface. The insulator can then be

considered semi-infinite.

The one-dimensional heat conduction equation is given by:

2 2
ST/ax = (pc/k)aT/3t (6.3)

where p is the substrate density, c the substrate specific heat, and k the

substrate conductivity. For a sinusoidal input:

T = Toexp(jwt) (6.4)

where To is time-independent, but can be distance-dependent, we have:

2 2
d T /dx = j(pcw/k)T (6.5)

or:

j 0/dx = jKT (6.6)

The solution to this equation is:

T0 = A exp-/jix + B ex C (6.7)

If we consider the substrate semi-infinite we have:

B = 0
 57

This gives a temperature time variation, at a depth x into the substrate of:

T = A(exp(-/(K/2)x))cos(t-/(K/2)x) (6.8)

The temperature attenuation into the substrate is therefore given by:

= exp(-/(K/2)x) (6.9)
T/Tx=O

and a similar expression occurs for heat flux attenuation.

The Kapton insulator used has nominal properties giving a value of

9.986 x 106 to the thermal constant pc/k. Thus, for a thickness of 25 pm

(1 mil), the temperature variation at the bottom surface is 1% of that on

the top surface when the frequency of the disturbance is 1 kHz, and less

than 0.5% at frequencies greater than 1.5 kHz. Therefore, for heat trans-

fer variations with frequencies greater than 1 kHz (rotor passing frequency

is approximately 6 kHz), the insulator can be considered semi-infinite and

a 1-D analysis can be used to obtain the varying heat fluxes.

6.3.2.1 One-Dimensional Heat Flux Analysis

The time-dependent, one-dimensional heat conduction equation was

presented in Eq. (6.3). An example of a method to determine the high-

frequency surface heat flux from the measured top surface temperature-time

variation will now be presented.

Equation (6.3) can be Laplace transformed to give:

d2
2=
dx 2 (pcs/k)T 9 (6.10)
dx

here T9, is the Laplace transformed temperature, and s is the Laplace

transform variable. Equation (6.10) can be integrated to give:

 58

T = Ae - (pcs/k)x + Be/(pcs/k)x (6.11)

Since the substrate is considered semi-infinite, the positive exponen-

tial term will be zero (B=O).

The transformed heat transfer rate at a depth, x, into the insulator

is given by:

* 3T
aT 9
Q = k

or

/p = /I Ae- Pcs/k x

or

Q =pck /I T (6.12)

The Laplace transformed temperature variation at the insulator surface

can be determined from the time history of the top surface temperature by

approximating it by the sum of a series of delayed ramps. Equation (6.12)

can then be used to determine the heat flux at the surface by inverting

this equation for each ramp, and then summing the delayed heat transfer

rates.

6.3.3 Calibration of the Insulator Thermal Properties

In order to determine the local surface heat flux from the measured

temperatures on the upper and lower surfaces of the insulator, the thermal

properties of the insulator and the insulator thickness must be known. It

is not desirable to assume that these correspond to the nominal values, so

measurements must be undertaken to determine the correct values for each

gauge.

For the direct measurement of heat flux at low frequencies, the

 59

surface heat flux is simply related to the temperature drop across the

insulator by Eq. (6.2). This equation requires a knowledge of the ratio of

insulator thermal conductivity to thickness. When high frequency heat

transfer rates are required, the density, specific heat, and conductivity

product is required (Eq. (6.12)).

Two calibration techniques can be used to obtain the required thermal

property "constants" for the direct and high frequency heat flux evaluations.

These will now be described.

6.3.3.1 Phase Shift Calibration

When a sinusoidal temperature variation is input to the top surface of

the insulator (by passing a sinusoidal current through the thin film temp-

erature sensor), there is an attenuation of the temperature variation as it

progresses further into the insulator. This is also accompanied by a phase

shift between the sinusoidal variation at a depth x, and the input varia-

tion (at x=O).

The temperature variation at a depth x into the insulator has been

presented in Eq. (6.7). For the semi-infinite case reported in Section

6.3.2, the phase shift at a depth x can be obtained from Eq. (6.8), and is

given by:

do = /pc/(2k) x

This can be re-arranged as:

S=/c /w/2 (d/k) (6.13)

for phase shift measurements across the insulator of thickness d.

Equation (6.13) therefore shows that, if the insulator is assumed

semi-infinite, the phase shift across the insulator varies linearly with
 60

the square root of the applied frequency.

The semi-infinite assumption may not be valid for all frequencies when

the bonded gauges are calibrated. Another analysis was undertaken to see

the effect of different bottom surface boundary conditions on the phase

shift across the insulator. A sensible boundary condition, if the gauges

were calibrated in air before application to the profile surface, is that

there is no heat flux lost through the rear boundary (since air is a good

insulator). In this case, the constant B in Eq. (6.7) is taken as non-zero

and a boundary condition of zero net flux is applied at x=d. This leads to

a modification of Eq. (6.13) to give:

2
do = do - tan-1 e kdsin( 2 kd) (6.14)
(6.13) l.+e cos(2kd)

The predicted variation of phase shift angle with frequency for Eqs.

(6.13) and (6.14) is presented in Fig. 6.3. These results were calculated

using the nominal insulator properties and thickness. It can be seen that

the effect of a different boundary condition is only apparent over a small

frequency range, and that the phase shift at 400 Hz is approximately 160

degrees. Of course, as mentioned previously, there is an increasing

attenuation of the temperature signal on the rear face as the frequency is

increased. This attenuation is also presented in Fig. 6.3. Although the

signal from the lower temperature sensor will drop off quite rapidly, there

should be sufficient signal over a wide frequency range to cover the linear

region.

The calibration detailed above gives the product of the two thermal

quantities needed to evaluate the low and high frequency heat transfer rates.
 61

6.3.3.2 Step Response

In this calibration, a step change in heat transfer rate will be

applied to the top surface of the insulator and the temperature variation

with time of this surface will be monitored. For this calibration, the

insulator is assumed semi-infinite, so the measurement time of the tempera-

ture must be such that this assumption remains valid.

The Laplace transformed equation relating surface heat transfer rate

to surface temperature rise is given in Eq. (6.12). For a step change in

surface heat transfer rate, the Laplace transformed heat transfer rate is

given by:

Q = Q /s
0
/

where Qo is the applied heat transfer rate per unit area. The Laplace

transformed surface temperature is therefore given by:

__ 1.5
T = /S

This results in a temperature time variation given by:

0
-

0t T = /
/ p ck

The temperature therefore rises parabolically with time. A knowledge

of the temperature time response and the applied heat flux enables the

thermal property, pck, to be found. If this calibration is performed in

vacuum, and a step in voltage is applied to the upper surface thin film,

then all the electrical power applied to this thin film sensor can be

assumed to pass into the insulator, thereby giving a value to %.

 62

6.3.4 Thin Film Temperature Sensors

Nickel thin film temperature sensors have been chosen for use with the

above bonded gauges to maximize signal output. Sets of various thickness

gauges have been manufactured as a check on the vendor processing, so that

the optimum film thickness can be chosen. These gauges have an active area

of approximately 0.2 x 1.0 mm, and have copper tags of low resistance (to

take the signals to the lead out wires) flashed with gold (see Fig. 6.4).

6.3.4.1 Thin Film Sensitivity

It is proposed to monitor the change in resistance of each thin film

by placing it in a Wheatstone bridge (see Fig. 6.5). Since the change in

film resistance is directly proportional to the change in temperature, the

output of the bridge will correspond to the change in film temperature.

The voltage that can be applied across the film is, however, limited by the

fact that the electrical heating of the film must be small in comparison to

the surface heat flux being measured. The electrical power dissipated by

the film is given by:

2
P = V /R (6.15)
f f

where Vf is the voltage applied across the film, and Rf is the film

resistance. If the width of the film is w, and its length is k, the

effective power per unit area, 41 is given by:

e = P/(wz) (6.16)

If the electrical power per unit area is the limiting factor, the

maximum voltage that can be applied across the film is given by:

V = / R /w (6.17)
max f /e
 63

For a film of given thickness, the film resistance is given by:

Rf = Kg/w (6.18)

where K is the sheet resistivity, giving:

V = /K (6.19)
max e

The maximum voltage that can be applied across the film is therefore

independent of the width of the film, and only depends on the length.

The maximum thin film sensitivity that can be obtained by placing the

thin film in a Wheatstone bridge (as shown in Fig. 6.5) can now be

calculated.

The measured voltage drop, dV, is given by:

dV = I2R2 - 1 (6.20)

and the total voltage by:

V=I (R + R + dR ) = I (R + R ) (6.21)
1 1 01 1 2 2 02

Manipulating the expressions above gives:

dV/V = ((R *R01 - R 1*R 02) + R2 *dR)/((R1 + R01 + d R)*(R2 + R 2)) (6.22)

The change in voltage between the start of the experiment and when the

sensor resistance has risen by dRl is given by:

R R dR
dV'/V = 01 2 1 (6.23)
(R +R +d )(R +R ) R
1 01 Ri 2 02 01

Assuming that dR1 << R, + R 0 1 gives:

R2R1 dR1
dV'/V = 2 01 1 (6.24)
(R1+R01)(R2+R 2)
R01

The voltage across the film is related to the voltage across the

bridge by:
 64

V /V = R 1/(R1+R 0) (6.25)

giving:

R dR
dV'/V = 2 1 .6.26)
f R +R R-*
2 02 01

If R 2 >> R 0 2 then:

dV'/V = dR /R 1 (6.27)

The resistance change of the film is related to the change in film

temperature by the expression:

dR, = R 01adT (6.28)

where a is the TCR of the film. This gives:

dV' = Vfa dT (6.29)

or, a bridge sensitivity of:

dV'/dT = Vfa (6.30)

The temperature sensitivity of the thin film gauges when placed in a

Wheatstone bridge can now be summarized. Equation (6.30) has shown that

the change in bridge output voltage for a one degree temperature rise is

given by the product of the voltage across the film and the temperature

coefficient of resistance (TCR) of the film. There is a maximum voltage,

however, that can be applied across the film in order to keep the electri-

cal power generated by the film small in comparison to the heat transfer

rate that is being measured (Eq. 6.19). Combining Eqs. (6.30) and (6.19)

leads to a thin film sensitivity given by:

dV'/dT = Z/k / a
e

where 2 is the film length, k is the film sheet resistivity, a is the film
 65

TCR, and e is the electrical power dissipated in the film. k and a are

functions only of film thickness (i.e., material properties), and e is

determined by the size of the heat transfer rate being measured, so once

the film thickness has been chosen, only changing the film length can

affect the film sensitivity.

In order to find the optimum film thickness to maximize the thin film

gauge sensitivity, a series of thin film sensors of different thicknesses

have been manufactured. These sample films were 1 mm long and 0.2 mm wide,

with copper leads flashed with gold (for oxidation protection), as shown in

Fig. 6.4. The measured sheet resistivity, TCR, and film sensitivity for

these 1 mm long films with an assumed electrical heating of 1 kW/m2 (approxi-

mately 1% of the expected mean rotor heat transfer rate) are given below:

Film Sheet
Thickness Resistivity TCR Sensitivity

Pm 2/square /*K mV/K

0.05 5.60 - 5.86 1.683 - 1.668 x 10-3 0.126 - 0.128

0.13 7.17 - 7.55 0.745 - 0.675 x 10-3 0.063 - 0.059
0.26 1.54 - 1.55 1.602 - 1.628 x 10-3 0.063 - 0.064

It can be seen from these results that a reduction in film thickness

from 0.26 pm to 0.13 pm results in a reduced TCR and increased sheet resis-

tivity. A further reduction in thickness to 0.05 pm results in anomalous

results where the sheet resistivity does not increase and TCR does not

decrease as expected. It was decided from these results that a thin film

thickness of 0.26 pm would be used for the manufacture of double-sided heat

transfer gauges suitable for use in the MIT turbine rig. The thicker gauge

was chosen because of its expected better durability.

The manufacture of double-sided heat flux gauges is being undertaken

 66

in two parts. The first part was the manufacture of double-sided gauges

with the same sensor size as the original single-sided test pieces. One

problem that has to be approached in the use of double-sided gauges is how

to connect instrument leads to the tags of the lower surface film which is,

of course, glued to the profile surface. This potential problem has been

successfully overcome by sputtering the lower surface tag around the edge

of the Kapton sheet and onto the top surface (as shown in Fig. 6.4).

The sensitivity of these double-sided gauges can, however, be improved

by serpentine meandering of the film in the 1 square mm sensor active area.

The original layout of the film and tags is also not suitable for ease of

measurement of heat flux in the MIT rig, so the thin film gauges that will

actually be used in the rig are shown in Fig. 6.6. The expected resistance

of each of these films for (0.26 4m film thickness) is 700Q, resulting in a

gauge sensitivity of 0.7 mV/K for 0.5 V applied across the film. The

expected temperature drops across the Kapton insulator on the rotor (10 to

30*K) would therefore result in 7 to 21 mV voltage differences that need to

be measured.

6.4 High Output Thin Film Gauges

These gauges have an insulator thickness on the order of 0.5 4m. Due

to the small thickness of these gauges, the attenuation of high frequency

signals from one side of the Kapton insulator to the other is small, and

the temperature difference can be used directly to infer the surface heat

transfer rate up to frequencies of approximately 40 kHz. The theory

concerning the relationship between the assumed and actual surface heat

flux will be developed, and the implications for gauge manufacture and use

will be discussed.
 67

6.4.1 Frequency Response

Two modes of analysis were used to determine the frequency response of

the insulator. In both of the cases, the rear wall temperature (T2 ) is

assumed constant. This is not unreasonable in that the profiles to be used

are aluminum and have a high conductivity. 2-D transient temperature

predictions on the rotor blade showed that the external wall temperature

rose by only a few degrees over most of the surface, and only about 10

degrees at the leading and trailing edges. Taking this temperature rise

into account will not, however, significantly influence the results presented

here. The modes analysed were:

1) Time response of the front face wall temperature to a step in heat

transfer rate.

2) Evaluation of the error involved in using Eq. (6.1) to determine the

amplitude of a steady state sinusoidal variation of surface heat

transfer rate.

If the insulator were of thickness d and conductivity k, the steady

state temperature drop across the insulator for an external heat transfer

rate would be

T - T
k
Q (6.31)
1 2

The material and its thickness should therefore be chosen to make the

thickness to conductivity ratio as large as possible. However, it is to be

expected that, the thicker the material, the longer it will take to reach a

steady state condition, and hence the lower the frequency range it can

operate over. The frequency response of the gauge will now be considered.
 68

6.4.1.1 Response to a Step Change in Heat Transfer Rate

The nomenclature used for this analysis is shown in Fig. 6.7. For the

run time of the experiment, the heat flow across the insulator can be

considered one-dimensional. The heat conduction equation is then:

2
a2T () T (6.32)
2 k at

where T is the temperature rise at depth x. Laplace transforming gives

2-
cS) T (6.33)
2 (Pk
ax

where S is the transform variable. This has a solution

x /pcS/k x
T = Ae + Be (6.34)

Applying the boundary condition that T 2 = 0 for all time gives

d
0 = Ae-/pcS/k d + Be/pcS/k

B = -Ae -2/p cS/k d

or

- -/p S/k x /pcS/k (x-2d

T = A(eVpcS/k x - e
)

(6.35)

The other boundary condition is a step change in upper surface heat trans-

fer rate at time t = 0. The transformed heat transfer rate at a distance x

into the surface is given by:

Q= -k
-k 3T
-

ax

A /= / se / x + e/pcS/k (x-2d)) (6.36)

at the surface x = 0
 69

= A /pk /s (1 + e-/p S/k 2d (6.37)

This must correspond to the imposed step change in heat transfer rate:

SAQ
i.e.,
QS

(6.38)

s3/2(1+e- p cS/k 2d)

The transformed surface temperature at the upper surface is given by:

-/vp cSA 2d
1, 1= A
-

Therefore:

A 1e-/pcS/k 2d
T = (6.39)
1 s 3/2
1+e
-/p cS/k 2d

This does not have a direct inverse transformation, but the lower exponen-

tial term can be expanded using the binomial theorem to give:

A0Q -_) cS/k 22nd A

T = +2 z (- )n
e,/p
e AQ (6.40)
1
1 s3/2 n=1 3/

Applying an inverse transformation to this gives:

T (t)
1
= 2/i + 2 nO (-1 )n 2v/t/I exp r- p_ n d
AQ n=2 k t

-
k k
i2n
j
erfc J/
k ]
t
(6.41)

where erfc is the complementary error function in the steady state

AQ = k/d T 1 s. This gives

 70

T(t) 2) 2
= +
2 (-1exp(- cnd
1 /p c/k d n=1 k t
S

(/p c/k nd)erfc(/pc/k nd//t) (6.42)

This equation therefore gives the time required for the insulator to reach

a steady state solution. This time will clearly be dependent on the factor

/p c/k d

and, as d increases, the time to reach steady state increases. Therefore,

an increase in insulator thickness results in a reduced frequency response.

However, such an increase will increase the temperature drop across the

insulator and hence improve the gauge sensitivity. There is, therefore, a

trade-off between gauge sensitivity and frequency response.

6.4.1.2 Steady State Response to a Sinusoidal Heat Transfer Rate

The nomenclature used in this analysis is shown in Fig. 6.8. Again,

we have:

2
==a
(C (6.43) a )
2 k at at
ax

since the heat transfer rate varies sinusoidally, the temperature will

also. We can therefore say:

3T
-= jwT where j =/-C
at

giving

2 jawT (6.44)
ax

This has a solution

 71

T = Ae~ + IIey ; Y = / = ,/a/2(1+j)= K(1+j)

we may also say

j(et+6 j(W t+62

A = Ae B = Be

-KX j(ut+6 -KX) KX j(Wt+62 +KX)

T = Ae e + Be e (6.45)

i.e., the temperature at a depth x into the substrate is the sum of sinu-

soids of different phase angles whose amplitude varies exponentially.

Using the assumption that T 2 = 0 for all times gives

-Kd j(W t+6 1-Kd) Kd j(wt6+6 2 +Kd)

O =Ae e + Be e (.6

Since at this point the sinusoids must be in phase, we have:

-Kd 62 + Kd
6

i.e., 62 = 6 - 2Kd

2
B = -Ae Kd
giving

K (x-2d) j( t+6 - 2 Kd+KX))

T = Ae e - e e (6.47)
)

the heat transfer rate at a depth x is given by

e
3T
ax

K(x-2d) j(wt+6 -2Kd+KX)

Q = AkK (1+j) (e KXe(Wt+61KX) e

and at x = 0

2
t+6
(eW ) - Kd j(wt+6 1-2Kd)
% =0 Akt (1+j) (e + e e (6.48)

Putting ej = cos $ + j sin

$
 72

Qx = Akr (1+j) (cos(W t+6 ) + j sin(t+6)

2 2
+ e Kdcos(W t+6 -2Kd) + je Kdsin(w t+6 1-2Kd)) (6.49)

Taking the real part of the above expression and utilizing the fact

that:

x== Q cos(Gt),

enables A and 61 to be found. Furthermore, the assumed heat transfer

rate is given by:

Qass = k/d (T 1 -T2 ) = k/d T (6.50)

This enables the ratio of the assumed to actual amplitude of the heat

transfer rate to be found. This is given by:

Q ass -2/iE
-= 1 1 -
- - 4e cos 2 (2V --
) + e
4/2B
+ 2e
2/B
(6.51
(6.51)
(1 + 2e /27cos(/2T) + e2-7)

where = /pc/k d/ and the phase lag between the inferred and actual heat

transfer rate variation is given by:

- 1 1 + 2e-'/2 sin(-/ 5 ) - e- 2,27

= tan --- (6.52)
1 - 2e -/7sin(-,/2T ) - e-2I

Equations (6.31), (6.42), (6.51) and (6.52) will therefore give the

trade-off between the temperature difference across the wall and the fre-

quency response of the gauge for different thicknesses of a given material.

6.4.2 Sensor Design Considerations

The gauge response can now be calculated using the results from the

preceding section for the insulator frequency response and Section 6.3.4
 73

for the sensor bridge behavior.

Over 50 insulator materials were surveyed for use in these gauges.

The principal criteria is that the temperature drop across the insulator be

as large as possible to maximize the signal output, i.e.,

AT d
K
Q

and that the material be compatible with thin film techniques. Three

seemed the most promising: silicon dioxide, silicon, and polyimide (Kapton).

The bulk properties used to calculate the insulator response are tabulated

below.

Material C(J/kg*K) k(W/m0 K) Density(kg/m 3

)
Silicon Dioxide 840 1.3 2200
Silicon 840 148 2330
Polyimide (Kapton) 1090 0.155 1420

Figure 6.9 illustrates the tradeoff between frequency response and

temperature difference versus insulator thickness. Clearly, the polyimide

properties are superior. For the 40 kHz target frequency response of this

program, a 0.5 pim thick layer is adequate. This is compatible with conven-

tional semiconductor spin coating techniques.

The results from the sinusoidal input analysis for 0.5 pm and 0.7 pm

of polyimide are presented in Fig. 6.10. At the blade passing frequency,

the inferred heat transfer rate is in error by only 0.03% for the 0.5 pm

polyimide.

The temperature drop across this thickness of Kapton is obviously much

less than for the bonded gauges, and estimated values are:
 74

Leading edge: T 1 -T 2 = 0.75*K

Mean value: Ti-T2 = 0.26 0 K

Lowest value: T 1 -T 2 = 0.16 0 K

for the turbine at the design Reynolds number. Thus, the insulator upper

surface temperature will be essentially the profile surface temperature.

During the course of the test, it has been predicted that the metal surface

temperature will rise by 2 to 3 degrees over most of the surface, with the

leading and trailing edge temperatures rising by, at most, 10 degrees.

These gauges will therefore not change the gas to wall temperature ratio,

and the heat transfer rates measured will essentially be the isothermal

heat transfer rates.

In order to detect temperature differences of this order, it is neces-

sary to have thin film temperature sensors with high TCR's and resistances.

The goal is to produce films with TCR's of the order 5 x 10-2, and resis-

tances of 2000 ohms. This will give a film sensitivity (in a bridge) of:

30 mV/ 0 K

resulting in film outputs the order of, or greater than those for, the

bonded gauges.

The analysis presented in this section enables us to summarize the

very thin gauge requirements as below:

e Frequency Response - requires that the thermal insulator between the

resistors be thin (0.5 pm). It also sets the maximum physical dimensions

of the sensor, hopefully no more than 0.5 mm square for active area.

The goal is 30-40 kHz frequency response.

* Sensitivity - requires that the TCR of the resistors be as high as

possible, the goal is 5 x 10-2. If a Wheatstone bridge arrangement is

used, the total resistance must also be high, 2 kilohm as a goal. Note
 75

that the sensitivity must be high because the temperature difference

0
across the insulator is small (~1 K) since it is so thin. Thus, there

is a tradeoff between gauge output and gauge frequency response.

e Bonding - The gauge (2 resistors and insulator) must be directly

fabricated on the surface since adhesive bonding is not allowed. This

is because an adhesive layer of any thickness (i.e., of order of the

thermal insulator, 1 [pm) would have just as large a temperature drop

across it, thus introducing errors in the measurement.

* Minimum Flow Disturbance - requires that the electrical connections to

the gauges be thin (25 pm or so). These also may have to be fabricated

as films. An additional requirement is that there be no abrupt

temperature discontinuities caused by the sensors. This implies that

the entire blade surface be covered by the thermal insulator.

* Coverage - We require that 10-30 of these gauges be fabricated on an

aluminum (or possibly steel) airfoil measuring typically 5 x 5 cm.

9 Reproducibility - Because each gauge will be individually calibrated,

the gauge to gauge reproducibility (resistance, TCR and insulator

thickness) need be no better than 20%.

* Alignment - It is important that the upper resistor active area be

closely aligned with that of the lower one (within 10% of the smallest

dimension) in order to reduce 2-D conductivity errors.

* Abrasion Resistance - The environment is relatively clean. Depending

upon resistor thickness, however, an overcoat (SiO or SiO 2 9 for

example) may be required for handling protection.

The requirements detailed above show that great care must be taken in

order to manufacture these very thin gauges, and this is likely to be

 76

expensive. It is proposed, therefore, to manufacture only one set of these

heat transfer gauges on a rotor, and compare the heat transfer rates from

these gauges with the bonded gauges. If the results from the bonded gauges

are acceptable, then they will be used as a relatively cheap and flexible

method for the evaluation of turbine heat transfer rates.

 77

7.0 INITIAL EXPERIMENTS

A test in the Blowdown Turbine Facility consists of two parts - a

manual preparation followed by an automatic sequence. The supply tank,

test section, and dump tank are first evacuated to less than 0.1 mm Hg.

For design point tests, the supply tank and valve are then heated. The

main valve is closed and the supply tank filled with the test gas mixture.

The turbine rotor is spun up past the chosen operating speed and the

rotor drive is then shut down. As the rotor speed slowly decays past the

operating speed, a preset digital tachometer initiates the automatic test

sequence.

An overall view of the test sequence can be seen in Fig. 7.1, which

shows the inlet (supply tank) total pressure, the turbine rotor exit total

and static pressures, the valve position, rotor mechanical speed, and the

brake current plotted as a function of time. As the valve opens, the

supply tank pressure begins to drop and the rotor exit pressures rise. The

eddy brake is switched off at approximately 1000 ms, terminating the test.

The small ripples in the supply tank pressure are acoustic waves at the

tank's longitudinal frequency. The slow rise in exit static pressure is

transducer drift. The bounce evident in the valve position trace is not of

concern since the motion does not carry the slider back past the full open

position.

The corrected weight flow and corrected speed are calculated from the

measured rotor speed and change in the supply tank pressure. Figure 7.2

shows these to be constant to better than 1% over a 400 ms period. (The

change in corrected flow reflects thermal drift in the supply tank pressure

transducer, which has since been replaced.)

Because of the high swirl in the exit flow from this stage (40 degrees),
the downstream flowfield is not totally decoupled from the back pressure

change in the dump tank. This increases the rotor back pressure earlier

than would be expected for nonswirling, choked flow in the exit throttle.

For these preliminary tests, the total pressure ratio changes 1% between

250 and 400 milliseconds (ms) from the start of the test and another 4%

between 400 and 600 ms. Deswirl vanes are being installed at the exit of

the rotor discharge duct in order to eliminate the coupling. This should

extend the constant pressure ratio test time to 600-700 ms into the test.

The average stage quantities measured at two throttle settings are

compared in Table 7.1 with a test of the same turbine in a conventional

aerodynamic rig. The agreement is quite good. The difference reflects

primarily transducer drift which can be reduced through further data

reduction and calibration.

Figure 7.3 illustrates the heat flux and static pressure distribution

as measured on the stationary rotor tip casing. The flow is seen to be

quite periodic (allowing production of the contour map without requiring

ensemble averaging).

TABLE 7.1
COMPARISON OF INITIAL TEST RESULTS

TEST IDENTIFICATION T3 T4 S.S. RIG

Throttle Position Full Open Partial N.A.

Mass Flow, In 8.85 8.89 8.95

(lbs/sec *76K/PSIA)

Corrected Speed, N//T (*K) 408 412 421

Stage Total Pressure Ratio 4.38 3.65 4.33

(Pin out)
Mid Span Total Pressure* 1.33 1.23 1.28
Tip Static Pressure

* Measured 4 chords downstream of rotor

 79

8.0 CONCLUSIONS AND SUMMARY

A transient turbine test facility has been designed, constructed, and

tested which can faithfully simulate the full range of present and future

high pressure turbine stages. The design intent has been to develop a tool

for research in unsteady turbine aerodynamics and heat transfer rather than

a facility for development testing. Nevertheless, the facility keeps the

important turbine non-dimensional parameters constant to 1% over 400 ms,

long enough for both extensive aerodynamic and heat transfer measurements.

The facility development has been relatively uneventful. Preliminary

tests show quite good agreement with data taken previously on the same

turbine at Rolls-Royce Limited.

Instrumentation development has been an integral part of the blowdown

turbine facility program. A new design heat flux gauge has been developed

specifically for the environment of this facility (essentially that of a

warm turbine rig) to permit both d.c. and a.c. heat flux measurements without

perturbing the airfoil boundary layer.

Now that the facility is operational, emphasis will shift to turbine

measurements with continuing work on instrumentation development. The

first task will be the time resolved measurement of NGV and rotor heat flux

distributions and rotor exit flowfield. Future work on the facility might

include flow visualization, implementation of blade film cooling, and

rotor-stator interaction studies.

 80

High Pressure Storage

Fly wheel
4 Reg.T
Heater

Expulsion Fly wheel

T
Bladder
Re g.I

T Compressor

Eddy Broke

cTh I

T-Turbine Stage

FIGURE 2.1: CANDIDATE FACILITY CONFIGURATIONS

(A) 150 ATM GAS STORAGE WITH FLOW HEATER
AND FLYWHEEL
(B) ROOM PRESSURE EXPULSION BLADDER AND
FLYWHEEL
(C) BLOWDOWN WITH COMPRESSOR POWER ABSORBER
(D) BLOWDOWN WITH EDDY BRAKE POWER ABSORBER
 81

6000 340

RPM

4000 - 260

-
0~
cr z
2000 180
I N

0 100
0 .2 .4 .6 .8 1.0
TIME (Sec)

FIGURE 2.2: TYPICAL PREDICTED BLOWDOWN TURBINE

MECHANICAL AND CORRECTED SPEED VARIATION
 82

80 4

PIN /OUT 0
(.i~
ft00
LU
60 3
V)
LU U)
a_
-- PIN
z
401 2 z
H m

20
0 .2 .4 .6 .8 1.0
TIME (Sec)

FIGURE 2.3: TYPICAL PREDICTED BLOWDOWN INLET PRESSURE

AND TURBINE PRESSURE RATIO HISTORY
 83

1200 Brake 1200

.o .-

Turbine
S 800 800 w
w
a 0
cr I
0 w
w 40040 <
m
m

w
0 0
0 .2 .4 .6 .8 1.0
TIME (Sec)

FIGURE 2.4: PREDICTED VARIATION IN TURBINE TOROUE

PRODUCED AND EDDY BRAKE TORQUE ABSORBED
 SECTION
TEST

TEST SECTION.
DUMP TANK
II V I1 co

SUPPLY TANK7
(21

VALVE
\ = =71

FIGURE 3.1: BLOWDOWN TURBINE FACILITY EXTERNAL VIEW

 N*-, 11- Nl- ' -, '*-, 11-11 I'll 1-1 '1-, '-~ ~ z
, '-' '*,-- '1I t,

Advanced
Military
ZFACILITY DESIGN OPERATING Engine
500 ENVELOPE
0FI01
(j
0
E 3 Takeoff* 0F 404
400 R B 211- 524
-

crQ.
w
MIT/RR/ONR
x 300 *ACE
VCE Supersonic*
Cruise
w
0 200 OE 3 Cruise
O-
w
0
100 *VCE Subsonic
0 Cruise
0

0 I
2000 I
3000 4000
COMBUSTION EXIT TEMPERATURE (*R)

FIGURE 3.2: BLOWDOWN FACILITY SIMULATION CAPABILITY

 86

Supply
Ta nk

To Turbine --

CLOSED

SL ppl y
T 3nk

Flow
[1-I

OPEN

FIGURE 3.3: MAIN ANNULAR PLUG VALVE SCHEME

 K
D M N

....
E. .

-........... n
.... .- -.

.. .. .
.

FIGURE 3.4: MAIN VALVE CONSTRUCTION DETAIL

 PT PT Pr PT
Pr
...... . . .

V
F
1~
VPD>
PD PD
P PT PT
*

dv > dv
ACCELERATE, d
dt
>0 COAST, dv
dt
0 DAMP, d
dt
(0

FIGURE 3.5: VALVE ACTUATION SEQUENCE

 Dump Tank
Chamber Y
PyIVy, Ty

Supply Tan k
Chambe r NGV Throat
P V
TN-A

r Chamber Chamber A MT
A E P,,V,, TS

A IE
Chamber F ChmrC
Chamber C

A = Area
Pil-- -Firing Volvo
h = Mass Flow Cylinder VGF VFc
P = Pressure G A
T= Temperature
V= Volume
Cylinder Supply
Chamber S 0 Reservoir
Ps Ivs I TS

GTL Test Floor

V_ IO5 ft 3

FIGURE 3.6: VALVE ANALYTICAL MODEL CONSTITUENTS

 90

I 0
0

8 160 20
Velocity

0
Stroke
6 120 0 -10

01
o Co

E 0
Pressure 03 0o in-
Ci
40 -0 10

Force

--- --- 20

0 40 80 120 160
Time (ms)

FIGURE 3.7: TYPICAL VALVE MODEL PREDICTION

 120
I I
6

I' Calculated

---- Calculated
--- Measured 80
4
\ V/

0 L

E,

(n

40

0
0 100 200
0 Time (ms)
0 100 200
Time (ms)

FIGURE 3.8: COMPARISON OF MEASURED VALVE BEHAVIOR WITH MODEL

(MEASURED PLOTS SHIFTED BY 30 MS)
 COOLING AIR
EXAUST TO
200 DEG. K 200 DG. KVACUUM TANK
80 PSIA VACUMAAN
INLET TANK SUPPLY
B D
500DEG. K 80 PSIA EDDY BRAKE
* BLEEDS
MAGNET STATOR

SLIP RING UNIT 10 HP. ELECTRIC DRIVE MOTC R

VALVE RAM
INLET VALVEAN AND
SPEED ENCODER

FIGURE 3.9: TURBINE FACILITY FLOW PATH

 DUMP TANK
COOLING AIR INLET BOUNDARY LAYER

TURBINE
EXHAUST EDDY BRAKE

FLOW PATH
TORQUE
METER

ROTATING

INSTRUMENTATION 10 HP
SLIP RINGS
DRIVE MOTOR

FIGURE 3.10: TEST SECTION DETAIL

 94

-1/2
Ta w
Tmax
Actual Be havior

"Induction Motc r " Model

T 2 (w/ w o
Taw = 2(W/
)
Tmax l+ (W/ w)
WO

SPEED

FIGURE 3.11: EDDY CURRENT BRAKE TORQUE VS.

ROTATIONAL SPEED
 Drum
Drum

ul

Coils

U CORE CONFIG I NDUCTION MOTOR CONFIG

FIGURE 3.12: CANDIDATE EDDY BRAKE MAGNET CONFIGURATIONS

 OLD DESIGN
NEW DESIGN

U CORE OLD

ADDED

NEW COILS
DRUM

RETURN IRON

FIGURE 3.13: ORIGINAL AND MODIFIED BRAKE MAGNET DETAIL

 'I

S
MAGNET
GENERATOR CO ILS (20)
250 VDC
800AMP .42 .6 n
150KW .003Hy
TOTAL
F ROM SHUNT
ROLLE SHUNT
CONT ROLLER 7A

A
I TO DATA
AQUISITI )N
SYSTEM
A

A- ISOLATION AMPLIFIERS (AD 289J)

S - VACUUM CONTACTOR (ITT RP900K)
V - VARISTOR (GE VSIIBA60)

FIGURE 3.14: BRAKE MAGNET CURRENT SWITCHING CIRCUIT

 98

800 I I
I

I
I

i
I

I
TOTAL GENERATOR CURRENT (Amps)
700

600

500

400

BRAKE COIL CURRENT (Amps)

300

200 BRAKE COIL VOLTAGE (Volts)

-

100

0 I I I I
I
0 0.2 0.4 0.6 0.8 1.0 1.2
T IME (Sec)

FIGURE 3.15: MAGNET SYSTEM ELECTRICAL PERFORMANCE

 99

1200
I I I I

1000 Full Reynolds

Design Point

800
New Magnet
Configuration

600
0
0-

400 Original Ma gnet

Configuratio n

200

0 I
0 20 40 60 80 100 120 140
Rotational Speed (Rev /Sec)

FIGURE 3.16: MEASURED BRAKE POWER ABSORPTION

 Throttle Expansion Tank
Valves IL

Main
Supply Tank Valv e
0
0

80KW
Heater

Working Fluid - 220 Gal

Dow Therm G
50 GPM Strainer
Max Temp- 500*F
100PSI 45PSI
Cit y
Cool Down Drain
Heat
Exchanger

FIGURE 3.17: FACILITY OIL HEATING SYSTEM

 Ar/R-12 Test Mixture To
Supply
I0atm (5kg/min)
------- Argon Tank

.......Liuid Freon N-

Choked
Gaseo us Freon Metering
Hot W ater Orifices
70
r SMtelw
700SLM Flow Meters 1425 SLM
20 atm
80 0 C Heated L iquid Freon Supply

100 kg
~ 23 (it _-

90*C

Freon Pressure 20 0 C

.
Regulator 5atm
Freon Boiler H-
0
Argon
Heater T 15kw
Water
Heater

Circulating
150 atm,20'*C Pump
From - - - - - - - - - --- - ---- ' ..................... K ...................1
From Liquid
High Pressure Argon Pressure
Argon Storage Freon Storage
Regulator

FIGURE 3.18: FACILITY GAS MIXING SYSTEM

 Vacuum Fast Acting In
Out Door Valve IOAtm, 500*K

E it Heat Flux . TEST SECTION

Throttle Test Specimen :100cm
Plate 15 cm Honeycomb Flow
10cm Endplates
Straightener

H--
0

Pitot Screens
Instrumentation
Access Flange (Typ)

216 cm

FIGURE 3.19: AUXILIARY INSTRUMENTATION CALIBRATION TUNNEL

 PERKIN- ELMER DATARAM MNEMRAC PDP -11/T0
3242 32MB RAM I TRECEE - PERATOR
COMPUTER MEMORY FRCIFE COMPUTER

COMPUTER ROOM-
BLOWDOWN TURBINE 70Meters
RS-232C 48 ANALOG
32 BIT
Serial INPUTS
DATA BUS
Li ne _______
35 MB/sec

AIM-65 MULTIPLEXERS MULTIPLEXERS

(6502 p. P) 1-4 5-8
0
KEYBOARD 16CHANNELS
SCOPE DISPLAY EACH
1 1 11
DISPLAY

SCHANNEL I
ANALOG I 64 BIT I 12 BIT D/A 1
TIMER 1 CLOCK/
INPUT CARD J DIGITAL I CONTROL/ DRIV ERCARD SEQUENCER
12 BITS 20 INPUT DISPLAY CAR D 2OHz- 2OOk Hz CARD
200kHz/CH 34 CARD CARD

LOCAL BUS

FIGURE 4.1: DATA ACQUISITION SYSTEM FUNCTIONAL LAYOUT

 MIT Blowdown Facilities
Data Aquisition & Analysis Computers

800/
800 1600
BPI Inter-Processor BPI
Link
PERKIN-
338 MB DEC 500KB/Sec ELMER 300MB
PDP- 11/70 3242

4MB
MAIN MAIN 8
-

LPiner
Plotter H-

MEMORY MEMORY Term inols

20 FPS-120B
Terminals A r r ay
I
Processor
I I

From DATARAM
Blowdown From
32MB RAM Blowdown
Compressor
Data Aquis. Turbine
Data Aquis.

FIGURE 4.2: DATA ACQUISITION/ANALYSIS COMPUTER SYSTEMS

 105

3.3 mm

450

KULITE
TRANSDUCERS

fL--

Cooling Water

FIGURE 5.1: FOUR SENSOR CYLINDER PROBE MEASURES

TOTAL AND STATIC PRESSURES AND TWO
FLOW ANGLES. FREQUENCY RESPONSE IS
D.C. TO 12 KHZ.
 --'u
#'
A - t. ~'

-t

s
74 wl4
-

H
0

. k-

+
-

FIGURE 5.2: SINGLE SENSOR TOTAL PRESSURE PROBE, 1.2 MM O.D1

n
 Hot
Wir es > /Wire Plane
I
\ N ~TT71
/A
4%604

1.5mn
a woos,

\\\\\I H
0

3mm Chok ed
Orif ice

To Vacuum

FIGURE 5.3: ASPIRATING PROBE MEASURES TOTAL TEMPERATURE

AND PRESSURE. FREQUENCY RESPONSE IS D.C. TO
20 KHZ.
 RADIAL
RADIAL
CIRCUMFIRENTIAL MOTION
OTION
MOTION

a
RADIAL DRIVE
STEPPING MOTOR

LINEAR DRIVE
MOTOR NOT SHOWN

//

CAVITY SEAL 100 MM

PROBE

FIGURE 5.4: STATIONARY FRAME, RADIAL-CIRCUMFERENTIAL TRANSLATOR

A) DOWNSTREAM VIEW B) SIDE VIEW
 SIDE VIEW
77

+
) 4- -.
4

LI

+
EAR VI EW
7

/
I, 7

/
/
/
10

FIGURE 5.5: SHAFT MOUNTED ROTATING

FRAME RADIAL TRANSLATOR
 110

HEAT FLOW TEMPERATURE

Ti SENSORS

d INSULATOR

ADHESIVE T

BLADE PROFILE SURFACE

(Not To Scale)

FIGURE 6.1: HEAT FLUX GAUGE CROSS SECTION

 111

1.0

.8 - No Adhesive

12pM Adhesive

0
.6
0

LL-
z .4 Oft---

.2
0 40 80 120 160 200
FREQUENCY (HZ)

FIGURE 6.2: EFFECT OF ADHESIVE THICKNESS ON

BONDED GAUGE FREQUENCY RESPONSE
 112

320
I I I
1.4 Semi - Inf in.
Adiabatic
240o
z
1.0
0
\Attenuation Phase
z
H 160 C,)
LU
w
.6 U)

800a

.2

I
0
0 8 16 24 32 40
SQRT (FREQUENCY, HZ)

FIGURE 6.3: PHASE SHIFT AND ATTENUATI ON VARIATION

WITH HEAT FLUX FREQUENCY FOR BONDED
GAUGE
 External Lead
Attatchment Area

5mm -----

-
.2mm

Note: Length And Width

To Different Scales

Copper Leads
5mm 1Opm Thick Upper Surface I cm
(Not To Scale) Temp. Sensor Kapton Sheet
2 5 pm
Thick
(Not To Scale)

Lower Surface
Temp. Sensor I+mm
(0.13pm)
8 cm

FIGURE 6.4: SENSOR TEST SPECIMEN GEOMETRY

 114

12 1
R2 Ri

dV V

R02 Rol + dRl VFilm

FIGURE 6.5: WHEATSTONE BRIDGE NOMENCLATURE

 1-15r

Ni

Cul

1.3mm

2cm Imm

2 5pm

-
, Ni

Cu Cu

Tabs

Sensor -04 .04

B) SENSOR & LEADS C) SENSOR DETAIL

13cm

5cm 4.6 cm

tinn nnu nfl nuu rt

Leading Edge
A) SHEET

FIGURE 6.6: HEAT FLUX GAUGE LAYOUT FOR NGV. THE SHEET OF
GAUGES IS WRAPPED ABOUT THE AIRFOIL, THE SENSORS
POSITIONED AT MIDCHORD, THE SOLDER TAPS ON THE
TIP SHROUD PLATFORM,
 116

t'O

AQ

t~o
t (0 t=o
-Ti
x/

dInsulator

T2 =constant

FIGURE 6.7: STEP RESPONSE CALCULATION NOMENCLATURE

Qo COS (wt)

I,

II
T /- d Insulator

T2 = constant

FIGURE 6.8: SINUSOIDAL RESPONSE CALCULATION

NOMENCLATURE
 10
I I I I

1.4
1.3
8 1.2
('1 1.1
E
1.0
3:
6 .9 POLYIMIDE
.8
0 .7
H
H

4 5.0 .6puM
xj 4.5
-4.0
3.5 SILICON DIOXIDE
3.0
2. 52. M--
2

SILICON
40 35 30 25 20)uM
0 I
0 10 20 30 40 50 60
FREQUENCY (kHz)

FIGURE 6.9: THIN FILM SENSOR SENSITIVITY AND FREOUENCY RESPONSE AS A

FUNCTION OF INSULATOR MATERIAL AND THICKNESS
 1.00

w
(n
Z
o .99 .5p m
CA
w

.7/z m
o Ho

.98
L.

.97
0 10 20 30 40 50
FREQUENCY (kHz)

FIGURE 6.10: THIN FILM SENSOR FREQUENCY RESPONSE VARIATION WITH INSULATOR THICKNESS
 I I

3.0 I
I I

2.5 Inlet Total Pressure (atm)

2.0

Eddy Broke Current (Amps x 10

1.5 Valve Position H

1.0
Rotor Mechanical Speed (RPS xI0-2)
Exit Total Press. (atm)
.5
Exit Static Pressure (atm)

0
0 400 800 1200 1600
Test Time (ms)
FIGURE 7.1: BLOWDOWN TURBINE FACILITY TEST RAW DATA. THE USEFUL
TEST TIME EXTENDS FROM 250 MS TO 650 MS.
 I I I I i I

9.5-
100
a)
a)
9.0
Corrected Weight Flow
0. 99

8.5.2(0
a.
98 Corrected Speed 4-
H
0

8.0
97

7.5
I I I I
I I
250 350 450 550
Time (ms)

FIGURE 7. 2: TYPICAL VARIATION OF CORRECT SPEED AND WEIGHT FLOW

 121

APPENDIX A
BLOWDOWN DYNAMICS

In this section we develop an analytical expression for the blowdown

of a fixed volume supply tank through a turbine coupled to an eddy brake.

A.1 Blowdown of Supply Tank

We assume an isentropic (adiabatic and reversible) expansion process

in the supply tank and an isentropic flow through the inlet duct. The tank

cross-sectional area is very much larger than the choking area A* so that

the tank may be considered a stagnation reservoir. During the expansion,

all properties are assumed uniform throughout the tank. A fraction, a, of

the test section mass flow is "by-passed" through the boundary layer bleeds.

The mass flow at the tank exit (station 1) may be expressed in terms of the

corrected flow at the test section inlet (station 2) as follows:

m = (1 a)m.1)

1
(1+a) m
____PA y RT T 2 P A2 (A.2)
(1+t 2 T2 (T2) TL TO 2(A2
) /yRT /y(P
TO
2 2 TO (2)

1+a ) T TO 2 (A.3)
T2 A 2 yRT
TO

where the last equality follows from the isentropic flow assumption. The

subscript T denotes stagnation quantities, while lack of the T denotes

static quantities, and number subscripts denote flow stations. Thus, PTO

and TTO represent the stagnation pressure and temperature within the

supply tank (station 0), and A 2 is the test section inlet reference area.

Observing that To ~ TTO and PO ~ PTO, and using the perfect gas law,

we have,
 122

2/ T2 1
m =
(A.4)
+Y 0 0 2
1 T2 2

For an isentropic expansion in the tank,

T(t) P(t) (A.5)

T(O) ~ p(O)

thus,

Y+ 1
pO(t) 2
m 2 yRT
m (t) = (1/y )(1+c )p() (A.6)
yRT0(0) A 2 P T2A 2 P (0)
T2 21

)
Conservation of mass requires,

m=-P CO)V --0)/ (A.7)

1 dt \
)

where V is the supply tank volume. Therefore

dp(t)dp~(t (1+c)A2(P (t)) 2 yRTo 2 yRTT (A.8)

dt yV PT2A2

where both 7(t) p(t) and the corrected mass flow are functions of time.
p (O)

If the turbine is choked during the run, the corrected mass flow will

be constant and a simple integral for p(t) is obtained:

-+ ) - 2 /T (1+x
( )A yR
dp= PT 2A YV dt ,I (A.9)
p

2 2 T2 1+c)A2 (A.10)
C
-

1
-Y P T2A 2YV

2
Now p (t=0) = 1, thus C = /-(1-y), and,
 123

ST(y- ) AA2/yRT 0O m2yRT , (Y-1

Y 2V PT A 2

Defining the blowdown time constant Tg by,

( 1+,x)A vy RT ,(0 yRT7

1 (Y-1) 2 0 m2 RT2
2 P T A 2(A. 12) T 9 Y
T - y 2V PA2

for the isentropic blowdown with constant corrected flow we have,

2
1+ t_ -1
p0 (t) - p 0 (O) (A. 13)

P (t) 00 0 1 + Tt (A.14)
= P (0)

0 1 9
T (t) S (0L) 1 + - -2 (A.15)
9
-

0 T (0 1

A.2 Corrected Speed

We now summarize the derivation of the corrected speed equation.

A.2.1 Equation of Motion

The shaft mechanical speed wj(t) is governed by,

2
1 dt (A.16)
2 dt It- b

Where I is the moment of inertia of the rotating system,

w(t) the shaft mechanical speed,

nt(t) the turbine power output,

and Rb(t) the power absorbed by the brake.

Other contributions to the net power balance are neglected (e.g., bearing
 124

friction).

A.2.2 Turbine Corrected Speed

Corrected shaft angular velocity is defined by,

Do,(t)
c (t) H Dw(t) (A.17)
cor -/yRTT2(t)

where D represents a reference diameter and TT 2 (t) the inlet total tem-

perature. wcor(t) provides a measure of the turbine tip Mach number. We

make the following manipulations to expresss w( t) in terms of W cor (t):

___ (__) T2 __t)

cor
cor =
co(t)
/y
~ T (t)
Y7O(T(t)2(A.18)
RT T2(0) TT2(t)
W0
()

We identify the quantity in braces as wcor(), and defining,

o(t) W 'cor(t) (A.19)

Wcor(O)

we obtain w (t)/wj(0) =0(t) Q (t). (A.21)

T2
A.2.3 ~~T2
Turbinet Power T o

The turbine shaft power, 2t(t), is given by,

fl (t) = P (0) D /yRT (t II (t) I (A.22)

where PT2(t) and TT 2 (t) are the total pressure and temperature at the

turbine inlet; D, a reference length; and Rcor(t) a turbine non-dimensional

or "corrected" power, which is a function solely of the turbine dimensionless

 125

operating conditions, themselves a function of time. Rearranging,

1
2t T(t) T (t) cor (t)
t~t = 2P90D, T2 (0 cor(0 P (0) T (- FO (0) (.3
T2 T2 cor

we identify the quantity in braces as I t(O) and define,

6(t) 2P(t) (A.24)

PT2(0

( TT 2 (t)\

0(t) I\TT 2 0 , (A.25)

if (t)
IL(t) cor ,(A.26)
cor

to obtain

it(0) 6(t) O(t) 11(t) (A.27)

I1t(t)

A.2.4 Eddy Brake Power

Utilizing the "induction motor" model for brake power Hb(t),

W(t)-2

Sb(t) = 2 H (t) 2, (A.28)

1 + I (t)

where 110 is the maximum torque power and wo the maximum torque speed,

we obtain,

22 1+[j (t) 2
b = E t(t) 0- t) 2(t) - (A.29)
b L1 + Ho 2
2(t) 2
Q (t)
 126

where

2 ' (t)
E (t) = (A.30)
b I (0)
0

and

S (0) ((t)31
0 0

Rearranging, we obtain the form to be used in the corrected speed equation,

_(o) 2 2 2
.[ 1- On(t) ' (t)
Ib~t F12 ( t)(1)2 2 0o
= E (t) 0 (t) 21+ . (A.32)
b 1 + 0 (t) Q (t)

When the excitation magnets are operating in their linear region, Eb

represents the excitation current ratio ib(t)/ib(o).

A.2.5 Corrected Speed Equation

Combining the above, we obtain the differential equation governing the

corrected speed of the turbine,

1 2 d 2 2 2 2 2
1 IW 2 (0) {0 (t)Q (t)} = , (0)6 (t)0(t) (t) - fl (0)E (t){o (t)o (t)}
2 tt b b

W()2 { (t) 2to

(t)} 2

+ [ (A.33)
22
2
(t)"Q (t)
L I +{t

which can be solved for the corrected speed wcor(t) = tcor(O) Q(t), given

the excitation history Eb = Eb(t), and the initial values, R t(O), w,(O) and

RbO). At t=0, note that 2(0), e(O), 6(0), Eb(O) and 11(O) are all unity.

The time dependences of 0(t) and 6(t) are a function of the blowdown

dynamics, and relations have been derived for the constant corrected mass
 127

flow case. Also observe that, for fixed corrected operating point operation,

1(t) = i(0) = 1.

The initial mechanical power itt(O) and speed c(0) are determined by

the conditions of the particular test, and the initial value of braking

power Rb(O) will govern the initial slope of the corrected speed response.

For, at t=O,

1 2 d 2 2
IW (0) -- {o (t)2 (t)} = (0) - (0) , (A.34)
2 dt t= t t

from which we obtain,

____ ' t(0) -

1
b( d) (A.35)
dt t=0 2 dt t=o

Blowdown Equations

For an isentropic blowdown at constant corrected flow, the inlet total

temperature and pressure are given by,

T (t) -2
T2 = 1 + (A.36)
T2Lg

2y
P (t) r~1
T2 = 1 t (A.37)
P (0) T g
T2 g

where Tg is the blowdown time constant. From our definitions of o(t) and

6 (t), Eqs. (A.19) and (A.20), observe that

e(t) = 1 + -] (A.38)
T9

6(t) = e(t) yA.31 (A. 39)

 128

A.3 Conditions for Constant Corrected Speed

The corrected speed equation may now be used to find the eddy brake

initial power Ilb(O) and excitation history Eb(t) which provides constant

corrected test conditions. We assume that the turbine is choked so that

Eqs. (A.38) and (A.39) apply for 0(t) and 6(t). Observing, therefore that,

2
do = (t)
(A.40)
dt
-

and requiring

- =0 ,
(A.41)
dt

Eq. (A.35) relating the initial condition yields,

2
n b
(0) =
t
(0) +
T
. (A.42)
g

Thus, to hold corrected speed constant as the test begins, the brake must

initially absorb the full turbine power plus the additional amount neces-

sary to slow the shaft speed to maintain a constant corrected speed. We

see that this increment depends upon the initial rotational kinetic energy

of the rotating assembly and the blowdown dynamics (rg).

We now use the corrected speed equation to find the required eddy brake

excitation, Eb(t). From Eqs. (A.38), (A.39), and (A.42), and observing that

at constant corrected conditions,

Q(t) = 2(0) = 1 (A.43)

n(t) = R(0) = 1 , (A.44)

the differential Eq. (A.33) reduces to a simple algebraic expression which

can be solved for Eb(t). After a little manipulation we obtain,

 129

2 - + 2

+
- ( )2 (t 2 -1 IW (0) 2

(t))T+-- b - ] , (A.45)
Eb(t) = + 2
b (0) 2 t2 +W(0
I~ (0)-r

which is the sought after result. Recall that Eb(t), for unsaturated

eddy brake magnets, represents the excitation current ib(t)/ib(O). A

plot of Eb(t) for a full Reynolds simulation in Argon/Freon-12 is shown

in Fig. A.1.
 130

1.0.-

0.8.

0.6

uj
N/
0.2-
ul-

0 0.1 0.2 0.3 0.4 0.5

Time (sec)

FIGURE A.l: EDDY BRAKE ExCITATION

HISTORY REQUIRED FOR CONSTANT CORRECTED
SPEED CONDITIONS
 131

APPENDIX B
EDDY CURRENT BRAKE THEORY

B.1 Load Requirements

Even under the transient flow conditions of the blowdown facility,

substantial power is produced by the test turbine. For a full-scale,

design Reynolds number simulation in Argon/Freon-12, the test turbine

generates approximately 1,078,000 Watts at 6190 RPM. Without a load, the

"free-turbine" acceleration would be sufficiently large to preclude the

simulation of steady-state operating conditions. In fact, for certain high

Reynolds number tests, the mechanical speed might even increase beyond the

turbine's safe operational limit. Furthermore, if the tests are to attain

fluid dynamic similarity with actual operation, the turbine "corrected"

speed, amoung several other parameters, must be held to within acceptable

bounds over the 0.5 to 1.0 second test duration. Thus, the blowdown turbine

must be coupled to a compatible power absorbing load which can match the

power production of the test turbine at the desired testing speeds, control

the test speed in accordance with the requirements of fluid dynamic

similarity, and absorb the total energy generated during the run time.

The transient nature of the blowdown flow imposes additional require-

ments upon the turbine load. To conserve power and to establish the proper

initial conditions for the tests, the turbine is brought up to operating

speed in vacuum, whereupon the main flow is established as rapidly as

possible; for the present facility, this is of the order of 50 milliseconds.

To assure a smooth transition, the load must engage in a time compatible

with the startup of the flow and hold the rotor speed "in-bounds" as the

mass flow builds up, thereby allowing it to lock upon and maintain the

desired corrected speed for the test.

 132

Furthermore, as a result of the blowdown of a constant volume reser-

voir, both the turbine inlet temperature and pressure will decrease during

the test. The "fixed operating point" turbine output power will consequently

decrease, and if turbine corrected speed is to be maintained constant, the

turbine mechanical speed must decrease in proportion to the corresponding

drop in sound speed. Thus, both the power absorbed by the load and its

mechanical speed must vary in a manner prescribed by the blowdown dynamics

if constant corrected conditions are to be maintained during the test.

Therefore, the load, in addition to matching the output of the turbine at a

given test point, should have a power vs. speed characteristic compatible

with the characteristics of the test turbine operating under blowdown

conditions. It is highly desirable that this load characteristic track the

turbine's in such a way that a passive "open loop" control can be achieved

without the need to continually monitor the speed and actively control the

load performance.

In summary, the blowdown turbine load should have the following

capabilities:

1) The ability to match the power output and mechanical speed of the

turbine over the desired range of test conditions,

2) A torque-speed characteristic which provides stable control of the

corrected test conditions,

3) A turn-on response compatible with the blowdown start-up transient,

4) A total energy dissipation capacity to meet test duration and over-

speed protection requirements in a physical package compatible with

that of the blowdown facility.

 133

B.2 Eddy Current Brake As Blowdown Turbine Load

The eddy current brake was chosen as the speed controlling load for

the blowdown facility because of its expected operating flexibility and

mechanical simplicity. It can meet the load requirements in a physically

compact package compatible with the size and layout of the facility, and

its braking torque can be selected by a simple adjustment of its excitation

current.

The torque vs. speed performance of the eddy brake is ideal for the

desired "passive" control of turbine corrected speed, and, since the braking

is purely electromagnetic with no physical contact or mechanically wearing

parts, its performance should be stable and repeatable.

The eddy brake is excited electrically; it can be turned on and off at

will with the proper high current switchgear. Its turn-on time constant,

determined by the excitation coil/magnetic circuit configuration, can be

configured to provide the transient performance required by the blowdown

facility.

The total energy dissipated by the brake appears as an internal heating

which must be stored within its structure. Although there are numerous

design trade-offs with regard to performance vs. the size and configuration

of the loss absorbing element, the eddy brake can be configured to meet the

energy dissipation requirement while providing acceptable torque vs. speed

performance.

Thus the eddy current brake can meet the criteria for the blowdown

turbine load while also providing to the facility a flexibility, not only

for a complete study of the present ACE turbine over a wide range of test

conditions but also in providing the blowdown facility with the capability

to match the load requirements of future test turbines.

 134

B.3 Eddy Brake Configuration

The eddy current brake is illustrated in Fig. B.1. The principal

components of the brake are: 1) the cylindrical, electrically conducting,

non-magnetic "loss" drum attached to the main shaft, and 2) the alternating

array of d.c.-excited electromagnets surrounding the drum. Cirumferential

motion of the drum through the alternating, radially oriented magnetic

field induces an emf which drives an axial flow of electrical current

within the drum. The eddy brake is fundamentally equivalent to the Faraday

disc generator except that the current, instead of being "picked off", is

allowed to flow in closed loops within the drum. As is the case for the

generator, the electrical power production gives rise to an "armature"

reaction; the rate at which electrical energy is produced must be balanced

(under steady state contitions) by the rate at which mechanical energy is

supplied to the rotating drum. Hence, a "braking" torque, opposing the

rotation, is developed and the mechanical work supplied appears as an I2R

heating of the loss drum. The torque may also be viewed as the result of

the Lorentz force which arises from the motion of a current carrying

conductor through a magnetic field.

The magnitude of the braking torque depends upon both the strength of

the applied magnetic field and the tangential speed of the drum. The

analysis presented in the following section will show that the torque is

proportional to the square of the applied magnetic field. We shall find,

however, that the speed dependence is somewhat more complicated and, as we

shall see, can be important to the overall performance of the eddy brake as

a speed controlling load.

As the shaft speed increases from rest, the braking torque is initially

proportional to speed. As the speed continues to rise, the torque peaks at

 135

a maximum value and then falls off with further speed increase. This

behavior results from the influence of the "induced" magnetic field,

produced by the eddy currents, upon the applied field. At low speeds, the

magnitude of the induced field is small compared to the applied field and

exerts negligible influence upon the braking process. But with further

increase in speed, it grows stronger, eventually attaining a magnitude

comparable with that of the applied field. Furthermore, the induced field

opposes the penetration of the applied field into the eddy brake drum.

Thus, the applied field is "convected" away from the drum and is further

weakened by the resultant stretching of its (solenoidal) field lines.

The torque-speed characteristic of the eddy current brake displays a

torque maximum as a result of this back emf effect acting on its rotor.

Typical torque and power performance curves for these "induced current"

machines are shown in Fig. B.2, plotted against shaft speed. At speed wo,

the torque reaches a maximum, Tmax, and the corresponding absorbed power

is H o = H (Wo) = Tmaxo. Also indicated on the plots is the general

shape of the curves at the limiting values of rotation speed.

The analysis will show that brake performance may be modeled by the

induction motor torque-slip characteristic,

T 2(w/w
)

T _ _ __ _ _

T 0 (B.1)
max 1 + (/
)

when the skin effect phenomenon can be neglected. (It will be shown that

this is usually true when w < w.) The power absorbed is then,

2(m/s )2
R
_=(w 0 )2(B.2)
II T o )2
o max o 1+(W/W
)

0
 136

This model is useful for machine design and analysis because Tmax and

Wo completely specify the model characteristic and can be simply related

to the machine configuration and excitation field strength. Thus, the

operating line is completely specified by wo and either no or Tmax*

For each test condition, the eddy brake operating point must follow a

time history prescribed by the blowdown dynamics if it is to hold turbine

corrected speed constant during data acquisition. The magnitude of the

applied field might be used to control operation of the brake as part of a

feedback control system. However, active control of the very large total

current (500-700 amps) required to excite the brake is difficult, therefore

the corrected speed will be controlled "open loop" by proper matching of

the brake characteristic and selection of its initial operating point.

From a turbine-eddy brake load matching analysis, a set of operating point(s)

on the normalized characteristic can be obtained which bound the required

eddy brake excitation history to that which can be provided through

adjustment of the passive ballast load resistor.

Rotating system stability is an important consideration in eddy brake

design and operation, for although the brake may match the turbine at the

desired operating point, the torque balance between the two must be stable.

The relative slopes of the torque-speed characterisics of the two machines

at the operating point will determine whether or not small perturbations

will grow or decay. If the design speed is larger than the brake maximum

torque speed, wo, turbine runaway becomes a possibility. In this case, a

small increase in turbine speed, for example, could increase the turbine

torque but would reduce the eddy brake torque resulting in an overall net

acceleration torque.

As we have seen, the eddy brake operating characteristic is uniquely

 137

specified by its maximum torque conditions Ho(Tmax) and jo. For a

given machine, no may be varied by adjusting the magnitude of the applied

magnetic field (up to the saturation limit of its core material) and is

thus an easily controlled variable for test to test adjustment. The

maximum torque speed wo, as we shall see from our model, depends on the

geometrical configuration of the drum and excitation magnets. For a given

machine, it is possible to change its value by adjusting the magnet gaps or

by reconfiguring the coil polarities. In practice, these adjustments are

difficult and time consuming to accomplish on a test by test basis. Thus,

it is important that the wo realized by the machine be suitable for the

range of test conditions expected.

B.4 Excitation Magnets

The excitation field is supplied by conventional U-core electromagnets

surrounding the outer surface of the loss drum. After passing across the

air gap (including the non-magnetic drum), the magnetic flux loops through

the annular return ring mounted inside the drum, thus completing the mag-

netic circuit. Both the U-cores and return ring are laminated to reduce

eddy current losses and are constructed from standard transformer stock

material with good hysteresis properties. This material saturates at

approximately 20,000 Gauss which represents, therefore, the practical upper

limit on core magnetic flux density.

The maximum torque and speed of the machine, hence its operating

characteristic, depend in part upon the pole to pole spacing and air gap

length of the magnets. However, these parameters also determine the

strength of the magnetic field passing through the loss drum relative to

its strength in the magnet cores (i.e., the amount of flux loss in the
 138

system), hence the maximum power absorbing capability of the machine.

The magnetic flux density in the loss drum will be less than that in

the core, partly as a result of the lamination stacking losses but more

significantly as the result of magnetic flux path leakage and fringing.

Both leakage and fringing depend critically upon the pole/gap geometry

which, in the case of the eddy brake, is dictated by the performace require-

ments and design constraints of the machine. These ,state that the air gap

be a significant fraction of both the pole face width and the pole-to-pole

spacing (pole pitch), with the result that both leakage and fringing must

be considered as possible sources of reduction of the loss plate field

strength.

To minimize flux leakage, the coils are wound "squat" and are mounted

close to the pole faces. This helps to reduce the pole-to-pole leakage

path area, hence the total leakage flux. Predicting the exact level of the

leakage is difficult due to the complex geometry of the configuration,

however it can be bounded by a worst case estimate.

The U-cores, assembled from standard transformer stampings, present a

constant area flux path around the "U". The total air gap flux, both

leakage and "working", must loop the excitation coils. Hence, essentially

all of the flux must enter and pass through that portion of the core

surrounded by the coils. Since the magnetic flux is a divergence-free

field, the flux density may be lowered significantly throughout most of the

remainder of the core by filling the hole of the "U" behind the squat coils

with ingot iron, thereby reducing both the reluctance of the core path and

its susceptibility to saturation (see Fig. B.3).

Fringing is more difficult to reduce because it depends upon the air

gap to pole width ratio, which is constrained by machine performance

 139

requirements. For the geometry of our brake, fringing and core stacking

alone will limit the drum flux density to about 60% of that in the core

pole faces. Therefore, before inclusion of the losses due to leakage, the

flux density available for braking would be limited by core saturation to

about 12,000 Gauss. To conservatively account for leakage, the brake is

designed for a maximum of 7500 Gauss applied to the drum.

B.5 The Loss Drum

The eddy current brake loss drum is a thin walled cylinder cantilevered

from the end of the main turbine shaft. The excitation magnets are wrapped

around its outer circumference and induce axially oriented current loops

within the drum material. The 1 2 R power dissipation in the drum might be

viewed as either the cause or the effect of the braking action.

Either way, the drum must have sufficient mass to absorb the total

energy load requirements of the blowdown, which will typically be on the

order of one to two megajoules. This requirement is dictated by the total

turbine energy produced during the start-up and active test time of the

experiment, plus that portion of the blowdown energy remaining after the

test which would contribute to turbine overspeed.

Since it spins with the turbine, the drum must be capable of both high

temperature and high stress operation. Since only the active portion of

the drum under the magnets will see uniform heating, thermal stresses must

also be considered. Each point on the drum will experience an N-pole per

revolution excitation (and possibly lower harmonics if there are pole to

pole magnetic field or loss drum nonuniformities), therefore the possibility

exists for adverse structural drum dynamics.

The drum electrical resistivity is a significant parameter which

 140

affects not only the power dissipation per unit volume of the drum but also

the maximum torque shaft speed. It is necessary that the resistivity be as

high as possible for good brake performance. In order of increasing resis-

tivity come copper, aluminum, carbon steel, non-magnetic stainless, and the

nickel superalloys. The stainless and nickel steels are the only ones that

possess the magnetic, structural, and electrical properties suitable for

the present application. Both were considered, but the increased structural

and thermal load capability of the nickel superalloy resulted in its final

selection.

B.6 Eddy Current Brake - Simplified Theory

Consider a non-magnetic, electrically conducting plate moving at

velocity v through a magnetic field B. An emf will be induced causing

"eddy" currents to flow within the plate. From Faraday's and Ohm's laws,

the induced current density is given by j = a(v x B), where a is the

plate conductivity. The interaction of the current with the magnetic field

results in a force opposing the motion, which per unit volume is given by

f = j x B. Energy conservation requires that a force must arise, for

mechanical work must be supplied to balance the rate at which power is

dissipated in the plate due to the flow of current across its finite elect-

rical resistance.

Now consider the linear eddy current brake configuration illustrated

in Fig. B.3. The "loss" plate moves through the magnetic field produced by

the array of alternately directed, rectangular poles of length 2a and width

2b. The applied field is assumed uniform under the poles, where it induces

a current density avB in the plate. Elsewhere, both the applied field and

induced current are assumed zero. The current flowing in the plate element
 141

directly under each pole is i = A2aavB. The electrical resistance of this

element, of cross-section A2a and length 2b, is 2b/ca2a. The power dissi-

ua4abB 2 v 2
pated per pole, H, is therefore , and must he supplied through

the rate at which mechanical work is performed upon the moving plate.

Thus, the force per pole acting upon the plate is

F = 1/v = oaA B v , (B.3)

p

where Ap =(2a)(2b) is the magnetic pole area.

This simplified model provides a good estimate of the eddy braking

force under the following conditions:

1) There is no fringing of the magnetic field, i.e., it is perpendicular

to the plate and uniform under the poles,

2) The induced magnetic field produced by the eddy currents is small

compared to the applied field,

3) The eddy current return paths contribute negligible power dissipa-

tion compared to the paths under the poles, and

4) The skin depth of the eddy currents is much larger than the plate

thickness.

For the turbine braking application, item 2) will be responsible for the

most deviation from this simple theory and will now be considered in more

detail.

B.7 Magnetic Reynolds Number - Influence of Induced Field on Braking Force

Suppose a magnetic field of magnitude Bo is imposed perpendicular to

the plate moving laterally at speed v. The induced current density, given

by j = G(E+vxB), is therefore of order GvBQ. From Ampere's law, curl

B = pj, the magnitude of the induced field Bi will be of order Ap j = pjAvB 0

.
 142

The ratio of induced field to applied field is therefore proportional to

the quantity paAv. This ratio is called the "magnetic" Reynolds number:

R 1 = paAv B./B , (B.4)

and it characterizes the ability of the applied field to penetrate the

moving conductor. At low values of RM (RM<<), the applied field passes

through the plate. At high values (RMl1), the applied field is convected

along the moving plate with very little diffusion into the plate.

This phenomenon can strongly influence the braking force produced

because it governs the penetration of the field, responsible for the flow

of the power dissipating eddy currents into the moving conductor. The

analysis which follows will show that its net effect is to place a limita-

tion upon the force-producing capability of the device as the magnetic

Reynolds number approaches unity. This represents a particularly important

consideration for the turbine braking application where, as a result of the

high power dissipation and shaft speed requirements, practical engineering

designs will, of necessity, be required to operate in the unity Reynolds

number regime.

B.8 Eddy Current Brake - Induction Motor Model

In modeling the eddy current brake, we follow the approach employed in

Refs. [B.1-B.3]. The physical configuration illustrated in Fig. B.4 is

replaced by the model configuration shown in Fig. B.5. To eliminate the

dependence of the field quantities upon the transverse y-direction, we

assume that the width of both the eddy plate w, and the magnetic poles 2b,

is large compared to the pole pitch T, and that all currents (both excita-

tion and eddy) flow along the y-direction. An edge effect correction
 143

will be introduced later to account for the return path current loops in

the eddy plate.

The excitation windings and salient poles of the physical machine are

replaced in the model by two infinitely thin conducting sheets, parallel to

the eddy plate and located at g. They are backed by a non-conducting

material of infinite permeability. The gap magnetic field is generated in

the model by surface currents assumed to flow on each of these "smoothed"

pole surfaces along the y-direction. The total current per unit length

along the x-direction is specified by the linear current density Ky (x,t).

The physical airgap Eg between the salient poles is replaced by an

effective airgap 2g as determined by Carter's air gap coefficients, Ref.

[B.4], to account for the interpole space and variable reluctance of the

actual configuration.

The field equations will be solved for a harmonic current distribution,

representing the fundamental component of the applied magnetic field. Solu-

tions may then be superposed to find the results for arbitrary periodic

excitations. The total current per unit length along the x-direction will

therefore be modeled by the sinusoidal distribution of amplitude KO, and

2
(spatial) period T:

ITx
K (x) = K sin -- (B.5)
y 0 T

representing the total current density from both sheets per unit length

along the x-direction.

The average normal component of the gap magnetic field, B, may be found

by applying Ampere's law to a rectangular loop of length T and height 2g,

H * dZ = i (B.6)
o tot
 144

where itot is the net current crossing the surface of the loop. Observinq

that H must vanish in the regions behind the current sheets where p

and that as a result of the symmetry of the excitation:

H (x) = -H (x+) (B.7)

z z

we have,

x+-r
(2g) 2H (x) = Ksin dx (B.8)
z 0 T
Ix

K T
H (x) = cosI- (.9)
z i(2g) 1

Therefore, in the gap where B = pH, we obtain,

B(x) = B cos I (B.10)

0 T

where

B 00 (B.11)
o n (2g/-r)

The excitation poles are centered at the zeros of the distribution and

in the model; each pole "face" occupies the entire region between alternate

peaks of the sinusoid. The equivalent amp-turns per pole, Ie, for the

sheet excitation is obtained by observing that each pole is composed of a

pair of adjoining strips of equal width T/ 2

, each carrying an equal but

oppositely directed current of magnitude,

T /2
2
1/2 0 K (x)dx = K 0/ -r (B.12)

per sheet. Assuming each pair to be connected at y = +b to form a single

turn loop, and noting that there are two such turns per pole, one from each

sheet, we obtain,
 145

I = K TIr. (1.13)
e 0

The excitation current, as observed from a reference frame fixed to

the current sheets, may be expressed in the form,

K(x) = K j exp(-jgx/T ) (B.14)

where the real part of this, and the following complex expressions, are

understood to represent the physical quantities. Assuming the plate moves

at velocity v in the positive x-direction and that v<<c, the excitation as

observed from the plate is:

K(x') = K j exp(-j n/T(x'+vt)) (B.15)

The field equations are then solved in the moving frame for a non-

magnetic plate of permeability p0, see Ref. [B.3]. If the heiqht 2g of

the air gap is small compared to the pole pitch T and skin depth 6skin'
the electric field induced in the gap (including the plate) is found to be

essentially independent of z. Adapting the solution quoted in Ref. [B.3]

to the geometry of our configuration, we obtain,

E (x') = E exp(-j 1T/-r(x'+vt)) (B.16)

y yo

where the complex amplitude

SvK
E = -j (B.17)
yo p aAv + ji 2g/T

contains the magnitude and phase-shift information. Defining a magnetic

Reynolds number RM based upon the thickness of the non-magnetic plate,

(B.18)
RM E P 3Av

and the ratio R,

 146

R
M
R if 2g/T (B. 19)

we express the complex amplitude Eyo (in magnitude phase form) by,

K
E o R exp(j$ (B.20)

)
yo
a + 2

where,

-1
tan R (B.21)

The induced gap field is therefore,

K
E =
o R
(3.22)
y a
2
/1+ R2

Applying Ohm's law in the moving frame, the induced current density in the

plate is found to be,

K
S = R [cos -- (x' + vt) (B.23)
a T
-

y IX
V 1-.n2,

The power dissipated in an elemental plate filament of length 2b is

then,

2 2b
2b dlt = (j dx dy) , (B.24)
a dxdz

Integrating over one pole pitch (having first transformed back to the

stationary frame)

2
I (x+T) +A/2 K 2
2b o R 2 x (B.25)
(- - ) dzdx
x -A/ 4A 21+R
-A/2cos

we obtain the power dissipation per pole,

 147

2 2
T K 2
2b o2R (B.26)
OAT 4 2
1+R

The first factor on the right is the total transverse resistance per pole

of the loss plate. The second factor represents the squared maqnitude of

an "equivalent" uniform and uni-directional current yielding the same power

dissipation per pole as the actual alternating induced distribution,

2 2
T K 2
2 _ o 2R
ie - 4.(B.27)
eq 4 1R2
1+R

Given in terms of the amp-turns per pole of the applied excitation, I,

2 2 2
2 n I 2R
ieq= 4 2 , (B.28)
1+R

These expressions will be useful in characterizing the design and operation

of the actual machine.

The braking force per pole, obtained from our power dissipation

expression, is given by,

2 2
H 2b_ 2b GaA TK 2R
F = - (B.29)
v oaA-r (ff 2 g/-) 4 1R2

This equation predicts a maximum force when R=1,

F
max
= F(R=l) = 2
cOAT
PO/T)
(iT 2g/-
LK2 (B.30)

at a maximum force Reynolds number RMO of,

RMO = R M(F max) = -T 2g/T . (B.31)

The power dissipated at maximum force is,

 148

Tj
2b T
10 = H (R=1) = 2b 1 2 (B.32)
o T[1 2J

where TKO/ 2 is the equivalent induced loss plate current per pole.

We infer from the form of the force expression that R is the signifi-

cant parameter characterizing the influence of the eddy current "back emf"

and observe that it may be interpreted as a magnetic Reynold's number based

upon the length scale (-rA/f2g); its significance is made apparent from the

solution of the field equations. If v0 is defined as the plate velocity

at maximum force,

2
v = v(R=1) = ,/T (B.33)
0 0G

it follows that R also represents the characteristic speed ratio,

R = = .~ (B.34)
T 2g/T v

In terms of the maximum force Fmax, power at max force no and speed

ratio R, the power and force relations therefore become

2(v/v )2
= 2 (B.35)
1 + (v/v
)

2(v/v
)

F =F 0 (B.36)
1 + (v/v
)

and since no = Fmax vo, we conclude that vo and either no or Fmax completely

specify the eddy brake performance characteristic within the assumptions of

the thin plate limit. We also observe that, in this form, the force

expression is identical to the normalized torque-slip relation of simple

induction motors, Ref.[B.5], hence the term "induction motor model" is

 149

applied to these equations.

We should recall that the model is valid when,

a) the applied field is normal to the plate and varies harmonically with

2
spatial period T in the direction of plate motion,

b) the plate thickness is small compared to the skin depth 6, and the pole

pitch T,

c) the air gap is small compared to the pole pitch, and

d) the air gap permeability is uniform, hence the use of a nonmagnetic

plate.

The model developed for the linear machine may be applied directly to

the rotating cylindrical configuration shown in Fig. B.1, where typically

the machine radius will be much larger than the pole pitch and air gap

spacing. The "loss plate" is a drum of radius r and axial length w,

rotating at an angular velocity w about its axis of symmetry. It is sur-

rounded by N alternating electromagnets, with rectangular poles of circum-

ferential length 2a and axial length 2b, spaced at a pitch T along the

circumference. Thus, NT = 2 r, with N an even integer.

The magnetic Reynolds number based on drum thickness A is,

R = p a Arw (B.37)

From our "induction motor" characteristic Reynolds number,

p aArj
2
R = = p a(AT/ g)rj, (B.38)
- (2 g/T) 0

the maximum force condition (R=1) may be used to obtain the angular velocity

at maximum torque,

T (2 g/T) N(2g)
W = = (B.39)
0 [ aAr 2 2
o p cAar
 150

from which we obtain,

R = W/W (B.40)

The total power at maximum torque for an N-pole machine is simply,

L = N(2b/cAT) (TK /2)2 (B.41)

0 0

and, using the NT = 2ff r constraint for the cylindrical configuration, we

obtain, in 1 2 R form,

= (rK ) (B.42)
o c1A 2lrr o

where we identify the first factor as the resistance of the active portion

of the cylinder to an axial flow of current and the second as the square

of the equivalent uniform current at maximum torque. Note that it is

independent of N.

The total power and torque for the N-pole machine are given by,

2(/ )2
R = 11 2 (B.43)
1+(W/W
0

T = T 0 (B.44)
ax1+(a/o )2
0

where

)(B.45)
(rK
= 2b 2
b
o aA2Tr 0

T =1 /W (B.46)
max o 0

and

= 1T(2g/-)
- (B.47)
o p aAr
 151

We must now remark that, so far, no account has been made for the

"edge" effects of a finite width plate. It has been assumed that the

induced current distribution is one-dimensional, alternating in sign in a

2
direction perpendicular to the motion with a spatial period of T. For the

actual plate, the current distribution forms return loops at the ends of

the loss drum. Thus, our force expressions are strictly valid for a plate

where these loops face zero resistivity and therefore do not contribute to

the power dissipation.

To account for these return loops, a resistivity correction factor k

is employed. It is computed by solving the field equations for the current

distribution in the plate and is a function of plate and pole geometry and

of the magnetic Reynold's number. An effective conductivity is then used

in place of the physical conductivity in the eddy brake equations.

References

B.1 Singh, A., "Theory of Eddy-Current Brakes With Thick Rotating Disc,"
Proc. IEE, Vol. 124, (4), April 1977, pp. 373-376.

B.2 Jackson, W.D. and Pierson, E.S., "Operating Characteristics of the

M.P.D. Induction Generator," IEE Conf. Rep. Ser. 4, 1962, pp. 38-42.

B.3 Bahler, W. and Van der Hoek, W., "An Eddy Current Coupling Employed as
a Variable Speed Drive," Philips Tech. Rev., Vol. 27, 1966, pp. 15-21.

B.4 Nasar, S.A., "Electromagnetic Theory of Electrical Machines," Proc.

IEE, Vol. III, (6), 1964, p. 1125.

B.5 Fitzgerald, A.E., Kingsley, Jr., C., and Kusko, A., Electric Machinery,
3rd edition, New York: McGraw-Hill, 1971.
 152

D.ru 11

Return

Coils

FIGURE B.1: EDDY CURRENT BRAKE GEOMETRY

 T P
ii

Paw --- 2 P0
1/ 2
Ta w /'
Tmax N- Paw
Behavior 2
P 2 (w /A 0)
06 P0 I+(w A 0o)2
Paw 2
"Induct ion Motor" Model
ra 2 (w/wo)
Tmax 1+ (w/wo) 2
1 10-
wo
SPEED SPEED
A) TORQUE VS. SPEED B) POWER VS. SPEED
(NOTE: THE P'S IN THIS FIGURE
CORRESPOND WITH THE H'S IN THE TEXT)

FIGURE B,2: EDDY BRAKE ROTATIONAL SPEED CHARACTERISTICS

 OLD DESIGN
NEW DESIGN

U CORE-, OLD
COILS

IL-

ADDED

NEW COILS
DRUM

RETURN IRON

FIGURE B.3: MAGNET DETAIL

 A Fixed Ref.
Frame Moving Frame

0 0 0o
Non-Mag. o 0
O
0 0 Z x
"Loss" Plate 0
0
0
0

---- Ib-V

3*4 2aH i~l w

0
00
ul

E xcit ation r VIEW A-A

Coils Pole
Pitch

FIGURE B.4: LINEAR EDDY CURRENT BRAKE CONFIGURATION

 Induced Current Distribution

j o R sin

R: v/ v
AZ\
= tan- R
L~rx V-
Hl

Sheet Current
id

K = Ko sin -i- x
~NZI~i~r

f-O

FIGURE B.5: INDUCED CURRENT NOMENCLATURE