Advanced Gasturbine (Agt) Technology Development Project: DOE/NASA/0167-12 Nasa CR-180891 Garrett NO. 31-3725

DOE/NASA/0167-12
NASA CR-180891
GARRETT NO. 31-3725(12)
ADVANCED
GASTURBINE(AGT)
TECHNOLOGY
DEVELOPMENTPROJECT
FINAL REPORT
Engineering Staff of
Garrett Auxiliary Power Division
A Unit of Allied-Signal Aerospace Company
March 1988
Prepared for
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
Lewis Research Center
Cleveland, Ohio 44135
Under Contract DEN3-167
for
U.S. DEPARTMENTOF ENERGY
Office of TransportationSystems
Heat Engine PropulsionDivision
WashingtonD.C. 20585
TABLE OF CONTENTS
Pa e
1.0 SUMMARY
1.1 Powertrain Development 1

1.2 Power Section Development 1
1.3 Component/Subsystem Development 2
1.4 Ceramic Materials and Component Development 3
2.0 INTRODUCTION
3.0 REFERENCE POWERTRAIN DESIGN
4.0 POWER SECTION DEVELOPMENT 13
4.1 Engine Mechanical Development 13
4.1.1 Foil Bearing 13

4.1.2 Thrust Bearing Development 19
4.1.3 Ball Bearing and Compressor Seal Design 22
4.1.4 Rotor Clearance Evaluation 23
4.1.5 Metallic Structure Design 25
4.2 Engine Rotor Dynamic Development 26
4.2.1 AGTI01 Rotor Dynamic History 26

4.2.2 Current Rotor System Configuration 31
4.2.3 Rotor System Test Results 31
4.2.4 Rotor Static Modal Determination Testing 32
4.2.5 Rotor System Analysis 32
4.2.6 Rotor StabilityAnalysis 33
4.3 Engine Development Evaluations 34
4.3.1 Metallic Engine Testing 36

4.3.2 Ceramic Structure/Metallic Turbine Engine Development 39
4.3.3 Ceramic Engine Development 42
4.3.4 Engine Performance Testing 45
4.3.5 Ceramic Engine Seal Development 45
5.0 COMPONENT/SUBSYSTEM DEVELOPMENT 47
5.1 Compressor Development 47
5.1.1 Compressor System Requirements 47

5.1.2 Aerodynamic Development 47
5.1.3 VIGV Design and Testing 56
5.1.4 Impeller Mechanic Design 57
5.1.5 Compressor Design for Improved Producibility 61
i
TABLE OF CONTENTS (Contd)
Pa e
5.2 Turbine Development 65
5.2.1 Aerodynamic Design 65

5.2.2 Turbine Aerodynamic Performance Testing 73
5.2.3 Metallic Dual Alloy Rotor 79
5.2.4 Ceramic Turbine Rotor Design 84
5.3 Combustor Development 88
Combustor Design 88
Combustor Development 90
5.4 Regenerator System 93
Ford Regenerator Development 94

Garrett Regenerator Development 124
5.5 Gearbox/Transmission 139

5.6 Controls and Accessories 141
5.6.1 Engine Electronic Control Unit 142

5.6.2 Engine Fuel Control 144
5.6.3 Variable Inlet Guide Vane Actuator 144
5.6.4 Electrical Accessories 144
6.0 CERAMIC MATERIALS AND COMPONENT DEVELOPMENT 147
6.1 Ceramic Materials Testing 147
6.1.1 Fast Fracture and Stress Rupture 147

6.1.2 Ceramic Compatibility Testing 149
6.1.3 Ceramic Interface Evaluation 170
6.1.4 RBSN Spalling Study 177
6.1.5 Laser Marking of Ceramic Materials 177
6.1.6 Slip Casting Rheology Studies 178
6.1.7 Turbine Diffuser Insulation Evaluation 179
6.2 Ceramic Turbine Rotor 182
6.2.1 Initial Evaluations 182

6.2.2 Bladed Rotor Fabrication 183
6.2.3 Hot Turbine Rig 187
6.2.4 Turbine Rotor Attachment 188
6.2.5 Cold Spin Testing 190
6.2.6 Rotor Blade Vibration 190
6.2.7 Rotor Stress Analysis 199
6.2.8 Rotor Life Prediction 200
££
TABLE OF CONTENTS (Contd)
Page
6.3 Ceramic Struetures 207
6.3.1 Component Design/Analysis 208

6.3.2 Rig Design/Development 255
6.3.3 Component Rig Testing 268
APPENDIX A AIRESEARCH CASTING COMPANY ADVANCED GAS TURBINE

(AGT) TECHNOLOGY DEVELOPMENT PROJECT FINAL REPORT
APPENDIX B FORD MOTOR COMPANY ADVANCED GAS TURBINE (AGT)

TECHNOLOGY DEVELOPMENT PROJECT FINAL REPORT
APPENDIX C PURE CARBON COMPANY ADVANCED GAS TURBINE (AGT)

TECHNOLOGY DEVELOPMENT PROJECT FINAL REPORT
APPENDIX D LIST OF SYMBOLS, ABBREVIATIONS, AND ACRONYMS
APPENDIX E REFERENCES
NASA C-168 FORM
iii
LIST OF FIGURES
F eure Title
1 AGT101 Program Schedule 7
2 AGT101 RPD Powertrain 8
3 AGTI01 Power Section 9
4 AGTI01 Drive System Schematic i0
5 AGT101 Installation- Side View ii
6 AGTI01 Installation- Front View 11
7 AGT101 Estimated Performance Curves 12
8 AGTI01 Specific Fuel Consumption 12
9 Basic Bearing Configuration 14
10 Foil Bearing Deflected State at 100,000 RPM Operating Speed 14
11 Film Pressure Distribution at Operating Speed 15
12 Foil Bearing Designs Compared 16
13 Foil Bearing Surface Coating Development 17
14 Foil Bearing Dynamic Properties Wear Rig 17
15 TiN Coated Ring (Top) Versus Au Coated Disk (Bottom) After

591 Starts at Ambient Temperature and 4.3 psid Load 19
16 Typical Results of Dynamic Friction Test 20
17 AGT101 Hydrodynamic Thrust Bearing Configuration 21
18 Thrust Bearing Design Detail 21
19 Thrust Bearing/Ball Bearing Load Share Characteristic 22
20 Total Analytical Heat Generation for AGTI01 Fluid Film

Thrust Bearing Cavity 22
21 AGTI01 S/N 001 Build 21 Fluid Film Thrust Bearing (Test 4-19-84) 23
22 Thrust Bearing Power Consumption Decrease with Oil Flow 23
iv
LIST OF FIGURES (Contd)
Title
23 Floating Ring Seal 23
24 Typical Turbine Axial Clearance Transient for Cold Start 24
25 Compressor and Turbine Clearance Loss Due to Thermal Growths 25
26 AGT101 Metal Structured Power Section 27
27 AGTI01 Engine Rotor System Configurations 28
28 AGTI01 Engine InitialRotor System Configuration Response 29
29 AGTI01 Engine Motoring Rotating Group Configurations Tested 30
30 AGTI01 Engine Aerodynamic Components Excite Subsynchronous

Motion 30
31 AGTI01 Rotor Dynamics Rig Impact Test Configuration 30
32 AGTI01 Engine Operational Rotor System Configuration 31
33 AGTI01 Engine Current Rotor System Configuration 32
34 AGTI01 Engine Rotor Mass and Stiffness Model 33
35 AGTI01 Engine Rotor Support Schematic 34
36 AGTI01 Engine CriticalSpeed Mode Shapes 35
37 AGTI01 Engine Unbalance Response/Support Displacements 36
38 AGTI01 Engine Unbalance Response/Support Forces 37
39 AGT101 Engine Whirl Speed Map 38
40 AGTI01 Engine Forward Mode Log Decrements 39
41 The AGTI01 Evolves to Ceramics 40
42 Ceramic Turbine Rotor Failed at 62,000 rpm 40
43 Strain Gaged Ceramic Turbine Rotor 41
44 S/N 002C - T4 Schedule Tested on February 16, 1984 42
45 Ceramic Structures Used in the 1149C (2200F) Test 43
V
Title
Ceramic Components Following the 100-Hour Endurance Run at

i149C (2100F) 43
47 Dual Alloy Rotor Showing Cooling Air Path 43
48 Engine Test Data Setup Model 45
49 Single Separator Seal Configurations 46
50 Flow Separator Seal Configuration 46
51 AGT101 Impeller Design Exit Loss Function Distribution 48
52 AGT101 Impeller Design Exit Deviation Angle 49
53 Compressor Inducer Leading Edge Blade, Air Angle Distributions 49
54 Compressor Impeller Relative Air Angle Distribution 49
55 Compressor Impeller Blade Surface Maeh No. Distribution - Shroud 49
56 Compressor Impeller Blade Surface Maeh No. Distribution - Mean 50
57 Compressor Impeller Blade Surface Maeh No. Distribution - Hub 50
58 AGT101 Impeller Design Minimum a/a'geom. 50
59 Impeller Normal Thickness Distribution 50
60 AGT101 Impeller Design Vector Diagrams 50
61 Splitter Blade Hub Leading Edge Movement 51
62 Splitter Leading Edge Tangential Position 51
63 Diffuser Deswirl Vane Analysis 51
64 AGT101 Diffuser and Deswirl Vane Assembly Axial View 52
65 AGT101 Diffuser Meriodional Flowpath 52
66 AGT101 Diffuser Vector Diagrams 53
67 AGT101 Modified Impeller Blade Loadings--Tip 53
68 AGT101 Modified Impeller Blade Loadings--Hub 53
vi
Title
69 AGT101 Modified Impeller Geometry 54
70 Compressor Rig 54
71 Full Stage Compressor Data, IGVs Open 55
72 Full Stage Compressor Data, IGVs = 40 Degrees 55
73 Full Stage Compressor Data, IGVs = 70 Degrees 55
74 AGT101 Inlet Guide Vane 56
75 AGT101 Compressor Inlet Guide Vane Test Rig 56
76 Inlet Guide Vane Flow Tabs 57
77 Inlet Guide Vane Test Results 57
78 AGT101 3-D Airfoil Elastic Stress at 100,000 rpm 58
79 AGT101 Impeller Airfoil Campbell Diagram 59
80 AGT101 Impeller Blade Vibratory Modes 59
81 AGT101 3-D Splitter Elastic Stress at 100,000 rpm 59
82 AGT101 Impeller Splitter Campbell Diagram 60
83 AGT101 Impeller Stress at 100 Percent Speed 60
84 AGT101 Impeller Isotherms - 29.4C (85F) Inlet Temperature 60
85 AGT101 Impeller Deformation at 100 Percent Speed 61
86 AGT101 Impeller Blade Vibratory Mode 1 62
89 AGT101 Titanium Impeller Campbell Diagram 63
90 Strain Gaged Impeller, Front View 63
91 Strain Gaged Impeller Blade 64
vii
vi re Title Page
92 Strain Gaged Impeller, Backfaee 64
93 Compressor Test Rig with Slip Ring Assembly Installedin Test Cell 64
94 AGT101 Titanium Impeller Campbell Diagram Excitation Source 65
95 Strain Gage Test Results 65
96 Tensile Properties of Forged ALCOA PM A1-Fe-Ce 65
97 ALCOA A1-Fe-Ce Forged Properties (1000-Hour Temperature

Exposure) 66
98 Tensile Properties of Aluminum Alloys 66
99 Stress Rupture Properties of Aluminum Alloys 66
100 HCF Properties of ALCOA A1-Fe-Ce 67
101 Impeller Blade Loading-Hub 67
102 Blade Loading-Mid-Passage 67
103 Impeller Blade Loading - Shroud 67
104 Off-Design Performance Characteristics as a Function of Design-

Point Exit Swirl for NASA-DOE Advanced Gas Turbine Powertrain
System Development Program 68
105 Turbine One-Dimensional Vector Diagrams 69
106 Turbine Design Summary 70
107 Baseline Turbine Inlet Duct Configuration 71
108 Turbine Inlet Strut Sections [Modified NACA 16-201 (tlc = 0.30)] 71
109 Turbine Inlet Hub and Tip Velocity Distribution for Baseline
Configuration 71
ii0 Ceramic Turbine Stator Profile Design 72
111 Ceramic Turbine Stator Velocity Distribution(Base, 19 Vanes) 72
112 Ceramic Turbine Stator Profile Design 72
viii
Title Page
113 Ceramic Turbine Stator Velocity Distribution (Linear Pressure
Surface, 19 Vanes) 72
114 Ceramic and Metal Turbine Flow Paths 73
115 Turbine Rotor Blade Angle Distribution 74
116 AGTI01 Rotor Radial Loss Distribution 74
117 Turbine Rotor Velocity Distribution 74
118 Exhaust Diffuser Velocity Distribution 75
119 Cold Turbine Test Rig 76
120 AGT101 Test Rig Installedin Test Cell 78
121 Baseline Ceramic Cold Turbine Rig Test Results, System Efficiency
Versus Pressure Ratio 78
122 Baseline Ceramic Cold Turbine Rig Test Results, System Efficiency
Versus Engine Speed 78
123 AGT101 Exhaust System Loss Characteristics 79
124 AGT101 Metal Dual Alloy Wheel - 13 Blades, Conical Astroloy Hub 79
l
125 AGT101 Dual Alloy Steady-State Temperature Distribution 79
126 Stress Distribution for Dual Alloy Turbine Wheel 8O
127 AGT101 Dual Alloy Steady-State Equivalent Stresses 8O
128 AGT101 Metal Dual Alloy Wheel 8O

l
129 Turbine Vibration Modes at Maximum Power 81
130 Dual Alloy Hub/Bladed Ring Assembly Cross Section 82
131 As-Cast AGT101 Bladed Rings and Machined Dual Alloy Wheel
Assemblies 83
132 Machined AGTI01 Dual Alloy Wheel Assembly--PM LC Astroloy

Hub and MAR-M 247 Blade Ring 84
133 Microetched Cross Section of AGTI01 Dual Alloy Wheel Test Specimen 84
ix
Title Page
Braze Closure After HIP Diffusion Bonding (Sample was Subjected

to Yo-Yo Heat Treatment) 85
135 Radial Turbine Design Flow 86
136 AGTI01 Ceramic Rotor Z-Section 86
137 Turbine Rotor Blade Angle Distribution 86
138 AGTI01 Rotor Surface Velocity Distribution (Shroud) 86
139 AGTI01 Rotor Surface Velocity Distribution (Mean) 87
140 AGTI01 Rotor Surface Velocity Distribution (Hub) 87
141 Piloted Premixed/Prevaporizing Variable Geometry Combustor 88
142 Combustor Configurations Investigated In Emissions Testing 89
143 Element Test Rig Emission Test Results 89
144 Current Configuration Ceramic Combustor 89
145 Duplex Airblast Fuel Nozzle 90
146 Simplex Fuel Nozzle 90
147 Delavan Nozzle Configuration 91
148 Pilot Combustor Proof-of-Concept (Metal Configuration) 91
149 Enlarged Pilot Zone Combustor 93
150 93
AGT101 "Regenerator System" Componenls
q
151 AGTI01 Regenerator System Components 95
152 Typical Regenerator Seal Schematic 97
153 Regenerator Seal Design Phase Comparison 98
154 Static Seal Leakage Comparison for Seal Design Phases I00
155 Inboard (Hot) Seal Crossarm Diagram Cooling Schematic 101
156 Middle (Support) Crossarm Diagram Temperature Distribution 102
X
Title page
157 Regenerator Seal Coating Wear Test Rig 103
158 3-D Thermal Stress Analysis 107
159 Regenerator Support Roller Locations 109
160 Regenerator Drive and Support System Bearing Designs 109
161 Diagonal Bond Configurations II0
162 Leak Paths and Pressure Notations IIi
163 Analytical Model for Dynamic Rig Leakage 113
164 Seal Diaphragm Temperatures at 0.4 Percent Cooling Flow 114
165 Seal Coating Delamination from Crossarm Shoe 115
166 Seal Systems Comparison 116
167 Analytical Model Evaluating Non-Diaphragm Leak Paths as a

Function of Component Distortions 117
168 Regenerator Core Axial Distortion Effects 118
169 Non-Diaphragm Seal Leakage Versus Uniform Clearance 119
170 Non-Diaphragm Seal Leakage Versus Uniform Clearance 120
171 Effect of Temperature on Seal Leakage 121
172 Regenerator LP Cold Rig 125
173 3-D Hot Wire Calibration Tunnel 126
174 Regnerator LP Cold Rig Preswirl and Screen Assemblies 126
175 Measured LP AP Data From Regenerator Cold Rig 127
176 LP Regenerator Cold Rig Upstream Diverter 127
177 LP Regeneration Cold Rig, Effects of Upstream Diverter 127
178 Regenerator HP Cold Rig 128
179 HP Regenerator Cold.Rig Flow Effects of Bonnet 129
x£
Title
180 Measured HP AP Data From Regenerator Cold Rig with Bonnet 129
181 Measured Analytical Flow Distortion Model 130
182 Measured HP AP Data From Regenerator Cold Rig 131
183 Measured LP AP Data From Regenerator Cold Rig 131
184 Capacity Rate Ratio and Integrated Mass Flux Versus Radial Position 132
185 Predicted AGTI01 Regenerator Discharge Gas Isotherms for the

Cruise Condition Based on Cold Rig Pressure Distortion Patterns 132
186 Predicted AGTI01 Regenerator Local Core Mass Velocities

for the Cruise Condition Based on Cold Rig Pressure Distortions 133
187 Effect of Linear Radial Distortion on Regenerator HP Effectiveness 134
188 Effect of Measured HP and LP Distortion on Effectiveness and

Fuel Flow 134
189 Hot Regenerator Rig Schematic 135
190 Schematic of Helium Seeding Technique for Measuring Regenerator

Leakage 136
191 Steady State Regenerator Discharge Temperature Distributions 137
192 Regenerator Seal Leakage 138
193 Regenerator Drive Torque 138
194 Phase V Regenerator Seal Drive Torque Versus Seal Pressure 139
195 Phase V Regenerator Seals Clearance Test 139
196 AGT101 Regenerator Seal Leakage 140
197 AGT101 Regenerator Core Torque 140
198 AGT101 Drive System, 141
199 AGT101 Transmission System 142
200 VSTC Test Results 143
xii
Title ea e
201 Engine Electronic Control Unit (ECU) 143
202 ECU Logic Diagram 145
2O3 Fuel Control Test Setup (P/N 3601500) 146
204 View A. Four-Point Flexure Strength of ASEA Si3N4_-2 1/2

Percent Y203 150
204 View B. Four-Point Flexure Strength of Ford RM-20, Hipped by

(Contd) ASEA, Cut From Rotor S/N 96 151
204 View C. Four-Point Flexure Strength of Kyocera SN 220M Sintered

(Contd) Silicon Nitride 152
204 View D. Kyocera SN 220M Flexural Stress Rupture 153

(Contd)
204 View E. Four-Point Flexure Strength of Kyocera SN 250M Sintered

204 View F. Kyocera SN 250M, Flexural Stress Rupture 155

(Contd)
204 View G. Four-Point Flexure Strength Comparison of Kyocera 250M

(Contd) SN 250M Cut From Rotors and Fabricated as Test Bar Specimens 156
l
204 View H. Stress Rupture Lives of Kyocera SN 250M From Rotors is

(Contd) Considerably Less Than That Fabricated as Test Bars 157
204 View I. Four Point Flexure Strength of Kyocera SN 270M

(Contd) Sintered Silicon Nitride 158
204 View J. Kyocera SN-270M, Flexural Stress Rupture 159

(Contd) I
204 View K. Four-Point Flexure Strength of NGK SN-73 Sintered
204 View L. Four-Point Flexure Strength of NGK SN-81 Sintered

(Contd) Silicon Nitr!de 161
204 View M. NGK SN-81, Flexural Stress Rupture 162

(Contd)
xiii
TitIe
204 View N. Four-Point Flexure Strength of NGK SN-82 Sintered

204 View O. NGK SN-82, Flexural Stress Rupture 164

(Contd)
204 View P. Four-Point Flexure Strength of N.GK SN-84 HiPped

204 View Q. Four-Point Flexure Strength of NGK SN-84 HIPped

(Contd) Silicon Nitride Cut From Rotors S/N 135 and S/N 136 166
204 View R. NGK SN-84, Flexural Stress Rupture of Test Specimens

(Contd) and Rotors 167
204 View S. Four-Point Flexure Strength of Kyocera SC201 SSC 168

(Contd)
204 View T. Four-Point Flexure Strength of Standard Oil SASC 169

(Contd)
2O5 Interface Conditions for 871, 1149, 1371C (1600, 2100, 2500F)
TIT Engines 170
206 Compatibility Test Condition Ill,Interface Material Combinations 172
207 Interface Test Apparatus 172
208 Static Coefficients of Friction Versus a Normal Load for RBSN and
SASC at Room Temperature 173
209 Friction Versus Relative Motion for Bare SASC and RBSN Under the
Point Contact Condition With a Normal Load of 25 Pounds 173
210 Coefficients of Friction for Oxidized RBSN Point Contact 174
211 Coefficients of Friction for Oxidized SASC Point Contact 174
212 Friction Versus Relative Motion for Bare SASC and RBSN Under the
Line Contact Condition With a Normal Load of 25 Pounds 175
213 Room Temperature Strength RBSN and SASC After Line and Point
Contact Tests 175
214 Calculated Peak Tensile Stresses for II00C Contact Tests of

RBSN and SASC 177
xiv
Title Page
215 RCG Coated and Machined HTP Insulation from Lockheed 180
216 AGTI01 Insulation Test Rig 180
217 Maximum Power and Idle Condition of Lockheed HTP-16 Insulation 181
218 Lockheed HTP-16 Cylinders After Cyclical Flow Testing 181
219 Simulated Rotors Were Used to Demonstrate Thick Section

Fabrication Capability 182
220 Hollow Shaft Design Improves Fabricability 183
221 Example of Incomplete Blade Fill 184
222 Rotor S/N 256 Failed During Spin Test in Mode Similar to
Previous Engine Failure 186
223 Increasing Consistency Evident in Cold Spin Test 187
224 IN907 Ceramic Test Sleeves 188
225 Initial Attachment Method Involved Solid Stub Shaft Rotor 189
226 Installation Difficulties with InitialAttachment Method 189
227 Installation Difficulties Solved with New Method 190
228 Internal Collet Attachment Method 190
229 Typical 3-D Views of the Finite Element Blade Vibration Model 191
230 Comparison of Inspected Z Sections (heavy lines) to TLs

at 2.54 mm (0.i0 inch) From Inducer Backface 192
231 Comparison of Inspected Z Sections (heavy lines) to TLs

at 3.68 cm (2.45 inch) From Inducer Backface 193
232 Campbell Diagram for SN 250M Rotors Showing 19/Rev Stator

Excitation Line 194
233 Consistency of Blade Frequency Response 195
234 Comparison of Holography Results with Predicted First

Exducer Mode 196
XV
Figure Title
235 Comparison of Holography Results with Prediced First

Inducer Mode 196
236 Comparison of Holography Results with Prediced Second

Exdueer Mode 197
237 InitialRotor Orientation Excited Exducer Tips (Bullen Test) 197
238 Rotor Re-oriented to Excite Inducer Tips (Bullen Test) 198
239 Cold Motoring Test Analysis Results 198
240 AGT101 Rotor Finite Element Stress Models 199
241 Maximum Power Thermal Distributions,T 4 = 1371C (2500F) 200
242 Partial Power Thermal Distributions,T 4 = 1227C (2240F) 201
243 Principal Stresses for Room Temperature 115,000 RPM Proof Test
Predicted from No-Fillet Model 202
244 Principal Stresses for Maximum Power, Predicted from

No-Fillet Model 203
245 Comparison of Concurrent and Independent Surface and Internal

Failure Distributions 204
246 Effect of Biaxial Tension on Fracture Risk According to Batdorf's

Criteria 204
247 Rotor Failure Probability for Partial Power Operating Conditions 205
248 Rotor Failure Probability for Maximum Power Operating Condition 205
249 Rotor Failure Probability for Partial Power Conditions Based on

Properties of SN 250M Rotors 206
250 AGTI01 Ceramic Components 207
251 Worst Case Light-Off Cycle (1282C) 209
252 Rig Cycle 871C (1600F) Static Structures 210
253 Effect of Scalloping Outer Diffuser 210
254 Outer Diffuser B1 and B2 Configurations 211
xvi
Title
255 Inner Diffuser Initial Configuration Stress Analysis Results 211
256 Inner Diffuser Scalloped Configuration Stress Analysis Results 211
257 Turbine Shroud/Outer Diffuser Critical Stress Areas

(Static Structures Rig Model) 212
258 Phase B Turbine Shroud Configurations 213
259 Static Structure Rig Light-Off Cycle 214
260 Static Structure Rig Ope1'ating Cycle 215
261 Turbine Shroud Critical Stress Areas (Shroud Screening Rig Model) 217
262 Screening Rig Cycle 13 218
263 Second Model Refinement to Shroud Fillet Radius 219
264 Normal Engine Light-Off, (Cycle 8) 222
265 Ramp Type Engine Cycle II 222
266 Turbine Shroud/Outer Diffuser Critical Stress Areas (Engine Model) 223
267 Turbine Shroud Stress Response for Ramp Type Cycle

(Engine Cycle 11) 223
268 Engine Heat Transfer Coefficients on Turbine Shroud/Outer

Diffuser Model 224
269 Design Change From Type B2 to Type A3 Outer Diffuser 224
270 Shroud Cone to Insulation Radiation 224
271 Shroud Configurations Investigated in Phase C Study 225
272 Development Engine Lightoff (Cycle 17) 228
273 RBSN Shroud Stress for Various Stages of Development 229
274 Transition Duct/Combustor Baffle/Turbine Backshroud Critical

Stress Areas 229
275 Static Structure Rig Light-Off Cycle 9 230
xvii
Title
276 Static Structure Rig Cool-down Cycle I0 230
277 A1 and A2 Configurations Combustor Baffle 232
278 Engine Cycle 15 (Light-Off to I149C Max Power) 234
279 Alternate Configurations of the Turbine Backshroud 237
280 Flow Separator Housing Strain Gage Test Correlation 237
281 The Original and Strutted RBSN Flow Separator Housing FEMS 238
282 Flangeless Flow Separator Housing Outer Cone and

Attachment Configuration (Type XI) 238
283 Flow Separator Housing CriticalStress Areas 239
284 Pressure Deflections at the Regenerator Seal for Types B1

and B2 RBSN Flow Separator Housings 240
285 Pressure Deflections at the Regenerator Seal for Types B1

and B2 LAS Flow Separator Housings 240
286 Pressure Deflections at the Regenerator Seal for Type X1

RBSN Flow Separator Housings 240
287 Steady-State Cruise Radial Temperature Distribution Along

the Flow Separator Housing Crossarm 241
288 Steady-State Cruise Thermal Deflections at the Regenerator

Seal for Type B1 RBSN and LAS Flow Separator Housing 242
289 Steady-State Cruise Thermal Deflections at the Regenerator

Seal for the RBSN Type B2 Flow Separator Housing 242
290 Type B1 Flow Separator Housing Thermal Stress Response

During Lightoff 242
291 Turbine Stator Cutback Vane Trailing Edge Redesign 244
292 Turbine Stator Finite Element Stress Models 245
293 Turbine Stator Finite Element Thermal Model 245
294 Metallic Stator Assembly Thermocouple Locations 245
xviii
u
Title Page
295 Typical Correlation Between Test and Analytical Turbine Stator
Thermal Response 246
296 Turbine Stator Temperature Distribution 20 Seconds After Light-Off 246
297 Turbine Stators Maximum Temperature and MPS Response

(at Mid-Span Trailing Edge) 247
298 Original Turbine Stator MPS Distribution 20 Seconds After Light-Off 247
299 Cutback Turbine Stator MPS Distribution 20 Seconds After Light-Off 247
300 Combustor Assembly Design 248
301 Combustor Assembly Finite Element Model 248
302 Combustor Assembly Steady-State 1371C (2500F) Max Power

Temperatures 249
303 Combustor Assembly Steady-State 1371C (2500F) Max Power MPS 249
304 Engine Start to Cruise (Cycle 22) 250
305 Spacer Sleeve (RBSN) Start Transient MPS 250
306 Nozzle Support (RBSN) Start Transient MPS 250
307 Combustor Support (RBSN) Start Transient MPS 251
308 Combustor Liner (SASC) Start Transient MPS 251
309 Ceramic Components Stackup Load Paths 251
310 First and Second Generation Diffuser Support System 252
311 Inner Ceramic Support Rocker Assembly 253
312 First and Second Generation Rocker/Eccentric Design Comparison 253
313 Slotted Versus Slotless Turbine Shroud Configuration 254
314 High Temperature Ceramic Seal Areas 254
315 Dual Piston Ring Seal Performance Benefit 256
316 Combustor Baffle/Backshroud Wave Spring Location 256
xix
Figure Title
317 Ceramic Component Qualification Sequence 257
318 Strain Gaged Regenerator Shield and Mechanical Screening Rig 259
319 Flow Separator Housing Mechanical Screening Rig 259
320 Turbine Shroud/Rocker Assembly Mechanical Screening Test Setup 260
321 Standard Oil SASC Stator Vanes in the Thermal Shock Test Rig 261
322 Inner and Outer Diffuser Screening Rig 262
323 Transition Duct and Baffle Screening Rig with Original (top)
and Final (bottom) Configurations 262
324 Turbine Shroud Screening Rig Original (top) and Final (bottom)
Configuration 263
325 Static Structure Rig Original (top) and Final (bottom) Configurations 265
326 1371C (2500F) Static Turbine Rig Original (top) and Final (bottom) 267
327 Mechanical Screening Component Test Results 269
328 Outer Diffuser Housing Design Evolution 270
329 Inner Diffuser Housing Design Evolution 271
330 Turbine Shroud Design Evolution 272
331 Thermal Screening Component Test Results 273
332 Turbine Shroud Piston Ring Seals 279
333 Turbine Shroud Piston Rig Seals 280
334 Single LAS Flow Separator Seal Configuration 281
335 Dual LAS Flow Separator Seal Configuration 281
336 Single vs. Dual Piston Ring Seals 282
337 Increased Loading Effect on Single LAS Seal - 5-28-86 282
338 Regenerator Shield Piston Ring Seal Configuration 283
XX
F gure Title Page

339 Regenerator Shield Seal Leakage over Operating Pressure Range 284
340 Cracked Dome On a Bladeless Rotor Casting After Drying 287
341 Part After Sintering Cracked Into Several Pieces Due to

Incomplete Drying 287
342 Green Vane Tip Showing Discolored Area 292
343 Flow Chart Showing Slip Casting Procedure 296
344 Plaster Base and Mylar Ring for Casting Solid Ingot 297
345 Ingot After Casting 297
346 Ford SRBSN Process 300
347 Oxidation Kinetics in RM-20 302
348 Strength After Oxidation of RM-I 302
349 Microstructure of RM-2 303
351 Strength at Temperature Data for RM-1, 2 and 3 304
353 Sintered Silicon Nitride Process for Fabrication of Advanced

Gas Turbine Rotors 306
354 Relationship of Viscosity Vs. Spindle Speed for Various

Concentrations of Dispex A-40 Deflocculant 307
355 Single Dip Wax, Plaster Base and Steel Retaining Ring 309
356 Rotor Pore Size Versus Spin Test Rotor Speed 317
357 Static Oxidation Test 323
358 Cornell Melt Front Advance 325
359 Stator Tool 326
xxi
Title Pa e
360 FillPattern With 0.010 Inch Eccentric, 0.006-0.014-inch

Diaphragm Gate Short Shorts (a)and (b),and Full Shot (c) 327
361 Vane Trailing Edges Typically Formed With Eccentric Gate

Good Fill(a),Knot Line (b)and Cold Packing (c) 327
362 Effect of Material Temperature on Vane Trailing Edge Knit

Line Flaws 328
363 Effect of Tool Temperature on Vane Trailing Edge Cold Packing

Flaws 328
364 Changing FillPattern With Material Temperature Concentric

Diaphragm Gate (0.015 in)Material Temperatures of 200F
(a) and 275F (b) 329
365 Relative Frequency of Vane Cracks as a Function of Material

and Tool Temperature 329
366 Short Hot FillPattern When Using Single Point Gate 331
367 Full Shots Using Point Gate Material Pull-Out in Gate Area (a)
and Solid Shroud Cracks (b) 331
368 Effect of InjectionSequence Timing on Material Pull-Out and

Shroud Cracks - Tool Temperature, 70F 332
369 Effect of InjectionSequence Timing on Material Pull-Out and

Shroud Cracks - Tool Temperature 90F 332
370 Frequency of Cracks Relative to Vane Numbers (Leading Edge) 334
371 Frequency of Cracks Relative to Vane Number (TrailingEdge) 334
372 Plot of Leading Edge Cracks Vs. InjectionTime 335
373 Plot of Trailing Edge Cracks Vs. InjectionTime 335
374 Strength and Thermal Expansion of Coming Code 9458 Lithium

Aluminum Silicate 339
375 Thermal Conductivity of Corning Code 9458 Lithium Aluminum

Silicate 339
376 Modulus of Elasticityof Corning Code 9458 LAS 340
xxii
Figure Title
377 Dimensional Stability of Corning 9458 LAS Material at Various

Times and Temperatures 341
378 Flow Separator Housing Pressure Test Fixture Shown Assembled

to Wooden Model of the Housing 341
379 Ceramic Flow Separator Housing Showing Strain Gages in Place 343
380 Ceramic Flow Separator Housing Mounted to Mechanical Test

Fixture 344
381 Ceramic Flow Separator Housing with Double Wall Bulkhead 345
xxiii
LIST OF TABLES
Table Title Pa e
1 Comparison of Foil Bearing Designs 16
2 Dynamic Properties Rig Test Results 18
3 Single Bearing Rig Evaluation of High Temperature Bearing Coatings 20
4 Steady State Ceramic Turbine Clearance Corrections 25
5 Ceramic Engine Turbine Axial Build Clearance 26
6 Operating Clearance For Ceramic Engine 26
7 AGT101 Impeller Design Aerodynamic Design Summary 47
8 AGT101 Impeller Design Geometric Design Summary 48
9 Impeller Properties 57
10 AGT101 Impeller Life Summary (-30 Material Properties) 61
11 Summary of Vibratory Blade Modes (_,, 0 RPM, 70F) 61
12 Turbine Design Point Cycle Conditions Sea Level, 29C (85F) Day,
Maximum Power 70
13 AGT101 Cold Turbine Rig Operating Conditions 77
14 AGT101 Turbine Development Test Summary 78
15 Dual Alloy Turbine Wheel Design Parameters 82
16 Combustor Data 92
17 Test Sample Temperature Limit 104
18 Regenerator Core Performance 106
19 Regenerator Matrix Thermal Properties 108
20 Calculated Elastomer Temperature 110
21 Leakage Air Flow Conditions 118
22 Flow Separator Housing Axial Distortion 122
xxiv
LIST OF TABLES (Contd)
Table Title
23 Non-Diaphragm Seal Leakage Analysis 123
24 Regenerator Analysis Summary 134
25 Regenerator Rig Test Results 137
26 AGTI01 Component and Material Summary 148
27 Fast Fracture and Stress Rupture Data was Measured for Many
Advanced Si3N 4 and SiC Materials 149
28 Ceramic Interface Test Summary 171
29 Predicted Contact Tensile Stresses For Point- and Line-Contact

Conditions 176
30 Flexure Strength and Fracture Origins of Laser Marked Test Bars 178
31 ACC Bladed Rotor Test Results 185
32 Analytical Results 194
33 Acoustic Blade Ring Results 195
34 Predicted Ceramic Blade Stresses 198
35 Rotor Fast Fracture Predictions 204
36 Phase B Shroud Configuration Study 214
37 Static Structure Shroud Cycle Study 216
38 Turbine Shroud Seal Radius Study 218
39 Material Sensitivityin Screening Rig 220
40 Engine Shroud Cycle Study 221
41 Phase C Shroud Configuration Study 226
42 Shroud Material Sensitivityin Engine 227
43 Transition Duct/Combustor Baffle Analyses in Static Structure 231

Rig Configuration
xxv
Table Title Page
44 Transition Duct/Combustor Baffle/Turbine Backshr.oud Analyses with 233

Rotor Effect
45 Transition Duct/Combustor Baffle/Turbine Backshroud Model Material 235

Study
46 Turbine Backshroud Alternate Configuration Study 236
47 Flow Separator Housing Pressure Stresses 239
48 Flow Separator Housing Steady-State Thermal Stresses 241
49 Flow Separator Housing Transient Thermal Strsses 243
50 Ceramic Component Rig Tests 258
51 AGTI01 Static Components Proof Test Cycles 264
52 Test Cycle for Initial 1371C (2500F) Testing in the Stator/Shroud 276
Screening Rig
53 Yield Point of Slip and Quality of Casting 286
54 AGT101 Rotor Discoloration Analysis 292
55 Discoloration Analysis of a Green Rotor 293
56 Particle Size Variation Within a Green Rotor 294
57 Slip Cast SRBSN Material Properties 301
58 Physical Properties RM-20 Material 305
59 Improved RM-20 Slip Preparation Procedure 307
60 Summary of Heat Treat Cycles and X-Ray Diffraction Results 312
61 Mean MOR Results for Heat-Treated Rotors 312
62 Life Improvements Resulting from Furnace Redesign 314
63 AGT RM-20 Rotor Spin Test Results 315
64 Flexural Strength Of Pressureless Sintered and

HIPped RM-20 Rotor Material 317
xxvi
Table Title Page

65 The Effect of Heat Treatment On Flexural Strength and
Secondary Phases In RM-20 Rotor Material 318
66 Stress Rupture Test Results RM-20 Rotor Material 320
67 The Effect of PSZ Addition to RM-20 Si3N 4 Material 322
68 FilletCrack Inspection Results 333
69 FilletCrack Inspection Results - Variables 336
70 Variable Investigationusing 0.015 X 3.0 Gate 337
71 Material Temperature Investigation 0.065 X 3.0 Gate 338
72 Stators Delivered to Garrett 338
73 Physical Properties of Coming 9458 LAS 339
74 AGT Flow Separator Housing Quality (Period 1980 - 1985) 345
xxvii
This Page Intentionally Left Blank
xxviii
1.0 SUMMARY
I.I Powertrain Development ponents of the system were all of a proven

technology.
The powertrain of the AGT101 is com-
prised of the AGTI01 power section, a split The Reference Powertrain Design (RPD)
path Variable Stator Torque Converter (VSTC), utilized the standard rear wheel drive layout
and a Ford-built four-speed Automatic Over- common to the automotive industry. A front-
drive (AOD) Transmission. The AGTI01 is wheel drive configuration was also shown as
intended for use in a 1364 kg (3000 pound) feasible.
vehicle, and is capable of a nominal 74.5 kw
(100 hp) and a specific fuel consumption of Performance of the RPD was projected
0.18 kg/kw-hr (0.3 ib/hp-hr). It employs a based on tested component performance and
radial turbine wheel and a centrifugal com- as-designed leakage flows, heat losses, and
pressor, mounted on a single shaft and sup- pressure losses. CFDC mileage was projected
ported by an air foil bearing and a high speed at 5.5L/100 km (42.5 mpg), and a 49 krpm idle
ball bearing. Maximum rotational speed is speed satisfies the driveability criterion of
I00 krpm, and idle speed is approximately 2.62 m (8 feet) displacement in 2 seconds'
time.
50 krpm.
The design of the AGT101 is based upon 1.2 Power Section Development
the use of high technology structural ceramics
Engine Mechanical Development
in the hot gas flowpath. Symmetry of the
ceramic structures was a primary design con- The gas lubricated foil bearing is located
sideration, as it provided more uniform at the hot end of the turbine rotor, between
stresses throughout each part and eased the the compressor and the turbine wheel. The
fabrication of components. It also made anal-
self-generated gas film requires no external
ysis by finite element methods a much simpler pressurization, and the bearing is tolerant of
task, as most work could be done using two- high temperatures as well as a degree of rotor
dimensional models.
misalignment and distortion. Once developed
for stability and load capacity, foil bearing
A continously variable transmission (CVT)
temperature capability and material compati-
with high reduction was required for use with bilty were developed. The bearing configura-
a single shaft turbine engine in an automotive tion in use at the end of the AGTI01 program
application. Power transmission is through a employed gold coated foils running against a
VSTC coupled with a high reduction planetary sintered silicon nitride (SSN) journal (the rotor
gearbox. This assembly is bolted direct to a shaft). This configuration offers excellent
Ford four-speed AOD transmission. Offsetting load capacity (24 ib at 75,000 rpm) as well as a
speed changes in the transmission and the low wear rate.
planetary gearbox produce a continuously vari-
able drive. The VSTC was tested early in the The rotating group support system also
program to generate a performance map, and employs a circular thrust runner, added to the
gearbox differential ratios were selected to engine configuration to increase damping and
operate the VSTC in the most efficient areas. control subsynchronous rotor whirl. This
Activities related to the development of the thrust runner is part of a pressurized oil-
power transmission portion of the powertrain lubricated thrust bearing. The compressor end
were deferred for industry sponsorship when of the rotating group is supported by a high
project funding levels were reduced and the speed ball bearing, and sealed by floating ring
project was redirected. Note that the com- seals.
i
Operational clearances were measured combustor performance, flowpath leakage, and
during metallic engine operation, and then regenerator durability.
corrected to determine ceramic engine clear-
ances. While some performance data was obtained
at 1149C (2100F) using the metallic (dual
Engine Rotor Dynamic Development alloy) turbine rotor, performance testing of
the all-ceramic configuration was limited by
During the course of the engine develop- the material properties of the rotor. The
ment, dynamic instabilities in the rotating performance data taken was used primarily to
group were addressed by isolating the rotor assist in seals development.
from gearbox excitations, inserting an oil film
thrust bearing into the design, and by making One seal area recognized as critical was
the tieshaft integral with the gearbox input that between the turbine shroud, flow separa-
quill shaft. These modifications allowed con- tor housing, and transition duct. The high
sistent operation to full speed of 100 krpm. temperatures in this area dictate the use of a
ceramic sealing element. The initial concept
Engine Development Evaluations for this area, which employed two seals sepa-
rated by a ceramic wave spring, was even-
The AGT101 test bed was developed in tually modified to three sealing elements with
three consecutive configurations. The first an increased axial load.
was metallic, with only the regenerator core
and regenerator seals employing ceramics. 1.3 Component/Subsystem Development
This configuration was used early in the pro-
gram for system development to 1600F Tur- Activities centered around the develop-
bine Inlet Temperature (TIT), and for rotor ment of the following major components and
dynamics, controls development, heat transfer subsystems: compressor, turbine, combustor,
coefficient determinations, rotor system eval- and regenerator.
uations, and regenerator development. This
configuration was also used to test a ceramic The compressor system consists of a single
turbine wheel to 97 krpm, and was employed stage backward swept wheel with 12 full and
as a proof test of ceramic turbine wheels prior 12 splitter blades, a three stage radial dif-
to their installation in the all-ceramic engine. fuser, and a set of 17 variable inlet guide
vanes (VIGVs). One design iteration from the
An intermediate configuration incorpo- original design yielded a compressor with the
rated ceramic structures in the hot gas flow- required ratio and flow, diffuser performance
path, but retained the metal turbine wheel and close to predictions, and a desirable efficiency
used a metal combustor with a thermal barrier envelope for the automotive application.
coating. This configuration was used to verify Although considerable work was done with
the design of the ceramic structures, and to Alcoa on the development of a forged powder
obtain performance data at 2100F. A success- metal aluminum compressor wheel, the backup
ful 100-hour endurance test was also run with titanium impeller was used for engine testing,
this configuration. due primarily to time constraints.
The third major_ iteration featured a ce- The AGTI01 turbine wheel is a single stage
ramic combustor and turbine rotor as well as radial inflow design, specifically for ceramic
ceramic structures. Operation during the materials. The turbine wheels used inthe 871C
AGTI01 program was limited to 1204C (1600F) and I149C (2100F) configurations
(2200F), due to material limitations of the utilized the ceramic design, but these metals
turbine rotor. Long term steady-state opera- wheels were cut back at the inducer tips to
tion at 2200F was demonstrated, although fur- reduce stress. Turbine stators were made in
ther development was indicated in the areas of two basic configurations, with an integral ring
2
fabricated of metal used successfully in the Testing was also done in a hot rig to
lower temperature 871C (1600F) configura- measure leakages in the regenerator system,
tion, while 19 individual segments were determine matrix effectiveness, measure core
required to make the higher temperture drive torques, and measure the cavity pressure
ceramic stators survive. at the core ID to determine the bleed orifice
size. It was seen during this testing that while
For the 1149C (2100F)design, a dual alloy the effect of core clearance on leakage at low
turbine wheel of MAR-M 247 and Astroloy was pressures was small, at higher pressures the
employed, although its use was discontinued leakage was very sensitive to clearance.
after a reliability proglem with the bond joint
betweenhub and blade ring was identified. 1.4 Ceramic Materials and Component
Development
The ceramic rotor was redesigned during
the course of the project to improve fabri- Ceramic Materials Testing- In the last two
cability. The changesmadeto the blade shape years of the AGT program, significant testing
did not significantly affect the efficiency of was done to characterize the advanced sin-
the design. tered silicon nitride and silicon carbide mate-
Combustor development looked at both rials. The high temperature properties of
variable and fixed geometries as a means of several candidate rotor materials were deter-
achieving the requirements of low emissions, mined to be inadequate to meet the rotor
multi-fuel capability, and good fuel economy. requirements. Data from as-fabricated test
The fixed geometry was determined to be bars showed adequate properties in four mate-
more feasible, for its lack of moving parts, rials: SN-82 and SN-84 from NGK-Locke, and
and at program's end, a redesign incorporation SN 250M and SN 270M from Kyocera. The
SN-82 and SN 270M materials could not be
a larger primary burn zone was favorably
evaluated. Nozzle development emphasized formed into the complex rotor shape, and
good atomization and reduced heat load, with testing of the material properties of the SN
a simplex design incorporating air cooling, air 250M rotors revealed significantly lower ma-
assist, and a center-body ignitor as the terial properties than indicated by the initial
favored arrangement at the close of the pro- test bar data. Similar testing of SN-84 rotors
ject. also showed lower properties than test bars,
but the reduction in properties was not as
Regenerator development was carried out significant, and so the SN-84 material appears
at Ford motor Company and at Garrett. The to be suitable for 1371C (2500F) engine test-
regenerator cores, supplied by Corning and by ing.
NGK-Locke, exhibited performance approach-
ing RPD goals. Critical to the successful Although large improvements have been
development of the regenerator system is the made during the AGT program, additional
design of regenerator seals. The performance work needs to be done to achieve better ele-
of these seals has been improved on a static vated temperature properties, fracture tough-
basis over the course of the program, ness, and shape capability.
approaching AGT goals, but performance and
durability have been hampered by distortions Ceramic Compatibility Testing-Stress and
at engine operating temperatures. Further temperature conditions at various engine
development of these seals if required before interfaces were identified, and a series of
RPD goals can be successfully met. tests was performed using test bars to evalu-
ate the sticking characteristic at these inter-
Activities at Garrett included studies to faces. In those cases where sticking was
determine the pressure profiles of the system, observed, retest showed that no sticking
evaluate means of controlling flow, and occurred when flame-sprayed mullite coating
predict effectiveness. was used at the interface.
Other testing was performed to investigate yses using a finer substructure model mesh
the effects of line contact and point contact were performed.
at component interfaces and to evaluate coef-
ficients of friction for reaction bonded silicon Stresses were analyzed for 3 modes of
nitride (RBSN) and sintered alpha silicon car- loading encountered during normal engine
bide (SASC). Also, a study was performed to operation; one pressure (full power) and 2
determine causes for spalling in RBSN, and thermal (steady-state and lightoff transient).
fabrication techniques were successfully Shutdown transient conditions were initially
modified to reduce the occurrence of spalling. considered but were insignificant due to core
thermal inertia.
Identification of components by laser
marking of serial numbers was introduced fol- Rig Design/Development
lowing a study which showed little or no
strength degradation in most materials. Only Major structural components, such as re-
lithium-aluminum-silicate (LAS) was unsuited generator shield, flow separator housing, tur-
due to excessive damage to a legibly marked bine shroud and rocker assembly components,
surface. are first tested in mechanical rigs to simulate
the pressure and contact loading encountered
Ceramic Structures in engine operation. These tests then are
followed by a proof test in thermal screening
Initial design criteria for AGT101 compo- rigs to simulate the maximum principal
nents were established empirically, with com- stresses encountered in engine operation. The
ponent testing providing feedback to the thermal cycles for the screening rigs are de-
analytical process and material selection. fined so as to simulate stresses which exceed
Component symmetry was pursued in the basic predicted engine stresses under normal start
design to eliminate circumferential element by 25 percent.
analysis. This reduced cost, and allowed for
rapid analysis and extensive design iterations. Screening tests are intended to eliminate
The only exception is the flow separator components with gross internal defects and/or
housing, whose function required an asym- smaller defects lo_ated in regions of high
metrical design. stress. Lower quality components will frac-
ture during screening, allowing only the high-
To eliminate extensive design optimization est quality parts to continue through qual-
on each material/component combination an ification and engine test. Results of these
evaluation was performed on the preliminary tests can also feed back into the empirical
configuration for each component to identify design loop. Significant examples of this are:
primary and alternate materials. Subse- the inner diffuser housing, which was rede-
quently, design optimization was performed signed from a shape similar to a birdbath to a
using a prime material candidate for each shape more like a dinner plate following fail-
component. After design optimization was ures in the screening rigs; the elimination of
achieved, alternate candidate materials again slots from the turbine shroud flange following
were evaluated. mechanical screening tests in which failures of
RBSN shrouds initiated at those slots; and the
Because of the lack of ductility in redesign of the turbine shroud to reduce a
ceramics, localized stress concentrations can- stress concentration in the seal fillet radius
not plastically redistribute stresses. Conse- following repeated thermal screening failures
quently, known stress concentrations must be with origins in that radius.
reduced or eliminated from the design wher-
ever possible. When component analysis The static structures rig test utilizes all of
revealed significant tensile stresses near the ceramic components in an assembly of the
known stress concentrations, additional anal- engine structure, without the rotor, and
checks the mechanical functioning of the developed and used for long-term at 1371C
assembly. Further, the static structures rig (2500F) to investigate high temperature ef-
evaluates the sealing capability, component fects on ceramics. To date, limited sticking
capability, axial and radial pilots, contact at 1371C (2500F) has been seen, and no failure
loading, and stresses at elevated tempera- has occurred due to sticking. Some interfacial
tures. The maximum operating condition for distress was seen between the transition duct
this rig was set at I093C (2000F), which and combustor baffle, but this can be elimi-
corresponds to the maximum operating tem- nated by using an alternating material stack in
perature of the turbine exhaust. Sufficient this area.
confidence was eventually gained in the
thermal and mechanical screening rigs that
A cold seal leakage rig was also developed
use of the static structures rig was discon-
in which the performance of three ceramic
tinued as a qualification rig.
sealing areas could be evauated: turbine
To evaluate ceramic components upstream shroud/flow separator (piston rings), regen-
of the turbine exhaust, another rig, capable of erator shield/flow separator (butt joint), and
operation of 1371C (2500F), was required. The regenerator shield/exhaust housing (piston
initial 1371C (2500F) milestone was achieved rings). The cold leakage data allowed a com-
twice in the stator/shroud screening rig in the parison of seal concepts, and measured the
spring of 1984, for short durations. A separate effects of load, concentricity, and pressure
rig, the 1371C (2500F) static rig, was later distortion on these sealing interfaces.
5
2.0 INTRODUCTION
This report is the final in a series of o Emissions less than federal standards
Technical Summary Reports for the Advanced
Gas Turbine (AGT) Technology Development o Ability to use a variety of fuels
Project, authorized under NASA Contract
The Garrett/Ford automotive Advanced
DEN3-167 and sponsored by the DOE. This
Gas Turbine has been designated the AGT101.
report has been prepared by the Garrett Tur-
The AGT101 nominally is a 74.5 kw (100 shp}
bine Engine Company (hereinafter referred to
as Garrett), and includes information provided engine, capable of speeds to 100,000 rpm and
operating at turbine inlet temperatures (TIT)
by the Ford Motor Company hereinafter
referred to as Ford), Standard Oil Engineered to 1370C (2500F) with a specific fuel con-
sumption level of 0.18 kg/kw-hr (0.3 pounds/
Materials Company (Standard Oil), and
horsepower-hour) over most of the operating
AiResearch Casting Company (ACC). The
project was administered by Mr. Thomas range.
Strom, Project Manager, NASA-Lewis The program was initiated in October 1979
Research Center, Cleveland, Ohio. This
and was to entail 68 months. Two phases were
report summarizes plans and progress from planned; Phase I was 45 months long and
October 1979 through June 1987. involved most of the component/ceramic tech-
nology effort and the early engine and vehicle
testing. The 23-month Phase II was scheduled
Project effort conducted under this con- to include completion of the all-ceramic
tract is part of the DOE Gas Turbine Highway
engine effort plus the vehicle deliveries and
Vehicle System Program. This program is evaluation.
oriented at providing the United States auto-
motive industry the high-risk,long-range tech- Budget reductions in mid-FY81 resulted in
nology necessary to produce gas turbine a re-orientation of the program. The vehicle,
engines for automobiles that will have reduced transmission and engine gearbox/transmission
fuel consumption and reduced environmental efforts were deleted. The program was then
impact. The intent is that technology result- oriented toward developing the long-range
ing from this program be capable of reaching high-risk technology of the AGT101 gas tur-
the marketplace by the 1990s. bine such that the auto industry can carry that
technology forward to production in the
1990's. Emphasis was placed on ceramics, gas
The advanced gas turbine, when installed in bearings, low emission combustion and im-
a Ford vehicle with a 1364 kg (3000 lbs) inertia proved component performance. The AGT101
weight would provide: gas turbine was envisioned as a test bed in
which to develop these technologies.
A combined federal driving cycle (CFDC) The program schedule with milestones is
fuel economy of 5.5 L/100 km (42.8 miles shown in Figure 1. Key milestones accom-
per gallon) based on Environmental Protec- plished include completion of the design
tion Agency (EPA) test procedures using review in January 1980, the first set of
DF-2. The AGT-powered vehicle shall give ceramic structures passing rig test to 1093C
substantially the same overall vehicle (2000F) in August 1983, ceramic structures
driveability and performance as a compar- engine test to 1149C (2100F) in February 1984,
able production vehicle powered by a con- first ceramic structure rig test to 1370C
ventional spark-ignition powertrain system (2500F) in May 1984, first all ceramic engine
6
test in January 1986, and completion of an 85- This final report summarizes the power-
hour all ceramic engine test to 1204C (2200F) train design, power section development, and
in October 1986. While the feasibility of component/ceramic technology development.
applying ceramics to a gas turbine was demon- Appendices include reports of progress from
strated, further improvements in the ceramic Ford, ACC, and Pure Carbon. Work performed
turbine rotor material properties were needed by Standard Oil is published in NASA topical
for further testing. report CR-180871.
=ISCAL YEAR
GOVERNMENT MILESTONES
CERAMIC DEVELOPMENT
STRUCTURES
TURBINE
REGENERATOR DEVELOPMENT
COMPONENT DEVELOPMENT
POWER SECTION DEVELOPMENT
GOVERNMENT 1. DESIGN REVIEW

MILESTONES 2. CHARACTERIZE ENGINE - BUILD 1
3. CHARACTERIZE ENGINE WITH CERAMIC STATICS
4. INITIATE ENGINE TEST WITH CERAMIC ROTOR
Figure 1. AGT101 Program Schedule.
7
3.0 REFERENCE POWERTRAIN DESIGN
This reference powertrain design (RPD) is pressure side of the ceramic rotary regener-
comprised of the AGT101 power section, split- ator. This feature provides increased thermal
path variable stator torque converter (VSTC) efficiency by minimizing heat loss. The par-
gearbox and Ford automatic overdrive (AOD) tially heated air passes through the regener-
transmission as depicted in Figure 2. The RPD ator core where it is further heated to approx-
is to be installed in a 1364 kg (3,000 Ibs) imately I060C (1940F) temperature (at idle),
inertia weight vehicle. and then to the combustor.
The AGT101 engine, shown in Figure 3 is The combustion system features a fixed
flat rated at 74.57 kw (100 hp) with a mini-
geometry, lean-burn, low-emission combustor.
mum SFC of 0.18 kg/kw-hr (0.3 Ib/hp-hr). The
Combustor air temperature increases to a
single-shaft rotating group is composed of a maximum of 1370C (2500F) with the addition
radial turbine, centrifugal compressor, and of fuel, is ducted to the ceramic stator, and
output gear supported by an air-lubricated
then expanded across the radial turbine rotor.
foil-journal bearing and an oil lubricated ball Turbine exhaust gases 1093C (2000F) maxi-
bearing. The maximum (steady state) engine mum temperature at idle) are ducted through
speed is 100,000 rpm, and idle speed is approx- the low pressure side of the rotary regenerator
imately 50,000 rpm. and out the engine exhaust at a maximum
temperature of 266C (510F) at maximum
Ambient air enters the engine through
power.
variable inlet guide vanes and passes through
the single-stage compressor. The compressed
air, at approximately 185C (365F), is routed All ceramic hot section structural compo-
around the full engine perimeter to the high nents are symmetrical except for one housing
687_TB-tM
Figm'e 2. AGT101 RPD Powertrain.
8
while the second path passes through the
planetary ring gear to a VSTC utilized as a
speed changing device. The variable output
speed of the VSTC is connected to the carrier
via a clutch that is open during vehicle idle
and closed during normal driving modes.
Speed ratio modulation across the VSTC
results in a variable output speed of the plane-
tary carrier, which in turn is geared directly
to the AOD transmission. Step ratio changes
in the AOD are phased with offsetting speed
ratio changes in the variable speed gearbox.
This produces a continuously variable drive"
line system which allows the engine to operate
at optimum speeds to deliver high efficiency
cruise power, acceleration power, or engine
braking power.
Preliminary packaging studies were com-

pleted for incorporating the RPD in Ford rear-
wheel "and front-wheel drive configurations.
93310-I
The rear-wheel drive configuration is the
887.-078-157
reference powertrain design (Figures 5 and 6)
Figure 3. AGTI01 Power Section. and can be installed in a 1364 kg (3,000 Ib)
inertia weight vehicle without modifications
that separates high and low pressure regenera- to the engine compartment structure. The
tor flow. The symmetrical design provides a same confuguration with a modified gearbox
more uniform stress distribution and eases and transmission was repackaged and shown to
fit a front-wheel drive vehicle.
manufacturing of the components.
Performance of the RPD is based on tested

Output power from the gas turbine is deliv- component performances and as designed leak-
ered to the vehicle via a variable high speed age flows, heat losses and pressure losses.
single-stage differential planetary gearbox in Using this data and analytical modeling this
conjunction with a modified four-speed Ford engine/gearbox-transmission and vehicular
AOD transmission. This combination results in
characteristics the projected combined federal
a continuously variable speed transmission driving cycle (CFDC) mileage and 2-second
required for the optimization of a single-shaft distance traveled versus idle speed is shown in
gas turbine vehicle system. Figure 7. As shown, an idle speed of 49,800
rpm satisfied the 2.44 m (8 feet) in-2 seconds
The planetary gearbox (schematically distance criteria for acceptable driveability.
shown in Figure 4) incorporates a differential The resulting fuel consumption is 5.5 L/100 km
in which power is split into two paths. Gas (42 mpg). The engine performance char-
turbine power is input at the planetary sun acteristics are shown in Figure 8 and depict
gear and extracted from the planetary carrier. the outstanding specific fuel consumption
One power path extends directly to the carrier (SFC).
9
KEY
Op:DIAM.RAL
PITCR ___ ACCESSORY
OH_VE
N = RPM RANGE T
T: GEAR
TEETH
INUMDERI
AOD = AUTOMATIC OVERDRIVE "aTr_+64T ,= 2.?. ,6,3
i 31T OVERRUNNINGCLUTCH
PLANETARY --: _ N = 1716 - 3684
CARRIER 74T VSTC
N =0- 65 8
, ,p..%"_l'lL,..j]___....
N1 : 55K-
I)p lOOK 35T MAIN OIL
_,,,,o_ .L"_ L_I P°"'

Op = 12
75T
AOO
AO0 INPUT SHAFT N = 0 - 2564
ISR = 0.85)
REARAXLE RATIO
/
PLANETARYRATIO = 11.2727 128 Dp]

X RATIO = 1.9176 112 Op]
Z RATIO = 1.04167 112 Op]
GB7.078-158
Figure 4. AGT101 Drive System Schematie.
10
_F POOR QUALITY.
687_7_150
Figure 5. AGTI01 Installation - Side View.
_,7_7,,= Figure 6. AGTIO1 Installation - Front View.
ll
LIFO0 km IMPGI ILB/HP-HR
0.70.
44
5.4 Ko/kW-HR
0.40" 065
5.5
43 0._'
0.35-
0._'
5.6
0.30 • 0._"
_ 5.7
0.43.
41 _ 0.25-
0.8
0._'
5.9
IFTi
020-
0.35"
0._-
\
0.15- 0.25
!
0 40 80 I00 120 KW
OUTPUT SHAFT POWER
!o -_
_ 3 G6/-O78.T62A
8 Figure 8. AGT101 Specific Fuel Consumption.
4 I I ! I
46 48 50 52 54
IOLE SPEEO,KRPM
Figure 7. AGT101 Estimated Performance

Curves.
12
4.0 POWER SECTION DEVELOPMENT
Power section development on the AGT101 the design of the ceramic structures, and to
concentrated on several key areas: gas foil obtain performance data at 1149C (2100F). A
bearing development, engine rotor dynamic successful 100-hour endurance test was also
development, and engine development evalua- run with this configuration.
tions. Development of the major components
of the power section is discussed in Section 5.0 4.1 Enffine Mechanical Development
- Component and Subsystem Development.
4.1.1 Foil Bearing
Experimentation with several different
foil, coating, and backing spring materials and The AGTI01 foil bearing design was re-
configurations eventually yielded a successful, quired to meet the following operating condi-
robust design in which gold coated foils run tions:
directly against a sintered silicon nitride jour-
nal (the rotor shaft). Operating speed 50,000 to i00,000 rpm
Maximum load 6 g for 0.2 seconds
Dynamic instabilities in the rotating group Maximum temperature 649C (1200F)
also required considerable effort to overcome.
In order to dampen the system and help to The operating range shown above repre-
control subsynchronous rotor whirl, the system sents the actual engine operating range, how-
as finally developed employs a circular thrust ever during engine starts and stops, the engine
runner as part of a pressurized oil-lubricated must operate at speeds of less than 50,000 rpm
thrust bearing. This and other modifications, and pass through 2 rigid body rotor critical
including the integration of the rotor tieshaft speeds. Additionally, the bearing breakaway
and the gearbox input quill shaft, finally al- torque must be below the engine starter stall
lowed consistent operation to full speed of 100 torque.
krpm.
In addition to the above, this application
The AGT101 test bed was developed in has imposed additional requirements on the
three consecutive configurations. The first foil bearing which are unique relative to other
was metallic, with only the regenerator core foil bearing applications developed at Garrett:
and regenerator seals employing ceramics.
This configuration was used early in program High unit loading at 1-g load. This bearing
for system development to 871C (1600F) tur- has a 30 kPa (4.3 psi) unit loading with a
bine inlet temperature (TIT), and for rotor ceramic turbine rotor or 39 kPa (5.6 psi)
dynamics, controls development, heat transfer with metal rotor. Previous high tempera-
coefficient determinations, rotor system eval- ture foil bearings were subjected to unit
uations, and regenerator development, this loads in the 7.5 to 14.5 kPa (I.I to 2.1 psi)
configuration was also used to test a ceramic range.
turbine wheel to 97 krpm, and was employed
as a proof test of ceramic turbine wheels prior Rotor excursions can exceed the calculated
to their installation in the all-ceramic engine. bearing sway space. The bearing may not
be hydrodynamic when the first critical
An intermediate configuration incorpo- rotor speed occurs.
rated ceramic structures in the hot gas flow-
path, but retained the metal turbine wheel and The AGT101 foil bearing was first devel-
used a metal combustor with a thermal barrier oped for stability and load capacity. This
coating. This configuration was used to verify work was performed through a combination of
13
analysis and testing in the single bearing rig. using the strain energy approach. The gas film
Bearing stiffness and damping were evaluated pressure distribution is then calculated based
during rotor dynamic development when minor on the deflected profile of the foils and shaft
changes in the bearing geometry were ef- speed. Convergence is achieved when the gas
letted to optimize these characteristics. film thickness matches the deflected foil pro-
Temperature capability and material compati- file for the same pressure distribution.
bility were developed once the basic geometry
was established. This work utilized the foil Figure 10 shows the model output for the
coating wear rig as well as the single bearing deflected height for the baseline bearing
rig to evaluate potential foil coatings. lightly loaded at i00,000 rpm. Take note that
The resulting foil bearing configuration is

shown in Figure 9 and described in the t.able
below: ARRIER
Number of foils 7
Foil thickness 0.15 mm (0.006 inch)
Foil arc length 29.2 mm (1.150 ihch)
Foil axial length 26.5 mm (1.045 inch)
Foil preform radius 29.2 mm (1.150 inch)
Foil material Inco X750
Backing spring web 0.14 mm (0.0055 inch)
thickness
Backing spring arc 13.3 mm (0.525 inch) FOIL
length
Backing spring 27.9 mm (I.I00 inch)
RACKING SPRING
preform radius
Backing spring Inco X750
material
BASIC OEARtNG CONFIGURATION
Bearing sway space 0.15 - 0.25 mm
6|7-0711-183
0.006-0.010" inch
Foil coating Au-Ni or Au-Co Figure 9. Basic Bearing Configuration.
Journal surface Sintered Silicon
Nitride 28 ¸
Journal diameter 34.3 mm (1.350 inch)
24
*Diametral clearance between the journal and
aE 20
foils with the foils and springs making all i,,,,,-
contacts in the carrier. 16
4.1.1.1 Foil Bearing Geometry Development •'-' 12,

w,,
Initial design of the foil bearing was con- 8

ducted analytically using a computer program
that was developed by Garrett using IR&D
funding. This parametric model was used to 0
study the influence of changing various bear- 0 I0 20 30 40 50 60 70 80 90 i00
ing geometric parameters. This analysis pro- ANGULARPOSITION.OEGREES
|lT_T&154
vides the simultaneous elasto-hydrodynamic

solution for an operating foil bearing. The Figure 10. Foil Bearing Deflected State
profile of the deflected foil is determined at 100,000 RPM Operating Speed.
14
most of the bearing deflection is taken in the ing torque is measured via a torque arm and
curvature of the foils and not in the backing load cell attached to the test bearing housing.
spring. This is because the backing spring is
used to help the bearing sustain high loads and Compatibility of the foil bearing with the
the model only simulates a lightly loaded con- rest of the rotor system was evaluated in the
dition. Figure 11 shows the pressure dis- rotor dynamics rig and in the engine test beds.
tribution for that bearing at operating speed. This work is discussed in the rotor dynamics
This model proved useful for establishing an discussion in Section 4.2.1.
initial bearing configuration and for screening
proposed bearing geometries before fabrica- Table 1 shows the progressive development
tion and testing. of the foil bearing geometry through the pres-
ent.
i .28
The "P" type foil is compared with the
1.24 reference bearing in Figure 12. This bearing
featured a redesigned backing spring and foil.
,.,.+1.20
The polyimide foil coating on the P type
1.1e bearing provided temperature capability up to
343C (650F), which was a 38C (100F) improve-
,_ !.12 ment over the temperature capability of the
I,,4,4
I',,I Teflon S coating used in the reference foil.
-" 1.08
=' 1.04 4.1.1.2 High Temperature Foil Bearing

Development
1.00
High temperature foil bearing development
0._ built on the basic "P" type bearing design.
Several different foil/journal coating systems
0.92
0 10 20 30 40 50 60 70 80 90 I00 were researched and evaluated to develop the
bearing soakback capability to 649C (1200F)
ANGULAR POSITION,OEGREES
and maintain the stability and load capacity
GII7-076-1_
goals. This effort resulted in a hardened gold
Figure 11. Film Pressure Distribution alloy foil coating that was compatible with
at Operating Speed. either a bare silicon nitride ceramic shaft or a
metallic shaft sputter coated with titanium
During geometry development of the foil nitride.
bearing there were two types of testing, static
spring rate testing and dynamic testing for Foil bearing surface coating development
stability and load capacity. Static spring rate was conducted according to the diagram in
testing was utilized to characterize the load/ Figure 13. During this process, a candidate
deflection of a stationary foil bearing assem- coating/journal combination was evaluated in
bly. The static spring rate, while not precisely the dynamic properties test rig. The dynamic
duplicating the dynamic spring rate, is the properties test rig, also called the foil bearing
benchmark parameter used in predicting bear- wear rig is shown schematically in Figure 14.
ing behavior. This rig tested the dynamic wear and friction
of the candidate surfaces at 427, 538, and
The Single Bearing Rig is propelled by a 649C (800, I000, and 1200F). After elimi-
small air turbine motor. This rig can spin the nating combinations with unacceptable wear
test shaft up to 110,000 rpm and impose static or friction, the single bearing rig was then
radial loads up to 222.4 N (50 Ibs) on the test used to evaluate surviving candidate systems
bearing at ambient temperatures. Test bear- for load capacity, power consumption, and
15
Table 1. Comparison of Foil Bearing Designs.
Bearing Reference Inereue Increase "P" Type Current

Parameter Bearing Web Thick FoU Thick Bearing Design Bearing
Teflon S Teflon S Teflon S Polyimide Au/NI

Foil Coating
Foil Material 302 CRES 302 CRES 302 CRES IN 718 IN X750
Foil Thiek, mm (inch) 0.15 (0.006) 0.15 (0.006) 0.178 (0.007) 0.178 (0.007) 0.15 (0.006)
Foil Length, mm (inch) 30.2 (I. 19) 30.2 (1.19) 30.2 (1.19) 30.2 (1.19) 26.4 (1.040)
Foil Radius, mm (Inch) 21.0 (o.85) 21.6 (0.85) 21.6 (0.85) 27.9 (1.100) 29.2 (1.150)
Spring Web, mm (inch) 0.127 (0.005) 0.15 (0.006) 0.15 (0.006) 0.15 (0.006) 0.14 (0.0055)
Spring Radius, mm (Inch) 17.78 (0.70) 20.3 (0.80) 20.3 (0.80) 22.9 (0.90) 27.9 (1.10)
302 CRES 302 CRES 302 CRES IN 718 IN X750

Spring Material
Journal Coating METCO 443 METCO 443 ME?CO 443 ME?CO 443 TiN or SSN
Sway Space, mm (inch) 0.279 (0.011) 0.259 (0.0102) 0.259 (0.0102) 0.203 (0.008) 0.203 (0.008)
Static Spring Rate, NN (lb/in) I000 188T 2358 2400 2400
Load Capaeity, N (It)) 53.4 (12) at 88.9 (20) at 88.9 (20) at 88.9 (20) at 106.7 (24) at
100,000 rpm 100,000 rpm 100,000 rpm 100,000 rpm 75,000 rpm
Turbine Vibration, mm (mil) 0.076 (3) at 0.038 (1.5) at 0.02 (0.8) at 0.02 (0.8) at 0.013 (0.5) at
100,000 rpm 100,000 rpm 100,000 rpm 100,000 rpm lO0,OOO rpm
PRESENT DESIGN ORIGINAL OESIGN
FOIL HOUSING-_h_ _ FOIL HOUSING-_
,Nco
x.,50
5U,3THATE ,'7, ,
_MATEH,AL
iM,L I I' [ /
TH,HNER JVPol (I'I'_ _/t'_1
FOIL I COATING:
LARGER PREFORM HI HER
:::_S
RAOmUS __/ ON 30_
TEFLON SS
COATING _ _--_j ,.
'_/_,
POLYIMIOE IIOOF SUBSTRATE
| TEMPERATURELIMITI ,,,. _ _-_._---r-_._,,_j \ i\ '___,,,,p / D-TYPE
l , LASER WELDED _ I I -- \
'NCO 719 MATERIAL _ _ .- -- ".,,. _,_1,_,. , _ j,.
TALLER, THINNER RIBS • \_ _" --\ [ _ I / I _ /
THINNER. WIDER WEBS _ J _ . I I • _ l_ I [•1 l
OETWEEN
SPR,NOS ,,8s " \ ix. I.L /
mAROER RADIUS
PREFORM j ....__ _ _ I _ I ,'1_ ",'f I / ",. /
50 PERCENT GREATER ,," /'_.,_,,_ % %J ' I- I_ } _ TACK WELD -- "'" -- I"
TOTAL THICKNESS %., _ -... _ \ / , _ _ ........
302 SS/
GeZ-OTS-ie7 MATERIAL
Figure 12. Foil Bearing Designs Compared.
16
OBJECTIVE:SURFACECOATINGSANO DISK ALSO TRANSLATES AXIALLY ON
MATERIALS FOR BEARINGS AXIS OF REVOLUTION. AMOUNT OF FORCE
TO 1200F
AT CONTACT SURFACE IS CONTROLLED. FLAT ROTATING OISK
_ t "JOURNAL" COATING APPLIED
[ CANDIDATECOATINGS] ONTACT SlOE
t [ DYNAMIC WEAR RIG TEST AT: ROOM TEMP
1 } 538C IIOOOFi
STEP I 427C
649C {{120OFi
80OF)
(
"FOIL" SPECIMEN SUPPORTED _ ,, ...... _... -
ON AIR BEARING AND
.'_R!N.-__N_-'- rLA, ",O,L" SPECIMEN
TOROUZ TO COONI[IIAC! _ _'I_TH C_]_'_'NG APPLIED
BOTATION MEASURED
SELECT • FOIL COATINGS COMPATIBLE

NOTE: TEST RIG ENCLOSEO IN OVEN FOR ELEVATED
WITH Si3N4 JOURNAL
IIIIT_T&IS TEMPERATURE TESTING
• FOIL/JOURNAL COATINGS
FOR METAL JOURNAL
Figure 14. Foil Bearing Dynamic Properties

Wear Rig.
and hot engine environments. Specimens were

t SINGLE BEARING RIG AMBIENT TEST FOR:
STEP • LOAD CAPACITY examined for coating wearthrough, excessive
2 ) • POWERCONSUMPTION
abrasion, and material transfer. Testing was
• DURABILITY
automatically stopped on a coating system
when the dynamic friction increased dramati-
cally, evidence of surface breakdown. The
matrix of coating systems evaluated are shown
t APPLY TO TEST BEARINGS QUALITY ANO MECHANICAL in Table 2.
STEP
3
i I FOR PERFORMANCEEVALUATION
N AN ENGNE INTEGRITY EVALUATION
Only three of the systems evaluated in the
dynamic wear rig showed acceptable wear and
friction. These systems were Au/SSN, Au/
TiN, and TiC/SCA systems. The photograph in
Figure 15 shows the Au/TiN test specimens
SELECT • I FOIL COATING after completion of the test series. The strip
FOR Si3N4 JOURNAL
chart in Figure 16 shows the speed and friction
• 1 FOIL/JOURNAL
COMBINATION FOR traces for a typical cycle for this material
METAL JOURNAL combination. The TiC/SCA coating system,
6DT,071i-I
M
initially developed for other Garrett high tem-
perature applications, was used as the baseline
Figure 13. Foil Bearing Surface of comparison for all coating systems evalu-
Coating Development. ated.
start/stop durability. Once coating systems The test procedure for single bearing rig
passed the single bearing rig evaluation, they evaluation of the candidate foil bearing sys-
were evaluated in an actual rotor system, tems was as follows:
either in the rotor dynamics rig or in an
engine. The dynamic properties wear rig tests Ten no load starts to 50,000 rpm. Measure
simulate the non-hydrodynamic operation of a bearing static friction before and after
foil bearing during starts and stops with cold starts
17
Table 2. Dynamic Properties Rig Test Results.
Foil Coatings
Yttria
IN?18 Stabilized Titanium Chromium
Journal Coatings Gold (Au) Boronized Zireonia (YSZ) Carbide (TIC) Oxide
Sintered SUieon Good atall Not tested Excessive abra- Not tested Not tested
Nitride (SSN) temperature sion at ambient
Titanium Nitride Good atall Not tested Not tested Not tested Not tested
(TIN) temperatures
1090 Chrome Coating failure Not tested Not tested Not tested Not tested
at 1000F
Karamite Good at all Not tested Not tested Not tested Not tested
temperatures
Yttria Stabilized Not tested Exeessive Not tested Not tested Not tested
Zireonia (YSZ) abrasion
at
ambient
Kaman SCA Not tested Not tested Not tested Good at all Coating
temperatures failure
O Ten l-g load starts to 50,000 rpm. Mea- perties rig testing. Table 3 summarizes the
sure bearing static friction before and resultsof this testing.
after starts
The initial testing with the METCO 105
O Remove bearing and visually examine for (plasma sprayed Alumina) and METCO 202
damage. If there is no damage, reinstall (plasma sprayed Yttria Stabilized Zirconia)
the bearing and measure static friction coated shafts with the soft and then with the
hard gold coatings gave a lead that the hard-
o Measure power consumption at loads from ened gold foil coating might be compatible
l-g to the maximum load attainable for 2 with the high density silicon nitride ceramic
seconds or more shaft. Subsequent testing in the dynamic
properties rig and then in the single bearing
o Remove bearing and examine rig proved this combination to be the optimum
system.
Replace bearing foils and shaft if neces-
sary. Measure static friction. Run 1000 1- The rig testing of the SCA/TiC coating
g load starts. Measure static friction,then system required some manipulation of the foil
examine bearing. bearing geometry to achieve the 85.3N (19.2
lb) load capacity. Initial bearing geometry
Single bearing rig testing was used to eval- was the same as the "P" type bearing de-
uate surviving candidates from dynamic pro- scribed in Table 1, however the load capacity
18
C,,,_:}E,:A:b P,'_'-, ,,
the desired load capacity. This geometry

change was carried over to the Au/SSN bear-
ing configuration.
The successful development of the Au/SSN

high temperature bearing system demon-
strated that bearing load capacity could not
only be maintained in making the transition
from low to high temperature foil coatings,
but also increased. This improvement was the
result of the use of a "thick" coating with a
relatively "soft", more compliant material
such as gold.
4.1.2 Thrust Bearing Development

90569-1
The AGT101 hydrodynamic (oil film) thrust

bearing was added to the engine configuration
to increase the available damping in the rotor
system and control subsynchronous rotor whirl.
A similar system was used successfully in
Garrett's CCGT25-1, which was a closed cycle
gas turbine engine with a straddle mounted
foil bearing rotor support similar to the
AGT101.
The drawing in Figure 17 shows the bearing

crossection used in the AGTI01 engine. A cir-
cular thrust runner on the engine shaft inboard
of the existing ball bearing. The engine rotor
thrust loads are transferred from the runner to
an oil pressure lubricated thrust washer that is
fixed to the housing. A splash lubricated
reverse thrust washer supports the runner in
the event of reverse thrust caused by com-
pressor surge or engine failure.
The ball bearing outer race is retained in a

90589-2
hydraulic mount and is restrained axially by
GB7_7&170
two springs, fore and aft. These springs pinch
Figure 15. TiN Coated Ring (Top) Versus the outer race with a 89N (20 Ib) load to
Au Coated Disk (Bottom) After 591 Starts prevent spinning. During engine operation, an
at Ambient Temperature and 4.3 psid Load. additional 178N (40 Ibs) of axial load is trans-
mitted through the ball bearing as rotor thrust
with the geometry and the SCA/TiC coatings compresses the forward bearing spring to pre-
system was only 58.2 N (13.1 Ibs). Bearing foil vent bearing skidding. The amount of this
thickness had to be decreased to 0.127 mm preload is controlled by the axial thrust bear-
(0.005 inch) to provide compliancy to attain ing clearance.
19
TIME--
I SEC 2 SEC 3 SEC 4 SEC
.._...._.__ _ . .
2000
; : ; i : : ! ] : i !"
O. .... " : ' ' ' : i _ I ' '_ ...... ' i t"
w,J
mL
[ ' ii :." _ : [:::'I -;b: :,. t .-.:':-'t-" :_::-::.::_= ,::::_.'-t:'l_.: .," "t',: "['-=,: :-.-:'T'-:_-:"V'-I
I ; ; ; , _ ; [ : ', [ ; ', , ', ',
1.0
I
! : ! ,"!: :".-]-_BREAKAWAYAND:.:;_::--
i ! :_ TRANSiTiON
TO _-]:_:-
r.
W,.
Z
RESIOUAL
TORQUE-4
IOOESNOTAFFECT_
....... _-_'-' _:-'1",- _--_ _i ABSOLUTE TORQUE_
M.#
:
.
;
: "
t -I..#.:::t,:-44;-I:I-jLE:=I:-=4_OYNAMIC:_-"
:." ?" :flT':._--:|::..'f_"_ to:_T':.: |:¢DIPTIflW:T'___---
-: -::I VALUES_L
__._
_4:'-_-'-,-----.-[-- ..!'_
0 --,
; 1 I i I ' ', I I I I _ F _
617-070-171
Figure 16. Typical Results of Dynamic Frietion Test.
Table 3.. Single Bearing Rig Evaluation of High Temperature Bearing Coatings
Foil/Journal System Durability Test Load Test
SCA eoated shaft, TiC sputter Heavy wear and high friction 85N (19.2) lbs for
eoated foils after 263 starts !2 see at 75,000 rpm
Sintered silicon nitride shaft, Unmeasurable wear after 1000 10?N (24) lbs for
>200 KNOOP gold eoated foils starts, rig failed after 2934 30 see at ?5,000 rpm
starts due to shaft failure
PVD TiC on shaft, >200 KNOOP Applied to dual alloy turbine

gold coated foils rotor and run on rotor dynamic
rig. Despite high sTnehronous
motion due to bladed eomponent_
bearing showed minimal wear
20
OIL SUPPLY PASSAGE
SECTIONA-A
THRUST WASHER
TYP 6 PLACES
THRUST RUNNER
THRUST WASHER
!
Figure 17. AGT101 Hydrodynamic
Thrust Bearing Configuration.
4.1.2.1 Tapered Pad Thrust Bearing Design
The oil film thrust bearing is a 6 pad, GB7-078-173

tapered step design. Design details are shown
in Figure 18. The design thrust load is 20
percent higher than that determined in the Figure 18. Thrust Bearing Design Detail.
thrust analysis to provide a reasonable margin
of safety. Inlet housing supply 551.6 to 1241 kPa
pressure (80 to 180 psig)
temperature
To assure proper alignment, the interface
seat surfaces of the washer and the thrust dependent
plate assembly are made spherical. This con- Metering orifice design 448 kPa (65 psig)
figuration will permit the thrust washer to pressure
align exactly flush with the mating thrust
Design inlet temperature 82 C (180F)
surface of the thrust rotor. This arrangement
also will sustain the alignment through small Maximum oil discharge 121C (250F)
displacements of the housing or the shaft temperature
during thermal excursions in engine operation.
Oil flow (Thrust bearing 4.16 L/rain
only) (1.1 gpm)
The thrust bearing is designed to utilize
Total oil flow (Thrust 5.3 L/min
ATF-Dextron II lubricant which conforms to
bearing plus hydraulic (1.4 gpm)
SAE Specification SAE-J-311B. The oil supply
mount)
is fed through the same drilled passages used
in the original AGTI01 configuration. From In the static condition the thrust runner is set
the main supply duct, oil is metered through a up with axial clearance between both forward
2 mm (0.080 inch) orifice to the bearing and reverse thrust washers. As the engine
through six holes in the thrust washer. The accelerates during a start and forward thrust
hydraulic mount is also fed from the main builds, the rotor assembly moves forward,
supply duct in parallel to the thrust bearing. compressing the spring which preloads the ball
Oil supply parameters are shown below: bearing. The oii film thrust bearing is not
21
{HPI
actually loaded until forward rotor movement KW
7.
decreases the forward bearing clearance suf- 5-
ficiently to develop the supporting oil film.
The curve in Figure 19 shows the rotor thrust 6 o
curve and the ball bearing/thrust bearing load

share as a function of rotor speed.
4_
MIN AXIALCLEARANCE
: _504 iN 5"
200"
FLUIDFILM
150" 4"
ffi
_50-
50.000 lO0.OOO
RPM 3"
0 __,,._._//////////////////////_ 2-
_50 BALL
REAHING
RANOLES
I. HALL
GEARING
AND
FLU'O
FILM
GEAR'N8
_!
FORWARD THRUST | SHARE FORWARD THRUST - 2-
NOTES:
I-
I. AXIAL CLEARANCE ' 0.004 IN
2. SPRING
RATE6250 LD/IN
GI/_T_It4
Figure 19. Thrust Bearing/Ball Bearing

0
Load Share Characteristic.
SPEEO, KRPM
4.1.2.2 Thrust Bearing Power Consumption
GOT.G78-175
Evaluation
Figure 20. Total Analytical Heat
Theoretical bearing power consumption is Generation for AGT101 Fluid Film
shown in Figure 20. An engine test was Thrust Bearing Cavity.
performed to evaluate this power consumption
and determine whether bearing scavenging was supporting evidence for the theory that exces-
sufficient. These test results,shown in Figure sive oil in the bearing cavity produced insuffi-
21 indicate that the apparent penalty due to cient scavenging and caused oilchurning.
the thrust bearing is much higher than pre-
dicted. It is believed that the higher test 4.1.3 Ball Bearing and Compressor Seal
results are an indication of insufficientscav- Desi
enging and the resulting churning that is 4.1.3.1 High Speed Ball Bearing Design
taking place.
The AGT101 ball bearing is a 15 mm 202
Another test was performed to evaluate series ball bearing designed to carry a maxi-
the effect of reduced oil flow on thrust bear- mum predicted radial load of 182N (41 lbs) and
ing power consumption. This test was con- a thrust load of approximately 1335N (300 lbs)
ducted at 70,000 rpm and the oilsupply to the at I00,000 rpm.
thrust bearing was separated from the hy-
The bearing was designed to meet the
draulic mount so that reduced thrust bearing
following conditions:
supply pressures would not affect the hydrau-
lic mount damping or stiffness. The results of
Speed 55,000 to I00,000 rpm
this test,shown in Figure 22, indicated that at
partial speed, the power consumption reduced Radial loads 42.3 to 182 N
substantially with reduced oil flow. This was (9.5 to 41 Ibs)
22
(HPI
KW
Thrust loads 311 to 1335N
3.0"
(70 to 300 lbs) 2.2,
2.9"
Life CFDC 3900 hours at 2.|-

2.8"
variable speed and 2.7-
load. I00 hours at
2.5-
2.0"
100,000 rpm and 1.335
0= L9,
kN (300 Ibs) 2.5"
2.4"
1.8-
Lubricant Automotive transmis-
2.3-
sion fluid 1.7- ENGINE SPEED: 70.000 RPM
LUBRICANT: MOBIL ATF 110

2.2"
LUBRICANT INLET TEMP: 66C (15OF]
Lubricant temperature 38 to 166C (100 to 1.8-
2.1
300F)
2.0 j i i i |
iGPMI 0.5 0.7 0.9 I.I 1.3
!
L/MIN
When the hydrodynamic thrust bearing was 2 3' 4' '5
OIL FLOW
added to the system midway through the pro- G|7_TLIT7
gram, the ball bearing axial load was limited

to 178N (40 Ibs). This reduction in bearing Figure 22. Thrust Bearing Power
load significantly increased bearing life be- Consumption Decreases with Oil Flow.
yond the design goal of 3900 hours.
4.1.3.2 High Speed Ring Seal Design
iHPI
KW
The high speed ring seal for the AGT101 is
14- shown in Figure 23. The seal has a double
10- 0 = HEAT GENERATION floating carbon ring design with provisions for
buffer air between the elements. Initial shaft
12-
clearance was designed at 0.013 mm (0.0005
inch) diametral, however after testing the
8 C, carbon consistently wore to 0.038 to 0.064 mm
10 _
(0.0015 to 0.0025 inch) diametral clearance.
Testing in the rotor dynamics rig and in the
engine revealed no significant leakage despite
8- this wear.
4.1.4 Rotor Clearance Evaluation

6- The rotor axial clearances in the AGT101
4
are critical from both a performance and
LOAD WASHER WAVE SPRING
4 _
117-TPHi (17-7PHi
OUTER SHELL
RETAINING RING
2_
MATING RING 1410/ (43,401
i I I I I
1440C1
"_f "__', _
5O 60 70 80 90 1O0
SPEEO, KRPM
G87_7_tT8
Figure 21. AGT101 S/N 001 Build 21

L'--
Gi?_7_17i
Fluid Film Thrust Bearing
(Test 4-19-84). Figure 23. Floating Ring Seal.
23
durability standpoint. Rotor clearances as- hub, and static structures. At a constant idle
sembled into the engine will vary depending on speed where pressure effects are minimal, the
the effects of thermal and pressure distortion turbine clearance decreased by as much as
on the rotor support structure and on the rotor 0.46 mm (0.018 inch).
shrouds. These effects were measured during
actual engine testing in the metal engine test Once thermally stabilized, the turbine
bed using capacitance probes mounted in the clearance changed very little as a function of
rotor shrouds with metallic rotors. speed. The curve in Figure 25 shows the
variation of axial turbine and compressor
4.1.4.1 Metallic Engine Clearance Testing clearance as a function of speed. The com-
pressor axial clearance decreased by 0.30 mm
Running clearances were measured on the (0.012 inch) as a function of engine speed.
compressor and turbine rotor during metallic Half of this change was attributed to "flower-
engine operation. These data included clear- ing" of the compressor rotor due to centrifugal
ance changes during low speed operation due effects and the other half was due to pres-
to light-off thermals and clearance changes as sure/thermal distortion of the adjacent struc-
a function of engine speed under essentially tures.
isothermal operation at 871C (1600F). The
engine build clearance is established when the Based on these test results, the build clear-
engine centerline is vertical and the support
ances for the compressor on the metal and
bearings have no radial static load. When the
ceramic engines was established on a go for-
engine running clearances are measured the ward basis. The rationale for establishing the
engine centerline is oriented horizontally, and
compressor clearance is shown below:
the rotor supports are radially loaded by the
weight of the rotor assembly.
Compressor Axial Clearance
The metal turbine clearance was strongly mm (inch)

affected by the start from a cold initial state.
The curve in Figure 24 shows how the turbine Thermal/Centrifugal 0.30 (0.012)
clearance changed as a function of time after
Rotor Droop Allowance 0.025 (0.001)
the initial engine start. This change is a
function of the rotor droop effect in combina- Worst Transient 0.051 (0.002)
tion with thermal effects on the rotor blades,
Worst Excursion 0.038 (0.0015)
Build Tolerance 0.025 (0.001)
IMILS)
O" Minimum Running 0.0?6 (0.003)

MM
CLEARANCETRANSIENT FOR COLD Clearance
Minimum Axial Build 0.52 (0.0205)

-0.I_"
Dimension
-I0.
_ -0.2N"
START TO OSITE
5O.OOO RPM, 871C [1600F)
OATAI
-15"
-0.38t - 4.1.4.2 Ceramic Turbine Clearance
-2'0 ' Analysis
The measured turbine running clearance

"0.535 " .25 | I I I
0 values from the metallic engine test noted

TIME, MINUTES above were corrected by comparing the calcu-
GII7@IEI-I/g
lated thermal growth of the metallic struc-
Figure 24. Typical Turbine Axial tures under the engine test conditions to the
Clearance Transient for Cold Start. ceramic structure thermal growth at the
24
desired operating condition. Further correc- The ceramic turbine build clearance was
tions were made for pressure induced struc- calculated based on the start effect noted
tural deflections resulting from differing during metallic turbine testing, plus similar
speedconditions. Table 4 showsthese correc- allowances used for establishing compressor
tions. build clearance. This information is sum-
marized in Table 5.
[MILS)
O- After the turbine build clearance was
MM established, corrections for running clearances

in normal engine operation were calculated.
On the basis of these corrections, the axial
operating clearances of the ceramic engine
were predicted. These values are shown in
-0.250 -
D -!0- _-_ COMPRESSORKNEE Table 6.
-0.125 - COMPRESSORAXIAL
..... TURBINE AXIAL
_' -0381 -15- STEADY STATE AT 871C (1600F)TIT. The analysis summarized above is con-
MAX TDT servative since it does not reduce rotor droop
E
allowance to compensate for the reduced mass

im _ _ Imiml _ Immm' amiJ inn' _ _imi, mmmD
-0.508 -20-
of the ceramic turbine rotor.
4.1.5 Metallic Structure Design

-0.635 -25
5o 1_
The initial design of the AGT101 metallic
ROTOR SPEED. RPM x 10.3 structures was consistent with the standards
667_7_160
and practices of the automotive industry. The
Figure 25. Compressor and Turbine design emphasis was on simplicity for en-
• Clearance Loss Due to Thermal Growths. hanced producibility and cost effectiveness.
Table 4. Steady State Ceramic Turbine Clearance Corrections.
TIT = 1371C
TIT = 1174C TIT = 1281C (2500F)
(2146F) Idle (2337F) Cruise Maximum Power
Parameter Evaluated 50 krpm 65 krpm 100 krpm
Pressure deflection -- +0.0076 mm +0.043 mm

(+0.3 Mils) (+I.7 Mils)
Thermal growth of turbine +0.033 mm +0.274 mm +0.323 mm

shroud and roeker assembly (+1.28 MiLs) (+10.8 Mils) (+12.7 Mils)
Thermal growth of compressor -0.023 mm -0.099 mm

backshroud (-0.9 Mils) (-3.9 Mils)
Thermal and centrifugal -0.038 mm -0.036 mm -0.053 mm

growth of turbine rotor (-1.5 Mils) (-1.4 Mils) (-2. I Mils)
Thermal growth of compressor +0.0127 rnm +0.0127 mm +0.0076 mm

(+0.5 Mils) (+0.5 Mils) (+0.3 Mils)
Net axial clearance change +0.299 mm +0.236 mm +0.213 mm

(+11.8 Mils) (+9.3 Mils) (+8.4 Mils)
25
Symmetrical design philosophy was used engine structural assembly. The other compo-
throughout the engine, and low cost, highly nents, the compressor housing, exhaust housing
producible materials were used for the metal- and inlet housing, are the main structures in
lie components where possible. the engine. These structures contain the pres-
surized gases within the engine, provide burst
The engine cross section in Figure 26 indi- containment for the rotating components, sup-
cates the metallic engine components in the port the engine at the mounting interfaces and
provide a reacting load path for the pressure
Table 5. Ceramic Engine Turbine loads generated within the engine.
Axial Build Clearance.
The design of the compressor housing and
exhaust housing was augmented by finite ele-
mm [ (inch) ment analysis to provide a stable support for
the engine ceramic and rotating components.
Thermal/Centrifugal/ 0.036 (0.015) Section stiffeners and gussets were designed
Rotor Droop into these components to minimize pressure
induced deflections.
Worst Speed Transient 0.0762 (0.003)
Worst Excursion 0.127 (0.005) 4.2 Engine Rotor Dynamic Development
Build Tolerance 0.051 (0.002) 4.2.1 AGT101 Rotor Dynamic History
Minimum Running 0.127 (o.oo5)
The initial AGT101 rotor system was sup-
Clearance
ported by an oil mounted ball bearing and a
Minimum Axial Turbine 0.762 (0.030) foilbearing. This initialrotor system configu-
Clearance ration, which includes a metal turbine for
Table 6. Operating Clearances For Ceramic Engine.
Reference Ceramic Ceramic

Metal Engine Shroud Axial Baekshroud
Engine Operating Condition Axial Clearance Clearance Clearance
mm (inch) mm (inch) mm (inch)
Build 0.762 (0.030) 0.762 (0.030) 0.508 (0.020)
Transient after light-off to 0.381 (0.015) 0.381 (0.015) 0.584 (0.023)

idle
649C (1200F) steady state idle 0.3048 (0.012 w m M
871C (1600F) steady state idle 0.645 (0.0254) 0.320 (0.0126)
1174C (2146F) steady state idle 0.605 (0.0238) 0.361 (0.0142)
1280C (2337F) steady state high 0.541r (0.0213) 0.424 (0.0167)

cruise
1371C (2500F) steady state 0.518 (0.0204) 0.447 (0.0176)
26
COMBUSTOR SEPARATOR
HOUSING
TURBINE
DIFFUSERS
RBINESTATORS
ROTOR
INLETHOUSING
COMPRESSOR
HOUSING
GB7-O78-181A TURBINE
SHROUD
Figure 26. AGT101 Metal Structured Power Section.
development purposes, is shown in View A of the frequency caused by rotor unbalance) and
Figure 27. The operational rotor system con- increased as the engine was loaded.
figurations with metal and ceramic turbines
are respectively shown in Views B and C of A waterfall plot of engine vibration during
Figure 27. acceleration, shown in Figure 28 clearly shows
the existence of gearbox excitation frequen-
Rotor thrust loads were transmitted to the cies.
support structure through the oil mounted ball
bearing. Rotor torque was transmitted to the A design was then introduced to isolate the
drive train through a sun gear mounted on the gearbox from the AGT101 rotor in order to
rotor shaft. The sun gear meshes with 3 eliminate gearbox excitation frequencies that
planet gears on a compound planetary gearbox may be causing rotor instability. To accom-
to provide a speed reduction of 37.143 to 1. plish this, the gearbox was separated from the
rotor by a splined quillshaft with the sun gear
Alignment and timing of the rotor mounted mounted on itsown bearings.
sun gear to the gearbox planet gears was found
to be important. Misalignment and mistiming Subsequent testing showed the near elimi-
contributed to the inability to repeat early nation of gearbox excitation frequencies.
successful engine testing. However, a subsynchronous frequency of large,
pulsating amplitude was present. As the rotor
Engine testing of the initial AGT101 rotor system speed increased, the amplitude of this
system showed that full speed operation could subsysnchronous frequency increased, causing
not be achieved because of high amplitude foil bearing failure consistently between
motion at the foil bearing that caused the 75,000 and 85,000 rpm.
rotor system to go unstable. This high ampli-
tude motion was subsynchronous in nature The cause of the detrimental subsynchro-
(meaning that its frequency was lower than nous rotor system excitation was identified
27
through a series of engine motoring tests. The placed by dummy components systematically.
engine was powered externally using a high F_gure 29 shows the sequence of rotor systems
speed air turbine starter so that the engine evaluated. As each of the aerodynamic corn-
bladed aerodynamic components could be re- ponents were replaced, the rotor instability
___ESSO# --'-->, .......
VIEW A
O.,_L
s.-- o,_.o..tEo / I _/ t
VIEW B
OIL FILM _ FOIL

T,,us, _._..// -,,_,- COLLET
-.L,.E. O,L
.0.... .E.,. / // \ / / J \
........ -
"_ COUPLIN6
uletl-zl VIEW C
Figure 27. AGT101 Engine Rotor System Configurations.
28
PLANET 2 X PLANET GEAR UNBALANCE
RING GEAR GEARS RESPONSE
RESPONSE RESPONSE
RESPONSE
i
2OO0
O FREQUENCY.HZ
697-078-189 A
Figure 28. AGT101 Engine Initial Rotor System Configuration Response.
th{'eshold speed increased. This trend is shown foil sway space, foil materials and foil coat-
in the curve in Figure 30. The conclusion ings were varied. Impact testing was also
drawn from these engine motoring tests is that conducted to determine the foil bearing damp-
detrimental rotor system excitation was aero- ing at operating speed.
dynamically induced.
Impact testing consists of striking the
Since aerodynamic excitation will always rotor system in the radial direction while it is
be present, an investigation was made of ways rotating, as shown in Figure 31. The rotor
to improve the stability of the rotor system. responds primarily in free vibration of the
first mode. The decay rate of rotor subsyn-
Testing indicated that rotor system motion ehronous whirl is an indication of the damping
during unstable operation was conical whirl in the rotor supports. The results of the
with motion at the foil bearing. A detailed impact testing showed that the foil bearing
investigation of the foil bearing was under- has a relatively low viscous damping coef-
taken to determine if the stiffness and the ficient of approximately 2.0 Ibf-sec/in.
damping of the foil bearing, which affect rotor
stability, could be significantly improved. By extending the length of the foil bearing
and reducing the sway space, the instability
Tests were conducted in which foil bearing threshold speed was increased by I0 percent.
length, foil thickness, backing spring stiffness, However, foil bearing stiffness and damping
29
o.. 0.0500- 2.00-1
o,.
E ,_: BUILD 45 !
could not be significantly improved while
. ,.oo.-,RL
RASELINEABEoRB
° o:o°,,o
E
0.0375-
maintaining an acceptable foil bearing thermal
environment. Therefore, a more stable rotor
1.2 T_
system would have to be obtained by improve- _ 0.0250-
ments in other areas. i "_'_ E>: BUILD 45C I
,_ j_._,_o BOMM,
,UBB,NE
The stability of the rotor system was im-
_ 0.0125. "_MMY COMPBE
'- 0.111111. ".®_ , ,_ ,_----?,
proved by changing the rotor support con- 60 85 70 75 DO85 _o 95 ,oo
figuration. The ball bearing was moved for-
ENGINE SPEED, RPM x 1000
617.078-191
Figure 30. AGT101 Engine Aerodynamic

Components Excite Subsynchronous Motion.
ACCELEROMETER
27.0 mm 11.0625 INI \

STEEL BALL \
BASELINE CONFIGURATION-- 45 POINT WHERE STEEL _ _,
BALL STRIKES _ r_
DUMMY ROTOR "_ .,.k_-"
i
CERAMIC TURBINE WHEEL -- 45A
0iLLMOEUANT
_D OEFORILN
G l I
IIII?-OTO-ZTA "_
Figure 31. AGT101 Rotor Dynamics Rig

Impact Test Configuration.
DUMMY TURBINE ROTOR -- 456
ward (toward the gearbox) and an oil film

thrust bearing inserted to provide additional
rotational stiffness and damping to the rotor
system. In addition, interference fit plating
was added to the quill shaft spline closest to
g
g the oil mounted ball bearing to cause the
splined connection at the sun gear to act as an
DUMMY TURBINE AND COMPRESSOR -- 45C
additional rotor support.
G87-078-190 A
Figure 29. AGT101 Engine Motoring Testing showed that this new configuration
Rotating Group Configurations Tested. was sensitive to gearbox-to-engine alignment,
30
but that careful assembly yielded consistent, 4.2.3 Rotor System Test Results
stable operation. The resulting operational
rotor system configuration is shown in Fig- The current rotor system configuration
ure 32. with the ceramic turbine has been tested
through the entire operating speed range and
4.2.2 Current Rotor System Configuration its rotor dynamic behavior recorded.
The current rotor system configuration is The ist critical speed occurs at approxi-
similar to the operational rotor system con- mately 12,000 rpm and the 2nd critical speed
figuration with the metal turbine except that is not seen because it is highly damped. A 3rd
it has a ceramic turbine with a collet attach- critical speed exists at approximately 82,000
ment. To simplify assembly, the plated rpm. This 3rd critical was seen in the test
splined quillshaft was removed and the tie- data between 70,000 and 90,000 rpm. Testing
shaft extended into the gearbox as shown in has demonstrated that these critical speeds
Figure 33. The extended tieshaft has the same are of no concern to the operation of the
stabilizing effect as did the plated splined AGTI01 current rotor system configuration.
quillshaft. The 3rd mode in the current rotor system
ORIGINAL, PACF, IS
DE POOR QUALITY
THRUST REARING
Figure 32. AGT101 Engine Operational Rotor System Configuration.
31
/
THRUST BEARING
\
\
SHAFT EXTENSION
GB7-078-193
Figure 33. AGT101 Engine Current Rotor System Configuration.
configuration is caused by the relatively low the bending critical speed would occur near
rotational stiffness of the toilet which has 82,000 rpm, which is the bending critical speed
axial slots cut in it for assembly. Also, the seen in current engine testing.
3rd mode rotor response at the foil bearing is
different in the horizontal and vertical direc- 4.2.5 Rotor System Analysis
tions indicating that the rotor system is whirl-
ing in an elliptical orbit. A rotor dynamic analysis was performed in
order to analyze the rotor system critical
4.2.4 Rotor Static Modal Determination speeds and the response of the rotor system to
Testing unbalance.
Static modal tests were performed on the The rotor dynamics model used in anal-
current rotor system configuration to verify yzing the current rotor system is shown in
analytical prediction and to correlate with Figures 34 and 35. The relatively low rota-
engine test results. tional stiffness of the collet is due to the axial
slots in this part and was calculated from
Static modal testing showed that a bending simple beam bending theory. The rotational
critical speed does exist in the operating speed stiffness near the curvic coupling was specifi-
range. If gyroscopic stiffening were addedp cally chosen such that the predicted bending
32
COMPRESSOR TURBINE
MASS: 0.0423 KG-SEC2/m t0.00237 LB/SEC2/INJ MASS: 0.0704 KG-SEC2/m _J0.00304 LO/SEC2/IN)
Ip: 0.000031 m-KG-SEC [0.0027 IN-LB-SEC ) Ip: 0.000041 m-KG-SEC 10,00354 IN-LB-SEC J
ID: 0.000026 m-KG-SEC2 [0.0225 IN-LB-SEC2) In: 0.000071 rn-KG-SEC
2 10.0_12 IN-LB-SEC2)
ECC: 0.0127 turn I0.0005 iN) ECC: 0.0127 mm (0.0005 IN)
A: 00EGREES /"'T'N
A: 180 DEGREES _
MASS MOOEL
r----f
r"----'--T-1 ' I
STIFFNESS MOOEL
STATION
NUMBERS
f J I I I I [ 1 I I
INCH -1.00 0.28 1.57 2.85 4.14 5.42 6.71 7.99 9.27 10.56
cm .2.54 0.04 3.98 7.24 10.52 13.77 17.04 20.29 23.55 26.82
AXIAL POSITION
TERMINOLOGY
= POLAR MOMENT OF INERTIA

GB7..O78-29A
= DIAMETRAL MOMENT OF INERTIA
ECC = ECCENTRICITY RAOIUS OF DISCRETE MASS
A = CIRCUMFERENTIALANGLE OF UNBALANCEECCENTRICITY
Figure 34. AGT101 Engine Rotor Mass and Stiffness Model.
critical speed matched the observed bending 4.2.6 Rotor Stability Analysis
critical speed of 82,000 rpm.
The dynamic characteristic of a stable
The rotor system mode shapes were deter-
rotor is to have low amplitude subsynchronous
mined by an undamped critical speed analysis
motion and the rotor supports are able to
and can been seen in Figure 36. The ist mode
positively dampen subsynchronous motion.
is conical whirl at the foil bearing and the 2nd
Improved stability usually occurs when there is
mode is whirl with more motion at the oil film
an increase in the forward mode log decre-
thrust bearing than at the foil bearing. The
ment, which is a calculated value from the
3rd mode shows the rotor flexing near the
rotor dynamic analysis. In the case of the
curvic coupling because of a relatively soft
AGTI01 with a ceramic turbine and collet
collet rotational stiffness.
attachment, the Ist and 3rd forward mode log
Unbalance response plots of rotor deflec- decrements are important.
tion and bearing loads are shown in Figures 37
and 38 and a whirl speed map is shown in The stability analysis of the rotor system
Figure 39 for reference purposes. Predicted shows that the difference between a stable
rotor displacements at the first and third and unstable AGT101 rotor system configura-
critical speeds are similar to measured rotor tion is predicted by a relatively small log
displacements. Although relatively large dis- decrement.
placements were predicted at the 2nd critical
speed, engine tests did not show large dis- Log decrement values for the Ist and 3rd
placements because the engine accelerated forward whirl modes have been plotted and are
rapidly through this critical and this mode was shown in Figure 40. Damping is present when
highly damped. the log decrement is greater than zero. The
33
TIESHAFT
175.126 kN/nlm _ _._ 112..985 m_N_/RAO

17.513 kN/mm _ ,?r,,_ _,.,,,,,,.. _ _ 112.985
1000m-N/RAD _ 175.126 kN/mm
(lOO.OOOLB/IN]_____ .
.....................
II OOO000 LB/INI _ L_
. < 1
I • .
000
IN-LB/RAOJ
<:_(1.000.000 LB/IN) COMPRE3SOR"'
'S,,N, ' O
ROTOR C)27
_ I, ,,_ IJ3Om-N/RAD
OIL MOUNT _ _ L,,_ 10.000 IN-LB/RAO
'_ I / OILFILM TUR ,NE
NON-REAM MAS.RL-__ 7-)7- THRUST 8EARING
175,513 kN/mm _].

1100.000 Le/INI _ (_ II.O00,O00 IN-LB/RAOi
BALL BEARIN6 ._
AND STRUCTURE,.._.__ F
STATION LOCATION SHEAR STIFFNESS ROTATIONAL STIFFNESS
11 COLLET 1,751,28 kN/mm 8553 m-N/RAO

(100,O00.O00 LD/INi [56,000 IN-LS/RAOI
12 COLLET 1.751,268 kN/mm 59,430 m-N/RAO

IIO0.O00.1XX) LD/INJ 1526,000 IN.LB/RAOI
23 CURVIC 1,751,268 kN/mm 15.948 m.N/RAO

COUPLING (100,000,000 LB/INI 1150,000 IN-LB/RAD)
OIL MOUNT PARAMETERS

GBI_7_30B
MOUNT LENGTH: (0.3 INCH) 7.62 mm

JOURNAL OUTER RADIUS: 10.685 INCH) 1740 mm
RAOIAL CLEARANCE: (0.00235 INCH) 0.0597 mm
OIL VISCOSITY: O.O00001 REYNS. 93C (20QFI
Figure 35. AGT101 Engine Rotor Support Schematic.
current AGT101 rotor system configuration gas flow paths, structural support systems, and
has demonstrated consistent stable operation rotor configurations were also similar, though
to I00,000 rpm. differing according to the specific require-
ments of ceramic and metallic components.
4.3 Engine Development Evaluations
The metallic configuration engine was the
AGT101 engine development testing aug- first operational AGT engine. Featuring all
mented the engine design effort through ver- metal structural and rotating components, this
ification of engine design. This program uti- engine was limited to 871C (1600F) maximum
lized three basic test bed configurations shown turbine inlet temperature. The metallic
in Figure 41, each representing different engine was used principally for system devel-
stages of ceramic engine development. These opment prior to ceramic engine testing. This
test beds were used toward the development engine was used for rotor dynamic develop-
of a high temperature engine configuration ment, engine controls development, heat
capable of achieving the reference powertrain transfer coefficient determination, rotor sys-
design goals. tem evaluation, and low temperature regen-
erator development.
All three engine test bed configurations
were similar in design and function, and all The ceramic/metallic engine was a hybrid
required the same facilitiesand energies for engine with a combination of ceramic and
testing. These test beds shared the same rotor metallic components in the hot power section.
bearing system and regenerator system. The This engine, was an intermediate step in the
34
ceramic engine evolution, featuring ceramic The all ceramic engine featured ceramic
engine structures, metallic combustor and a structures, ceramic combustor, and ceramic
metallic dual alloy turbine rotor. The devel- turbine rotor. The operating temperature of
opment of ceramic structures as part of the this engine was limited by the stress-rupture
engine system was performed in this engine. properties of the turbine rotor material.
MODE 1: CONICAL WHIRL AT THE FOIL BEARING

1.00
TOTAL _ ]
n_ 20 PERCENT COMPRESSOR
_i
_",_._-_- I,oPERCENT o: DUTEH

SHAFT
__ _i_ - ° I S.E. o FOIL BEARING
--- -2.00
/
-0.72 0_14
SE. 4.42
_ 5.71 6.99 8.27 9.50
o_T,ESRAFT
10.84
-0.50 t THRUST BEARING AXIAL POSITION, IN
]
79 PERCENT STRAIN ENERGY {S.E.JIN FOIL BEARINGSUPPORT
MODE 2: PRIMARY WHIRL AT THE BALL BEARING
,oo _ I
/_ I _ -_ o_T,ESRAFT
_ 0.50- 26 ST BEARING FOILBEARING

,_' i
,= 0.50 '_" , , , _ ,
_,, -2 O0 -0.72 0.57 1.85 3.14 4.42 5.71 6.99 S.27_ , 1.56 10._.
"_ I _ TURBINE
| AXIAL POSITION,IN
-+.o.5o
1 ,.,,_
| 46 PERCENT
| 26 PERCENT S.E. IN THRUST BEARINGSUPPORT S.E,
t II PERCENT S.E. IN FOIL BEARINGSUPPORT ROTOR
MODE 3: SHAFT FLEXURE NEAR THE CURVIC COUPLING

1.00
0.50- FOIL REARING S.E.

'
OUTER SHAFT [_ -=TIESHAFT
47 PERCENT ° = OUTER SHAFT
0.00
't
,_ -+.,2 0:5_ ,:B5 s ,AL
pO4SI_S'2' " "
THRUST BEARING
-0.50 COMPRESSOR
12 PERCENT S.E. IN THRUST BEARINGSUPPORT

2 PERCENT S.E. IN FOIL BEARING SUPPORT
GOl-O76-31A
Figure 36. AGT101 Engine Critical Speed Mode Shapes.
35
ROTORSUPPORTS
OIL FILM FOIL

-...,I,.- BALL BEARING THRUST BEARING --I-- BEARING
I
I
I
I
.... I- .................
I
I
I
I
% .... T ........ _
-
I
-{ ......... { .....
I I
I I
l I
I I
F_ .J ........
-t- ........ L_ ........
I
I
I
I I
I
........... I ........ I ........ I- ........

{ I i I I
I I I I I
I I t I
l I I I
....... ........ ...... fi__ , -'--,

I 1 I
I l I
I I I I I
0 { i { I {
0 20 40 60 80 IOO 120 140
RESPONSESPEED. RPM X I0 "3

GD7-07_32
Figure 37. AGTIO1 Engine Unbalanee Response/Support Displacements.
Turbine inlet temperatures of 1204C (2200F) operation, fuel control operation, and rotor
were achieved. This engine was used to eval- dynamic development.
uate the characteristics of all ceramic engine
operation.
Because the ceramic and metallic engines
shared similar operating systems, the metallic
4.3.1 Metallic Engine Testing
engines provided a low risk test bed to evalu-
ate new design features before they were used
The AGT101 metallic engine saw extensive
in the ceramic engine. Examples of this work
use during the first five years of the program.
were the testing of new ECU software, light-
these engines compiled a total of 270 hours
off evaluations of new cornbustor configura-
and 720 starts. This experience was accumu-
tions, the evaluation of a new regenerator
lated during two types of evaluations, for
drive control, and evaluations of a new engine
engine mechanical development, and for ce-
air starter system.
ramic engine development. Basle mechanical
system development include the following ac-
tivities, engine operating procedure develop- Because the metallic and ceramic gas flow
ment, engine control unit (ECU) operation, foil paths were similar, metallic engine testing
bearing operation, regeherator drive system was used to improve the analytical methods
36
ROTOR SUPPORTS
--*-- BALL SEARING _FOIL BEARING

N (IbfJ
266.89- 60-
222.41 - 50 ..................
---1 ........................
177.93- 40 .................. -4 .........
,,0
133.45- 30 ......... 1- ........ J- ........
88.96- ......... I-
I
I
44.48 -
O-
0 20 40 6O 8O 16O 120 140
RESPONSESPEED, RPM X 108

GD7-OTO-33A
Figure 38. AGT101 Engine Unbalance Response/Support Forces.
employed in the design of ceramic compo- rotors have been pretested to high speeds and
nents. Thermocouples were installed in the 871C (1600F) in a metallic engine before test-
metallic engine flow path to collect data on ing in a ceramic structured engine.
the heat transfer characteristics of the engine
flow path. This information was used to Significant in the development work per-
calculate heat transfer coefficients along the formed in the metallic engine were rotor dy-
flow path. Extrapolations were made based on namic engine testing and ceramic turbine
these data to simulate the ceramic engine blade vibration evaluations. Engine rotor
condition for improved analysis of ceramic dynamic development was discussed in Section
components. 4.2.1, ceramic turbine blade vibration evalua-
tions are discussed below.
Ceramic turbine rotors were evaluated in
the metallic structured engine. Ceramic tur- Turbine Blade Vibration
bine blade vibration was noted as a potential
problem when a ceramic rotor failed during As a result of early ceramic turbine rotor
testing in a metallic engine. Turbine blade tests performed in a metallic structured en-
variation was evaluated by testing an instru- gine, turbine blade vibration was identified as
mented rotor in a metallic engine. Ceramic a potential rotor failure mode.
37
(9 IST BACKWARD 0 1ST FORWARO _ 2NO BACKWARD [] 2NO FORWARO
A 3RO BACKWARD )_ 3RU FORWARD
40"
/
0 20 40 60 8O tO0 120 140
SPIN SPEED, RPM X 103

687-078-34
Figure 39. AGT101 Engine Whirl Speed Map.
In those tests of S/N 001 engine, early vibration tests were conducted. In the first
generation sintered silicon nitrlde ceramic test, cold strain gages were located at the
turbine rotors were evaluated. These rotors point of peak stress for the fundamental vibra-
contained several defects which were removed tion modes, on the inducer and exducer por-
through extensive blade shroudline and back- tions of the ceramic blades as shown on Figure
face recontouring, and hand finishing. Con- 43. Then the rotor was installed on a mag-
sequently, the rotors did not conform to print netostrictive shaker and vibrated at the blade
dimensions. The first rotor was installed in natural frequencies to determine strain-to-
S/N 001 engine and operated to 97,300 rpm at failure.
871C (1600F) turbine inlet temperature before
a failure occurred. The second rotor, shown in Following this test, a similar straln-ffaged
Figure 42, failed at 62,000 rpm. Analysis of rotor was installed in a metallic structured
the rotor excursions during the failures indi- engine. This engine was then motored, forcing
cated that a sudden imbalance was imposed on ambient temperature air through the turbine
the turbine rotors which caused them to de- section, powering the engine through the blade
part from a previously stable orbit, in the resonant speed range.
second test the rotor failed at a speed known
to excite blade resonance. The results of this test, reported more
fully in Section 6.2, indicate that these newer
A test plan was developed to evaluate the generation turbine rotors have the room tem-
blade vibration problem. Two different blade perature strength and toughness to withstand
3B
0 1ST FORWARD • 3RD FORWARO
0.6-
0.6-
0.4
\
\ \
02'
i,----(9
0 20 40 60 80 100 120 140
SPIN SPEED. RPM X 103

G97 078-35
Figure 40. AGT101 Engine Forward Mode Log Decrements.
blade vibration induced by the passing of the operation with 1149C (2100F) turbine inlet
blades through the stator wake. Analysis temperatures.
indicates that the cold condition is much more
severe with larger driving forces; however this Principal evaluations with this engine ad-
can be avoided in engine testing by limiting dressed the design verification of the ceramic
the speed of the engine without a flame in the structures, and for transient thermal strength
combustor. during starting. In addition, this engine was
used to induce the maximum pressure loads on
4.3.2 Ceramic Structure/Metallic Turbine the ceramic structures through operation at
Engine Development maximum speed and the collection of 1149C
(210OF) engine performance data.
The ceramic structure/metallic turbine
engine configuration was an intermediate step Analysis indicated that the highest stresses
in the evolution of the AGT101 engine. All of occurred in the ceramic structural components
the ceramic structures had been developed to during the few minutes immediately following
the engine test stage in the ceramic structures engine start. During early testing, precise
rig, however the ceramic turbine and ceramic control was exercised over the engine thermal
combustor were not yet ready for engine test- conditions and extended thermal transient
ing. A dual alloy turbine rotor and the metal- schedules were used during the start to mini-
lic diffusion flame combustor, slightly modi- mize the thermal shock to the ceramic com-
fied and incorporating a thermal barrier coat- ponents. The start transient for the first
ing, were used on an interim basis to achieve 1149C (2100F) engine test took 52 minutes to
39
CERAMIC PARTS
871C 11600FI 1140C (21001=) 1371C {25GOF)
METAL CERAMIC STRUCTURES/ ALL CERAMIC

METAL TURBINE
- 6117-0711-11_
Figure 41. The AGT1Ol Evolves to Ceramics.
ing, the start transients were shortened to 15

seconds to reach 982C (1800F) with a 6-minute
hold before increasing the turbine inlet tem-
perature to I149C (2100F)
4.3.2.1 Endurance Testing
Significant testing of the 1149C (2100F)

engine configuration was conducted to evalu-
ate the long term durability of the engine
ceramic structures at I149C (2100F) turbine
inlet temperature. The ceramic components
used in this engine came from a variety of
manufacturers using several ceramic mate-
rials. The cross section in Figure 45 shows the
origin of these components.
The purpose of the endurance test was to

evaluate the durability of the ceramic com-
ponents in the engine environment under long
667..078-193 e88o85-i
term steady state conditions. This engine was
Figure 42. Ceramic Turbine Rotor operated in the test facility for 100 hours at a
Failed at 62,000 rpm. simulated cruise condition in two cycles. The
first cycle was 80 hours long and was termi-
reach 1149C (2100F) as shown in Figure 44. nated when an oil line in the test facility
Confidence increased as better ceramic mate- burst. A precautionary teardown revealed
rials were available and component designs some fractured turbine stators and local coat-
improved to reduce thermal stresses. By the ing delaminations of the regenerator seals.
end of the 1149C (2100F) configuration test- These components were replaced and the
engine was reinstalled in the test facility and
4o
ORIGINAL PA'CE IS
0t_ P.OOR QUAL/T'E
DRIGD, rAU PXCg Ig
DE _OOR QUALIT._
INOUCER
STRAIN GAGE
EXDUCER
STRAIN GAGE
GB7-078-184 97095-4
Figure 43. Strain Gaged Ceramic Turbine Rotor.
the remaining 20 hours of testing was per- ing this series of tests, improvements in inter-
formed on the engine. nal engine and regenerator seals were also
evaluated. The ceramic components endured
The ceramic engine components were in the full pressure conditions without failure.
Problems were encountered concerning the
very good condition after the completion of
the test. These parts are shown in the photo- durability of the metallic components oper-
graph in Figure 46. Neither excessive oxida- ating at the elevated temperature and high
speeds.
tion nor sticking at the ceramic component
interfaces were noted. The regenerator seals
4.3.2.2 Dual Alloy Rotor Integrity
did show significant distress though the core
and drive components were in good condition.
The dual alloy turbine rotor was designed
for short term operation at 1149C (2100F)
The successful completion of this test indi- turbine inlet temperature and was utilized to
cated that the ceramic structural engine as- enable ceramic engine development on an
sembly could endure the rigors of engine oper- interim basis until a ceramic turbine rotor was
ating conditions for extended periods. Con- available. The rotor was designed with a
fidence was gained to proceed with engine forged powder metal Astroloy hub hot-
testing to high speeds where component pres- isostatic-press (HIP) bonded to a directionally
sure loading and interface stresses were at solidified cast MAR-M 247 blade ring. Shown
their maximum level. in Figure 47, this rotor had provisions to cool
the bore of the rotor with air bled from the
Several full and near full speed tests were compressor seal, enabling operation at 1149C
conducted in the 1149C (2100F) engine. Dur- (2100F).
41
c IF)
1200
"70
2O0O
1000m SPEED -50

-55
1500 •
f
-50
60O
1000 {
-30 --
D 400_ _..
- 20
500-
200m
I0
PREHEAT -
0 I I I I I I i I I I I I I I
O 5 10 15 20 25 30 35 40 45 50 55 60 65 70
TIMEIMINUTES)
TEST DURATION = 58 MINUTES

687-076-185
Figure 44. S/N 002C - T4 Schedule for the First Ceramie Engine Test.
These rotors were used in all 1149C Both ultrasonic inspection techniques and vis-
(2100F) engine testing, including the 100-hour ual methods were used to examine the rotor
endurance test and the various runs to maxi- bond joints. These inspections indicated that
mum speed. During the course of this testing, most of the rotors had detectable bond joint
two dual alloy rotors failed. These rotors had defects, though whether these defects could
a predicted stress rupture life of 500 hours at cause a rotor failure was not known. More-
100,000 rpm with 1149C (2100F) turbine inlet over, the review disclosed that there was no
temperature. The first rotor failed due to effective proof test or NDE technique that
stress rupture of the east blade ring after 100 could positively weed out defective rotors
hours of testing at 67,000 rpm, and a short from the good rotors. As a result, the use of
time at 95,000 rpm. This failure was an dual alloy rotors was discontinued in future
indication that the blade crystal geometry was engine testing due to the uncertainty of their
less than optimum, resulting in a less than integrity.
predicted stress rupture life. The second dual
alloy rotor failure occurred during engine test- 4.3.3 Ceramic Engine Development
ing at 68,000 rpm and 982C (1800F). This
rotor experienced a separation of the blade The ceramic engine configuration incor-
ring from the hub. An examination of the porated an entirely ceramic hot power section,
rotor remnants revealed that the HIP bond including ceramic engine structures, ceramic
joint was discontinuous in several areas. combustor, and ceramic turbine rotor. The
evolution of this engine was the progression
This experience caused a review of the from the ceramic structure/metallic turbine
other dual alloy rotors in the AGT inventory. configuration as the ceramic turbine rotor and
42
Evaluation of the ceramic engine configu-
KEY MATERIAL SOURCE ration was initiallyapproached from a durabil-
I MAS -- MAGNESIUM ALUMIMUM NGK-LOCKEE ity standpoint. Engine performance was re-
SILICATE garded as secondary to insuring that the
2 LAS -- LITHIUM ALUMIMUM CORNING GLASS
ceramic structures were sound, that the
SILICATE WORKS
3 ROSN -- REACTION BONDED AIRESEARCH
ceramic turbine had sufficient life at temper-
SILICON NITRIDE CASTING CO ature, and that this system would operate as a
4 SAME AS 3 ABOVE
5 SN-50 SILICON NITRIDE NGK-LOCKE
6 SASC -- SINTERED ALPHA CARBORUNDUM

SILICON CARBIDE (STD OIL)
j_,_. _, ,,,, ,, , . £'
i_, "=. . _; _ j
Figure 46. Ceramic Components

Following the 100-Hour Endurance
GB7-078-186A
Run at 1149C (2100F).
Figure 45. Ceramic Structures Used in

the 1149C (2200F) Test.
ceramic combustor were developed to the

engine testing phase.
Significant testing of the ceramic engine

configuration included the activities noted
below:
o Long term steady state testing at 1204C

(2200F)
o Combustor performance and carbon form-

COOLINGAIR PATH
ing tendencies
GDT-O7B-IM
o Interpath gas leakage

Figure 47. Dual Alloy Rotor Showing
o Regenerator durability Cooling Air Path.
43
unit. The development of performance en- the ceramic component interfaces was not a
hancing components such as interpath seals problem. These structural components were
and the regenerator system, was performed in reused on a subsequent test.
their individual rigs until basic enginedurabil-
ity was established. This engine was rebuilt incorporating sev-
eral improvements in the combustor fuel noz-
Ceramic enginedurability testing was per- zle and regenerator system. The combustor
formed to identify the weak areas in ceramic fuel nozzle was thoroughly flow checked and
engine design and to demonstrate the capabil- tested in the combustor rig for carbon forming
ity of the ceramic engine system. To this end, tendencies. The regenerator core was fabrica-
the ceramic engine was tested steady state at ted from an improved MAS material with
1204C (2200F) and 60,000 rpm for 85 hours reduced thermal expansion. In addition the
until a failure of the ceramic turbine rotor inner regenerator seal was modified to relieve
stopped the test. The results of this test potential thermal interference at the cross-
provided much valuable information on the arm to outer peripheral interface. The test
durability of the ceramic components used in cycle was similar to the 85-hour test except
the engine. that a periodic shut down would be made to
pull and inspect the fuel nozzle for plugging.
The cause of the engine failure was traced
to partial plugging of the air assist ports in the After 12 hours of testing, another failure
fuel nozzle. The plugged ports allowed two occurred. This time the turbine rotor experi-
unatomized fuel streams to impinge with the enced a hub burst, most probably as a result of
wall of the pilot combustor, forming deposits premature stress rupture. The stress rupture
of hard carbon. When the carbon deposits properties of specimens cut from similar
broke loose from the combustor wall, they rotors were inconsistent and lower by orders
traveled down the gas path and impacted the of magnitude than the stress rupture proper-
turbine rotor inducer blade tips causing dam- ties of test bars of the same material. Fur-
age to the turbine and subsequent engine fail- ther rotor material development is required to
ure. resolve this problem.
The engine regenerator system showed The regenerator system fared much better
severe thermal distress. The magnesium during the 12-hour test. The regenerator core
aluminum silicate (MAS) regenerator core had showed minimal distress and thermal stress
indications of high thermal stress with thermal cracking was not significant. The regenerator
cracks emanating from the outer rim toward seals, however were again distorted. Follow-
the center. The regenerator seal wearface up work at Ford was performed to resolve the
coating on the hot side seal had delaminated seal distortion problem. This subject is dis-
about the inner periphery and the shape of the cussed more thoroughly in Section 5.4.1.
seal crossarm was significantly distorted. This
damage occurred during the course of the test Ceramic combustor light-off performance
and was not a result of the failure. has been marginal. During steady state oper-
ation at high temperatures, this combustor has
Follow-up testing in the combustor rig had provided consistent, reliable, stable operation.
indicated that the key to preventing fuel noz- However further development is required to
zle fouling was to maintain proper nozzle face improve the combustor system start and low
temperatures. Indications were that excessive temperature performance.
temperatures in the air-assist passages of the
nozzle caused fuel in those passages to coke The design deficiencies of this combustor
and plug the passage. have been identified and design improvements
have been made. This modified combustor
The ceramic structures survived the test design, reported in Section 5.3, has been fabri-
intact. Oxidation was minimal and sticking at cated in metal, tested in the combustor rig,
44
and tested in the metallic structured engine. 8RA_
r_ SPZEO
° ,oo._RPM
Engine testing of the enlarged bore combustor oYNo
_ ' lc°.
M_ESs°R 4TURBINE
has shown dramatic improvements in start and
low temperature performance.
_,L. '_ T3.6 I
4.3.4 Engine Performance Testing ,b "1 J TURBINEPISTONRING (7.86%)
I _.......................... "I"1 [--I= ,TERATIVE
Engine performance testing was limited ...........

;......................
............ SOLOT,ON
OVERBOARD FOIL BEARING12.0%] _.
during the AGT101 project due to limitations
in the capabilities of the ceramic materials .... LEAKAGE= PERCENTLEAKAGEFLOW T3"I
and design development of the engine internal
seals. What testing was performed was done
at development conditions, that is, at condi- TEST OATA CALCULATEO
VALUES
tions less severe than the RPD, for the pur- DYNOPOWER 22.42 kW [30.07 HP] 22.43 kW (30.08 HP)
pose of seal development. SHAFTPOWER 32.4 kW (43.4 HPI
PI 97.168 kPa (14.096 PSi] 97.188 kPa (14.096 PSi)

Performance testing was performed at var-
TI 29C (85FI 29C 185FI
ious engine operating conditions for use in
T3.0 229C 1445F) 241C 1466F)
determining the leakage levels of the engine
cycle. This data was used in conjunction with 261C (502F) 260C [501FI
the engine performance model to establish the T3.R 7,5QC{13B3F) 762C (1404F]
overall leakage of the engine and determine 1170C (2139F) 1170C 12139F)
T4.1
the equivalent engine performance at sea level
T5.1 860C (1597F]
standard conditions. Using the measured data
TB.O 36QCI680FI 347C 1058F)
as a starting point in the engine performance
model, the internal leakages are varied until a7.0711.194
the best data match is obtained. During this

iterative modeling exercise, energy and mass Figure 48. Engine Test Data Setup Model.
balance is maintained.
duct is recognized as critical to the overall
Performance data was obtained at 1149C leakage performance of the AGT101 engine.
(2100F) turbine inlet temperature in a ceramic Because temperatures in the region will ap-
structured engine with a dual alloy turbine proach the 1371C (2500F) TIT at full power,
rotor over a range of speeds from 60,000 to use of ceramic sealing elements is required.
I00,000 rpm. Limited component instrumen-
tation downstream of the compressor could be The concept employed for the first ce-
accommodated due to the ceramic structures. ramic engines involved two sealing elements
Based on this approach, the internal leakage separated by a ceramic wave spring, as shown
was 21 percent. The 21 percent leakage level in Figure 49. The wave spring concept was
is consistent with a standard seal level perfor- originally the result of observed distortion of
mance of 32.2 kw (43.2 shp). The engine RBSN rings during nitriding. The wave pattern
configuration was far from optimum, but the generated as the rings drooped over the furni-
test represents the baseline performance from ture in the furnace was recognized as having
which efforts can be addressed and compari- potential application as a spring.
sons made. This model and the data match is
summarized in Figure 48. This design was found to have shortcomings
in terms of sealing capacity and especially
4.3.5 Ceramic Engine Seal Development repeatability. The effectiveness of the seals
depended greatly on their ability to seat them-
The interface between the turbine shroud, selves under a pressure load, with the wave
the flow separator housing, and the transition spring serving little purpose other than to
45
DUAL FLOW SEPARATOR SEALS WiTH
LP
TORGINE SHROUO SEAL
HP
TURGINE BACNSHilOOO / R'P OUAL SEAL RINGS

TURBINE "SHROUO WAVESPRING _ LAS SEAL RING
HP
_v
ilTiT_IN
Figure 50. Flow Separator Seal

Figure 49. Single Separator Seal Configuration.
Configuration.
ments between the turbine shroud OD and the
position the seals during startup. Sealing flow separator ID (a large pressure drop area
capability was often improved by physically where leakage represents a direct loss to the
jarring the engine external housing. This is an power cycle). As in the previous configu-
indication that the ability of the wave spring ration, tight-fitting custom-ground LAS seals,
to reliabily seat the sealing elements is mar- with no end gap, could be used for engine
ginal at best. operation up to 1204C (2200F) TIT. Higher
operating temperatures required the use of a
sintered silicon nitride or RBSN rings, with an
To improve the sealing capacity and relia-
overlapping and gap to accommodate the dif-
bility of this system, modifications were made
ferent thermal growth rates of the rings and
to ceramic hardware as shown in Figure 50.
the flow separator.
Specifically, the wave spring (now made of
sintered silicon nitride) was made smaller and Testing was performed to evaluate the
installed between the turbine backshroud and improvement which the updated design af-
the combustor baffle. The seal pocket was forded. This is discussed in detail in section
now able to accommodate two sealing ele- 6.3.2.6.
46
5.0 COMPONENT/SUBSYSTEM DEVELOPMENT
Activities centered around the develop- o Stage Efficiency, 0.805

ment of the following major components and Total-to-Total
subsystems: compressor, turbine, combustor,
and regenerator, o Rotational Speed I00,000 rpm
5.1 Compressor Development These performance goals are quoted at

maximum power; however, as previously
5.1.1 Compressor System Requirements stated, the AGT101 was designed using a "sys-
tems design" philosophy wherein a distinct
The AGT101 compressor system is com- recognition of CFDC duty cycle has been
prised of a single-stage backward swept cen- considered. Therefore, since a significant
trifugal impeller having 12 full blades and 12 portion of operation is at low power levels,
splitter blades, a three stage radial diffuser during the course of development, maximum
with one stage of diffusion vanes and two power performance goals were compromised
stages of deswirl vanes, and an upstream set with more important program goals, i.e.,fuel
of seventeen variable inlet guide vanes (VIGV). economy over the duty cycle.
Performance predictions for this compressor
system at maximum power operating condition
5.1.2 Aerodynamic Development
are as follows:
o Mass Flow at 29C (85F) 0.388 kg/sec 5.1.2.1 Compressor Aerodynamic Design
sea level (0.8542 lb/sec)
The aerodynamic and geometric impeller
o Stage Pressure Ratio 5 design summary is shown in Tables 7 and 8.
Table 7. AGT101 Impeller Design Aerodynamic Design Summary.
Inlet Total Temperature 302.6C (544.7R)

Inlet Total Pressure 99.5 kPA (14.431 psia)
Speed 100,000 rpm
Flow Rate 0.387 kg/see (0.8542 lb/sec)
Tip Speed 570 m/see (1872 fps)
Impeller Efficiency, Total/Total, No 89.67 percent
Dump Loss
Specific Speed 57.2
Shroud Inlet Relative 1.125
Maeh Number
Shroud Relative Velocity Ratio (PD 0.600
Analysis) (0.529 Caps)
Shroud/Mean Meridional Velocity Ratio 1.25
(PD Analysis)
Average Exit Blade Deviation 3.27
Inlet Aerodynamic Blockage 0.98
Exit Aerodynamic Blockage 0.90
Average Slip Factor 0.9488
47
Table 8. AGT101 Impeller DesignGeometric DesignSummary.
Inlet Hub/Tip Ratio 0.464
Inlet Hub Radius 1.52 cm (0.600 in)
Inlet SweepAngle (Meridional Projection) 7.47 Deg.

Exit Radius 4.45 cm (2.146 in)
4
Exit B-Width 0.39 em (0. 154 in)

Blade Number 12/24
Average Exit Blade Angle 49 Deg.
Exit RakeAngle 0
Axial and Radial Clearance 0.076 mm (0.003 in)

Inlet ShroudNormal Thickness 0.381 mm (0.015 in)
Inlet Hub Normal Thickness 0.635 mm (0.025 in)
Exit Shroud Normal Thickness 0.381 mm (0.015 in)
Exit Hub Normal Thickness 1.32 mm (0. 052 in)
Aerodynamic boundary conditions used in o._5

the detailed blade shape design are shown in _--
Figures 51 and 52. These conditions include
impeller exit crosspassage distributions of the
f
entropy function, which relates total pressure _ 0.95o
to total temperature and blade deviation --
angle. The entropy function (Q) and aero-
dynamic blockage (Aeff/A_e o) were distri- _
buted linearlythrough the irhpelleralong each '_
streamline from the leading to trailingedge.
Deviation was distributed parabolically from
_ o945 \
the point of inception to the exit according to \
an established method.
The blade/air angle relationship at the in- o

ducer leading edge is shown in Figure 53. For HUB $HOOUO
completely subsonic flow conditions, the im- PERCENT FLOW

pact-free incidence relationship, corrected for GB7-0?i1-197
compressibility, is normally used. However,

proceeding outward to the inducer tip, roach Figure 51. AGT101 Impeller Design Exit
number exceeds 1. For this region, blade Loss Function Distribution.
incidence was modified (from impact free)
according to transonic axial compressor blade Final air/blade angle distributions are
design rules (toward 0-degree meanline inci- shown for shroud, mean, and hub streamlines
dence). Higher blade angles resulting from in Figure 54. Resulting blade suction and
this correction will be advantageous for part- pressure surface velocities are shown in Figu-
speed operation. res 55 through 57 for the shroud, mean (50-
48
10
65
i i I i
18 -- AIR ANGLE
/ 60
\ .... 9LAOE ANGLE
i°\ / \
/
/
,I
/
/ i \ \
\
_J
/ -__ 4o \ _MI __._
EAN
_'"
/
"/,, '
/
O
0 2O 40 60 60 100 \
\
/,i
RUB SHROUD _f
PERCENT FLOW
3O
6117.07666196 •
0 20 4O 60 60 IO(_
Figure 52. AGT101 Impeller Design Exit PERCENTMERIDIONAL DISTANCE

Deviation Angle. 11117.076.290
Figure 54. Compressor Impeller Relative

Air Angle Distribution.
RELATIVE AIR ANGLE
62 \ ..-.
i
-'60 r_ f
1 %
58
\ -- \
\
oc \
56 % \
f /" IMPACT FREE \ \
,-/ BLADEANGLE % \
54 t \ \
O.6 O.8 1.0 1.2 1.3 [INI \ \
\
\
2'0 (o cM 0.4
\
SHROUO
0.3 T
RADIUS
6DT.076.Ig9 O 20 40 60 80 100
Figure 53. Compressor Inducer Leading PERCENTMERIDIONAL DISTANCE
Edge Blade, Air Angle Distributions. 697.07666201
Figure 55. Compressor Impeller Blade

percent flow), and hub streamlines. These Surface Maeh No. Distribution - Shroud.
velocity distributions compare favorably with
previously designed, successful compressors of procedure requires that the local value not be
similar pressure ratios. less than 1.04, a condition met for allstream-
lines.
To assure adequate impeller flow area at
design point conditions, a check of available The final main blade normal thickness dis-
inducer throat area was made using a ratio of tribution is shown in Figure 59. This distri-
local geometric (blade-to-blade) area to the bution was required to satisfy blade stress and
area required for sonic flow conditions at vibration criteria.
leading edge stagnation conditions. A mini-
mum value curve of this parameter across the Impeller vector diagrams are shown in Fig-
inducer is shown in Figure 58. Current design ure 60.
49
MM {INCNI
1O
f - /°°5-
•_o.8 \ HUB
i 1.0
''l1 0.04-
o.o+- f
_ 0.6
,,,, o.J=:
O.OI-- SHROUD
MEAN _
0.4 + O.O0 I I I l I
0 20 40 60 80 I00 O 20 40 60 80 I00
PERCENTMERIDIONAL
DISTANCE PERCENTMERIDIONAL
DISTANCE
601.OIE-_ GII7-078-205
Figure 56. Compressor Impeller Blade Figure 59. Impeller Normal Thickness
Surface Maeh No. Distribution- Mean. Distribution.
0.8
63.98 OEG
f
• / VM = 188
i551
HIS
Ipsl
f_ 1319 fpsl
f J VM : 97 MID DE_ 58.86
f
W T = 344 M/S (1229 fPsl WT = 160 M/S
SHROUD 1524 fPSl
INOUCER LEAOIN6 EOGE RUB
{W T = RELATIVE TANGENTIAL VELOCITY)
HUB
O.2
51.45 OEG
O 20 40 60 80 I00
6il7-07i1-203
PERCENTMERIOIONAL
DISTANCE VM : 110
(362
M/S
fPsl
5,.DE,
Figure 57. Compressor Impeller Blade
W T = 170 M/S W T = IN M/S
Surface Maeh No. Distribution - Hub. i55G Ipsl 1522 fPsi
SXROUO IMPELLER EXIT

HUG
|.3 _
II07-07&3_
Figure 60. AGTI01 Impeller Design

|.2--'
Vector Diagrams.
/|" geomimass
= 1.086
Impeller splitter aerodynamic and stress
i avg
analyses were completed. To reduce a locally
1.1-
high splitter leading edge stress, the hub lead-
ing edge was moved axially forward from the
original position to make the edge more radial
(Figure 61). The circumferential position of
the splitter leading edge then was shifted to
1.0
I I I I | match the 50-percent mass flow streamline
20 40 50 80 100
across the entire blade height. This meanline
PERCENTMASSFLOW
shift was accomplished as shown in Figure 62
Gii7-076-204
for each streamline calculation. The reposi-
Figure 58. AGTI01 Impeller Design tioning was performed to eliminate a potential
Minimum s/a'geom. loss penalty associated with a non-uniform
50
CM (INCH)
2.2- flow split which might result if the splitter

blades were left at the mean tangential posi-
tion between main blades. This technique has
2.0--
S.O- been successfully used on previous Garrett
high pressure ratio compressor designs.
1.8-
Diffuser Design
|.8-
4.8- Based on the reference study and sub-
sequent design analyses, the radial diffuser
1.4- SHIFTED SPLITTER LE_ design consists of three vane rows. The first
row is a standard 2-D vane islanddiffuser with
a converging end wall followed by a cascade of
I°2_
two deswirl vane rows. During the analysis, a
tradeoff was made between the first stage
1.0-- (vane island) diffuser area ratio and the tan-
dem deswirl vane diffusion factors. The ef-
2.0 4
0.8--
0.6
1 0.2
I
ORIGINAL SPLITTER LE fect of varying the first deswirl vane exit
angle on the diffusion factor of each row is
shown in Figure 63. The firstdeswirl vane row
AXIAL DIMENSION, INCH
o ,o 2]o s_o
AXIAL DIMENSION. CM
o.82
GB7-078-207 z
Figure 61. Splitter Blade Hub 0.81 ,f

0.80 I
Leading Edge Movement.
== 1.5
CURVE EVEBYWHE
NORMAL TO MAiN \ _-"_\\\\\",_ i.o
GLAGEMEAR',NE
\ E 1.0
ROW 1
_-'_ _ ,'_'-[ FMEAN TANGENTIAL o.e
_.<_-\\\'_. _" I \ POSITION BETWEEN .- ROW 2
0.4
, I\ I"';-_;DE,'BEO
S,L,TTEB 0.2
ON STATION _
LINE_ J 0.8
• I_l .......... -_ n ISATISFIES
I',\--5G-50 MASS _ .MAIN BLADE 0.7
0.8
Ii\ ,uc,
ONSOB,ACE 0.5
'_ 0,4
0.3
3.0 VANE ISLANO

L DIFFUSER AREA
RATIO = 2.471,
-_ i ,W 2.0 EXIT MACH
ROW 2 _-
o,= __ STATION LINE
NO = 0.225
_. _ MERIOIONAL COOROINATE 1.0
10 20 3O 4O
MERIOIONAL COORDINATE, rn
ROW 1 EXIT ANGLE [AIJ

gg1_lg.ZB
GD7_7_20_
Figure 62. Splitter Leading Edge

Tangential Position. Figure 63. Diffuser Deswirl Vane Analysis.
51
inlet Mach Number was 0.225, a value pro- CM ilNI
5.0-
viding the vane island diffuser with an area
12.0.
ratio of 2.59 which is consistent with Garrett
4.5. R : ]1.163 CM 14,397INI
experience.
R 9.911CM 13.902INI
I0.0. 4.0.
Further design iterations resulted in a dif-
fusion system having a convergence over the I 3.5-
R = 6,712 CM 13.233INi
radial vane island of 0.05 cm (0.020 inch) from
&0-
the leading to trailing edge, and a trailing 3.0.
0.376CM [0.146INi OIA
edge thickness of +the vane island of 0.178 cm R : 5.860 CM
12.307
(0.070 inch) and 34 vanes. An axial projection 2.5-
8.0-
of the diffuser-system is shown in Figure 64 34 VANES
_.0,
and meridional flowpath in Figure 65. Dif- |
O0 1.0 2.0 3.0 40
fuser vector diagrams are shown in Figure 66. X -- AXIS.EN
o_0 71o ,'.o o_o 8_o ,o'o
5.1.2.2 High Flow Modification X-AXIS. CM
GlT,O7|.210
Based on initial test data, the impeller was

found to be low on work input and flow capa- Figure 64. AGT101 Diffuser and Deswirl
city. To restore these deficiencies back to Vane Assembly Axial View.
(INCHI
2ND DESWIRL VANE EXIT
n = 11,17 CM (4.397 INI
4.6-
.p--.-- FLOW ANGLE = 00EGREES
MACH NO = 0.105
CM 68 VANES
4.2
1ST DEGWlRL VANE EXIT
I0.0" R = 0.91 CM 13.901 INJ
FLOW ANGLE = 32.58 DEGREES
3.8-
MACH NO = 0.141
34 VANES_
3.4-
DIFFUSER EXIT
O.O ¸ R = 8.21 CM 13.233 IN)
FLOW ANGLE = 47.03 OEGDESS
3.0-
MACH NO = 0.225
34 VANES_,__
2.6"
DIFFUSER INLET
2.2- R = 5.86 CM 12,307 INI
", FLOW ANGLE = 72.8 DEGREES
\ MACH NO = 0.853
1.8-
', iMPELLER EXIT
R = 5.45 CM (2.145 IN]
4.0"
1.4"
1.0" SPLITTER LEAOING EDGE
2.0
0.6
1o o ,io 3o
i
-2.0
IINCHi
-4'.0 -2'.o o;o 21o 4;o 8'.0 :M

AXIAL OIMENSION
GB7_7_211
Figure 65. AGT101 Diffuser Meridional Flowpath.
52
72,81 BEG
1.2"i_/I
I.I -_
VR = 107.3 M/S
(351.9 fpsl
VT = 346.6 M/S (1137.7 fpsi
OIFFUSERLEAOING EDGE (VANE ISLANOi _ 0.8
.=.
54.07 OE6
_ 0.6
• _ ,4 (151.8 Ipsl
(192.9 fPSi 37.65 OE6 (164 Vfpsi

R = 46.3 M/i
Vn = M.8 M/S __ VR = 50.0 M/S
VT = 62.9 M/S V T = 38.5 M/S

1272.2 Ipsl (126.3 Ipsi
IST OESWIRL VANE 2ND DESWIRL VANE 2NO OESWIRL _ELLER iMOOELEO FROM TEST 1 OATAi
LEAOIN6 EDGE LEAOINGEOGE VANE EXIT 0.2
O 10 20 30 40 50 60 TO nO gO 100
iilT-OTO-2]
2
PERCENT MERIOIONAL OISTANCE
667.076.213
Figure 66. AGT101 Diffuser Vector Diagrams.
Figure 67. AGT101 Modified Impeller
design levels, an analysis was conducted to Blade Loading's--Tip.
evaluate several exclusive and related modifi-
cations. Working within available hardware
constraints, modifications that added 0.02 cm
(0.052 inch) to the impeller tip radius, 0.019
cm (0.048 inch) to the inducer shroud line and 0.6
0.007 cm (0.018 inch) to the exit blade height
0.5
were analyzed to increase flow capacity and
work. A comparison of the modified impeller
blade loadings and loadings based on Test 1
(reference paragraph 5.1.2.3) data is shown in
Figures 67 and 68 for the shroud and hub
0.2
respectively. The loadings shown for this Test
i
1 impeller were modeled using tested losses
O. ------ MOOIFIEO IMPELLER INIOH FLOWEO AND TIP EXTENOEDI
and deviation. Figure 69 shows the modified "_ BASELINE iMPELLER [MOOELEO FROM TEST t OATA)
impeller geometry. I I I I I I I I l I
0 10 20 30 40 50 60 70 80 90 150
Based on the modified impeller geometries,
the stationary shroud and diffuser also were PERCENT MERIDIONAL DISiANCE
007-076.214
modified. Since Test I data indicated that the
diffuser minimum loss occurred at design inci- Figure 68. AGT101 Modified Impeller
dence, the original diffuser design criteria was Blade Loadings--Hub.
retained. In addition, since the match ob-
tained between vane rows I, 2, and 3 showed
good performance, a decision was made to 5.1.2.3 Aerodynamic Testing
modify only the Ist vane row to accommodate
the larger impeller and retain the 2nd and 3rd Compressor development testing was con-
vane geometries as designed and tested. To ducted on test rigs suitable to aerodynamically
accomplish this, vane row I was increased in and mechanically evaluate the impeller, dif-
height, and the vane height was linearly con- fuser, VIGVs, inlet plenum and compressor
verged to match vane row 2. discharge airflow path.
53
lINCH]
2.3,
MODIFICATIONS i_
2.1 DIFFUSER
CM _'__! , DRIVE TURBINE
5.0- 55.8 MM i2.146 TO 2,198 INCHi _

1.9- i EXTENDEDTiP SHROUD
54.5 TO
RECONTOURED II |_
INCREASEDEXIT VANE HEIGHT H
1.7- 3.91 TO 4.37 MM 10.154 TO //
0172,NCH_ //
MODIFIED //
1.5-
fTEST _J/ VIGV
1.3 "
I BASELINE /
1.1-
G87,078-216
0.9- Figure 70. Compressor Rig.

2.0-
0.7- from transitioning compressor discharge flow

to regenerator high pressure inlet. In addition,
O.5 , the inlet air filter and associated ducting (not
-0.1 o'1 o'3 o'5 o'.7 11, 13 pNCHESl shown) can be accommodated to verify aero-
dynamic design intent and again establish any
detrimental flow effects resulting from the
! I CM installation.
O.O 3,0
MERIOIONAL DISTANCE The rig incorporates the capability for con-

Gli7..07ll-215 trolling and evaluating critical compressor im-
peller running clearances based on capacitance
Figure 69. AGT1Ol Modified Impeller probe measurements at selected stations along
Geometry. the impeller meridional flowpath. The clear-
ance control mechanism is based on a design
As discussed in paragraph 5.1.5 powder concept successfully demonstrated on other
metal aluminum materials were evaluated dur- Garrett programs. This mechanism features
ing this program. However, due to timing an Acme thread, used to axially adjust the
constraints titanium was used during this test position of the impeller under dynamic opera-
program. tion. This is accomplished by fixing the inter-
nal thread and turning the external thread,
Compressor Test Rig thus forcing the externally threaded members
(impeller shafting) to move axially. Clearance
The compressor test rig, is a straddle- control capability a}lows thermal stabilization
mounted bearing configuration utilizing the of the test rig and dynamic stabilization of the
AGT101 engine ball bearings. Power is de- impeller prior to recording data such that
rived from a drive turbine system composed desired clearances can be evaluated. In addi-
primarily of the Garrett Model TV81 turbo- tion, this mechanism will allow performance
charger hot flowpath components. Actual evaluation of larger than desired clearances
AGT compressor stage hardware (VIGVs, im- and/or duplication of engine conditions for
peller and diffuser) is utilized in the rig to performance comparisons.
permit detailed compressor design perfor-
mance mapping. As shown in Figure 70, power Compressor test rig instrumentation was
section ducting has been incorporated to identified such that detailed evaluation of
establish any flow phenomena effects resulting individual and system compressor components
54
can be achieved. Instrumentation is of a size 0,60-
consistent with the small stage dimensions and

0.75,
of sufficient number and spacing to provide
realistic average values for all parameters _'_ 0.,0.
z,w_
measured.
w N 0.65-
I
Two compressor stage tests were con- 0.65" w_mTEST 1
ducted on the AGT101 compressor. During
0.65 --TEST 2A
Test 1, compressor flow and pressure ratio
were below design intent. Based on Test 1 3,50"
results and subsequent analysis of test data, a

modification to the impeller was accom-
plished. Test 1 data showed that diffuser
performance was essentially as predicted, thus 70"
modification to the diffuser was accomplished

only as required to accommodate the impeller i! 2.OO-
1.50-
modification. Figures 71 through 73 show the ",% : I 70
4O 50 6O
comparative results of Test 1 and Test 2A. 1.00
Performance mapping was completed at o, o'.2[

o;s0.'4oi_o:5Jo:7
o;s
RATING STATION INLET CORRECTEDFLOW. (LBM/SEC)
speeds from 40 to i00 percent'of design cor-
rected speed and IGV settings of 0, +20, +40, _., o'.3
+60, and +70 degrees. Axial clearance for KG/SEC
GB7-078-218
Test 2 was controlled to within 0.076 to
Figure 72. Full Stage Compressor Data,
IGVs = 40 Degrees.
0,80
0.50
,-,I
0.70
] 0.70 " " '_ ""
_--- _ TEST I
0,60 Z
0"65t
_ 0.55 0.75.
0.50
TEST 2A
0.65
100 "_k"_ 2.60 TEST I

4.50
4.50 2.40 TEST 2A
_' 3.50 2.20
70 - - ,_,_,,
-
2.50.
_ 3.00
I,- ,,_
2.50
1.5050.-..,---.._,,,_,,
-,,_
2.50'
'_ 1.50
1.50
"'150
50
ILBM/SEC]
0.10 0,30 0.50 0.70 0.90
I I KG/SEC 0.15 0.25 0.35 0.45 ILBM/SEC]

0.1 0,4
RATING STATION INLET CORRECTEOFLOW 0_1 0.'2 KG/SEC
GII7-0711-217 STAT,ON
RAT,NO CORRECTEO
_NLET _'OW
607..076-219
Figure 71. Full Stage Compressor Data, Figure 73. Full Stage Compressor Data,
lGVs Open. IGVs = 70 Degrees.
55
0.102 mm (0.003 to 0.004 inch) as designed. STATIONARY
STRUT
Test data indicates the following:
X I =0 X=O
o Design pressure ratio and flow were
achieved / r P,VOTPOINT
o Diffuser performance was very close to

i
predicted levels
6674)78-220
A highly desirable efficiency envelope has

been achieved, which has the peak stage Figure 74. AGT101 Inlet Guide Vane.
efficiencies occurring at the lower end of
the speed range
Compressor performance characteristics

!
demonstrated during Tests 1 and 2 indicate
that the compressor was performing as desired
when considering an automotive duty cycle
dl,
application, i.e., low power/high efficiency.
This results from the_fact that the vast major-
ity of driving time occurred over the CFDC
between 48- and 70-percent engine speed.
5.1.3 VIGV Design and Testing
5.1.3.1 Variable Inlet Guide Vane Design

Analysis
The VIGV is an articulated design (fixed

leading edge, movable trailing edge) using a
NASA 63-0012 series thickness distribution
G87-078-221 P72492.]0
modified to a maximum thickness of 13.5
percent of chord for the articulated portion of Figure 75. AGT101 Compressor Inlet
the vane. The fixed leading edge uses 26 Guide Vane Test Rig.
percent of a NASA 63-0010 modified for a
maximum thickness of 13 percent of chord. Initial testing identified objectionable
This design (Figure 74) is based on a similar acoustic noise levels (in the audible range) in
articulated guide vane successfully being used excess of 130 dB when the IGVs were closed to
on several engine compressors under develop- 50 degrees or higher. This phenomena was
ment at Garrett. Results of tests conducted similar in nature to previous Garrett experi-
on these similar designs have been used in the ence on another articulated radial IGV system.
off-design compressor performance for Although the exact mechanism generating the
AGT101 cycle calculations. tones is not well understood, it is believed that
the IGV-induced swirl transfers flow energy to
5.1.3.2 Inlet Guide Vane Testing acoustic energy.
Development testing on compressor stage This problem was resolved, based on prior
hardware encompassed inlet guide vane (IGV) Garrett experience, by introducing a minimal
acoustic evaluation. The IGV test rig is shown disturbance in the flow field. Flow tabs, as
in Figure 75. shown in Figure 76, were installed on the fixed
56
t30 [ I !
FLOW RATE = 0.131 LS/SEC I
IGV ANGLE = 65 OEGREES
110- _-
- t20_,
tOO-
--JTNOUT
F=W
TABS
gO'v--WITH FLOW TABS
0 Iooo 2o00 3001

FREQUENCY, HZ
GD7_7_223
GB7-078222 4
P72492-8 Figure 77. Inlet Guide Vane Test Results.
Elastic three-dimensional stress and fre-

Figure 76. Inlet Guide Vane Flow Tabs.
quency analysis was performed on the final
blade configuration. The blade was optimized
leading edge of two IGVs. These flow tabs
to provide good aerodynamic performance
dramatically reduced the tone amplitude. Fig-
while maintaining stress levels necessary to
ure 77 presents the results of IGV testing with
and without flow tabs. assure mechanical integrity. Blade vibrations
primarily are confined to the inducer region of
5.1.4 ImpeLler Meehanieal Design low specific speed impellers. The short ex-
ducer blade height makes this blade region
5.1.4.1 Stress Analysis
insensitive to aerodynamic excitations. To
The selected long term impeller material is assure an inducer region vibratory margin,
a powder metal aluminum (section 5.1.5). Due stress levels were limited to 137.9 MPa (20
to the availabilityof property data for the PM ksi),thereby p_rmitting vibratory stress levels
alloy,forged 2219 T6 Al material characteris- of 34.5 MPa (5 ksi) with 2219 T6 AI. Plots
tics were initiallyused for stress and life showing stress results from a three-dimen-
analyses (titanium material was used for sional finite element analysis are presented in
engine test). Impeller properties are provided Figure 78. This analysis was performed with
in Table 9. zero displacement boundary conditions at the
Table 9. Impeller Properties.

I
Total Weight, kg (lbs) 0.249 (0.549)
Blade Weight, kg (lbs) 0.015 (0.034)
Ip kg-m-see 2 (lb-in-see 2) 1.61 x 10 -5 (0.0014)
Id kg-m-see2(lb-in-see 2) 1'.15 x 10 -5 (0.0001)
Hub Area, cm 2 (in 2) 7.645 (1.185)
CG cm (in) (Forward of Blade Root Exit) 0.704 (0.277)
Kinetic Energy kg-m (in-lb) 907.0 (78700)
Density g/era 3 (lb/in 3) 2.80 (0.101)
57
217.2 MPa 191.7 MPa
(31°5 KSII 127.8 KSI]
10
SUCTION SIDE PRESSURE SIDE

EQUIVALENTSTRESS (KSI}
EQUIVALENTSTRESS (KSII
104.1 MPz 137.2 MPa

[15,1 KSII (19.9 KSI]
G87-078-224
Figure 78. AGT101 3-D Airfoil Elastic Stress at 100,000 rpm.
blade/hub interface. Calculated blade fre- A 3-D elastic stress and vibration analysis
quencies are shown in Figure 79 and the first was performed on the final splitter configura-
four mode shapes in Figure 80. The first tion. Results are provided in Figures 81 and
vibratory mode is at 3.5E (where E represents 82. The splitter stress levels are similar to
the rotational speed) at 100-percent speed. the levels in the main blade. Frequency
Experience has shown that impeller blades are analysis of the impeller splitter blades shows
relatively insensitive 1_o distortion related potential interferences of the first and second
excitations when the first mode frequency is vibratory modes with the IGV excitation. This
above 3E. The first three blade modes will not analysis also shows physical location on the
have interferences with 17E IGV wakes in the airfoil where the maximum vibratory strain
operating range. With the relatively long flow would occur for each of these modes. Since
space between the IGVs and blades, higher these modes occur at part speed, the blade
mode excitation is not expected. steady-stress level and metal temperatures
will be significantly lower than the levels that
Exducer region stress levels have to be exist at the 100-percent speed mechanical
controlled to provide adequate rupture life. design point. Using the stress and tempera-
The combination of 614 m/sec (1872 ft/sec) tip ture values that exist at the peak vibratory
speed and 50-degree backward curvature re- strain location for the predicted speeds of the
sults in a maximum elastic stress at the exit interferences, the analysis shows that the al-
of 214 MPa (31 ksi) and an adiabatic wall lowable vibratory stresses of the first and
temperature of 190C (374F). second modes respectively are 51.7 and 69.0
58
17E INLET
GUIDEVANES R
30-
L__x /
25"
IOE
_20-
4TH MOOE 8E NORMALMOOESHAPE

NORMALMODESHAPE
7E MODE1 FREQIHZ] = 5063 MOOE2 FOEQ(HZJ= 10,531
3RD MODE
6E
2NOMOOE
10- 5E
4E
1ST MODE
5
2E
IE
O
0 25 50 75 100 125
SPEED,PERCENT
667_7&225
Figure ?9. AGT101 Impeller Airfoil

Campbell Diagram.
NORMALMOOESHAPE NORMALMODESHAPE
MODE3 FREQ(HZ) : 12.930 MODE4 FREQ{HZ]= 15.341
MPa (7.5 and 10 ksi). This allowable vibratory
stress should provide adequate fatigue margin, GB7-07D'2"Z8
particularly since the vane to rotor spacing is

more than 1.5 times the chord length of the Figure 80. AGTI01 Impeller Blade Vibratory
IGVs. Modes.
195.1MPa
Elastic stress analysis was completed for i28.3 KSI)
the impeller hub. Results of this analysis are
shown on Figures 83 and 84 along with the VIEW2
!
VIEWJ
isotherms for the 29.4C (85F) day design point.
These stress levels are consistent with Garrett SE:200 / I SF = 2.00
RPM = 100,000
experience on similar aluminum impeller de- R RPM=,OO.OOO
/ /
signs. A life summary is presented in Table
I0.
The axial and radial displacements of the 167.5MPa.,-_._/

blade, taking into account maximum tempera- 124.3KSII
ture 100-percent operating speed and centri-
fugal stiffening is shown in Figure 85. PRESSURESIDE SUCTIONSIDE
EQUIVALENT
STRESSMPa (KSI] EQUIVALENT
STRESSMPaIKSI]
5.1.4.2 Impeller Holography Tests Results G67_78.227
An aluminum (2219-T6) and titanium (Ti-6- Figure 81. AGTI01 3-D SplitterElastic
4) impeller were mechanically evaluated to Stress at I00,000 rpm.
59
173.9C
I
5TH MOOE
4TH MOOE
t
i
I
3ROMOOE l
2NO MOOE (INLET6UIOEVANES(
I
1ST MOOE 8E
8E
4E
2E
25 50 75 100 125
SPEEO.PERCENT
137.8C
116C 1280F)
Figure 82. AGT101 Impeller Splitter 1240FI
Campbell Diagram.
15
687-078-230
Figure 84. AGT101 Impeller Isotherms -

29.4C (85F) Inlet Temperature.
MPR
137 XSII 5.1.4.3 Impeller Strain Gage Test Results
The titanium impeller was instrumented
with I0 strain gages. Placement was based on
201 holographic and acoustic test results. Gages
I0 2', were placed on full and splitter blades with
two gages on the impeller backface. Figures
3O 90 through 92 show the strain-gaged impeller
prior to test. Figure 93 shows the compressor
rig _n the test facility with the slip ring
assembly installed.
303.4 MPR
687-078-Z29 EFFECTIVE STRESS 144 KSI) Data obtained during strain gage testing,
showed an excitation at 73,000 rpm (Figure 94)
Figure 83. AGT101 Impeller Stress at induced by a 4 per rev source (inletdistortion).
100 Percent Speed. The vibration was noted on five full blades at
two different occasions (Figure 95). The max-
determine vibratory blade and disk character- imum strain level measured was less than
istics. Holography and acoustic testing was 500pc. This correlates to a maximum blade
conducted, and these results then were com- stress level of 181.3 MPa (26.3 ksi) for the
pared with previously predicted analytical titanium impeller and 113.1 MPa (16.4 ksi)for
results. Table 11 summarizes these efforts. the aluminum impeller. Stress values for the
Figure 86 through 88 show the predicted vibra- aluminum impeller were ratioed by the modu-
tory mode shapes and the holography results lus of elasticityfor the two materials. These
for selected frequencies. Figure 89 presents a values compare reasonably well with the pre-
Campbell diagram for the titanium impeller. dicted stress values. Maximum strain levels
In general, test values agree closely with measured on the splitterblades were less than
analytical predictions. 150_e.
6o
Table 10. AGT101 Impeller Life Summary (-30
Material Properties).
O Material - A2219-T6 AI Development

Impellers
Burst Ratio 1.50
BR - 0.85 (UTS-3a)
aAve Tan
O Bore LCF Life (Starts to 10096 speed > 10,000 cycles

W/O Transient Thermal Stress)
O Stress Rupture Life Based on Estimates >100 hours

from Elastic Stress levels (3-D Creep
Analysis Has Been Initiated)
O Airfoil Allowable Vibratory Stress 34.5 MPa (5 ksi)

(Inducer Region)
0.13 MM Table 11. Summary of Vibratory

j,__-_{
I--,00o53
,,) Blade Modes (A_, 0 RPM, 70F).
Blade Predicted Acoustie Holography

Mode ISOVIB (Average) (Average)
. _ ,o.oogB,Nj],/ {] L
"1JI 1 22,174 24,877 25,277
•--"-] I
-.-- {0.0056INI
0.,4M. j/ // /1
,//"
2 10,340 10,427 10,411
3 12,341 13,108 13,138
4 15,172 16,858 17,127
5
J. ,
17,194 18,467 18,681
6 18,201 22,256
7 22,174 24,877 25,277
5.1.5 Compressor Design for Improved

Preducibility
To improve the long term producibility of

the impeller two items were considered,
OEFORMATION material and blade design. Early producibility
studies indicated emerging aluminum materi-
G67_)79-231
als could be considered for the impeller, how-
Figure 85. AGT101 Impeller Deformation ever due to availability, titanium was re-
at 100 Percent Speed. commended for testing. Following test, a
61
MODE1 FREQ[HZ) = 4681
MODE2 FREQiHZ) = 10,340
i
R
-- X NORMALMODESHAPE
l NORMALMODESHAPE
Ge7-079-233 "_p _,
Figure 86. AGTI01 ImpeLler Blade
Vibratory Mode 1.
Figure 87. AGT101 Impeller Blade
redesign of the impeller was initiated to im- Vibratory Mode 2.
prove producibility. The efforts regarding
these studies are discussed in the following Results show (Figure 97) that the ductility
paragraphs. improvement goal was successfully met.
Tensile strengths, as predicted, are lower than
5.1.5.1 Material Evaluation - Powder Metal the baseline data but continue to exceed
Aluminum AGTI01 design goals. Properties at elevated
temperatures, particularly above 288C (550F),
Material characterization and evaluation were observed as less sensitive to TMP vari-
of the ALCOA AI-Fe-Ce powder metal (PM) ables, as evidenced by the convergence of
aluminum continued. Baseline screening tests property levels at 343C (650F). Comparing
on sub-size pancake forgings indicated the the tensile strength of PM AI-Fe-Ce with two
material had the capability to exceed the of the high-strength aluminum alloys com-
design goals (Figure 96). Based on these early mercially available (A2219-T6 and A201-T7),
evaluations, component size pancake forgings shows it to be superior even after the 1000-
were procured 11.4 cm (4.5-inches) diameter hour exposed condition (Figure 98). Stress
by 7.6 cm (3 inches) thick. These forgings rupture capability of the alloy also is signifi-
were thermomechanically processed (TMP) cantly better than the two commercial high
specifically to improve alloy ductility. strength alloys(Figure 99).
52
34E
DIFFUSER
48
MODE 3 FREQ (HZ) = 12,341
MODE 15
17E
32
NORMAL MODE SHAPE
0 o
O 25 5O 75 ! O0 125
SPEEO, PERCENT
100,000 RPM AT TEMP

100 PERCENTRPM = 100,000
GB7-O78-23S
Figure 89. AGT101 Titanium Impeller

Campbell Diagram.
Figure 88. AGT101 Impeller Blade

Vibratory Mode 3.
High cycle fatigue (HCF) properties of the

AI-Fe-Ce alloy at room temperature and 232C
(450F), shown in Figure I00, tend to indicate
good HCF life capability. Stress levels of
207 MPa (30 ks_ at room temperature resulted
in runout at 10 ° cycles while cycles exceeding
107 were obtained at stress levels of 172 MPa
(25 ksi) at 232C (450F).
5.1.5.2 Straight Line Element Design
Impeller redesign activities were done on

the straight line element (SLE) blade defini-
tion to address near-net-shape powder metal GB7.-078-236
4 P73934-4
die forging. The basis of the redesign was to
maintain the performance levels demonstrated Figure 90. Strain Gaged Impeller, Front View.
63
Figure 91. Strain Gaged Impeller Blade.
GB7-078-239
Figure 93. Compressor Test Rig with

Slip Ring Assembly Installed in
Test Cell.
during Test 2A (Reference paragraph 5.1.2.3),

while approximating the arbitrary blade shape
using straight line elements. Duplication of
the Test 2A impeller exit blade angle distribu-
tion could not be fully accommodated using
SLE methodology. For example, to keep the
same hub and shroud blade angles in the exit
region resulted in an average exit blade angle
less than Test 2A impeller. This lower exit
blade angle gives rise to a lower slip factor
and higher work for the same tip diameter.
Therefore, the tip radius was reduced from
2.198 to 2.182 inches to compensate for the
P7393,4-3 reduction in average exit blade angle from
50.71 to 47.97 degrees. Blade loadings are
shown in Figures I01 through 103. All other
Figure 92. Strain Gaged Impeller, salientdesign features were satisfactorilypre-
Backface. served using the SLE technique.
64
loo.OOO RPM AT TEMP
100 PERCENTRPM = lO0,O00
IKsll
DO
MPa
17E 5O0-
\
AS-FORGED
=
IOE
9E
BE
16
6E
5E
4E
3E
2E
IE IOO'
O.
0 25 5O 75 100 125 10
SPEED, PERCENT
607-078-240
0
75 200 400 5O0 800 1000 IFI
Figure 94. AGTI01 Titanium Impeller
Campbell Diagram Excitation Source.
! I
MEASURED 1DO 50O C
LOCATION BLADE STRAIN (_EI
1 2 311 TEST TEMPERATURE
I 6 229 GB7.078-242
t 8 493
Figure 96. Tensile Properties of Forged
2 3 224
ALCOA PM AI-Fe-Ce.
2 9 291
5.2 Turbine Development

Ti 60.7+_78.6 MPa (8,8±11,4 KSII
AL 37.9_+50.0 MPa 15.5±7.1 KSli
The AGT101 turbine is a singlestage radial
Ti 73.1+_75.8 MPa (10.6_+11.0 KSII
AL 45.2__.47.6MPa 16.7±6,9 KSI)
inflow design. It included a ceramic stage and
a metallic stage. The metallic stage was a
duplication of the ceramic design but fabri-
cated in metal. To achieve operating temper-
atures, a dual alloy fabrication technique was
employed. The following sections discuss the
design and development of the turbine compo-
nents. Details of the ceramic turbine mechan-
ical analyses are given in paragraph 6.2.7.
5.2.1 Aerodynamic Design

{ Ti TITANIUM
Ti 111.4±66.9
AL 71.7+_42.1 MPs
MPa |10.4+_6.1
(16.6-+9,7 KSII
KSi} I AL ALUMINUM
Three maximum power design point vector
GB7-078-241 diagrams were evaluated with the engine driv-
Figure 95. Strain Gage Test Results. ing cycle. The three turbine designs were
65
fKSli IKSII
8O 15
MPa
500-
80 i ''%"
70 TM \ 7O
! • MOFIf)IEOTMP
\i
I ' -- 6O
5O
%.
_ %/1
BASELINE
DATA I0
0
• -......t"
_"-_ BASEL
DAT
4O
DESIGN
13O
MODIFIEDTMP
3O
MOOIFIEO
TMP
/\ _2o
2O 2O MODIFIED
TMP- --
i ', I BASEL
I0
I0 10 I DAT
O
o "[ .... "'"lT
0
75 200 4OO 000 80O 1000 IFI 75 200 400 600 8iX) 1000 IF) 75 200 400 600 800 1000IFI
i !
,_o _ c 1Do _ c ,_o _ c
TESTTEMPERATURE TESTTEMPERATURE TESTTEMPERATURE
GD7-07i1-243
Figure 97. ALCOA AI-Fe-Ce Forged Properties (1000-Hour Temperature Exposure).
[KSII
90
I
MPA
509-
70'
co,! ! [
5O Mra IKSII
698.5 100
ALCOAAI-Fe-Ce
I
4O ALCOAAI-Fe-Ce
[1000-HOUR
EXPOSURE) (FORGEOl
k=-
3O i (FORGED)
A-Z219-TB
69.O 10
\'
20 ¸
_'_ A-2219-T6
100.
•....-"A-2O1-T7 A-2Ol-T7
10
0
T = RUPTURE
LIFE. HOURS
75 200 400 600 800 1000 {FI
! I
16 17 18 19 20 21 22 23 24 25 26 27
100 50O C (T + 4601120 + LOGtJ X 10.3
TESTTEMPERATURE GB7_7&2_
6D7_7&244
Figure 99. Stress Rupture Properties

Figure 98. Tensile Properties of Aluminum Alloys. of Aluminum Alloys.
66
IKSll
MPA
tOO TENSIO'N/TENSION R = 0 A = I
600.
80
24C 175F1__.
60
=ffi 1,3
Z
0 SLE BLADE
40
Z
1.2
;..,--...
2O 232cf450FI
100- 0 AGT TEST 2A
1.1
0
lo2 io_ Io4 ,o5 I@ Io7 loS
1.0
CYCLES TO FAILURE NF
O.S
687_T&21
0.8
Figure 100. HCF Properties

of ALCOA AI-Fe-Ce. = 0.7
0.6
0.7 O SLE BLADE
- .'._.._
0.5
O,§ _ AGT TEST 2A _T_
0.4
Q 0.3
_. 0.4
:>"-'_ _ ,,=" f PERCENTMERIDIONAL OISTANCE
• 0.3 G|7_?B-249
0.2 Figure 103. Impeller Blade

Loading- Shroud.
0.1
o ,o 3o ;o
PERCENT MERIOJONALOISTANCE identified as -9.8 (low), -27 (high) and -37
GB7..078-247 degree (maximum) rotor exit swirl angle de-
signs. As shown in Fignre 104, examination of
Figure 101. Impeller Blade Loading-Hub.
the original optimized -9.8 degree case
LI
SLE BLADE showed significant performance decrement at
1.0 0 AGT TEST 2A
engine idle, and since "system design" objec-
tives are to maximize engine performance at
0.9 low power while maintaining acceptable
vehicle acceleration capabilities, the -27
0.8, degree exit swirl design was selected for the
baseline ceramic turbine. Vector diagrams for
maximum power and idle are presented in
_ 0.7.
Figure 105 (-27 degree exit swirl design). Also
ffi 0.8,
shown is the maximum power vector diagram
0.5,
for the 1149C (2100F) AGT101 metallic rotor
configuration.
0.4.
The maximum power design point cycle

conditions for the selected -27 degree rotor
,EnCENT
.E.iO,ONAL
B,ST.NCE exit swirl design are presented in Table 12.
G87,0_'9-2U
The turbine system station designations are
Figure 102. Blade Loading-Mid-Passage. shown in Figure 106.
67
+ 70_
÷60 0.60
÷ 50
>VIA _ MAX
.40 0.50
= V/Ac_HIGH .J
_=÷30 d/A LOW ¢[
I.-
+20 o.4o
.J
E uJ
I--
+10
.J
_"
X
Wd
0 o3o
O -10
I.- X)"
wl-
O
_" -20
O I o. o I-
-30 --d i
O
_10:
o._o _
O.O0
0.88
>-
U
Z
uJ
LL
o..6 1HIG
I,I,
LU
.I
<
I-
O 0.84
I-
-<
LU
I-
L4
). 0.82
OPERATING PE RCENT KW
CONDITION P/P)T-SYS N_ I NOMINAL
hp
0.80 IDLE 1_34 58.5 II 3.0

. 2.2
I
I
© MAXIMUM
THROTTLE
CRUISE
1.88
1.78
5,_.0
67.3
I
i
18.0
13.4
_ o.78 15.7
( 4-28 100.0 135.0
I 21.0
J I POWER
MAXIMUM 100.7
0.76 I
1.0 2.0 3.0 4.0
P/P}T-T' SYSTEM TOTAL PRESSURE RATIO ( Q - -)

[5.1 )
GB7-078-282
Figure 104. Off-Design Performance Charaeteristies as a Function of

Design-Point Exit Swirl for NASA-DOE Advanced Gas Turbine Powertrain
System Development Program.
68
II,
{AI CERAMIC MAXIMUM POWER CONDITION (9J CERAMIC IDLE CONDITION
a = 72 35 . _ = 72.57 ,
"qJ_t_ = " R,:'_ 2.6356 IN /3- 4435
•+. O,oj'Z__,+_ I_,._/3 -15.48 , ._
,,i+++,.3>
"_ - o.+,
_'-.IV"j,,,,+++.;, . _ jW,A,, :o.m
........ I _=r + " +I 32 "qll_/ L"'_ /3 = .60.02
v,,,,%,= ,.,.=:,+, I t. /" • _ I I ____. ,-.. +m.o
o,__ @ -
WU/A'cr - --0.786
TI N r= 1644K [2959.7R1 TIN " 1440K 12604.0R1

N " 100,000 RPM N = 55,000 RPM
P/P)T'SYS " 4.25 P/P)T"'SYS " 1.325
'q'r-sYS = 0.868 _T-SYS " 0.810
% v"e . 0.421 w,_ . 0.263
(CI METALLIC TURDINE (DUAL ALLOY( MAXIMUM POWER CONOITION
+ o,. 73,06
_" ---:--J-'_ ./ +, -1889

u/A_. - 0.9405 ,__-_ •
+,+,.
Vx/A" " 0 407 . RRM • 1.3494
P/P)T-T" 3.S_ wU/A"

a - -o.72?
_r-T " o.eTe
Wg..__..__ . 0.417
6 ALL ANGLES IN DEGREES
68t-07_283
Figure 105. Turbine One-Dimensional Vector Diagrams.
._ 69
Table 12. Turbine Design Point Cycle Conditions Sea Level, 29C (85F) Day, Maximum Power.
Dual Alloy
Ceramic Metallic
TIN, Turbine Inlet Total Temperature K (R) 1644.3 1422.1

(2959.7) (255r.71
(P/P)T-T, Stage Total-to-Total Pressure Ratio 4.000 3.866
Wg/9/6, Stage Inlet Corrected Flow, kg/see (Ibm/see) 0.191 0.189

(0.421) (0.4171
Wg, Mass Flow, kg/sec (lbm/sec) 0.3?3 0.378

(0.82311 (0.8325)
AH, Stage Specific work kJ/kg (Btu/lbm) 88.318 85.292

(37.9?0) (36.669)
PTIN, Stage Inlet Total Pressure, kPa (psia) 473.235 e. 449.380

(68.637) (65.177)
qt-t, Stage Total-to-Total Efficiency 0.901 0.892
N. Rotational Speed, rpm 100,000 100,000
Nl/9, Corrected Speed, epm 41863.3 45015.6
U t, Rotor Inducer Tip Speed, m/s (ft/sec) 701.04 640.08

(2300.0) (2100.0)
APT/PT)Exhaus t, Total Pressure Loss From Rotor Exit-to-Regenerator Inlet 0.0607 0.0312
APT/PT)Inle t, Total Pressure Loss From Turbine Duct Inlet-to-Stator inlet 0.010 0.010
ctExlt, Exit Swirl, Degree -27.0 -16.7
5.2.1.1 Stator Inlet Flowpath
The turbine inlet duct design was com-

pleted as shown in Figure 107. Three modified
NACA 16-021 airfoildesign struts have been
selected to reduce strut tip blockage and thus
decrease tip loading (Figure 108). The re-
sulting inlet duct velocity distributionis shown
in Figure 109. A 2-D boundary layer analysis,
which verifiesthat flow through the inlet duct
isstable, was performed.
5.2.1.2 Stator
ROTOR The stator vane profile design objective

was to maximize stator inlet loading (which is
CENTERBOOY
then relieved with end-wall contouring) and
SYSTEM TOTAL EFFICIENCY.
minimize exit loading from the throat region
,rTT-SYS, (_). Q =0.868
to the trailingedge. This approach minimizes
total vane loading and, therefore, secondary
607-07S-284 flow losses. Nineteen vanes were selected
based on turbine rotor blade vibration anal-
Figure 106. Turbine Design Summary yses.
70
0.110
CM IN
5,0 0.1oo
12.0 •
0.090-
I0.0 - 4.0
3 - STRUTS
% ___ STATOR 0,080 -

8.0.
3.0
== 0.070-
6.0, - ROTOR
!
2.0
0.080
4.0. _ COMBUSTOR
EXIT PLANE U6
I _ 0.050
/
2.0, ----- FINAL MOOiFiCATIOI_
_ O.04O-
O- 0
-2.0 -1.0 0,0 1.0 2.0 3.0 4.0 (INCHESI
0.030-
G|1.010-2117 ,xi,t COORO,NATE 0.020-

; _HUB CENTER BODY
' POINT
0.0 STAGNATION
Figure 107. Baseline Turbine 0.0 Ii0 21o 30 40 5'o o'o 7o 8;0
Inlet Duet Configuration. MERIOIONAL DISTANCE, INCHES
, i i t
0 50 I00 150 200
MD, mm
GB7-078.289
mm INCHES
Figure 109. Turbine Inlet Hub and Tip

5 0.2- C = 10.744 mm
Velocity Distribution for Baseline
Configuration.
SECTtON A-A
Two stator vane profile designs were com-
•_ (T,P
NES,ONI pleted. The baseline design is based on a
o_. o14 016 INCH trailing edge thickness (normal) of 0.508 mm
i
(0.020 in) and a fillet radius of 0.762 mm
0
(0.030 in). The design is consistent with
mm INCHESj previous turbine designs used for the specific
10- o.4-1 speed correlation (i.e., no additional aero-
C = 31.623 mm |1,245 INI
dynamic performance penalty is required).
o2-1 This profile, presented in Figure ii0, was
based on the selected vane number of 19.
Vane suction and pressure surface velocity
O- 0.0 I
distribution is presented in Figure 111 with the
end-wall contour established from the inlet
.o.2-I duct optimization. The second vane profile
i SECTION B-O was designed to facilitate tooling for an injec-
i INUe,EG,ONJ tion molded integral vane ring. Trailing edge
-10" -0.4 !
o o_ O;A O]S 1.4
thickness was increased to 0.762 mm (0.030 in)
INCHES
and the fillet to 1.220 mm (0.050 in). In
0 1IO i
20 !
30 mm
addition, the entire pressure surface and a
6B7-078-2"68 portion of the suction surface were modified
to straight lines, as shown in Figure 112. The
Figure 108. Turbine Inlet Strut effect of a linear pressure surface on the vane
Sections [Modified NACA 16-201 velocity distribution is presented in Figure
(tie = 0.30)]. 113.
71
LINEAR PRESSURE SURFACE
LINEAR SUCTION--_ _---__
LINEAR PRESSURE-% /_b, \
/ _// \ ,= 7o.o,,7.
/_t_. R= 90.6043 mm R= 70.9117 mm
/ _/f \ t3.567! iN) \

/'----1t f4, . _1 ,N) 12.70,s
,NI
/ j// o._,,,,, \ \
y/OR
I
I /
///
/ I
T_ESS
19 VANES
= 0.508 mm [0.020 'NI \\
_--
/
//I
///
,/_ T.E. NORMAL THICKNESS
: o.761
,. Io.o3o
"-
,N,
\
\ \
\
6D7.076-Z92
GO7 -076-2'D0
Figure 110. Ceramic Turbine Stator Figure 112. Ceramic Turbine Stator
Profile Design. Profile Design.
I i I I I i I I
bINLE T _ 11.684 mm
1 ] n_9o..o_.. \ l R:----,'. 1
o
I.O[ / |3.5671INI I
13.5671INI
_ 10
19 VANES / R 709117 mot INLET FLOW ANGLE = O DEG
/ /Io VANES ' INLET FLOW ANGLE = OEG
/ IR = 70.0117 mm I --1 .,,_-"--
J _12.7918IN] I
•-"l F"-o.2s7s
iN _'.'_
5 o.s
_, 0.5
/ _ _--_-O_,G,N_L LOAO,NG
- - _- LINEAR' PRESSURE
G _ "''" "-""
O ! SURFACE
O 10 20 30 40 50 60 70 80 90
0 10 20 30 40 50 80 70 80 90
mlT471-MI
M7.071i-ZI3
Figure 111. Ceramic Turbine Stator Figure 113. Ceramic Turbine Stator
Velocity Distribution Velocity Distribution (Linear Pressure
(Baseline, 19 Vanes). Surface, 19 Vanes).
5.2.1.3 Rotor The turbine rotor blade angle distribution

for the hub, mean, and shroud streamlines are
Figure 114 shows the meridional view of shown in Figure 115. Calculated radial rotor
the ceramic and metallic rotor design. Since loss distribution as a function of blade height
the metallic rotor blading geometry is identi- is shown in Figure 116. These calculations are
cal to the ceramic, the metallic rotor is consistent with past design and Garrett empir-
achieved by a simple tip cutback to reduce ical experience wherein rotor exit survey data
stresses in the rotor. shows that the radial loss characteristics for
72
b
CERAMIC ROTOR METALLIC ROTOR

FLOW PATH FLOW PATH 12150F)
Ut = 701.04 M/SEC Ut = 640.08 M/SEC

75 - 3.0 (2300 FT/SECI. -. (2100 FT/SECI
_ b = 6.5405 mm
• _ _ 10.2575 IN] ,,
2.5 ? I / , -.i,,,I _ b = 7.2136 mm
/ I It " z ', ', {0.2840 IN]
aT= 50.9442 mm It lii I \
12.6355 IN) \ J I %
\ nT = 61.1226 rnmi
50 2.0
"i 1.5 -?N. = . m

--r-/-l-t
I / I \
--rot R = 43.5127 rnm
RS= 40._1m,. _ ,.7131,It RS= 40_ ",.mI \ I1713,,Hi
'1?'"', ,,.i,H,......
I
25 1.0
/I "---.--._..__At,
- /I "----...._J
',,
0.5 / i-- l1 ._._INI --{ I = 21.3537 mm i J- 11.55 Ill --I I1 = 21.3537 ran1
/ ......... [0.8407 ,Hi L -- -- [0.8407 INI
0 0 / i I I I I I i i I i i
0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0
Z. IN Z, IN
I I I l Ij I
0 _ 25 50ram 0 25 50nlm
Z. mm Z. mm
687-078-294
Figure 114. Ceramic and Metal Turbine Flow Paths.
radial turbines are similar. That is, in the hub 5.2.2 Turbine Aerodynamic Performance
region the losses are low (due to lower loading) Testinl_
and progressively increase to about 80-percent
blade height, and then are fairly uniform to The test rig, shown in Figure 119, is a
the tip. Velocity distributions for the rotor double overhung assembly supported by two
are shown in Figure ll7. The velocity distri- angular contact ball bearings. During the
bution for the ceramic and metallic rotors are turbine and rig design effort, an investigation
similar at shroud, mean, and hub streamlines was conducted concerning off-design perfor-
except in the inducer area where the higher mance mapping, particularly accurate perfor-
inducer loading of the metallic rotor is evi- mance measurement at the idle-cruise regime
dent. of the CFDC. A summary of the investigation
and subsequent impact on the cold turbine test
5.2.1.4 Exhaust Diffuser rig design is outlined below.
For the exhaust diffuser, additional calcu- The aerodynamic analysis has shown cold
lation stations were added to the rotor flow turbine rig power requirements at maxi-
solution, since the velocity distribution into mum power of 19.3 to 20.1 kw (26 to 27 hp)
the exhaust diffuser is strongly dependent on with a temperature differential across the
the rotor exit radial swirl and associated loss rig of 98-107C (208-224F). The power
distribution. Both uniform and non-uniform level drops to 0.56-2.38 kw (0.75-3.2 hp) at
inlet conditions were evaluated. The current idle-cruise and the temperature differenti-
configuration is shown in Figure 118 for an al decreases to 3-19C (37-67F) for these
area ratio of 1.75. conditions. The kw/rpm ratio (hp/rpm)
73
b
CERAMIC ROTOR
13 BLAOES.
AXIALLENGTH: 39.370 mm (I.5,5 INI 13 BLADES,
Ut = 701.04 M/SEC 12300 FT/SECI
70 I I -I I i _, TIN = 1371C [2500F}
m" 60 _-SHROUD STREAMLINE- A_IAL LENGTH = 39.370 mm 11.55 IN]
.J f
¢3 50 I MAXIMUM POWER
Z f
<_ 40 /,-
w " 30 METALLIC ROTOR
,< (3 20 IMOO IJ
ma 10 13 BLADES
Ut = 040.08 M/SEC 12100 FT/SECI
% TIN = 1149C (2100F)
AXIAL LENGTH = 39.370 mm (I.55 IN)
7O MAXIMUM POWER
I I I I
60 _ MEAN STREAMLINE --
-i
so l I I SHROUD STREAMLINE
40 f4
1.10 i
_SUCTION
!
uu" 30
1.00
10 -"
0.80
_ 0 _
0.60
7O <
I I I I ' • 0.40
uJ" 60 ----HUB STREAMLINE--
-I LU
50 i >
Z
f F- 0.20 SURFACE _------ J •
<_
_luJ
4o -'>l
w
2O ¢ 0.00
' 0 -0.20
0 10
MEAN STREAMLINE
20 30 40 50 60 70 80 9O lO0 1.00
PERCENT MERIDIONAL DISTANCE o_
I
gi1_178-_6 Lu
m 0.80 I
Figure 115. Turbine Rotor Blade Z 0.60
Z f .._
Angle Distribution. (J
< 0.411
SHROUD 1_
ql uJ
> 0.20
../"
k-
3 0,00
8O
i,r
-0.20 J
I-- AVG O CALCULATED

FROM Pexit, Tel HUB STREAMLINE
1.00
LU u,
'I" w J
UJ 0.80
<
40 z
_n 0.60 o-
i x
WITHOUT T E % < 0.40
2O "LOCKAC
E w
> 0.20
HUB 0
WITH
BLOCKAGE 'I
T E
_
,
r-
<
BJ 0.00
0.92 0.94 0.96 0.98 tOO -0.20

0 10 20 30 40 50 60 70 80 90 100
Q = • "_lsICp
GB7.078-297 % MERIDIANAL DISTANCE
gllT-OTli-ZI6
Figure 116. AGT101 Rotor Radial Figure 117. Turbine Rotor Velocity
Loss Distribution. Distribution.
74
MAXIMUM POWER CONDITION
(Ae/A i = 1.75, L/AR = 5.31)
Z - AXIAL LENGTH• mm
SURFACE MACH NO
50 75
0.56
5.70 I
G"
,* Jl'°'" ,11
•
5.50 _UMP -- .j -
DIFFUSER
0.50 It IIN2I - EXIT "
5.00 125
CM2
10 t _---.._
4.50
-9O
4.00 .I_
DIFFUSER
-8O
3.50 _--AREA RATIO_ --
_ 0.38 /I - 1.75 I 0,
"'..
',..
,," b,_ k /- 75
I- AelAi - 1.75 I
"',',
b
oL' I
Un"
0.,jK
__ UNIFORM
PT.TT
NONUNIFORM
AT RMS
iNLET
RADIUS
INLET
CONOITIONS
_.
N_
I/ L . • l °= I ,; jL
50
"---.,_
CONDITIONS
\
°20 / "5oRs = 11.713 INI
0.16 _..__. _ _ ---I_ _ 25
7,0
0.13 [
0 10 20 30 40 50 80 70 80 90 100 RR = 21.3537 mm'm'm'm'_
°'e° 1.r" 2.0 2.5 3.0 3.3 3.5
10.8407 INI
PERCENT MERIDIONAL DISTANCE Z - AXIAL LENGTHIIN)
697,078-299
Figure 118. Exhaust Diffuser Velocity Distribution.
ratio varies from 3.728 x 10 -4 (5 x I0 -4) - The high speed, direct-coupled, torque
at maximum power to a low of 2.237 x meter represents the instrumentation
10 -5 (3 x I0 -5) at the idle condition (see with accuracy both for high and low
Table 13) power mapping (mfgr., Torque Meters
Ltd., Northhamption, England)
O Preliminary results of the heat transfer
The platinum resistive temperature
analysis of the cold turbine test rig indi-
sensors provide secondary or backup
cates heat transfer losses could introduce
measurement of performance and pro-
significant errors. Revisions were intro-
vide an important reference for the
duced to the test rig design to provide
temperature surveys at various points
insulation both internal and external to the
in the turbine system.
rig, and to introduce conduction heat
transfer barriers where possible o The combination of direct coupled torque
meter and platinum resistive temperature
O The survey of the industry for precision sensors as backup provided a desirable per-
instrumentation confirmed the availability formance measurement system to assure
of platinum resistive temperature sensors design goals and to properly assess CFDC
and direct-coupled, high speed electronic mileage estimates during the cold turbine
torque meters with the desired accuracy mapping program.
75
COMBUSTOR EXIT TT -6 THERMOCOUPLES
@ - 3 ROSEMONT PROBES
PS " 8 KIEL PROBES
TRANS,T,O,
O°CT PS HUB SHROUO
@TO@
ROTORSRROUO
@,o@
PS HUB L
ROTOR EXIT PS 8 KIEL PROBES
0
@
EXRAUST.O_.FUS_ PS HUB SHROUD
_TO_
O,_F,SER._T
BE_ PS HUB SHROUD
REGENERATORiNET TT 9 THERMOCOUPLES
J
;!: oo,
/
o.,,.,& ....,.o.,o
® -3 ROSEMONTPROBES
PS" SHROUO _, _._ _ TEST

RIG_ DEABeox
ROTOR CLEARANCEPROBES 4 RADIAL PROBES " 5_ TORQUEMETER
4 AXIAL PROBES
"ASSUMED
BEARINGTEMPERATURE.
MECHANICALIN_RUMENTATION
VIBRATION. OIL FLOW.
OIL TEMPERATURE
TORQUE
TORQUEMETE_)
MASS FLOW SONIC NOZZLE
©
GB7-076-299
Figure 119. Cold Turbine Test Rig.
5.2.2.1 Test Procedure bine rotor. Capacitance probes were used in

all planes (radial, axial and backface) to assess
During mechanical checkout tests, the rig rotor running clearances during aerodynamic
was configured with a "dummy" rotor for tare mapping.
loss calibration tests. The dummy bladeless
rotor was designed with identical mass and The torquemeter was configured with the
inertia characteristics of the bladed rotor. appropriate shaft depending on the selected
Test rig operation was accomplished to full mapping range. Prior to each data run, a dead
speed using the dynamometers in the driving weight calibration was performed. Data re-
mode and under full vacuum to eliminate cords were obtained following an iterative
dummy rotor disk friction characteristics. procedure wherein inlet temperature at a de-
sired pressure ratio was adjusted to obtain a
The test rig then was assembled with the rig discharge temperature within several de-
AGTI01 871C, (1600F) machined Astroloy tur- grees of cell ambient, thereby minimizing heat
76
Table 13. AGT101 Cold Turbine Rig Operating Conditions.
4m
Maximum Cruise Cruise Maximum
- Parameter Power No. 1 No. 2 Throttle Idle
P/PT-System 4.263 1.787 1.639 1.676 1.341
TIN, C (F) 151.3 70.8 63.6 63.1 47.4

(304.3) (159.5) (146.5) (145.5) (117.4)
TOU T, C (F) (Assumed) 26.7 26.7 26.7 26.7 26.7

(80.0) (80.0) (80.0) (80.0) (80.0)
g.
N, rpm 52,000 30,804 28,379 24,870 25,849
NI/O, rpm 41,863 28,194 26,251 23,025 24,507
W,/'8/6, Corrected 0.192 0.171 0.164 0.171 0.117

flow, kg/sec (lb/see) 0.423 0.377 0.362 0.376 0.259
W, Physical Flow, 0.1594 0.0524 0.04590 0.04459 0.02667

kg/sec (lb/sec) (0.35141) (0.11556) (0.10120) (0.0983) (0.0588)
Reynolds Number, 1.0 x 105 38,128 34,130 33,150 20,600

Re, = W/RtP
PTIN, kPa (psia) 102.153 33.92 30.61 28.61

(14.816) (4.92) (4.44) (4.15) (3.52)
PTEXIT, kPa (psia) 24.13 20.00 18.68 17.07 18.098

(3.50) (2.75) (2.71) (2.476) (2.625)
kw 19.95 23.26 1.70 1.630 0.5569

(26.76) (3.119) (2.28) (2. 186) (0.7468)
kwlrpm (hplrpm) 0.00037 0.000075 0.000060 0.000067 0.000022

(0.0005) (0.00010) (0.000081) (0. OOOO9) (0.0O003)
Note: Gamma effects not included will alter results by few percent.
transfer effects. Speed then was set based on Figure 120 shows the unit installed in the
inlet temperature, and inlet pressure (for a test facility.
given pressure ratio) was adjusted to obtain
the desired turbine Reynolds number. This A series of cold rig tests were conducted
procedure was followed for all performance to aerodynamically evaluate the baseline
tests and data records obtained after system ceramic configuration (as represented by the
stabilizationfor approximately 10 minutes. Astrology rotor). Baseline ceramic cold rig
77
....... - #,._,_ ,- i_i/
0.00
1985
0.86
0.84
P-
OPERATING
_ o.82
CONDITIONS % N/,yl8
O IDLE I 53.4
__ 0.80
O IDLE 2 58.6
_" 0.78
CRUISE 1 62.4
I"1 CRUISE 2 67.0

• 0.76
r-l, FLAT RATED 87.0
0.74 O MAX POWER 100.0
__ MAX
0.72
POWER___
OESIGNt
0.70
3 4 5
COLD PRESSURERATIO
607-078-303
Figure 121. Baseline Ceramic Cold Turbine

Rig Test Results, System Efficiency Versus
607-078-302 Pressure Ratio.
...... 1985 PREDICTION {NASA CR-1651751
.... 1979 PREDICTION
Figure 120. AGTI01 Test Rig
0.00 _"_0_ 1982 COLO RIG TEST | _n = 0.028
Installed in Test CeIL
- I sl t TECHNOLOGY
L/'_"_-._--4------_----'4"-_ "''" _[_I IMPROVEMENT
0.84 ........ 19791985
test results are directly applicable to the

1371C (2500F) AGT101 design. _ 0.00
wi
Figure 121 shows turbine system efficiency z 0.76

characteristics as a function of system pres-
sure ratio over the operating conditions. The
system is rated, for cycle analyses purposes, _" o.72
4O 5O 00 7O 00 9O 1DO
from the combustor exit-to-regenerator LP RPM x 1000
inlet. The overall system efficiencies at the G87.07&304
cycle operating rpm and pressure ratios are

cross plotted in Figure 122. Table 14 presents Figure 122. Baseline Ceramic Cold Turbine
a summary of the AGT101 turbine develop- Rig Test Results, System Effieieney
ment. Versus Engine Speed.
At idle, an 84-percent efficiency was Table 14. AGT101 Turbine

achieved compared with the 79 percent pro- Development Test Summary.
gram goal. Cruise performance was 84.5 per-
cent versus 84.3 percent predicted. At maxi- Efficiency
mum power, an initial design efficiency goal 1371C (2500F)
of 83.7 percent was projected; test results rProgram
show that 85.2 percent was achieved, slightly Test Goal
below the program goal of 86.5 percent. (1982) (1985_
Idle 0.840 0.790
An exhaust system loss coefficient is pre- Cruise 0.845 0.843
sented versus diffuser inlet (rotor exit) swirl Maximum Power 0.852 0.865
78
angle in Figure 123. Results of baseline MM liNI
ceramic tests (indicated by square symbols) 75, 3.0
are compared with metallic turbine test data
(circles) and an average curve. The additional
data confirms that the exhaust system loss
2.0
characteristics are low and decrease with in- 50.
creasing swirl. This significantly increases the

turbine system performance at the engine idle
and cruise conditions where swirl angles are 55 25. I.O liliil
degrees and 40 degrees, respectively.

Ill l
II I
O.O
OiO 110 210 310 410 [INi
I---to)
_=. I II I I 1,, J i
1-2.0 --AVERAGE EXHAUST_SYSTEM
OLOSS SYMBOLS '-T'-- GRID
i_._V CHA_ACTERISTZS
l ! 0 TEST
1 (METALUCI_L.__ Z-SECTION
r, I.oi ill I
.- ....
ill/ /
n_-_x_ B I
-o-.,,,
I .LJ
CERAM_C_
I --
I I I
Gll7-076-306
IO 20 30 40 50 60 70 80
Figure 124. AGT101 Metal Dual Alloy
a M, DIFFUSERINLET AVERAGESWIRL ANGLE.OEGREES Wheel - 13 Blades, Conical Astroloy Hub.
617-079-305
The predicted temperature distribution at

Figure 123. AGT101 Exhaust maximum power is shown in Figure 125. Idle
System Loss Characteristics. and maximum power equivalent stress distri-
butions are shown in Figures 126 and 127
It should be noted that the original off respectively. Blade vibration analysis results
design predictions were based on increasing are shown in Figures 128 and 129. Table 15
diffuser losses with increasing swirl angle and summarizes the dual alloy rotor mechanical
thus yielded a low value of turbine system properties.
efficiency (0.762 percent) at idle. The marked
decrease in exhaust system loss contributes
significantly to the baseline ceramic turbine
system efficiency (0.840).
J---- 927 117061
\ N : IO0,OOO
RPM
5.2.3 Metallic Dual Alloy Rotor _. TIT : 1149C [210OF}
_ TEMPERATURE IN C IFI
_ 9_ ISOPLETH INTERVAL = 55.5C [100El
5.2.3.1 Dual Alloy Mechanical Analyses
Thermal and stress analyses were con-

ducted for the AGT101 (1149C, 2100F) dual-
alloy turbine rotor at maximum power
(100,000 rpm) and idle power (55,000 rpm). /' mi1,=l
' -- /
/649112001 i 427 800
I,,
I \ _ / / 5,141
110001 371 17_I
Turbine inlet temperature is limited to 1059C _. \ \ ([ I _tij I
(1930F) relative blade temperature to maxi-
mize life. The rotor is comprised of a MAR-M
247 cast blade ring and a powder metal astro- |ll4ll_OI
loy hub HIP bonded as shown in Figure 124.

The rotor is bore cooled using one percent of Figure 125. AGT101 Dual Alloy Steady-State
total mass flow. Temperature Distribution.
79
IDLE CONDITION
• N = 55.000 RPM VIBRATION FREQUENCIES
o_ • TIT = 1689C (1993F) 100"/, RPM - 100,000
• COOLING FLOW RATE = I% X WA CAMPBELL DIAGRAM

55% RPM 55,000
KSI MPa
/ _ '_ f_ 120) 137995

/, _ \ / A fsoi _.s42
48
_r,-.,. " _ _ _ 1401 275.790
/ _ \ _, _ _ (68J 344737
_' _ ]SOPLETH INTERVAL = 68.9 MPz [10 KSI)

4O
_: 32
U 9E
Figure 126. Stress Distribution for

.u
lg
24
/ ,
I
I / I /10E
Dual Alloy Turbine Wheel
16 jf'_SE
KSI MPa
I 6E
ilOi 68.947 I 5E
12Oi 137.805
1301 2OO,942
f40i 275.790
(50i 344.737
(68i 413.685
i I I i _ !
i701 482.633
0 25 50 75 100 125
fool 551.550
fOOl 62O,528 SPEED, PERCENT
Gli7471i_10
(IDOl 669.475
(168,51 1.2'99,681
N = 100,000 RPM Figure 128. AGT101 Metal Dual Alloy

Wheel
TIT = 1149C {210OF}
_,,__1_ PLETH INTERVAL =1168; KMIPI' 5.2.3.2 AGT101 (1149C_ 2100F_ Dual Alloy
Turbine Rotor Fabrication
The turbine inlet temperature for AGT101

b (1149C, 2100F) engine development imposes
\ ot.ma x = 1299.7 MPi 1168.5 KSI] unique materials requirements for the turbine
rotor fabrication. One material cannot satisfy
all mechanical properties requirements. The
Figure 127. AGT101 Dual AHoy limiting conditions for airfoils are stress rup-
Steady-State Equivalent Stresses. ture and creep, and tensile strength and hub
low cycle fatigue.
Use of the dual alloy rotors was discon- Rather than compromise the turbine tem-
tinued following a rotor failure. Subsequent perature capability and performance, a mate-
careful inspections of the bond joint between rials mix or dual alloy fabrication concept for
blade ring and hub led to the conclusion that the radial turbine wheel was preferred.
the bond joint was not sufficiently reliable to Materials selected for the dual alloy wheel
continue the test program with the dual alloy fabrication are a cast MAR-M 247 bladed ring
rotor. with directionally solidified (DS) inducer blade
3o
ISOPLETH INTERVAL = 2
RPM = IO0,OOO
I-.-zDELTA[YI Lc OELTAIY)
MODE I FREO = 7679 Hz MODE 2 FREQ = 15448 Hz
OELTAIYI L., OELTAIY)

MODE 3 FREQ : 19364 Hz MODE 4 FREQ : 22709 Hz
l
k
OELTAIY) mL._,z DELTAIYi
MODE 5 FREQ : 25818 Hz MODE 6 FREQ = 31660 Hz
G87-078-311
Figure 129. Turbine Vibration Modes at Maximum Power.
81
Table 15. Dual Alloy Turbine Wheel Design Parameters.
Inlet temperature C (F) 1149 (2100)
Number of Blades 13
Material (Blade Ring/Hub) MAR-M 247 (DS)/Astroloy
Weight kg fib) 1.73 (3.82)
Ip kg-in 2 (lb-in-see 2) 0.001024 (0.0092)
Characteristics at Max Power (100_000 Rpm)
Tip Speed m/see (ft/see) 640.08 (2100)
Avg. Tangential Stress MPa (ksi) 441.264 (64)
Burst Margin 1.49
Energy Joule (lb-in) 56876.6 (503,400)
S-R Life (Hrs) at R = 49.784 mm (1.96 in) 103
tips for higher temperature capability and

powder metallurgy (PM) Astroloy for hub con-
struction. The mated hub blade ring geometry
in shown in Figure 130.
Dual Alloy Wheel Assembly
The Astroloy was purchased as HIP powder

consolidated logs. The TRW Prototype
Foundry developed the DS blade casting tech-
nique.
The hubs and bladed rings were machined _ PM LC AsTRoIny
as mated pairs to a 5-degree tapered fit

(Figure 130). Borazon (cubic BN) grinding
wheels were used to arrive at a surface finish MAR-M 247DS
typically about RMS 62 (Figures 131 and 132).

601-076-312
Past experience at Garrett shows that the use
of other grinding wheel compositions (ie,
A120 3 or SiC) results in foreign particles Figure 130. Dual AHoy Hub/Bladed
becoming imbedded in the work surfaces. Ring Assembly Cross Section.
82
ORIC_N.IL PA-G_ 1_
OF. POOi<. QUALITT_
77848-4
Figure 131. As-Cast AGT101 Bladed Rings and Machined Dual Alloy Wheel Assemblies.
Exposed edges of mated assemblies were Solution I079C (1975F)/4 hrs/rapid argon
brush nickel 0.005-0.010 mm (0.0002-0.0004 cool
inch) plated and all parts were then cleaned plus S71C (1600F)/8 hrs/AC
using Garrett developed procedures. These 972C (1800F)/4 hrs/AC
include a sequence of steps starting with vapor 649C (1200F)/24 hrs/AC
degreasing followed by MEK solvent, HCf solu- 760C (1400F)/8 hrs/AC
tion, distilled water, and Freon rinses. The
mated assemblies were hermetically sealed
using a vacuum brazing technique. Bond Zone Evaluation t
A representative macroetched cross-

The first batch of five dual alloy assem- section of an AGT101 dual alloy wheel, shown
blies were diffusion bonded during exposure to in Figure 133, indicates the presence of a
a HIP cycle consisting of 1204 *14C for four continuous bond. Some braze alloy penetra-
hours at 103.42 MPa (2200 ,25F/4 hrs/15 ksi) tion is detected and a view of an unetched
at the Howmet Turbine Components Corpora- section of the bond zone shows the absence of
tion facilities. After bonding, all five assem- porosity due to the Kirkendall effect or other
blies were heat treated using the following causes (Figure 134). The area of braze pene-
standard Yo-Yo procedure developed for LC tration will be removed during final machin-
Astroloy: ing.
83
Q
5.2.4 Ceramic Turbine Rotor Redesign
The rotor aerodynamic design is the deter-

mination of the blade geometry at hub, mean,
and shroud (and other streamline locations as
required), which satisfiesthe optimum veloc-
ity diagram and distribution consistent with
the work output requirement. The radial rotor
is designed through a number of steps, which
are given in the flow chart in Figure 135.
To improve ceramic manufacturing cap-

abilities, the baseline rotor was redesigned
with increased blade thickness and modified
blade shapes, as shown in Figure 136. The
redesigned rotor preserved the original blade
angle distribution to match the baseline veloc-
ity diagram. The rotor blade angle distribu-
tion for the hub, mean, and shroud streamlines
are shown in Figure 137.
zzs4n-2 Blade surface velocity distributions, calcu-
lated from the axisymmetric flow solution at
the shroud, mean, and hub streamlines are
Figure 132. Machined AGT101 Dual Alloy shown in Figures 138 through 140 and compare
Wheel Assembly--PM LC Astroloy Hub and baseline design rotor loadings. The figures
MAR-M 247 Blade Ring. illustrate the hub-to-shroud variations in the
surface velocities and the blade loadings, as
well as the variations along the flowpath. The
blade is more heavily loaded along the shroud
than elsewhere because of lower solidity.
Loading near the rotor inlet where the flow is
nearly radial, is considerably higher than the
loading near the rotor exit where the flow is
nearly axial. A eomparison of blade loadings
between the redesigned and baseline rotor
shows that the redesigned rotor loading at the
exit eonverges to a single value (which was
lacking in the baseline design). Theoretically,
this means highe_ work extraction in the rede-
signed rotor. HOwever, the redesigned rotor
loading indicates some diffusion downstream
of the throat when eompared with baseline
rotor design. The performance penalty due to
the surface diffusion probably offsets any
benefit that could be elaimed due to loading
closure. Thus, no efficiency changes in the
redesigned rotor are expected.
Figure 133. Microetched Cross Section Detailed stress analysis of the ceramic
of AGT101 Dual Alloy Wheel Test Specimen. turbine rotor is discussed in paragraph 6.2.?.
_A_,o.t.,,\L " .... "

84
O_ £OOR QUALITY
• , .... • ;..:
• '_ " D • " "
• a "°* • e'IP _
" L° "
• q e
" go
BRAZE
• " 2 ..
"' _ "s _ o ALLOY

• ," " o PENETRATION
. . .
.. °, .,.
• .
..
.OF POOR QUALITY
° •
UNETCHEO 50x
CAST MAR-M 247
Pl LC ASTROLOY
KALLIN(S ETCNANT
GD7-076..317
Figure 134. Braze Closure After HIP Diffusion Bonding (Sample was Subjected
To Yo-Yo Heat Treatment).
85
1
] ASsu
IuAOe
_mLEOv_xa_mORAL
G_MEtRT l j-"]__ """_::: ° '"_":'_ .... /
-I ,..AM j+ SHROUOSTR_
RANSMIT5 INFORMATION
AT CALCULATION 13 BLADES
AXIALLENGTH
STATIONS
= 39.37 mm 11.55 INCH)
j OLAOE PROFILE O_FINITION I

._ 80-
/ "ICI_
._vls[
--
_.;;'o';.......
• nlRTRI|IITtO-
I
|
At¢ALCULATIO.
TRANSMITS INFORMATION
°:
I "
MEAN STREAMLINE
Y
I SLAOE CHANNEL _
I FLOW ANALYSIS I I_ 0
SATISFACTORY _
F--
/
•" 6R-
DISTRIBUTION _-
_- i '?"_R"_+
°'
HUBSTREAMLINE S
._,T,o, j
I,-,._.., I
GITl_ 0
rERCENT
MERIOIONAL
OISTANCE
Figure 135. Radial Turbine Design Flow. 617471-326
2.70- Figure 137. Turbine Rotor Blade Angle

Distribution.
Z._41
-
2.46- t
MO DIFI EO-,-.,,..,,. I t
BLAME _ I tI
//i
Z.34-
Z22-
Z.lO-
/ 1.0
SHROUDSTREAMLINE
-- ORIGINAL ROTOR .- ---

iii
REOESIGNEDROTOR_
t.g-
_ORIGiIIAL
o_ o.e SUCTION SURFACE -j"
.- ._"
,,
.//_'7/j
SLAG[
I1
1Jll-
; O.D;.
m O.l
_+_,+
I %/
1.74-
0.2, _'_/'_
IJIZ -
10 20 30 40 50 60 70 80 90 100
i PERCENT MERIDIONAL DISTANCE

0+2 GDt-O7B-327
Z. INCR[S
GB7_)78-325
Figure 138. AGT101 Rotor Surface

Figure 136. AGT101 Ceramic RotorZ-Section. Velocity Distribution (Shroud).
:+er, t_-j+/\L +_+,,+r--_

86
i.E POOi_ QUAL, IT_
1.0"
m ORIGINAL ROTOR
o.a.
REOESIGNEO ROTOR MEAN STREAMLINE f f _"_,_
,," f "
0.6- /"" _ $__'t _
._ o.4. _ .... _ J__._ _'/"
0 I0 20 30 40 50 90 70 80 90 I00
PERCENTMERiDIONALDISTANCE
GB7_78-328
Figure 139. AGT101 Rotor Surface Velocity Distribution (Mean).
ORIGINALOOTOR
HUB STREAMLINE
_m_ flEOESIGNEO
ROTOR
TE
PRESSURE
0
O I0 20 30 40 50 90 70 90 90 190
PERCENTMERIOIONALDISTANCE
gl7_7_329
Figure 140. AGT101 Rotor Surface Velocity Distribution (Hub).
87
5.3 Combustor Development
5.3.1 Combustor Design /
5.3.1.1 Variable Geometry
Emphasis on gaseous emissions, multifuel

capability, and fuel economy in an automotive
duty cycle created unique requirements for
the AGT101 combustor. The combustor prob-
lem statement included the requirement to
operate on gasoline and diesel fuel, with alter-
nate fuels designated by the NASA project
manager. Emissions goals were set as follows:
Oxides of Nitrogen, NOx 0.24 gm/km

(0.40 gm/mile) MAX POWER IDLE
Total Unburned 0.25 gm/km PM/PV LENGTH, CM IINI 1.1-5,1 10.44-2.01 0.36-1.7 (0.14-Q681
REFERENCE VELOCITY. M/S fFPSI 78.2 (25QJ 17.2 156.5i.

Hydrocarbons, HC (0.41 gin/mile) PM/PV RESIOENCE TIME. MS 0.15.,0.7 0.2-1.0
Carbon Monoxide, CO 2.1 gm/km PM/PV AIR. PERCENT Wa 59.1 17.1
(3.40 gm/mile) NOZZLE A/F 20.0 35.4
Total particulates 0.12 gm/km PM/PV F/A 0.0234 0.0183
PRIMARY ZONE _, MS 1.33 5.0

(0.20 gm/mile)
LOADING PARAMETER, 0,003 (0.0197) 0.00098 (0.00621
KG/M3-KPA-SEC ILOM/FT3-ATM$EC)
GII7 070 ZSO

The ceramic rotary regenerator, designed to
recover waste heat, creates combustor inlet
temperatures up to 1093C (2000F), which Figure 141. Piloted Premixed/Prevaporizing
make it necessary that ceramic combustor Variable Geometry Combustor.
components be capable of withstanding the
extreme inlet and discharge temperatures. evaluation utilized an element test rig with
CO 2 injectionto determine the fuel/air mixing
In response to these requirements, and performance of various combustor configura-
based on experience with a 46-month Variable- tions. A number of different combustor geom-
Geometry Combustor Development Program etries were evaluated for fuel-air mixing with
jointlysponsored by the U.S. Air Force Propul- the best configuration being a radial/axial
sion Laboratory and the U.S. Naval Air Propul- swirler combination. Analytical modeling of
sion Center, a variable geometry concept was this configuration showed streamlines and
examined for the AGT combustor. The initial temperature fields to be very uniform. The
variable geometry design included an axial element test rig was then used to test concen-
pilot combustor in-lin#with the main combus- trations of CO, CO 2, NO, and HC emissions.
tor to assistin cold starting. Figure 141 shows Four different radial/axial swirler combina-
the piloted premixed/prevaporizing variable tions were evaluated and are depicted schema-
geometry design with itsperformance specifi- tically with airflow splitsin Figure 142. Con-
cations for idle and maximum power condi- figuration No. 1, with a fullradial and no axial
tions. swirler was the best all-around performer in
emissions testing. Correcting the data to
Concurrent with the variable geometry ef- engine temperatures and airflow rate based on
fort, a fixed geometry design was evaluated published lean-burn flame tube data revealed
empirically and analytically to determine the that CO and NO x were below design levels
possibilityof meeting emissions goals without (Figure 143). Based on these results,the lean-
the associated complexities of variable geom- burn combustion concept was deemed capable
etry. One of the tests conducted in this of meeting emissions requirements and was
88
CONFIGUIqATION NUMBER 1 CONFIGURATION NUMBEq 2
CONFIGURATION HUM6ER 3 CONFIGURATION NUMBER 4
FULL AXIAL HALF AXIAL

5&i PERCENT 25.0 PERCENT
HALF RADIAL
I " I. " 5h8 PERCENT FULL RAOIAL l
_.__10,5 PERCENT L_"_ ' 8-9 PE RCEN T
GD7076-251 [__j____ "PRIMARY ZONE

,'R,,,,,,,
JETS
Figure 142. Combustor Configurations GB7 078 _'._3
Investigated In Emissions Testing.

Figure 144. Current Configuration
Ceramic Combustor.
MAX POWER
FLAT RATER MAX POWER
construction which helps minimize thermal
r--_
'OLE
r--_
CRUISE i_ ______1100PERCENT SPEEO
stresses. The actuators and sliding fits
required with variable geometry are elimi-
nated with fixed geometry, creating an inher-
ently more reliable design. The ignitor is
mounted on the centerline of the fuel nozzle,
CONFIGURATION
RAOIAL AXIAL providing reliable start capability with cooler
5201 I FULL NONE
operating temperatures to prolong ignitor life.
I--I 2 HALF NONE
z,ooJa E_] 3 HALF FULL The combustor is composed of a primary burn-
CO El. GM/KG FUEL
am 4 FULL HALF
ing zone immediately followed by a dilution/
quenching zone. The primary zone is acti-
vated by 8 primary jets which carry 22 percent
HC El. GM/NG FUEL

"]
10
i1 J_
of the total airflow. A small axial swirler
mounted on the O.D. of the fuel nozzle adds a
45-degree swirl to 14 percent of the total
flow. The quench air enters approximately
El -- EMISSION INOEI
2.54 cm (I inch) downstream of the primary
697-078-252 .
zone through a 30 degree radial swirler. The
Figure 143. Element Test Rig Emission swirler carries 64 percent of the total airflow
Test Results. - and is designed to penetrate and mix the
products of combustion with relatively cool
chosen for development using ceramic compo- quench air. This minimizes the dwell time at
nents. The design objective was to create a high temperature when NO x is produced and
uniform, lean fuel-air ratio that minimized lowers the pattern factor at the turbine inlet.
NO x and oxidized CO and HC with high com- All of the combustor components are made of
bustor inlet temperatures. sintered silicon nitride (SSN) except for the
eombustor liner which is sintered alpha silicon
As seen in Figure 144, the all ceramic lean carbide (SASC) and the fuel nozzle, which is
bdrn combustor is a simple, multiple piece made of Hastelloy X.
89
5.3.2 Combustor Development
5.3.2.1 Fuel Nozzle Development COOLING AIR__

FUEL----_-.___._,__ _
Throughout the design evaluation of vari-

I__D_. AIR ASSIST
able and fixed geometry combustor configura- AIR ASSIST
ilN AOJACENTTUDEI
tions, fuel nozzle development proceeded with
emphasis on improved atomization and re- FLAME
CENTERSOOYj
duced heat-load. A duplex-type nozzle from AND
AXIAL SWIOLER ;_:._OLING AIR
The Delavan Corporation was initially selected
for use in the variable geometry eombustor.
The 50-ram. (2-inch) diameter duplex nozzle is
shown in Figure 145. As depicted in the
g|1_711.2e, r_
figure, three flow swirlers were incorporated;

an outer swirler, inner swirler, and central
swirler. Airflow entered through the outer Figure 146. Simplex Fuel Nozzle.
swirler to"atomize the secondary fuel while air
through the inner swirler atomized the pri- adaptability to ceramic fabrication. Unfortu-
mary fuel. Central swirler air flow helped nately, both exhibited fuel coking problems
enhance the radial fuel-air mixing intensity in during rig testing. In order to solve the
an effort to produce a more uniform fuel-air problem, air cooling was incorporated into the
distribution. The nozzle was fabricated using simplex design, along with air assist for en-
a modular construction so that various swirler hanced fuel atomization. Fuel injection ori-
angles could be investigated. Although the fice diameters were also increased from 0.3 to
duplex nozzle performed acceptably in rig 0.46 mm (0.012 to 0.018 inch) for greater
testing, a simplex design, shown in Figure 146, particle tolerance. Additional development
matched its performance and offered the ad- relocated the ignitor to the axial center line
vantages of lower production cost along with of the nozzle. This configuration produced
good light-off results and lowered the ignitor
operating temperature, however, coking was
OUTER still occurring during rig testing at high com-
AIR
bustor inlet temperature.
Garrett continued to address the coking

problem by developing a modular fuel nozzle
ignitor assembly as a replacement for the
simplex. The modular construction was de-
signed to allow installation of extensive in-
strumentation used to gather temperature and
pressure data inside the nozzle. Like the
PRIMARY
simplex, the modular Garrett nozzle contained
air assisted fuel injection and air cooling sys-
FUEL tems. Fuel injection was concentrated at the
ECONDARY nozzle ID, near the ignitor, and fuel was
directed toward the centerline of the combus-
tor, creating a locally rich zone for ignition.
GD7.076-254 After initial rig testing of the Garrett proto-
type demonstrated good light off and steady
state performance, the Delavan Corporation
Figure 145. Duplex Airblast Fuel Nozzle. was asked to design and fabricate a fuel
90
nozzle based on this design which would also
be easy to manufacture. One of three nozzles
FUEL NOZZLE __ '_
delivered by Delavan was heavily instrumented
with thermocouples and pressure probes to
AIR COOL _/_ //_PNSULATION I_
assess the potential of coking and fouling
problems. Testing at most severe conditions
showed the nozzle was sufficiently cooled to
prevent coking and led to the selection of the I_ _ _ i I J.i___
Delavan nozzle for all subsequent engine and
rig testing. The current configuration Delavan 1.1 MM ___ ._ii_+ J L"_II - -
nozzle, shown in Figure 147, has been used (O._I4MM, _---_ 3.5 MM __.----'I "_. .........
suecessfullly in light-offs with JP-4 and gaso- ,,ST

X':-',--J.
"` ,N.J__
l.J +
hol and with some success using DF-2. Future
effort will be directed at improving the noz-
L..... , i J ] )- }
zle's ability to atomize DF-2.
I _" AIR ASSIST G874)78-257
PLENUM Figure 148. pilot Combustor Proof-of-

.1 / NOZZLE
Concept (Metal Configuration).
/ A,,
ceramic components shown in Figure 144.
This represents the latest configuration com-
bustor for the all-ceramic engine. Data which
defines combustor geometry and performance
are listed in Table 16. The configuration has
been used successfully in hundreds of hours of
rig and engine operation. Testing has demon-
strated its capability to operate on DF-2,
AIR ASSIST FUEL ORIFICE
, ORIFICE JP-4, and gasohol. CO and HC emissions are
GeT-OTB-Z58 well within program goals and the ceramic
hardware has demonstrated excellent reliabil-
Figure 147. Delavan Nozzle Configuration. ity in rig and engine operation. Further im-
provements are required in reducing No x emis-
5.3.2.2 Ceramic Combustor Development sions and DF-2 cold start performance. These
improvements are being pursued through an
After selection of the lean-burn, center analytical/empirical approach involving 3-D
body ignition combustor concept, hardware combustion flow analysis, nozzle testing, and
was produced in the configuration shown in continued combustor rig and engine testing.
Figure 148. The arrangement included a metal This approach has resulted in a combustor
pilot combustor welded onto the fuel nozzle iteration which contains an enlarged primary
which proved the viability of the design zone developed to improve DF-2 combustion
through testing, but was not capable of 1371C and cold start performance. The configura-
(2500F) discharge temperatures due to mate- tion, shown in Figure 149, is currently fabri-
rial property limitations. After successful cated in metal to facilitate instrumentation
completion of 1149C (2100F) rig testing, the and geometry changes, and is also designed to
metal pilot combustor was replaced with a be easily produced with ceramic materials. In
sintered silicon nitride pilot combustor which a metal engine test, the combustor demon-
became a part of the spring-loaded stack of strated light-offs with perfect consistency
91
Table 16. Combustor Data.
Parameter Units AGT101
Configuration Can
Combustor Length cm (In) 30.5 (12.0)
Diameter at Primary Zone em (In) 4.8 (1.9)
Length/P.2. Diameter 2.8 (6.3)
Combustor Airflow kg/s (Ibm/s) 0.729
Combustion Efficiency at Design Point Percent 99.9
Airflow Split
Primary Zone Percent 22.18
Dilution Zone Percent 63.41
Swirler/Injeetor Percent 14.38
Inlet Static Pressure kPa (psia) 394 (57.1)
Combustor Total Pressure Loss Percent 3.6
Inlet Total Temperature C (F) 975 (1787)
Average Total Exit Temperature c (F) 1371 (2500)
Pattern Factor <0.09
Combustor Surface Area cm 2 (in 2) 897 (139.1)
Combustor Volume em 3 (in J) 1627 (99.3)
Liner Wall Material Sintered
Silicon
Nitride
Overall F/A Ratio
Rich Blowout 7.100
Design Point Steady State 0.0118
Lean Blowout at Design Point 0.001
Primary Zone F/A Ratio of Design Point 0.0569
Fuel Injector Type Air Assist
No_ Fuel Injectors 1
Design Point Fuel Flow g/s (Ibm/hr) 3.9 (31.1)
Minimum Fuel Flow for Lightoff at g/s (lbm/hr) 0.34 (3.0)
Start Limit
Maximum Viscosity for L/O at Start Limit eentistoke 12
Turndown Ratio 21
Tolerance to Diffuser Exit Swirl Angle _-degrees 12
Types of Fuel Used DF-2 JP-4
Gasohol
Heat Release Rate at Cruise j/hr-kPA-m 3 (Btu/hr-atm-ft 3) 9.6 x 108
(2.61 x 106)
Aerodynamic Loading at Cruise kg/m3-kPA-sec (Iblft3-see-at m) 7.5 x 10 -5
(0.015)
Reference Flow Area cm 2 (in 2) 67.1 (10.40)
Inlet Math No. 0.06
Primary Zone Reference Velocity m/s (ft/see) 43.2 (141.6)
Residence Time at Cruise ms 5.4
Emissions at Cruise
Nox gm/kg fuel 13.5
CO gm/kg fuel 2.4
HC gm/kg fuel 0.21
Number of Ignitors 1
92
RING
GEAR DRIVE
RING GEAR
SEPARATOR
REGENERATORCORE
ROUSING
MATRIX
IROTI
SEAL
OUTERI
SEAL
G67_78-258
EXHAUST
Figure 149. Enlarged Pilot Zone Combustor. HOUSING
SUPPORT
SUPPORT ROUSING
using both JP-4 and DF-2. It also demon- ROLLERS
strated excellent temperature ramping char- G67_7_3,57
acteristics which easily meet performance
goals. Testing of the metal and ceramic
versions of this configuration will continue to Figure 150. AGT101 nRegenerator System n
characterize its performance. Components.
5.4 ReEenerator System trix assembly. These drive and support

components, radially position the ceramic
The term "Regenerator System" involves matrix and drive it in a rotary motion.
the iterative design/analysis/test/evaluation
of the following components as defined below 0 Regenerator Drive and Support system
and as shown in Figure 150: Bushings- The bearing material developed
for the drive pinion and support rollers.
O Xegenerator Core Matrix- The circular
0 Regenerator Seals- The combination of
ceramic honeycomb waste heat transfer
matrix. static metallic diaphragm seals and
ceramic coated dynamic seals.
O Ring Gear Elastomer- The silicone based
Hot Side Seal- Inner regenerator seal
polymer used to bond the metallic drive/
composed of two peripheral "C" shoes
support ring gear to the ceramic core and a crossarm
matrix.
Cold Side Seal- Outer regenerator seal

O Ring Gear - The cast iron ring gear used to composed of a single peripheral "C"
drive and support the ceramic matrix in a shoe and a crossarm
rotary motion.
0 Regenerator Seal Coatings- The ceramic
O Drive Pinion and Support Rollers- The rub coating applied to the metallic regen-
pinion gear designed to mesh with the ring erator seal shoe substrate that provides
gear while support rollers circumferential- high temperature dynamic sealing against
ly support the ring gear and ceramic ma- the ceramic core matrix.
93
O Flow Separator Housing- A ceramic hous- Hot regenerator rig design/testing/evalua-
ing that divides engine compressed air tion-seal leakage evaluation with helium
from turbine discharge exhaust gas and injection
provides the platform for the inner regen-
erator seal. o Exhaust housing distortion analysis/test
evaluation
0 Exhaust Housing- The metallic housing
that divides engine compressed air from o Drive and support roller bearing/material
turbine discharge exhaust gas and provides tests
the platform for the outer regenerator
seal. 5.4.1 Ford Regenerator Development
O Ring Support Housing- The metallic hous- The primary AGT101 program objectives
ing that provides the platform for the flow for the regenerator system were to obtain an
separator housing. effectiveness of 92.9 percent, a total pressure
loss of less than 7.5 percent at the maximum
Regenerator system development activities power operating condition, and a seal system
were conducted at Ford and at Garrett Tur- capable of operation at 1090C maximum tem-
bine Engine Company (Garrett). Specific com- perature with 3.6 percent leakage. In addi-
ponent development activities to each com- tion, the design and development of a durable
pany are listed below and are discussed in drive and support system was required. This
detail in paragraphs 5.4.1 for Ford Regenera- report summarizes the design and development
tor Development, and 5.4.2 for Garrett Regen- efforts from October 1979 through June 1987
erator Development, respectively. toward achieving the program objectives.
Ford Development Activitie_ By using the extensive operating experi-

ence in the Ford 707 gas turbine engine, the
o Regenerator matrix performance evalu- initial regenerator system for the AGT101
ation program was designed to be capable of operat-
ing at a maximum system temperature of
o Drive and support system design 980C (1800F). The system temperature re-
quirement was subsequently increased to
o Ring gear elastomer configuration design I090C (2000F). Utilizing a regenerator matrix
core with a wrapped, thin-wall, corrugated,
o Regenerator seal design sinusoidal fin from Coming Glass Works, the
regenerator effectiveness was projected to be
o Regenerator system durability testing 90 percent at the maximum power operating
condition. Therefore, an increase of 2.9 per-
Garrett Development Aetivitie_ cent in effectiveness was required. In addi-
tion, regenerator seal system leakage at this
o High pressure (HP) and low pressure (LP) condition was estimated at 7.2 percent based
gas regenerator flowpath design on comparable seals in the Ford 707 engine. A
significant reduction in leakage was required
Regenerator matrix flow pattern to meet the AGT objective of 3.6 percent.
matching between HP and LP regenera-
tor matrix inlets The AGT101 regenerator system compo-
nents are shown in the photograph in Figure
o Flow separator housing 151. This system consists of the regenerator
o Exhaust housing seals, which sandwich the regenerator core.
The core in turn is supported radially and
o Regenerator Seal Platform Design for min- rotated by the regenerator drive system com-
imum distortion/co-planar engine operation ponents; a pinion roller, a fixed roller, and a
94
ORIVE PINION REGENERATOR
SEAL
/ SPRING ROLLER
ASSEMBLY
FIXED ROLLER REGENERATOR

CORE/RING
ASSEMBLY GEARASSEMBLY 72652-8
GB7-O78-330A
Figure 151. AGT101 Regenerator System Components.
spring roller. The development of these com- system durability. These design criteria were
ponents are described in the following para- applied to the following areas of the new AGT
graphs. seal system:
5.4.1.1 Regenerator Seals I) Weldless diaphragm system

2) Rubbing shoe size
The primary purpose of this task was to 3) Diaphragm fabrication technique
develop new seal designs to achieve the pro-
gram leakage objective of 3.6 percent of total The "weldless" designation refers primarily
engine airflow at the maximum power operat- to the fact that the diaphragm system is
ing conditions. Six distinct phases of seal independent of the rubbing shoe system. This
design were investigated analytically and ex- design demonstrated a proven record of reli-
perimentally to develop the final design. ability and performance in the programs men-
tioned above, and had demonstrated uniform
Design criteria for the initial AGT seal seal loading and coating wear characteristics.
(Phase I) were derived principally from seven With the additional surface contact of the
years of experience in developing the Ford 707 diaphragm on the rubbing shoe, the weldless
seal, which involved thousands of hours of seal provided the additional benefit of friction
engine testing in various performance and dur- damping which is essential for maintaining
ability cycles. This development was followed shoe contract on the regenerator in the dy-
by an additional 218,000 regenerator-hours of namic operating modes.
engine tests in a joint Ford and DOE/NASA
program to develop ceramic regenerators, The relationships of width and thickness in
which demonstrated over 4,000 hours of seal both the crossarm and peripheral shoes were
95
also derived from the 707 seal. The inboard Seal Leakage Development
(hot) seal assembly is positioned vertically to
the flow separator housing by means of a Starting with the Ford regenerator seal
retaining rig. Peripheral seal component stiff- technology as a baseline, further reduction of
ness had to exceed the critical buekling seal leakage was a development process in
stresses to prevent radial distortion when the which seal systems were initiallyanalyzed via
limits of constraint (due to thermal growth) computer programs for shoe loadings and dia-
were reached, but also must have sufficient phragm clearance throughout their operating
flexibility to conform to the profile of the range. Seals were then fabricated for evalua-
regenerator core. Wider shoes tend to reduce tion in the static seal leakage fixture at room
leakage by creating a better labyrinth effect. temperature. Concepts that demonstrated
Since low leakage was a prime objective for leakage reduction in the static seal rig were
the AGT regenerator system, the peripheral then tested under hot operating conditions in
shoe width (190 ram) is equivalent to the 707 the regenerator test rig and engine. The rig
system. The crossarm shoe width is 38.0 mm and engine results were used to update the
compared to 41.1 mm for the 707. computer analysis and refine the design of the
next generation of seals. Such an approach
The diaphragm system, which is the com- was successful in the development of seal
pliant member between the shoe and engine systems for other engines at Ford.
structure, was also based on 707 technology.
The diaphragms were formed from foil mater-
ial 0.0762 to 0.203 mm thick. The material A typical regenerator seal cross-section is
that had proven durability, Inconel X750, was illustrated in Figure 152. The seal working
selected for the initial fabrication. height (SWH) represents the total clearance
for the seal assembly resulting from the in-
In order to design a low leakage regenera- stalled clearance minus the axial movement of
tor seal system, the following design consider- the adjacent engine structures and regenerator
ations required investigation" core at specific engine conditions. The dia-
phragm clearance (C) represents the amount
1) Control of engine structures that affect of clearance from the sloid height of the
seal operating clearance. assembly. Regenerator system build clearance
is defined as the total diaphragm clearance
2) Diaphragm design to optimize seal load for (2C) that exists for both seals at engine
minimal leakage without excessive coating assembly. The total above shoe force (FT)
wear and drive torque requirements. imposed on the rubbing shoe, consists of the
mechanical loads from the deflection of the
To assist in the development of a low diaphragm elements in addition to the air
leakage seal systemp the following test rigs pressure force. The combined forces acting
were utilized: above the shoe can be represented equiva-
lently as a force (F T) acting at a specified
I) Static seal leakage rig to monitor design distance (D U) from the inner edge of the
changes. rubbing shoe, as illustrated on Figure 152.
2) Test fixtures to establish seal assembly The primary engine structures that control
mechanical load characteristics. regenerator seal working height are the flow
separator housing and exhaust housing. Axial
3) Hot test rig to evaluate system leakage thermal growth and distortion of these com-
and coating wear characteristics. ponents directly affect the SWH.
96
engine testing, supported by additional loading
and thermal analysis.
I FLOW SEPARATOR Analytical Regenerator Design Methodology

HOUSING
T t
c HETAINER __'_
_OIAPHRAGN UPPER In order to analytically determine the dia-
swH I _- _f _ _MiOOLE phragm load and stress as a function of clear-
,, tu
RP LOWE, ance, the diaphragm load characteristics were
_, --RETAtNEH
required input to the "core position" analysis
computer program, which performed a summa-
tion of forces and moments imposed on the
RUBBING SHOE regenerator. Once the operating plane was
I P WEAR FACECOATING determined, the variation in force and stress
distribution for each component of the regen-
erator seal system was defined.
REGENERATORCORE
The main objective for the regenerator
seal system design was to achieve sufficient
contact force (F C) with minimal variance at
EQUIVALENT FORCESYSTEM HP
f 1 each seal component location to provide
effective sealing with coating wear at a uni-
I T i form and an acceptable rate. Contact force
LP r | II
(F C) is defined as the net load resulting from
the above shoe diaphragm mechanical and
pressure forces (F T) that are partially offset
LP_ _ --"- -----,-
..,..,,, ?E U II by the undershoe pressure force distribution
I (F U) that exists between the seal wear face
coating and the regenerator matrix face (Fig-
HP
ure 152). The undershoe pressure distribution
FCONTACT = FC = FT- FU is a function of the labyrinth effect due to
GI7076.'L'_I leakage ch_,racteristics between the shoe and
regenerator core.
Figure 152. Typical Regenerator In order to minimize contact force on the

Seal Sehematie. LP carry-over portion of the crossarms while
ensuring contact on the HP carryover portion,
To minimize this problem, the flow separa- optimization of the location of the retainer
tor housing was redesigned to be positioned with respect to the LP edge of the rubbing
and supported off the outer casing. This shoe (W I) was required. For the initial AGT
change resulted in acceptable variations in regenerator seal system, the crossarm retainer
SWH for both the metallic AGT and ceramic was positioned 17.8 mm for the LP carryover
AGT engines and allowed the same diaphragm portion and 2.54 cm for the HP carry-over
system to be used on both engines. position. The significance of the estimated
contact force levels for the reference seal
The AGTI01 seal design evolved through a system was established by the coating wear
series of configurations, or phases, shown in pattern of the actual seal hardware after rig
Figure 153. The initial design, Phase I, was and engine testing. Seal design changes were
based on prior Ford experience combined with suggested by the leakage test data and visual
extensive analysis. Subsequent designs were examination of the coating contact pattern.
generated as a result of static leakage testing, Input data to the core position analysis was
hot regenerator rig testing, and AGTI01 modified to simulate the proposed design
97
IXi IP]
PHASE CROSSARM SECTION PHERIPHERALSECTION COMMENTS
HiGH LIFT OiAPHRAGMS

ON X AND P
X - SAME AS i
P - LOWLiFT OIAPHRAGMS
I I I I PLUS SPRING
X - STIFF iNTERNAL SPRING

Ill P - STIFF INTERNAL SPRING
I I 1 I RETAINERSPRING
X - THIN OIAPHRAGMS.LOW RATE SPRING

IV
I I I I P - THiN OIAPHRAGMS,LOW RATE SPRING
X - LOW LiFT ONE-PIECEDIAPHRAGM
I I I I P- SAME AS IV
X - COOLINGLOUVERSAND HOLES
VI BARRIERCOATING ON SHOE
I P - SYMMETRICAL
DIAPHRAGMS UPPER AND LOWER
G|7_7_331
Figure 153. Regenerator Seal Design Phase Comparison.
changes and the results were compared on a metric studies in the static leakage rig indi-
relative basis to the baseline regenerator seal cated the need for increased loading of the
system. In this manner, the comparative re- lower sealing diaphragm against the metal seal
sults served as a guideline to provide direction shoe in order to reduce leakage at low pres-
to additional design changes. This procedure sure levels.
was used in previous engines for optimizing
the regenerator seal system design. The Phase III regenerator seal diaphragm
design evolved from parametric studies in the
The initial Phase I seals evaluated in the static seal leakage rig to minimize leakage at
static seal rig at the design clearance ex- low pressures.
hibited leakage slightly above the initial ob-
jective of 7.2 percent of total engine airflow The most significantchange was the addi-
at the maximum power operating condition. tion of a supplementary spring mounted on the
Subsequent testing at Garrett indicated mar- retainer. Engine and rig testing indicated
ginal seal loading and subsequent high leakage substantially improved leakage over Phase I
at part power engine conditions. and II,though seal loading was excessive.
The Phase II seal configuration featured a A parametric test series was conducted on
spring diaphragm inserted between the two the Phase Illconfiguration in parallel with a
sealing diaphragms, and provide the seal load- parametric analysis. This work evaluated the
ing required to reduce seal leakage. Para- effect of seal loading on seal leakage.
98
Through optimization of load distribution, it In addition to these design improvements,
was found that seal loading could be substan- the quality of the prototype hardware was
tially reduced without sacrifice of leakage. improved through the procurement of new
tooling and improved processes.
The main objective for the Phase IV seal
system were to preserve the leakage charac- An interim quality control standard was "
established at that time for the static seal
teristics associated with the Phase Ill design
with a further reduction in mechanical load leakage rig to ensure that low leakage seals
would be sent to Garrett for engine and test
characteristics to reduce drive torque require-
rig evaluation. A 0.006 m3/s static leak test
ments. Additionally, the Phase IV seal in-
requirement per seal (inboard or outboard) at a
corporated cooling passages which used a con-
7.0 mm seal working height (SWH), was estab-
trolled amount of leakage to maintain the
lished. This represented the state-of-the-art
diaphragm temperatures within the limits of
at this time for the Phase V seals. Hot and
diaphragm material capabilities.
cold seals that totaled 0.011 m3/s leakage at
7.0 mm SWH in the Ford static test rig _vere
Through another series of parametric tests
evaluated in the hot regenerator rig at Garrett
and analysis, further improvements in leakage
and mechanical loading were attained. using the helium tracer technique and indi-
cated a leakage of about 5.5 percent.
The mechanical load characteristics for
Additional testing in the static seal rig
each of these seals were measured. The load
with symmetrical upper and lower outer per-
characteristics indicated the Phase IV design
ipheral diaphragms and equal length retainer
was an effective compromise between the
legs demonstrated a potential 20 percent re-
Phase [ and Phase Ill designs. Consequently,
duction in static leakage. This concept was
the actual hardware substantiated the analyti-
adopted as the basis for the next generation
cal predictions.
seal design, which was designated as Phase VI,
as illustrated on Figure 153. In addition, the
Rig and engine testing indicated that in the
Phase Vl design can accommodate more cool-
dynamic thermal environment of a regenerator
ing air for the inboard (hot) crossarm seal
system, drive loads for the regenerator were
compared to Phase V.
higher than the Phase Illdesigns. Leakage was
equivalent to the Phase Ill configuration, Additional features of the Phase Vl seal
though the allowable SWH tolerance was much configuration included a crossarm thermal
more broad with the Phase IV seal. barrier coating for improved seal diaphragm
cooling and bore pressure relief slots in the
Based on the test results for the Phase IV cold seal inner periphery.
seal design, a new design iteration (Phase V)
was initiated. The design objectives for the Phase Vl seals qualified with less than
Phase V seal (Figure 153) configuration were 0.011 m3/s leakage at 7.0 mm seal working
to further decrease the leakage and drive height. These hot seals were also made with
torque requirements. In order to accomplish reprocessed, used shoes. Based on engine
these objectives, the Phase V design featured exhaust flow at full power conditions and
an integrated crossarm diaphragm which corrected for ambient temperature, a static
eliminated the inner secondary diaphragm that seal rig leakage of 0.004 m3/s per seal was
was used in previous designs. Tests in the estimated to be adequate to meet the program
static seal rig indicated a significant portion objective of 3.6 percent for the regenerator
of the total leakage occurred at the inside and system. Therefore, the combination of the
outside corners, in addition, the integrated low leakage Phase VlB cold seal (0.004 m3/s)
crossarm diaphragm had a reduced lift angle, with one of the Phase VIB hot seals (0.005
which would reduce the mechanical load char- m3/s should nearly meet the program objec-
acteristics and drive torque requirements. tive in an engine.
99
b
Seal Design Summary PHASE iV PHASE V 1084
The initial AGTI01 regenerator seal sys-

tem was based on prior experience with the
Ford 707 gas turbine engine combined with
extensive analysis. The systematic reduction 0.010. V A 1985
of seal system leakage toward achieving the

program objectives was described. Design 0.006 __....__,_._ ,_'_'_I_HAS E
methodology was based on a development o/| _ PHASE

v,A,088
approach consisting of six unique phases.
(7.71
Design changes were incorporated based on 17.01 17.11 SEAL WORKING HEIGHT. MM (INCH)
static and dynamic test rig evaluations supple- G|7-O78-333A

mented by analysis that provided guidance for
each design iteration. The initial design, Figure 154. Static Sea/Leakage
which was very sensitive to operating clear- Comparison for Sea/Design Phases.
ance, was capable of achieving the original
leakage level of 7.2 percent at design clear- to flow between the upper and lower dia-
ance. Subsequent designs systematically re- phragms. A transfer tube at the corner
duced leakage to 50 percent of the initial allowed the air to continue through the high
level. In addition, sensitivity to operating pressure side of the inner periphery. The
clearance was practically eliminated for amount of cooling air flow is dependent on the
clearance variations within 0.63 mm of design. pressure differential that exists in the chan-
The curves in Figure 154 illustrate these nel. For discharge holes in the middle dia-
trends during the AGT101 regenerator seal phragm at the center of the HP periphery, the
development effort. Seal wear-face coatings pressure differential (less than 1.7 kPa) estab-
were changed where necessary in the final lishes the minimum rate of cooling air flow.
design to accommodate the design temper- For maximum cooling, holes in the retainer
ature levels. (Figure 155) allow the air to discharge into the
center hole cavity, which is vented to the
Seal Diaphragm Cooling exhaust duct. Use of this discharge location
would require the incorporation of a restric-
To help establish the environment for the tion at the opening of the funnel on the
hot side seal diaphragms a detailed finite crossarm end diaphragm to throttle the cool-
element model of the AGT flow separator ing flow required.
housing was generated and a heat transfer
analysis was performed. Based on this analy- A three-dimensional heat transfer analysis
sis, the temperature level at the inner circular program was modified to analyze the seal
portion of the RBSN or LAS FSH will be 1040 cooling system (Figure 155). To minimize the
to 1065C for a gas inlet temperature of number of nodes and elements, three axial
1090C. The linear segments that connect the locations were selected (X=O, 130, 254 ram),
inner and outer circular sections of the FSH which corresponds to cooling flow inlet at the
were estimated to operate at 930C. end of the crossarm, transfer location from
linear to peripheral, and cooling flow exit
In order to accommodate the I090C regen- location, respectively. Convection coef-
erator system operating temperature associ- ficients were estimated based on cooling air
ated with the all ceramic engine, the regener- flow rate and cross-sectional area of the
ator inboard (hot) seal crossarm required dia- channel. Since the cooling flow exists on both
phragm cooling to keep the metal diaphragm sides of the main support leg of the middle
temperatures below 930C. This design (Figure diaphragm, the heat transfer coefficient can
155) had a funnel incorporated into the cross- be maximized. By incorporating an appropri-
arm end diaphragm, which allowed cooling air ate number of openings (slots or louvers) in the
i00
axial direction in the middle diaphragm, the
flow can be interrupted to prevent boundary
layer build-up and promote localized turbu-
lence, which increases the localized heat
transfer coefficient. In addition, convection
coefficients between the upper diaphragm and
flow separator housing were also estimated
based on leakage.
As illustrated on the model (Figure 155),

the portion of the middle diaphragm, which is
in contact with the upper and lower dia-
phragms, represents a direct conduction path
between the FSH and the rubbing seal shoe.
Consequently, a thermal barrier coating such
as zirconium oxide was required on the back
side of the rubbing shoe to reduce temperature
in this region and minimize cooling air re-
quirements. The inner circular portion of the
FSH also required a" thermal barrier coating,
though in practice, this was not attempted due
to development considerations of this LAS ce-
ramic component.
The design concept was evaluated for cool-

ing flows varying from 0.0004 to 0.0014 kg/s,
which corresponds to 0.25 percent (minimum)
and 0.85 percent leakage, respectively. For
each cooling flow, the convection heat trans-
fer coefficient was increased 20 percent to
simulate interrupting the flow.
A X 25.4 CM A
[10 INI The middle diaphragm temperature distri-
AIR INLET
bution for axial locations of 130 mm and 250
/ mm as a function of cooling flow are illus-
HEAT TRANSFER MOOEL trated on Figure 156. The barrier coating
thickness for the seal substrate and FSH inner
circular section were both 1.3 ram. The
minimum cooling flow (0.25 percent) was suf-
ficient to keep the middle diaphragm tempera-
ILASI
ture below 927C (1700F) for the straight
portion of the crossarm (X = 0 to 130 ram), as
CTG.
shown in Figure 156. In order to cool the
SHOE IINC060tl _ i °
circular portion of the crossarm (X = 130 mm
COATING //
to 250 ram) to the design temperature level,
VIEW A-A
Gel_xO3,14 7 the cooling flow requirement was 0.85 per-
AIR EXIT cent. The original leakage objective of 3.6
percent included a 0.25 percent allowance for
cooling the hot seal crossarm diaphragm.
Figure 155. Inboard (Hot) Seal Crossarm Based on this analysis, an additional 0.60 per-
Diagram Cooling Schematic. cent leakage was required to sufficiently cool
i01
the circular portion of the crossarm dia-
phragm.
The cooling analysis shows that the middle

WL = 0.no KG/MIN diaphragm is coldest at the outboard end of
,,_-- #L = 17.5 WIM2-K
the crossarm and progressively increases until
it reaches a maximum at the center of the
,INTERRUPTEOFLOWI _//_ _ f//
inner peripheral section. Core position analy-
sis shows that the seal clearance in this region
will be about 16 ram, and the average dia-
phragm stress level at this clearance, as esti-
mated by MENTOR II, is about 139,000 kPa.
Examination of Rene 41 creep data shows that
BARRIERCTG - Zr02 +Y203
0 0.0_4 10.0531 40 [0.1801 0+25 (MINI SHOE & FSH 1.0. this will cause some local yielding. This local
[] 0.046 10.1021 50 i0.2001 0.48 THK. , 0.0_C iN yielding will be confined to the spring in the
<> 0.040 IO.lOBi 80 102401 0.52
0.056 tO.124i 91 lOCi peripheral section of the crossarm where the
0 0.062 i0.160i 130 i0.5201 0.65
metal temperatures are highest. It is doubtful
IFI
that any yielding will occur in the straight
2060-1 X = 12.7 CM 15 INi section of the crossarm where the tempera-
II. / tures are-much lower. As the spring yields the
local stress is reduced to the point where
. _- ........ further creep occurs at a very low rate. The
L¢. spring will still provide containment for the
cooling air and it will still separate the dia-
L
phragms so they can be pressure loaded. It is
I - zgL: difficult to believe that any significant in-
crease in leakage will occur.
_0- In summary, the regenerator seals have

been developed through six distinct phases and
I_0
OIAPHRAGMLOCATION now appear capable of meeting the program

IFI leakage objectives. In addition, they have
shown acceptable levels of drive torque and
2000 1 X = 25.4 CM (lO INi
the capability of operating over a wide range
10_.
1660 4 of seal working heights with little change in
-'__ _ _" _-- /_"" '_"_ _ _" OOJECTIVE leakage. Cooling air is required to keep the
diaphragm temperature of the hot seal to
acceptable levels. The cooling air require-
ments are modest and any diaphragm creep
that occurs will be limited to a local region at
the center of the inner, hot peripheral dia-
phragm. This creep is not expected to result
in any significant increase in leakage.
I.
GOT_Z&_ 5.4.1.2 Regenerator Seal Coating

Development
The primary concern for achieving a 1090C

Figure 156. Middle (Support) Crossarm regenerator system was the upper temperature
Diagram Temperature Distribution. limit of the hot seal crossarm wear face
102
coating I-ll2. This was the only Ford proprie- determined. Periodic measurements of sample
tary seal coating that did not have docu- thickness determined the seal coating were
mented durability at the higher AGT crossarm rate.
temperature conditions. In order to establish
the upper temperature limit for the 1-112, an In order to determine the upper tempera-
existing test rig was modified to increase the ture limit of the 1-112 coating to be assured
maximum test temperature from 815 to I090C that it had I090C capability and to be confi-
(Figure 157). The primary modification was dent that its friction and wear characteristics
replacement of the metal test sample holder could be controlled during processing, the fol-
with a ceramic holder. lowing parameters were investigated:
The stationary test sample was loaded 1) Chemical composition and particle size
against the rotating test core by a compres- distribution of the coating powder.
sion spring. A lever arm, which was pinned to
a center shaft, applied the torque reaction 2) Spraying procedure for deposition of the
force against a calibrated load cell. By know- coating powders on the metal substrate.
ing the applied force (P), torque reaction force
(FR) and physical dimensions of the sample, 3) Interface coatings to accommodate the
the sliding coefficient of friction (]_) could be difference in thermal expansion between
the [-112 coating and the metal substrate.
TORQUE LEVER ARM

4) Substrate serrations to form a mechanical
TEST STAINLESS I:_:l_l

k bond attachment of the coating.
TEST CORE CORE SCREWS ,_ K II Fourteen samples were tested for more
RETAINER \ / [//_\_,J _.L_P:_ than 2000 total hours in order to evaluate the
above parameters.
All of the 1-112 powder evaluated was

processed by Ferro Corporation. In order to
determine the optimum process for the [-112
\_,_.._,,._,_a" I // /HOLDER SPRING CATCH powder using the equipment and facilities at
((..._[ _ / AND LOCK
Ferro Corp., three different batches of powder
_ SEAL
COAX,NO
were processed. Each of these batches were
TEST SHOE
TEST RIG SCHEMATIC within the Ford specification for chemical
composition and particle size distribution.
The primary evaluation of the 1-112

powder and plasma spray process was depen-
dent on coefficient of friction and wear rate
as measured in the seal wear rig. For the Ford
707 gas turbine engine, the acceptance cri-
FR teria for the I-I12 coating were as follows:
1) Coefficient of friction (l_): less than 0.35

at operating temperature and 68.9 kPa load
687-078-338 2) Wear rate (w): less than 0.020 ram/100 hr.
Figure 157. Regenerator Seal Coating Since the loads in the AGTI01 regenerator
Wear Test Rig. seals were approximately the same as those in
103
the 707 seals, a coefficient of friction less The effectiveness of the deep serrations
than 0.35 and a wear rate of less than 0.020 are clearly illustrated (Table 17). Tempera-
mm/100 hr. should result in acceptable drive ture capability appeared to be increased 55 to
torque and seal life in the AGT engine based 166C compared to a non-serrated sample.
on 707 experience. Tl_ese same criteria were
used, therefore, to establish the acceptability Based on visual observations of hot flow
of the I-i12 coatings that resulted from varia- path components tested in the initial hot
tions in composition, particle size, spraying structures rig and ceramic engine at 1090 to
procedure, etc. I150C turbine inlet temperatures (TIT), the
1-85 coating for the hot seal outer peripheral
For the 14 samples evaluated in the wear shoes appeared to be marginal. Although the
rig, which consisted of the three different copper based 1-85 did not delaminate, soften
batches of powder as well as variations in or experience significant wear, trace amounts
spraying process, the 1-112 coating demon- of copper were found to be the cause of
strated friction and wear characteristics well sticking between other ceramic components.
with the Ford acceptance criteria for exposure In order to increase the upper temperature
up to 1090C and 68.9 kPa maximum load. limit, 1-151 was evaluated as a replacement
for the 1-85 coating on these shoes. Based on
To evaluate the substrate bond coating, wear rig tests at Ford in the early 1970's, the
samples sprayed with commercially available 1-151 (zinc-oxide) had 112C higher tempera-
METCO 443, METCO 447, and GE Superalloy ture capability compared to 1-85 (copper
in place of NiCr were evaluated. In order to base).
evaluate the effectiveness of the deep serra-
tions, additional samples without serrations The initial seals with 1-151 installed in the
were also tested. A summary of the tempera- ceramic engine at Garrett experienced delam-
ture capabilities of the various test combina- ination in the 1-151+I-39 interface bond re-
tions are listed on Table 17. Some improve- gion. An evaluation of the 1-151 coating was
ment was evident with METCO 443 and conducted which established that the coating
METCO 447, which appeared to increase the powder used was not within specification.
temperature capability approximately 28 to Since the coating vendor had difficulty pro-
56C. The GE Superalloy appeared to satisfy ducing this coating, it was decided that the
the I090C objective. 1-151 would not be pursued.
Table 17. Test Sample Temperature Limit.
Maximum Total Test Time,

Temperature, C Hrs
Interface Without With

Bond Coating Serrations Serrations Without
Serrations ! With
Serrations
NiCr (Standard) 920 - 950 980 - 1010 288 1108
METCO 443 870 - 900 1010 - 1040 45 135
METCO 447 160
GE Superalloy 920 - 950 1090 + 100 200
104
As a replacement for 1-151 peripheral seal the 1090F regenerator inlet temperature re-
components, modified versions of 1-85 with quirement, the regenerator matrix must have
improved plasma spray parameters as well as sufficient thermal stress capacity. Durability
1-112 peripheral components were assembled is dependent on matrix material stability at
for engine evaluation at Garrett. Testing seal elevated temperatures for extended time.
assemblies with improved 1-85 or 1-112 in the
ceramic engines did not cause sticking be- During the AGT101 program, regenerator
tween ceramic components. As expected, seal cores have been supplied by Corning Glass
assemblies with 1-112 in place of 1-85 did, Works, and NGK-Locke. The Corning cores
however, have higher drive torque require- are fabricated as a two-sheet wrapping pro-
ments. cess with aluminum-silicate (AS) material.
NGK-Locke utilize the extrusion process with
For the final AGT seal configuration magnesium-aluminum-silicate (MAS) material.
(Phase VIB), the best combination of wear face Geometry and performance for each of these
coatings that would satisfy the 1090F regener- core configurations are compared in Table 18.
ator inlet temperature requirement was as
follows: The Coming aluminum-silicate (AS) wrap-
ped thin-wall sinusoidal matrix satisfied the
1) Hot (inboard) seal: I-ll2 on all compo- pressure drop objective but is 1.0 percent low
nents in effectiveness. Corning was provided addi-
tional funding to fabricate samples of an ex-
2) Cold (outboard) seal: Improved 1-85 on all truded AS matrix with a rectangular cell shape
components except for 1-112 on the cir- and 0.089 mm material wall thickness to meet
cular section of the crossarm. the AGT101 regenerator effectiveness and
pressure drop objectives. The wall thickness
In summary, two of the Ford proprietary wear in the sample provided was thicker than antic-
face coatings (I-85 and 1-112) have been ipated. The estimated performance for this
adapted for the higher temperature require- matrix geometry (Ext. 1) was 0.8 percent
ments associated with the AGT program. Im- lower effectiveness than the original matrix
proved plasma spray parameters have pre- geometry with equivalent pressure drop. Even
vented the 1-85 wear face coating from de- though the extruded rectangular matrix is
laminating at the higher temperature condi- more efficient than the sinusoidal geometry,
tions of the outboard seal assembly. 1-112 the thicker fin wall more than offsets any
combined with GE superalloy bond coating on expected gain in effectiveness. Further devel-
a serrated inboard seal crossarm substrate has opment of the extruded Corning matrix would
satisfied the 1090C temperature requirement. have extended beyond program termination.
5.4.1.3 Regenerator Core In order to achieve the improved effective-

ness required, new tooling was developed by
The AGTI01 regenerator core was evalu- NGK. NGK has fabricated full-size cores with
ated in terms of core performance, thermal three different matrix geometries, labeled
and mechanical stress, and material thermal C0.3, C0.7, and C0.9 and defined in Table 18.
stability. In addition, tooling was developed for a fourth
fin geometry (C0.8). The C0.7 matrix geome-
The design performance objectives for the try is an isosceles triangular fin with 0.089
AGT regenerator are 92.9 percent effective- mm wall thickness. The C0.8 and C0.9 struc-
ness with less than 7.5 percent pressure drop tures, which have different aspect ratios, are
at 100 percent gas generator speed. To meet rectangular with a nominal 0.114 mm wall
the 3.6 percent leakage objective, the thickness. Regenerator core performance can
through-wall porosity leakage in the core must be projected based on estimating the heat
be less than 0.5 percent. In order to satisfy transfer and pressure drop characteristics of
105
the fin geometry utilizingshuttle rig test data Corning material. NGK took two approaches
for similar geometries. The predicted per- toward resolving this problem. First, they
formance for the four NGK fin geometries are developed coatings for the inner and outer
listed on Table 18. Estimated performance circumferences, in addition to impregnating
predictions indicated the program perform- the matrix walls. This method was not effec-
ance objectives could be attained with the tive in reducing porosity in the thin wall C0.7
C0.7 fin geometry, provided a more dense or C0.8 matrixes. Then, they fabricated full-
material could be developed to reduce exces- size cores with their dense, less porous revised
sive wall porosity leakage. MAS material designated MAS-A.
Due to the greater wall porosity of the NGK has demonstrated significant im-
original NGK MAS regenerator matrix mater- provements in modifying their standard mag-
ial, core leakage across the crossarm and nesium-aluminum-silicate (MAS) material in
through the inner and outer circumferential order to achieve the porosity leakage objec-
was higher when compared to the more dense tive. The improved MAS-A incorporated in-
Table 18. Regenerator Core Performance.
Wall Cell Porosity

FIN Thickness, Density, EFF, P/P, Leakage,
ID mm Cells/cm 2 % % %
l
Goal @ N E = 100 Pereent 92.9 7.5 0.5

Min. Max. Max.
Coming:
Wrapped Sinusoidal T14.2 0.064 203.0 91.9 0.20
Extruded Rectangle Ext.1 0.127 161.9 91.1 0.20
NGK:
Extruded Isosceles C0.3 0.140 142.6 90.1 6.0 0.30
Extruded Isoeeles C0.7 0.076 234.0 93.6 7.3 1.7
Extruded Rectangle C0.8 0.109 215.4 93.3 9.2 1.0 (est.)
Extruded Rectangle CO.9 0.119 172.8 91.4 7.2 0.60
Extruded Rectangle CO.9* 0.109 190.0 92.1 7.7 0.40
NOTE: NGK leakage based on GE-cordierite coating and impregnation.
*MAS - A material with no coating or impregnation.
106
creased shrinkage during firing, which reduced
wall porosity from 38 percent to 28 percent.
t" z
Based on static seal leakage data for a full-
size core with the rectangular matrix param-
eter (C0.9), the estimated leakage at full
power conditions was 0.40 percent, which sat-
isfied the porosity leakage requirement. In DISTANCE BETWEEN
addition, the matrix wall thickness was re_ SLOTS = 2G
duced from 0.119 mm to 0.109 mm with a

corresponding increase in cell density from
170.5 to 186 cells/cm 2. Further development
of the C0.7 matrix with the MAS-A material
would result in significantly improved core
performance.
The regenerator thin wall AS matrix, made

by Corning Glassworks, was examined for
thermal stress using an axisymmetric stress 3-D MODEL
AGT REGENERATOR
STRESS ANALYSIS
analysis (2-D) for the cruise condition. This
represents the worst case from the thermal
stress viewpoint since the largest matrix ther-
mal gradients are present in this operating
KP= IPSII
N''N:l, GII
000TS,DED,,,ET
I
LPIN= II [] CENTEGHOL I
mode. The analysis showed the operating
1305-
tangential thermal stresses in the hub and rim
do not exceed 413 kPa. This stress is well
within the 1380 kPa capability of the material, l S<DTS -I
_ _ LOCALIZED STRESS
so that no problems would be expected in \ \ TNAT
OCCURS
regenerator matrix life with AS material at % APWGOX_._CM
I090C (2000F) operation.
=i i 7000 1015-580.670725
"li
A three-dimensional thermal stress analy-
\\ NEAR
RDT
COR
,ACE
0O
sis (SAP Ill) was completed (Figure 158) on a 435-
M_S regenerator matrix made by NGK. This
matrix contained a number of tapered slots
extending from the inside diameter of the
rubbing seal shoe on the hot face to the matrix
outside diameter on the cold face. Previous
-J 14,5- I ! ""-
I
|
analysis and experimental work had shown that ,l= , 111 ill= II li lillllil'l=lllll II III1 Ira Ill ' I
these slots were close to the optimum design 5 tO 15 20 25 30 35 40 45
for reducing tangential thermal stress. The

m74BTO-,33T
objective of the 3-D analysis was to determine
the number of these slots required to reduce
the stresses to an acceptable limit. The Figure 158. 3-D Thermal Stress Analysis.
results of this analysis indicated that 28 slots
were required and that sufficient thermal stress) considerably less than unity have accu-
stress capacity would be expected with NGK's mulated over 3500 hours of engine test with
MAS material at the cruise operating condi- relatively minor rim thermal cracking. This
tion. Engine operating experience with MAS was attributed to the fact that most of the
cores in the 707 engine has shown that MAS rim is at low stress levels and the cracks tend
cores with thermal stress safety factors (mod- to further stress relieve the rim. These cracks
ulus of rupture/maximum calculated elemental usually terminate below the rim where the
107
matrix is in compression. As the safety factor material has a high thermal stress capacity
increases, the potential for the formation of even though the AS is considerably less stable
rim cracks decrease. Since the safety factor than the MAS materials.
for this NGK core with 28 slots is approxi-
mately 1.0, thermal rim cracking is not ex-
Two full-size cores fabricated with the
pected to be a problem.
new improved MAS-A material were received
from NGK near the end of this program. Ring
The high temperature dimensional stability
gears were bonded to them for engine and test
of the NGK originaland improved MAS mater-
rig evaluation. One of these cores accumu-
ials were evaluated by periodic measurement
lated a total of 96 hours in Ford development
of specimen length change as a function of
test rig,with 53 hours at I090C (1994F) expo-
exposure time at 1000C (1832F) and II00C
sure. This core demonstrated improved
(2012F) for 1008 hours (Table 19). In addition,
thermal stress capacity compared to cores
the thermal expansion behavior between room
fabricated with the original MAS material.
temperature and 1000C (1832F) was deter-
The stress relief slots effectively eliminated
mined for each material before testing and
radial thermal cracks at the outside diameter
after completing the prescribed testing sched-
of the core on the hot face.
ule. As a reference, the Corning AS material
data generated in 1980 is included (Table 19).
Both NGK materials exhibited equivalent sta- In summary, regenerator cores have been
bilitywhen compared to reference data gener- supplied by Coming Glass Works. and NGK-
ated in 1977 for the NGK matrix that demon- Locke. The Corning core, which is fabricated
strated 2500 and 250 hours at 800C and 1000C, as a two-sheet wrap structure with aluminum-
respectively,in the Ford 707 engine. silicate material, is approximately 1 percent
below the effectiveness objective. The most
The significantdecrease in thermal expan- recent NGK extruded core with their improved
sion for the 1986 NGK materials compared to magnesium-aluminum-silicate material is
the 1977 material provides a significant de- approximately 0.7 percent below the effec-
crease in thermal stress, which allows a sub- tiveness objective. Both cores have sufficient
stantialincrease in dimensional stabilitytoler- thermal stress capacity for I090C (1994F) sys-
ance. This is the main reason the Corning AS tem operating temperature.
Table 19. Regenerator Matrix Thermal Properties.
Coefficient
Dimensional Change of Thermal
(ppm) After 1008 Hr. Exparlsion Thermal Expansion (ppm)
Exposure (X106/C) at 1000C
i|
After 1000C 1 After ll00C

NGK Material 1000C 1100C (20-1000C) Before Test Test Test
MAS (1977) -80 +400 1.5 1480 1380 1250
MAS (1986) +60 +380 0.63 620 900 840
MAS-A (1986) -60 +400 0.59 580 900 740
Coming AS +300 +700 -0.05 -50 -400 -450
108
5.4.1.4 Drive and Support System
The regenerator mounting and drive system

is a derivative of the Ford 707 engine support
system. Support roller locations have been
optimized, as shown in Figure 159, based on SHOULOER
the interacting forces of the system as well as
ring gear flexure. This optimum configuration
ORIGINAL REVISED
minimizes the load concentrated on the pinion ROLLER6EARING ROLLERBEARING
and support rollers and provides a durable
system consistent with prior Ford engine ex-
perience.
Development efforts were concentrated on

improving the graphite bearing design. The
original design, which was based on the Ford SOLID GRAPHITE STEEL-BACKED
PINION BEARINGS PINION 66ARINGS
707 engine design that successfully replaced
the grease packed ball bearing system, is G67_78-339
shown in Figure 160. The revised design

features the following improvements: Figure 160. Regenerator Drive and
Support System Bearing Designs.
1) Elimination of costly shoulders on the orig-
inal roller and pinion bearings.
2) Total compressive support for the bearing
by encompassing them with metal.
Fp 3) Simplified roller support design.
30
4) Steel backed pinion bearings can be press
fitted without preheating the housing,
which was required for the original design
when the carbon bearing was press fitted
t directly to the housing.
5) Multiple bonded graphite buttons, which

have been highly developed for other appli-
FS cations, are utilized as the thrust surface
FF for the pinion assembly.
35_ ''__ 0 FSy FR
FT
Test experience in the hot regenerator rig

ALL ANGLES WT {NO FRICTIONI and engine have indicated no significant prob-
IN GEGREES
lems.
TORQUE: 271 Nm
WT = 9 Kg
FR : GEAR SEPARATING FORCE : 39 KO There are two key areas involved in the
FT : GEAR DRIVING FORCE = 107 elastomer bonded ring gear to regenerator
TI = PHERIPHERALSEAL REACTION FORCE : 36 KO
matrix design for the AGT application that
F2 : CROSSARMSEAL REACTION FORCE : 22.7 KO
FS : SPRING ROLLERFORCE : 23 KO fall within Ford regenerator system experi-
FF : FiXEO ROLLERFORCE : 91 KO ence. These are: elastomer operating tem-
Fp : PINION ROLLER FORCE= 29 KO
perature and compliancy. One of the key
G87-078-338
elements for the elastomer bonded ring gear
Figure 159. Regenerator Support design is to minimize the temperature of the
Roller Locations. elastomer. Utilizing a two-dimension heat
109
transfer program, the results indicated that,
by locating the ring gear near the cold face of
the regenerator, the elastomer temperature
will be comparable to those in the Ford 707 OlAGONALBOND CONFIGURATION
engine (Table 20). Consequently, this elasto-
mer has had no difficulty meeting AGT engine I
cycle conditions. /
• I._. RING GEAR
Table 20. Calculated Elastomer i""
Temperatures. @_ ELASTOMER
Configuration Temp, C ORIGINAL
Ford 707 Standard 288
Ford 707 Undercut Experimental 343
AGT AS Metro/Hwy 187
AGT AS Wide Open Throttle (WOT) 302
AGT MAS- Metro/Hwy 231

I
AGT MAS - WOT 344
ELASTOMER
From the standpoint of compliance, the

CORE,,
equivalent modulus of the elastomer must be
reduced to an acceptable level to accommo-
REVISEO
date the marginal radial compressive strength
of the Corning AS matrix material. Since the
metallic ring gear grows much more than the
low-expansion matrix material, an elastomer GB7-O78-34OA
attachment with a high modulus or low com-

pliance will transmit a large outward radial
force to the rim of the core. Ford experience Figure 161. Diagonal Bond Configurations.
has shown that if the ring gear is heated while
the elastomer is being cured, a compressive High torque levels generated by Phase IV
pre-load can be applied to the rim of the core. seals operating at tight clearances caused ring
If the compressive strength of the matrix is gear separation from two NGK cores by shear-
low, the amount of preload will be limited. A ing the elastomer. In order to improve elas-
low modulus or highly compliant elastomer tomer shear strength, the diagonal bond con-
configuration is desirable to solve this prob- figuration cross-section was modified as shown
lem. Several candidate configurations were in Figure 161. Application of the GE cordie-
evaluated based on prior Ford experience in rite coating to the outer peripheral surface
the 707 engine. A three-dimensional finite created a very hard layer, which withstood the
element analysis of the Corning AS matrix additional radial force associated with the
required the highly compliant diagonal bond modified design. Subsequent cores were
elastomer design configuration as illustrated bonded with this configuration for evaluation
on Figure 161. at Ford and Garrett without incident.
ii0
In summary, the three-point support drive up to 650C (1202F) and durability was evalu-
system for the AGT engine which is a deriva- ated at temperatures up to I090C (1994F).
tive of the Ford 707 system, has not indicated Leakage evaluations were performed in
any significant problems in limited rig and metallic structured rigs with flow paths simi-
engine testing. Attachment of the metal ring lar to the engine configuration. The Garrett
gear to the ceramic regenerator matrix with rig attempted to isolate the regenerator sys-
the diagonal bond configuration, which has a tem from the outer system seals by welding or
high degree of compliance, has minimized the brazing the structural assembly together,
drive system reaction forces imparted on the
eliminating many of the interpath seals. This
core.
rig incorporated a separate flow path for the
high and low pressure sides of the regenerator
5.4.1.5 Regenerator System Development
system. Leakage was measured by seeding the
Rig high pressure inlet gas with helium and sensing
the helium concentrations in the low pressure
AGT101 regenerator system rig testing has
side gas discharge. The Phase V-A configura-
been performed during the AGT101 project at
tion seal was tested in this rig. Leakage was
both Ford and at Garrett. The purpose of this
measured at the 6 percent level at the
testing was to evaluate the regenerator sys-
elevated temperature condition of 650C
tem under controlled conditions for develop-
(1202F). Continued development was con-
ment of the regenerator system components
ducted in the Ford rigs.
and evaluation of the regenerator system op-
erating characteristics. Regenerator system Shown in Figure 162, the Ford rig was
leakage was evaluated at inlet temperatures similar to the Garrett rig, though it could
SEALS
TURBINE
P3.6
COVER_ RINGS
PISTON_
RINGS_
[{]WELOED JOIN1
REGENERATI
SEALS REGENERATOR
CORE
GB7-0711-341
Figure 162. Leak Paths and Pressure Notations.
iii
accommodate either metallic or ceramic 2) The entire velocity pressure in each of the
structures. The leakagesfor each of the non- restricting areas is lost.
regenerator system seals were calibrated in
cold, static testing and corrected for high 3) Reynolds Number effects are ignored.
temperature conditions. For measurement of
overall rig leakage at operating temperature, 4) Temperatures at the various points in the
this rig utilized a high temperature orifice flow path were inputted from rig test data.
measuring section between the regenerator
high pressure discharge and rig combustor in- 5) The flow areas used to eliminate the dif-
let. To simulate the pressuredrop across the ferent leaks were determined from mea-
turbine this rig incorporated a valve which surements made in the cold, static rig in
seated on the turbine shroud. The valve the Build-Up Laboratory.
featured a power screw adjustment that
allowed the valve position to be changedin the 6) Other flow areas were estimated from
test cell for precise simulation of engine pressure drop measurements made in the
operating conditions. The operating tempera- dynamic rig.
ture of the rig was limited to 590C due to the
materials used in the orifice section. The 7) All flow areas, except that of the valve,
regenerator leakage was determined by sub- were held constant in all calculations.
tracting the system leakagesfrom the overall•
This model correlated well with the test
Assembled in the ceramic configuration, data taken from the rig at inlet temperatures
this rig was characterized for leakage in cold up to 590C (1094F). The regenerator system
static testing, and overall leakage measured leakage at the hot, cruise condition was pro-
up to the 590C temperature limit of the jected at 4.0 percent using the same model.
metallic interpath orifice section. This data This was close to the program objective of 3.6
was further reduced through analysis to deter- percent• If a Phase VIA hot seal and a low
mine how the regenerator seal leakage com- leakage Corning core had been used, it was
pared to the program goal of 3.6 percent. anticipated that the leakage objective would
have been met.
An analytical model was developed to aid
in understanding the different sources of leak- Durability Testing
age. The flow paths used in this model are
illustrated in Figure 163. All the high pres- After completing the seal leakage evalua-
sure, cold air leaks (P3.0 and PS.I) were tions, elevated temperature testing was con-
consolidated together and simulated by one ducted to evaluate the high temperature dura-
leak, as were the high pressure, hot air leaks bility of the regenerator system. This testing
(P3.6 to P5.1). The unique high pressure, hot was conducted in the ceramic version of the
air leak, (P3.6 to P4.1) which is present in this Ford Regenerator System Rig without the in-
rig but not in an engine, was also included. terpath orifice section installed. This rig was
The components and leak paths _,ere defined capable of operation of regenerator inlet tem-
by a set of flow restrictions, as illustrated in peratures up to the system design goal of
Figure 163. Each restriction is an orifice 1090C (1994F).
having a specified flow area. The computer
model iterates on this network of restrictions During this test program the rig was as-
until it converges on a solution in which the sembled 13 times, accumulating 238 hours of
sum of the pressure drops of the individual testing, including 73 hours at I090C (1994F)
components equals the available pressure drop, regenerator inlet temperature. This testing
P3.0 to P6.0. The following assumptions ap- contributed to the design and development of
ply: the inner regenerator seal cooling configura-
tion and evaluated improved materials and
1) The flow is compressible. designs used in the regenerator core assembly.
112
ANALYTIC MOOEL
P3.0 TO P5.! LEAKS

le,,
P3.6 TO P5.1 LEAKS
P3.6 TO P_ LEAKS
_Y
,,. °.u.,. &Eke- L, ,OE
INLET
ORIFICE
,E...E.,. .C. .C. .,.E.T.
MEASURING
ORIFICE
ANALYTICAL SIMULATION OF MODEL

WITH FLOW RESISTANCES
OOWNSTREAM
AIRFLOW FLOW
INLET HP SlOE MEASURING TURBINE CONTROL LP SlOE
ORIFICE REGENERATOR ORIFICE COMDUSTOR NOZZLES VALVE REGENERATOR
P3.0 P3.6 P4.I P5.1 P6.O

MAIN le,
FLOW
P3.6 TO P4.1 LEAKS
•._ _ _ ._ "_V% "v_-_MV_. f _

P3.6 TO P5.1 LEAKS
P3.O TO P5.1 LEAKS
Gil7.078-342
Figure 163. Analytical Model for Dynamic Rig Leakage.
The regenerator seal cooling system was configuration of the regenerator rig. Long
developed through an iterative process. Seal term durability of the seal assemblies were
diaphragms were instrumented with thermo- evaluated in tests planned for 50 and 100
couples and installed in the rig for each test. hours. Significant problems such as regenera-
The resultant temperature data was recorded tor core thermal cracking, seal coating delam-
and evaluated, providing valuable information ination, and shoe distortion were discovered
for seal design improvements such as the ap- and evaluated during extended durability test-
plication of thermal barrier coatings to the ing.
seal shoes, finned support diaphragms etc..
The curves in Figure 164 show typical data The regenerator core thermal cracking
from this testing. problem was related to the thermal expansion
of the MAS regenerator core material and the
Additional information was generated as a strength of t]pe matrix. In cores assembled
result of endurance testing in the ceramic from extruded segments, the strength and
113
C IFI
190( 6"_,
i,,m
!000, (_. _'_ "3.
%
900. _b• %
IBOO _• •
800.
•i
1400
700" 1300 i , , , ,
NOMINALT5.I = I093C 1200OF)
C 1200 I I I I ' L METALTEMPERATURE
I00O- IFI
1800
1700 <.,),_
8OO- 1500 _"_ _'= _ _ BUILDCLEARANCE

=
C_ _ I__ 1.87 mm10.074IN]
", .c o *3.B
1300 'm3
700,_ %% _ 0 1 _- --.0
-,_ THERMOCOUPLE
#1
THERMOCOUPLE
_2
1200 i I _k
NOMINALT5.I = 962C 11800Ei %% O-----C] THERMOCOUPLE
#3
5001 1100 I I I I _ I
C IFI LOCATION
OF
1400 _ _ OIAPHRAGM
SPRING
1200 C_ '._%
"" I I I r',,.
t_O
NOMINALT51 = 843C 11550F) •

800 t I I I t 0.4/0 COOLING
FLOW
30 40 50 80 70 60 90 IO0 120
r3J) liNHGA)
gO/-Ol&343
Figure 164. Seal Diaphragm Temperatures at 0.4 Percent Cooling Flow.
114
compatibility of the bonding cement was also Subsequent investigation in the regenerator
a significant factor. Small matrix cracks form rig evaluated the coating delamination at tem-
on the outer periphery of the hot face of the peratures up to 1090C (1994F). Seal design
core after several hours of durability testing. modifications were made to remove restric-
In the most severe cases, the cracks would tions to seal thermal growth, however the
propagate with time and temperature across distortion was evident in follow-up testing.
the face of the core to the center, generating
large leaks, and core debris. The problem was Regenerator System Sum mary
solved through proper location of stress relief
slots in the outer rim of the core and the use Based on test data and analysis accumu-
of improved MAS-A material from NGK. lated at both Ford and Garrett, the current
status toward achieving the three main pro-
The distortion of the regenerator seal gram objectives are as follows:
shoes, and delamination of the seal coating
during I090C (1994F) operation are related I) 92.9 percent regenerator core effective-
phenomenon which were discovered during the ness with less than 7.5 percent pressure
last phases of the AGT regenerator seal devel- drop at WOT engine conditions.
opment effort. The photograph in Figure 165
shows the crossarm shoe and how the coating Regenerator cores supplied by Corning
delaminated from the inner periphery of the Glass Works and NGK-Locke, Inc. are ap-
shoe wearface. Inspections of the seal indi- proximately 1.0 percent and 0.7 percent
cated that the circular inner periphery was below the effectiveness objective, respec-
distorted to an oblong shape and that the tively, with a nominal 7.5 percent pressure
coating was lost from the distorted portions of drop. NGK has developed a thinwall isos-
the shoe. Characteristically, the seal shoe celes triangular structure that will exceed
was also bowed axially toward the coated side. the objective provided a more dense mate-
SHOE CROSSARM
I __
Figure 165. Seal Coating Delamination from Crossarm Shoe.
115
rial can be developed to have acceptable 5.4.1.6 Non-Diaphragm Seals
wall porosity leakage.
As an alternate approach to cooling the
2) Regenerator System capable of operating inboard (hot) seal crossarm diaphragm, a feas-
at 1090C (1994F) regenerator core inlet ibility study was initiated to eliminate dia-
temperature. phragms from this seal, as shown in Figure
166. The core would run against a wear face
Coming cores fabricated with AS material coating applied directly to the flow separator
and NGK cores extruded with their im- housing (FSH). The key requirements for a
proved MAS-A material have demonstrated successful non-diaphragm seal are low friction
the thermal stress capacity and matrix and wear coatings combined with minimum
thermal stability required for I090C engine leakage gaps resulting from compatible plat-
conditions. form deformations of the core and FSH. The
work plan for this study consisted of the
Ring gear attachment to the regenerator following:
with a compliant elastomer system design
and the associated drive and support sys-
tem has also demonstrated the system CURRENT SYSTEM
temperature objective.
HP
Outboard (cold) seal assemblies with all
1-85 coated substrates, except for 1-112 on
the circular portion of the crossarm, have
LP [-- I_:A_H_-_OHAGM SYSTEM
achieved the temperature objective.
I-II2 COATING
Although inboard (hot) seal assemblies have

demonstrated the capability of the 1-112
coating to operate in a 1090C (1994F) NON.-OLAP'NOAG
M SYSTEMS
environment, distortion of the present sub-

strate materials limits this component to a
1040C (1904F) system temperature. Sub- CO,ER
__ ,./I ,.
stitution of 316 stainless and Haynes 230 "_/-/JUJl .,.-,'M
material for 430 stainless peripheral and
INCO 601 crossarm substrates, respective-
ly, should achieve the 1090C (1994F) tem-
.,T,NO GOI ':gE H
APPLIEO OIRECTLY HOUSING
TO HOUSING
perature objective. U7-078-348
3) Maximum regenerator system leakage of

3.6 percent at WOT engine condition. Figure 166. Seal Systems Comparisom
Based on static and dynamic leakage mea- I) Thermal analysis

surements in the Ford test rigs and sup-
ported by analytical models, the regener- a) Three dimensional finite element analy-
ator seal system appears to have achieved sis of flow separator housing (FSH) dis-
the leakage objective. In addition, the tortion performed at Garrett
Corning and NGK cores have also demon-
strated minimal porosity leakage that is b) Two dimensional finite element analysis
essential for achieving the system leakage of the regenerator matrix distortion
objective. performed at Ford
116
c) Labyrinth seal leakage analysis for re-
generator core - FSH gap performed at
Ford
2) Attachment of ceramic coatings to the

FSH
3) Evaluate coating wear and friction char-

acteristics in the seal wear rig,
An analytical model based on labyrinth seal

theory was developed to evaluate non-dia-
phragm seal leakage as a function of regen-
erator core and flow separator housing (FSH)
distortions. The inboard (hot) seal consists of
six leak path components, as illustrated in
Figure 167. Analytical expressions were de- SEAL
veloped for both the circular and linear com- LENGTH,
COMPONENT CM (IN.)
ponents.
I 67.5 (26.61
The hot face thermal distortion for the 2 87.5 126.61
3 27.4 110.81
original Coming (AS) and NGK (MAS) mater- 4 27.4 110.8)
ials was previously evaluated for cruise and 5 10.9 [4.3)
6 10.9 (4.3)
full power conditions, which corresponds to
engine speed of 60 percent and I00 percent,
respectively. The results, shown in Figure
168, indicated that the Corning material had
acceptable distortion variance (0.0152 ram),
whereas the original NGK material variance
(0.076 ram) was unacceptable for a nondia-
phragm seal system.
The obiective for the AGT regenerator EE7_Ig_I8
system leakage is 3.6 percent Inaximum at the

maximum power operating condition. Based Figure 167. Analytical Model Evaluating
on static leakage data and analysis, the total Non-Diaphragm Leak Paths As a Function
leakage consists of the following: of Component Distortions.
I) Inboard (hot) seal assembly - 1.9 percent

2) Outboard (cold) seal assembly - 1.2 percent In order to evaluate seal system sensitiv-
3) Regenerator core carry-over loss = 0.25 ity, the initial study consisted of evaluating
percent seal component leakage as a _nction of a
4) Inboard seal diaphragm cooling allowance = uniform gap at cruise and maximum power
0.25 percent operating conditions, which are listed on Table
21. The results at cruise, shown in Figure 169,
Therefore a non-diaphragm inboard seal and maximum power, shown in Figure 170,
will have a design objective of 2.15 percent demonstrate the powerful influence of leakage
(1.9 percent seal + 0.25 percent cooling) as a air temperature. As expected, the leakage for
replacement for the present diaphragm seal components i and 4 are minimal since pressure
syst em. drop at these locations is very small (less than
117
•,)_ p.OOR QUALIT_'_
CORE RAOIUS
CM 5 I0 15 20
I I I J
MM tONi I 2 3 4 5 6 7 8 9
O- O I I I L 1 J
0.001•
005.
0.002 •
0003- NE = 50 PERCENT: UNSHADED
NE _ 100 PERCENT:SHAOEO
0.10--
0 CORHING1946ll
0.004.
_, NOK ICO.31
_ 0005,
_ 0.15-
0._
0,067-_
0.20-
0 000._
0.009-1
0.25-
O.OlO_
li17]Tt,141
Figure 168. Regenerator Core Axial Distortion Effects.
Table 21. Leakage Air Flow Conditions. 10.28 kPa). Conversely, the leakage for Com-
ponent 2 is substantially higher than Compo-
nents 3, 5, and 6 even though pressure drop is
Component Po PN ] TO approximately the same.
Speed = 60% W = 0.161 This is directly attributable to the much

1 205.5 204.1 99 lower temperature of the compressor dis-
2 205.5 104.8 99 charge air compared to the discharge air
3 204.1 101.3 1063 through the high pressure side of the regenera-
4 104.3 101.3 1090 tor. Component 2 leakage represents approxi-
5 204.1 104.3 1063 mately 70 percent of the total leakage at both
6 204.1 104.8 1063 engine conditions (Figures 169 and 170). Leak-
age through component 3 is more than the
Speed = 100% W = 0.377 crossarm (componen_ 5 and 6) due to the
1 506.1 504.7 243 following reasons:
2 506.1 111.7 243
3 504.7 101.3 914 1) The seal width of the crossarm (38.1 ram)
4 111.7 101.3 965 in twice the width (19.0 ram) of the peri-
5 504.7 111.7 914 pheral seals
6 504.? 111.7 914
2) For a wrapped process regenerator (Corn-
DO: Leakage Inlet Pressure, KPaA ing Glass) the crossarm will have a geo-
PN: Leakage Exit Pressure, KPaA metrical advantage due to a substantially
TO: Leakage Inlet Temperature, C higher number of sealing points (laby-
W_. Engine Air Flow, kg/sec rinths).
o
118
SYMBOL COMPONENT
0 OUTER HP "C"
INNER HP "C"
[] _ OUTER LP "C"
0 INNER LP "C"
© 0+0 CROSSARM
0 TOTAL
O
NE = 61) PERCENT
W = AIRFLOW = 0.161 KG/S (0.355 LB/SEC}
IINI 0 0001 0002 0003 0.004 0005
I I i I I
(MM] 0 0.025 0.05 0.075 0.10

CLEARANCE
GD7_7&348
Figure 169. Non-Diaphragm Seal Leakage Versus Uniform Clearance.
To further demonstrate the importance of ring support interface, which results in zero
air temperature, the leakages at cruise and axial distortion of the outer ring of the FSH.
full power were estimated based on ambient Consequently, the distortions predicted are for
air as a function of uniform clearance (Figure the inner ring relative to the outer ring. In
171). In this case, the leakage with ambient addition, to simplify the initial analysis, dis-
air was approximately twice as high as leakage tortion due to temperature and pressure were
based on engine temperatures. evaluated separately. Distortion due to tem-
perature was evaluated at engine cruise (N E =
The next study consisted of evaluating seal 60 percent). Conversely, distortion due to
leakage for the FSH distortions predicted by pressure (isothermal) was evaluated at full
Garrett from their 3-D finite-element anal- power conditions (N E = i00 percent).
ysis, which is summarized on Table 22. For
the current Garrett model, one of the bound- A total of thirteen combinations were
ary conditions assumes axial constraints at the evaluated and summarized on Table 23. The
119
SYMBOL COMPONENT
<)
[] OUTER LP "C"
& iNNER HP "C"
OUTER HP "C"
0 INNER LP "C"
(b 0+0 CROSSARM
0 TOTAL
A L 77 KG/S 10832 LB/SEC]
On
I
PNI o 0.001 0+002 0.003 0.004 0.005
! I i . I
[MMI 0 0.1)25 0.050 0.075 0.10
G87+07114140 CLEARANCE
Figure 170. Non-Diaphragm Seal Leakage Versus Uniform Clearance.
maximum component gap is predicated by the baffling the compressor discharge flow around
contact point that results from meshing the it, the distortion is reduced significantly. In
axial distortion of the core with that of the addition, distortion of an RBSN housing due to
housing (FSH). If the distortion of the core is pressure is reduced significantly with the in-
greater than the FSH, then a gap will exist at corporation of struts.
the outer ring (components 1 and 2) and leak-
age will be substantial. Conversely, when the
axial distortion of the core is less than the Based on these preliminary results, the
FSH, then the gap at the outer ring is zero and leakage of a non-diaphragm inboard seal sys-
leakage will be minimized. tem was projected to meet the objectives,
provided the clearance between the core and
For the thermal distortions without pres- FSH at the outer ring was minimized. To
sure at cruise conditions, leakage was calcu- ensure this condition is met, the axial distor-
lated for both the current configuration (H = tion of the core should be less than the axial
28.39 W.m2C) and the simulation of a metal distortion of the FSH inner ring (components 3
baffle (H = 5.68 W/m2C) around the outer and 4). In addition, to more accurately predict
perimeter of the FSH. Initial results at seal leakage, axial distortion of the FSH due
Garrett indicated that, by allowing the mount- to the combined effects of pressure and tem-
ing flange of a RBSN housing to heat up by perature must also be determined.
120
A composite coating on a LAS ceramic
wear rig substrate consisting of a pre-heat
treated coating of Ford ATLAS glass frit,
THE EFFECT OF TEMPERATUREON
SEAL LEAKAGE which provides an intermediate bond coating,
and I-I12 plasma sprayed wear face coating
was subjected to I090C (1994F) furnace expo-
sure without failure. The sample accumulated
SYMBOL NE. PERCENT 63 hours at temperatures up to 980C (1796F)
0 6O in the seal wear rig against an LAS core
[] 100
without apparent difficulties. The coefficient
UNSRAOED: AMBIENT 21C {7OF) of friction varies between 0.16 at 980C
SHADEO: ENGINE TEMP. (1796F) and 0.22 at 815C (1499F) with negligi-
ble wear. A friction coefficient below 0.3 is
considered excellent.
Another LAS sample with S-77 material

applied without the Ford ATLAS intermediate
bond coating accumulated 120 hours in the
seal wear rig at temperatures up to 980C
(1796F). The coefficient of friction varied
between 0.15 and 0.20 with negligible wear.
Minor chipping and delamination occurred
after the 980C (1796F) exposure.
Since flow separator housings made from

silicon nitride were being considered in the
AGT program, S-77 coating was sprayed on a
silicon nitride substrate. It accumulated 75
hours in the wear rig at temperatures up to
870C (1598F) without distress. The coeffici-
ent of friction varied between 0.08 and 0.13.
The test was terminated when the ceramic
sample holder failed while attempting to run
the 980C (1796F) condition. A casting mold
I I '1 I was fabricated to form reaction bonded sili-
(INI 0 0.0005 0.0010 0.0015 00020 0.0025
con-nitride (RBSN) blocks as a replacement
! ! I
MM 0 0.025 0.05 sample holder. One of these blocks was
CLEARANCE
sintered and machined into a sample holder.
G07-078-350
Garrett provided Ford additional funding to

Figure 171. Effect of Temperatu_on coat a thick-flange LAS flow separator hous-
Seal Leakage. ing with the special "non-diaphragm" seal
coatings. The thick flange separator housing
The evaluation of candidate coatings and (FSH) with the plasma sprayed Ford wear face
method of attachment to LAS samples was coating (S-77) was held at I090C (1994F) for
investigated. Based on furnace exposure and 24 hours in a kiln without difficulties. A cold
seal wear rig testing, two wear face coatings seal was fabricated with thicker diaphragms to
developed at Ford continue to show .promise. increase the mechanical loading to compen-
These coatings are zinc-oxide based and desig- sate for the diaphragms eliminated from the
nated as 1-112 and S-77. hot seal. This modification was required to
121
1125
Table 22. Flow Separator Housing Axial Distortion.
Flow Flow Film Distortion (ram)

Sep Coeff Loc #3 Loc #4 Loc #5 & #6
Speed Sep
% Mat'l WIM2C Max Min Max Min Max Min
Config
60 LAS Non 31.2 0.147 0.143 0.147 0.147 0.139 0
60 LAS NON 6.24 0.099 0.083 0.099 0.101 0.094 0
60 RBSN NON 31.2 0.272 0.241 0.267 0.246 0.254 0
60 RBSN NON 6.24 0.0914 0.0813 0.0889 0.0864 0.089 0
60 RBSN Strutted 31.2 0.099 0.038 0.099 0.066 0.099 0
100 RBSN NON 0 -0.046 -0.005 -0.099 0.046 0.046 0
100 RBSN Strutted 0 0.018 0.0051 0.005 0 0.005 0
100 LAS NON " 0 0.069 -0.0457 0.0457 -0.069 0.046 0
I00 LAS Strutted 0 0.038 0.015 0.015 -0.005 0.015 0
697-078-351
122
Table 23. Non-Diaphragm Seal Leakage Analysis.
9
\ _ .91-- _ +AF.S.H
LEAKAGE
08JECTIVE: 1.25 PERCENT MAX.
Max
Component Gap, mm
Core Film Total Leaka_e
Speed, FSH FStI Core Dishing Coeff, Loc Loc Loc Loc Loe
4% Contig Matl Type mm WIM2-C 1 2 3 4 6 (kg/s) (%)
60 No strut LAS Coming 0.03048 5.678 0 0 08.04 71.12 50.8 0.000612 0.38
60 No strut LAS Coming 0.03048 28.39 0 0 114.3 116.84 88.9 0.001421 0.88
60 No strut LAS NGK 0.2286 5.678 147.32 147.32 16.002 18.542 119.38 0.011338 7
60 No strut LAS NGK 0.2266 28.39 104.14 104.14 20.32 22.86 88.9 0.007256 4.5
6O No strut RBSN Coming 0.03048 5.670 0 0 58.42 58.42 43.18 0.000506 0.37
6O No strut RBSN Coming 0.03048 28.39 0 0 238.76 231.14 177.8 0.003930 2.4
60 No strut RBSN NGK 0.2286 5.078 147.32 147.32 8.382 6.35 114.3 0.011187 6.9
80 No strut RBSN NGK 0.2280 28.39 0 0 40.64 33.02 30.48 0.000226 0.14
6O Strutted RBSN Coming 0.03048 28.39 0 0 63.5 60.96 35.56 0.000453 0.28
100 No strut RBSN Coming 0.04064 0 114.3 114.3 88.9 35.50 83.82 0.021467 5.7
100 Strutted RBSN Coming 0.04064 0 17.78 17.78 16.51 5.08 12.7 0.001511 0.39
100 No strut LAS Corning 0.04064 0 152.4 152.4 20.066 7.874 12.192 0.033713 0.9
100 Strutted LAS Coming 0.04064 0 20.32 20.32 40.64 12.7 20.32 0.002418 0.64
G87-078-352
123
ensure that the core will contact the FSH. velocity measurements in the regenerator core
The FSH and associated cold seal with the mapping test.
thicker diaphragms were sent to Garrett for
evaluation. The hot wire technical capability required
for testing was developed in a specially de-
In summary, preliminary analysis of the signed calibration tunnel (Figure 173). The
non-diaphragm seal system indicates feasibil- calibration tunnel was utilized to cafibrate a
ity provided the clearance between the core total of eight 2-D and two 3-D hot wire
and FSH is minimized. To ensure this condi- probes for use in the regenerator core survey
tion is met, the axial distortion of the core test. Additionally, an in-house hot wire probe
should be less than the axial distortion of the repair capability was developed to support the
FSH. Based on furnace exposure and seal wear test.
rig testing, two wear face coatings continue to
show promise for bonding to the FSH. Analytically predicted turbine discharge
conditions had to be duplicated for the dif-
5.4.2 Garrett Regenerator Development fuser testing. Specifically, turbine exit swirl,
radial pressure profile, and corrected flow
Two cold flow regenerator test rigs, low rates were duplicated for idle, cruise, and
pressure (LP) and high pressure (HP), were maximum power engine conditions. Hardware
designed and fabricated at Garrett in order to: configurations to meet these requirements are
shown in Figure 174. Preswirl assemblies for
a) Determine circumferential and radial pres- idle (+55 degrees), cruise (0 degrees), and max-
sure profiles imposed by engine geometry imum power (-27 degrees), were fabricated
on the regenerator core along with various percent open areas screens
to generate the analytically predicted 5-per-
b) Evaluate flow profile control mechanisms cent radial pressure gradient.
c) Analytically predict the regenerator LP Rig Test Results

installed effectiveness based on the test
The main objectives of the LP cold flow rig
profiles obtained in Steps 1) and/or 2) tests were as follows:
Construction of an analytical flow and heat Determine circumferential and radial pres-
transfer model of the regenerator system in sure distortion profiles imposed on the LP
3-D form was completed and is discussed in inlet core face by engine flowpath geome-
paragraph 5.4.2.3. try
5.4.2.1 Regenerator Low Pressure (LP) Investigate and evaluate different flow in-
Cold Rig fluencing devices and the effect on overall
regenerator performance
The LP regenerator cold rig design con-
cept, Figure 172 is based on engine/model As previously stated, hot wire anemometry
similarity in areas of flow path geometry, techniques were employed during initialtest-
regenerator matrix frontal area, pressure ing. However, due to the geometry of the 3-D
drop, and engine corrected mass flow in the hot wire probe sensing elements, wherein
turbine discharge plane. Rig objectives in- discrete (matrix cell) velocity jets could im-
clude baseline turbine diffuser performance pinge on an individual probe element, the
mapping and regenerator flow pattern map- results of testing were highly suspect. There-
ping. Instrumentation requirements for the fore, hot wire anemometry data was used to
diffuser performance are shown in Figure 172. ascertain flow trends within the plenums, and
Special instrumentation (hot wire anemom- core pressure drop data was used during later
etery) was required for accurate low level testing for analysis.
124
1 PS TAP NO
(HOLE AND SLOT

PREPARED ON 1 PS TAP NO 46
P/N 3808128)
PREDRILLED
ON P/N 3601112_
1 Ps TAP NO 47
12 Ps TAPS NO :]4
(HOLE PREDRILLED THRU NO 45
ON P/N 3608127)
EQUALLY SPACED
S PT BRA HEADS
NO 20T J NO 25
- EQUALLY SPACED
tHOLES 4 TAPS
REGENERATOR
ARE PREPARED ON
CORE (LP| P/N 3608124
SWAGE LOK
6 REQUIRED
1 PS TAP NO 49_
(HOLE & SLOT _ 0.21

0.19
PREPARED ON _ E
TO CENTERLINE OF
r *
COBRA HEAD
TYP 6 PLACES
;UST DIFFUSER
19 PS TAPS NO 1
\k 8 C
THRU NO 19
tHOLES FOR STATIC
PRESSURE TAPS ARE

PREORILLED ON
P/N 36081241
19 PS TAPS NO 50
EXTEND PS TUBES
THRU NO 68
STRAIGHT BACK FOR
(HOLES FOR STATIC
18.00 MIN LENGTH PRESSURE TAPS ARE
PREDRILLED ON
NO SEALLANT P/N 3608125)
REQUIRED AROUND
PS TUBES PS TAPS NO, 1
THRU NO 12
EQUALLY SPACED
CIRCUMFERENTIALLY
PS NO 50 THRU NO 61 EQUALLY SPACED
CIRCUMFERENTIALL Y
PS NO62
NO 13
/ NO 84
Ps NO
m
/
- P$ NO 6S
/ PS NO 67
-- PS NO 68
Figure 172. Regenerator LP Cold Rig.
125
/
./
• /
i
Figure 173. 3-D Hot Wire Calibration Tunnel.
8 PRESWIRL VANE A$SY 10 FLOW CONDI'TiONER

MAX POWER PLA'[E
PRESWIRL VANE ASSY i!_

....
IDLE "_ ::
Figure 174. Regenerator LP Cold Rig Preswirl and Screen Assemblies.
126 oi,"i-'.oOi<
QUA//2.M
A series of tests were conducted at simu- drop) for the LP core face imposed by engine
lated turbine rotor exit swirl angles and re- flowpath geometry at the simulated cruise
spective flows to determine the effect of swirl condition.
on regenerator LP inlet pressure distortion.
Swirl angles of +55 degrees (idle), 0 degrees Several flow influencing devices were eval-
(cruise), and -27 degrees (maximum power) uated to determine the potential of reducing
were induced at the rotor exit plane. Concur- LP inlet flow distortion. Downstream devices
rent 3-D analysis showed that circumferential (baffles) showed limited potential for reducing
distortion of comparatively large magnitude LP inlet flow distortion. An upstream
has a negligible effect on regenerator effec- "diverter," Figure 176 showed good potential
tiveness when gas/matrix heat transfer and for redistributing the flow, as noted in Figure
core rotation effects are considered. 177. The parameter U/VRE F seen in Figure
177 is based on 3-element anemometer data
Tests also were conducted at various mass
flows [hence Mach number and plenum head Q
(ev2/2g)] and 0 degrees swirl. These tests 3.8 CM
were conducted to determine effects of mass UPSTREAMDIVERTER [1 5
TURBINE
flow variations on regenerator LP-inlet pres- DIFFUSER
sure. Distortion magnitude remains approxi- -,< EXiT
"\ "_"
o
mately linear with inlet Q as mass flow rates
increase.
Based on the above test results and the _J

z
fact that the aggregate fuel consumed over
CFDC in the upper 60-percent power range is I
negligible, all remaining cold flow pressure

distortion testing (LP and HP) was conducted
6BT_TO_IS0
at cruise conditions.
Figure 175 presents the observed inlet Figure 176. LP Regenerator Cold Rig
pressure distortion (local cold core pressure Upstream Diverter.
kPa (IHH20i
2.0
1.0
'rEST.7 I[
O-tO_ II
,.,co,.oO,,,A
cRuisc,,,ol,,o,s I CRUISE CONDITION _,.._
CIRCUMFERENTIAL l _, I I
1.5 -- WITHOUT OIVERTER __ 7"-
3.5_[- POSITION. DEGREES--
Ii CRUISE CONDITtON
WITH OIVERTER
/ /
!
I ! I
!
I
25 RAW TES _J
PHI 4 5 | 7 O 5.0 6.0 7.0 6.0
IIN) 4.0
i I I
I I I
CM 10 15 CM 10 i5 2O
RADIUS
RADIUS
U7.07_358 667.078-,180
Figure 175. Measured LP AP Data Figure 177. LP Regenerator Cold Rig,

From Regenerator Cold Rig. Effects of Upstream Diverter.
127
and is the ratio of local to average axial 5.4.2.2 Regenerator High Pressure (HP)
velocity at core inlet. Cold Rig
The HP side regenerator cold rig (Figure

For the ceramic engines, placement of the
178) has the same criteria for engine simi-
upstream diverter requires ceramic fabri-
larity as the LP side of the cold rig. The
cation tooling changes. Therefore, it was
objectives of this rig are to match the flow
decided to await hot regenerator rig test re-
patterns of the LP side such that the maxi-
suits prior to any further evaluation or adop-
mum heat transfer is optimized. Flow control
tion of this upstream diverter for AGT101
techniques for this purpose include various
application.
hole patterns and radial web stiffeners in the
exhaust housing and internal recontouring of
Data obtained from LP and HP cold rig the housing. Here again, a 3-D hot wire probe
testing were used to predict regenerator per- was installed immediately downstream of the
formance in the engine as described in"para- core for direct low level velocity measure-
graph 5.4.2.3. ment. Additional instrumentation was in-
CLEARACRYLIC
/
_X
COMBUSTOR
REGENCORE /" INLET
G87-078-393
Figure 178. Regenerator HP Cold Rig.
128
cluded for the determination of the combustor Figure 180 shows the HP regenerator cold
annulus inlet velocity profile such that an rig test data used for subsequent analysis.
optimum shape could be determined for the Testing was conducted both with and without
regenerator exit/combustor inlet guide vane the HP-inlet housing (exhaust housing) and
configuration. A uniform circumferential flow confirmed that HP-inlet "flow distortion is
profile is desired at the combustor inlet to dominated almost solely by hot face flow
assure the best flame control and primary/ separator housing geometry (downstream bon-
secondary mixing in the combustor. Effects of net).
this guide vane selection on the regenerator kPa IINH201
flow profile were also determined. IJ /
NP COLO SiS OATA CllLCUMF'ERENTIAL
TEST I(.10 POSITIONS.OESREES
0.44
Regenerator HP Rig Testing 12-12,E) o--340
CRUISE CONOITIONS o-- 310

o-- 270
,_-- 2"30
Main objectives for the HP cold flow rig 1.1
0.42
tests were the same as the LP cold flow rig
tests.
o--200
Testing was initiated to determine HP inlet O.4O ]J ¢
flow distortion at engine cruise operating con- • |

ditions. Radial distortion (from hub to shroud) I
was measured for a bare, untreated HP flow- SAW TEST OATA
15 I
path.
liNI 4 5 S 8
t I I
Several flow influencing devices (bonnets) CM 10 15 20
were evaluated on the downstream HP flow- RADIUS
path utilizing hot-wire anemometry. Figure G87-078-363

179 shows the dramatic influence of one of the
bonnets on HP-flow conditions in terms of HP- Figure 180. Measured HP AP Data
discharge velocity ratios (U/VREF). From Regenerator Cold Rig With
Bonnet.
5.4.2.3 3-D Finite Element Analysis of

Regenerator Cold Rig Data
Regenerator Analysis
To predict regenerator effectiveness based

on cold flow test data, an analytical model
g, l using 3-D finite element techniques was used.
The core matrix was divided into 10-radial
stations, 20-axial stations and 360-rotational
increments. Core matrix geometry and pres-
0.5 i sure drop/heat transfer characteristics were
4.0 4.5 5.5 5.5 8.0 a.5 7.0 7.5 LO program inputs.
[INJ
I I t
CM IO 15 20 Care should be exercised when comparing
RADIUS quoted regenerator effectiveness associated
GB7-070-362
with engine performance results presented
herein. Engine performance includes effects
Figure 179. HP Regenerator Cold Rig of regenerator seal leakage and, therefore,
Flow Effects of Bonnet. represents the mixed flow effectiveness for
129
the entire regenerator system. The results core matrix, are determined using local prop-
described here refer only to core effective- erties (i.e.,local Reynold's Number, viscosity,
ness. Mass flows used in the 3-D analysis for etc). The analytical model also includes
both the LP and HP flowpaths have been effects of local radial and axial conductive
adjusted for predicted seal leakages. heat transfer to neighboring elements. Indi-
vidual elements are analyzed for heat transfer
Analytical Model along five paths (four core conduction and one
gas convection path). Time integration is used
Regenerator core effectiveness was ana- to arrive at a steady-state heat transfer solu-
lytically predicted by imposing measured or tion that is consistent with this pressure dis-
prescribed core inlet face pressure distortion tortion profile imposed at the HP and LP inlet
profiles on the rotating 3-D heat transfer planes.
mode. Figure 181 shows the principle features
of the analysis wherein matrix geometry, Sample Procedure
Colburn number characteristic, Fanning fric-
tion factor characteristic and core rotational As an aid in explaining the test and analyt-
speed are program inputs. Heat transfer and ical procedures, the following paragraphs
core pressure drop calculations, within the describe the steps involved in obtaining an
analytical prediction of regenerator perfor-
mance in the engine. Since the cruise condi-
tion is of most interest, it is the subject of the
P = ,_lr, el
example case.
% Step 1: Cold Rig Testin_
Cold rig tests were performed using the LP

and HP cold flow rigs. The salient corrected
flows (and therefore the Mach numbers) pre-
dicted for engine operation at cruise were
supplied to the respective rigs. The corrected
P" o2 - RAmAL
CONOUCnON
OFCORE flows simulated were LP inlet and HP dis-
charge since these are the locations generat-
O3, Q4 -- AXIAL CONDUCTIONOF CORE ing the bulk of the pressure distortion.
_o_HGAS -- HEAT TRANSFERWITH FLOW
Data were taken in the form of core pres-
_ Oi 04 sure differential. Data for the LP rig were
O3__ o:,,HGA
S
obtained by stagnating small core areas [0.64
_HGAS cm (0.25 inch) diameter] from the exhaust
side. The upstream stagnation pressure was
. SINGLE FLOW CHANNEL-- LP SIDE then measurable through the core matrix.
i_ m• • •
Static pressure data from the HP cold rig were

obtained in similar fashion from the HP inlet
"" "''''''''LP GAS core face.
"" ---CORE
Assuming that the core face stagnation
pressure is essentially unaffected by the in-
K, AXIAL STATION strumentation measurement technique, the
GB7.078-364 axial velocity through the core corresponds to
only about 0.25 mm (0.01-inch) H20 velocity
Figure 181. Measured Analytical Flow head (approximately one percent of the peak-
Distortion Model. to-peak distortion level).
130
Figures 182 and 183 describe the raw cold Step 2: Test Data Preparation
rig AP (local cold core pressure drop) data for
the HP and LP rigs. Note the excellent The Ap data shown was scaled slightly
agreement between the theoretical (calculated upward [by the ratio of P(c_,/_-o) 2] to match
from the core matrix Fanning friction factor the latest predicted engine hot face corrected
data) AP for uniform flow and the average flow and Mach number at the cruise condition.
measured Ap in both cases.
This step is required solely because engine
predicted stage points had changed slightly
kPa IIN H Ol between test and analysis stages.
1.0
RP COLO. 'RiG OATA CIRCUMFERENTIAL I ]
0.44 -
TEST..,0. U.,2-aO POSmO,._.
DEGREES Step 3: Analytical Input
I -2 OONNiET e-- 340 [ I
,.,,,,,. '0/ 6 = 0.35 LD/SEC c-- 310
I t o- 2,0 I / The scaled HP and LP pressure field data
1.7
"_" 0.42 - were input to the 3-D analysis as hot face
pressure boundary conditions. The analysis
THEORE L FOR allows a uniform pressure drop bias to pass the
0.40 . I _ _ _ FLOW AT TEST CONOITIONS I required flow rates. However, the variable
portion of the pressure profile is preserved.
The gross level of pressure drop is a result of
core matrix pressure drop while the variable
I lAW TE OATA (profile) portion is the result of flowpath
,,, !
[INJ 4 0 O 7 0 geometry and flow redistribution patterns.
I t f
CM I0 15 20
IL4OIUS Step 4: 3-Dimensional Analysis

6117-19711-_5
The analysis was run with the above pres-

Figure 182. Measured HP AP Data sure profiles as boundary conditions and the
From Regenerator Cold Rig. predicted engine flow, temperature and pres-
sure data as input. The analysis iteratively
arrives at a heat transfer solution and HP- and
kPa ItNH20i
LP-flow profiles, which satisfy the heat trans-
AO
LP COLD RIG OATA CIRCUMFERENTIAL i ] fer, energy balance, and pressure drop bound-
TEST 11-7 POSITION. OEGREES I i ary conditions.
s.,s.e_ i o-_ I [ I
1.0 V I CRUISE CONnITiONS I o- so _ I I
Step 5: Interpretation of Analysis
![....
.7
| THEORETICAL L_P FOR-
_ONIFOflM FLOW AT TEST
|
_
_
---]
_
-3
I
Figure 184 shows the analytical prediction
of capacity rate ratio (CRR) and LP and HP
mass velocities (F = pV), which result from the
imposed
tions.
(test) pressure fields and flow condi-
The mass velocities and CRR are
,, I ,,wTE,,- L L j circumferentially
station in Figure
averaged
184. Note
at each
that CRR
radial
is
IINI 4 5 O 7
I I defined here as the ratio of LP-capacity rate
CM 10 15 2O
divided by HP-capacity rate and is not neces-
RADIUS
Gll7-07_366
sarily equal to CMin/Cma x. Figures 185 and
186 summarize the analytically predicted
Figure 183. Measured LP AP Data From cruise discharge face gas isotherms and mass
Regenerator Cold Rig. flux contours.
131
RATIO
iFi = C
ko/s.m 2 [LiIM/S.FT2|
240 116 IT40 949

ANALYTICAL PiiEOiCTiON OF
290 140 1840 1004
1.2[-la00 I CRUSE COND T ON US NG "-1" r' I"
340 171 1890 1032
,,peExsun[a
OlSTOiiT_OR
LEVELS
! ..!. _ _-- -- _- 39O 199 1940 1080
AVG T3 = 1060C [IO4OF)
44O 227 1990 I068

L _ '_" I Ir_Cp p 490 254
5.0 - 1.0 _ _ _CAPACITY RATE RATIO • M0 202
--X,; i 500 310
T 5 = iO03C [200OF]
T2 = IOX 1218Fi
--_ _ ,,,,.f6 = ii¥ : _/AL LD/S-FT2
TEST DATA WI
ANALYSIS I AVG TT : 173C (343FI
%_. BERAGE Cllii • IJml
2..O L4 4 5 S _ _
PNI
t I J
CN 10 15 29
RADIUS
Sit74711..367
Figure 184. Capaeity Rate Ratio and

Integrated Mass Flux Versus Radial
Position.
Observe that even nominal distortion gen-

erates a significant (55.5K/inch) radial dis-
charge gas temperature gradient at the LP and
TEMPERATURES IN C
HP carryover point (about 170- and 350-de- 270"
GBT_7_3M
VIEWEO FROM COMBUSTOR ENO
grees angular position). The idealized ease (no
pressure distortion) yields essentially radial
isotherms (no radial gradient) except for slight Figure 185. Predicted AGTI01
asymmetry introduced by the erossarm seals. Regenerator Diseharge Gas
Isotherms for the Cruise Condition
The analysis predicts a HP-side tempera- Based on Cold Rig Pressure
ture effectiveness of 0.9662 with the imposed Distortion Patterns.
distortion levels. This compares with the ideal
effectiveness of 0.9779 in the absence of pres- engine operating line from idle to maximum
sure distortion. The predicted loss in effec- power. The penalty, in terms of fuel flow, was
tiveness due to distortion is, therefore, 0.0117 then correlated to engine output power as
at the cruise condition. No significant effect described in the following section.
on performance is anticipated from the slight
predicted increases in core pressure drop. Analytical Results
These small increases in pressure drop are
manifestations of residual flow kinetic energy Table 24 summarizes the regenerator anal-
which is already considered lost in the engine ysis results. Cases 1 through 7 and 14 are
performance predictions. hypothetical cases using prescribed distortion
profiles. Cases 9 through 13 and 15 through 18
Analysis of distortion effects was similarly are evaluations with measured distortion pro-
accomplished for operating points along the files based on LP and HP testing. Initialwork
132
hub to shroud. Note the difference in LP and
LBM/S.FT 2 = KO/S-M2
HP sensitivity.
0.5 2.4
MASS VELOCITY IN KG/S-M2 (LB/S-FT 2] 0.6 2.9
Early measured LP-cold flow data for the
0.7 3.4
0.8 3,9 low cruise condition was analyzed with a uni-
0.9 4.4 form (ideal) HP inlet flow. Case 8 results,
° 1.0 4,9
shown in Figure 187, indicates that the distor-
TEST OATA i
tion measured on the LP rig yields an effec-
tiveness degradation of (0.0043) equivalent to
a linear radial LP inlet distortion profile of
ANA_ 0.14 kPa (0.56-inch) H20.
Cases 9 through 18 summarize the analysis

for idle, cruise, and maximum power using
current 2500F engine cycle statepoints. Cases
,-.-1 . I +? --0"
13 and 14 can be used to evaluate the effects
of LP-eircumferential pressure distortion.
Case 14 is a case with a hypothetical circum-
ferential pressure distortion of 0.55 kPa (2.23
inches H20) (positive gradient in the direction
of rotation). This magnitude is two to ten
times greater in peak-to-peak magnitude than
the observed cold flow radial distortions, and
yet results in a negligible (A - 0.0001) deg-
radation of effectiveness. Other cases (HP
270° side and negative gradients), not included in
GeT-07e.369 VIEWEDFROM COMBUSTOR
END Table 24 show equal or lesser impact. Cir-
cumferential distortion of any reasonable
Figure 186. Predicted AGT101 magnitude is not considered to impact regen-
Regenerator Local Core Mass Velocities erator performance.
for the Cruise Condition Based on Cold
Rig Pressure Distortions. Cases 9, 12, and 16 are traditional 1-D
analyses and are provided for comparison with
(cases 1 through 8) was conducted using early Cases 10, 13, and 17, which are 3-D analyses.
engine cycle statepoint data corresponding to Cases 11, 15, and 18 are analytical evaluations
a low cruise 11.2 kw (15 hp) condition. Sub- using distortion profiles measured during LP
sequent work (cases 9 through 18) was per- and HP cold flow testing. Both the LP and HP
formed using current 1371C (2500F) engine inlet flowpaths are untreated (no baffles or
statepoints. flow diverters) and the magnitude of distortion
is corrected from cold test conditions to en-
Results of ease studies showed good agree- gine__ope_ating conditions by the ratio of
ment between the traditional 1-dimensional P(_Je/6)'.
(l-D) solution and the 3-dimensional (3-D)
finite element solution for undistorted cases. Cases 11, 15, and 18 results are utilized in
the construction of Figure 188, where effec-
Figure 187 presents results of cases 2 tiveness and fuel flow penalties are projected
through 8. These case studies, with hypotheti- as a function of engine horsepower. These
cal distortion were configured to evaluate and projections are determined through the use of
quantify the effect of radial inlet distortion engine performance and CFDC computer
profiles. The prescribed inlet distortion pro- models. Since the majority of fuel consumed
file for these cases is linear with radius from over the CFDC occurs in the 0 to 29.8 kw (0 to
133
Table 24. Regenerator Analysis Summary.
Effectiveness
ATIATM A x)
Engine LP Core Inlet HP Core Inlet
Case Power Distortion Distortion EHp ELp
1 Low Cruise 1-D Ideal 1-D Ideal 0.9823 --

2 Ideal (3-D) Ideal (3-D) 0.9837 0.9473
3 0.42 in-H20 Radial** Ideal 0.9810 0.9445
4 1.00 in H20 Radial** Ideal 0.9734 0.9359
5 1.95 in-H20 Radial** Ideal 0.9607 0.9212
6 Ideal 0.21 in-H20 Radial** 0.9814 0.9449
7 Ideal 0.976 in-H20 Radial* 0.9648
8 AP Map 8-18-80 Ideal 0.9794 0.9426
9 Idle 1-D Ideal 1-D Ideal 0.9849 --
10 Idle Ideal (3-D) Ideal (3-D) 0.9841 0.9632
11 Idle AP Map 8-18-80" AP Map 8-12-80" 0.9808 0.9603
12 Cruise 1-D Ideal 1-D Ideal 0.9750
13 Cruise Ideal (3D) Ideal (3-D) 0.9779 0.9430
14 Cruise 2.23 in-H20 Ideal (3-D) 0.9778 0.9429
Circumferential
15 Cruise AP Map 8-18-80" AP Map 8-12-80" 0.9662 0.9299
16 Max 1-D Ideal 1-D Ideal 0.9506
17 Max Ideal (3-D) Ideal (3-D) 0.9549 0.8996
18 Max AP Map 8-18-80" AP Map 8-12-80" 0.9421 0.8858
*AP maps scaled from cold flow rig for appropriate corrected flow
**Linear variation of (P SHROUD-PHUB) in inches H20.
ENGINE NET POWER, KW

CASE
Z C'SE" , F !
0 I0 20 30 40 50 60 70
I i I I I I ! [
- oN .CAS_
i__
24 i 5
I UNTREATEG FLOWPATliG
2.o / _.---- ---..... _
0.97 ; i \ - "_ • •
_ I [ i I _, I _TEO
I I ] ] t \ [ _ lip SlOE mEAL
P
2_
I I i i : Ire SJpE
mEAL , T",.C_E5 tJ.
[INHzOI 0 0.2 0.4 OJ 0.9 1,0 12 1.4 |J 1.9 2,0 T[ST DATA ANALYSIS
kPa
I
1.0 0.4
I
0 I t I o
0 IO 2O _ 4O 5O 8O 7O 60 on ioo
HYPOTHETICAL LINEAL RADIAL OISTORTION
ENGINE NET POWER. {HORSEPOWERI
OF CORE iNLET VELOCITY HEAD [PSHROUO -- PHURJ
G87.078371
GB7_78.370
Figure 187. Effect of Linear Radial Figure 188. Effect of Measured

Distortion on Regenerator HP HP and LP Distortion on Effectiveness
Effectiveness. and Fuel Flow.
134
40 hp) range, the expected fuel consumption
REGENERATORCORE
penalty for an engine without regenerator dis- ASSEMBLY
tortion treatment is predicted to be approxi-
DISCHARGEFLOW
mately 3.5 percent. (COLO GA
5.4.2.4 Regenerator Hot Rig

IHOT GASI F_
The regenerator hot rig objectives are

summarized below:
O The hot regenerator rig is designed to

measure regenerator seal leakage by the
use of an inert gas (helium) seeding/detec-
tion system _L_ _ HP INLET FLOW
ICOLD AIR
HP OISCHARGEFLOW _,_l-_.._,_
The regenerator hot rig will be utilized to iHOT AIRJ /
optimize flow distortion treatment geom- OPTIMIZED HP
DRAZED
etry a;,.m-m FLOW OIVERTER METALLIC
STRUCTURES
The matrix effectiveness will be deter-

Figure 189. Hot Regenerator Rig
mined at optimum core rotational speeds Schematic.
for idle,cruise, and maximum power simu-
lated conditions. Seal leakage will be
all seals not required for unit assembly. Braze
factored into the core performance
joints replace seal joints so that leakage from
the HP side of the rig to the LP can only occur
o Core drive torque will be determined under
the following conditions: across the static side of the regenerator seal,
flipper seal, and the dynamic face of the
regenerator seals.
- New seal breakaway torque
New seal break-in interval (ideal con- Hot Regenerator Test Results
ditionsfor approximately 10 hours)
The regenerator hot test rig was assembled
- Core drive torque during engine start and installed in the test facility. During rig
sequencing Builds 1 and 2, minor assembly and laboratory
problems precluded testing. Build 3 of the rig
- Steady-state torque at idle, cruise, was completed and testing initiated. With the
maximum power engine conditions. NGK core and first-generation Ford regenera-
tor seals, static leak checks showed excessive
O Verify matrix Ap/p (HP/LP) at idle,cruise, leakage (=25 percent of compressor inlet
maximum power simulated engine condi- flow). The unit was disassembled and a series
tions of vacuum leak tests were conducted. Several
areas were noted as contributors with a signi-
Measure cavity pressure at core ID under ficant leakage found in the NGK core. This
idle, cruise, maximum power simulated core exhibited excessive through wall porosity
engine conditions, and size-bleed orifice to resulting in approximately 40 percent of the
LP discharge to limit pressure in this area total cold leakage noted earlier. Subsequent
testing of the Corning thin wall cores exhib-
The regenerator hot rig design is shown in ited negligible leakage. Modifications were
Figure 189. The rig was designed to minimize incorporated to the test rig and the Corning
leakage paths (at seal locations)by eliminating core installed for Build 4.
135
Build 4 was completed and the unit instal- accurately to approximately 0.2 percent of the
led in the test cell. Initial operation of the rig HP inlet flow.
was conducted at ideal conditions [equivalent
to 870C (1600F) TIT] for approximately 4 Due to requirements of the 871C (1600F)
hours to break in (or glaze) the seals prior to engine test program, testing was limited to an
performance mapping. idle condition only. Additional testing will be
conducted following engine testing.
Further testing wasconducted at simulated
engine conditions. Instrumentation was lo- At points near idle conditions, leakage was
cated upstream and downstream of the HP and measured at 0.26 kg/min (0.576 lb/min),
LP core faces. Thermocouple rakes were approximately 5.2 percent of test compressor
located in equal area planes at five radial flow. Figure 191 shows the LP and HP core
locations, spaced approximately 35 degrees exhaust temperature profiles at the simulated
apart. idle condition. The tested temperature effec-
tiveness of the HP side of the regenerator was
Dynamic seal leakage was determined 95.9 percent at idle conditions (Table 25).
using a unique helium flow seeding technique
depicted in Figure 190. Helium is injected and Figure 191 shows that the radial tem-
perature gradients of the discharge gases at
the carryover points (170- and 350-degree
arcs) are minimal indicating insignificantflow
distortion effects.
NORMAL
ENGINE
Temperature effectiveness of the regener-
I TURBINE ENVELOPE
ator system is shown to be one percent lower

I than the goal. This deficiency may be solely
I
the result of the additional interpath leakage.
I COMOLJSTOR
I
Ford fabricated three sets of seals for test

INJECTION COMPRE$ 0 1
evaluation:
ii _
'lP
HELIUM MASS o Baseline Phase IV seals
REGENERATOR SPECTROMETER
GB7-078+373 Phase IV with increased inner peripheral

seal loading plus tapered thickness hot
crossarm shoe
Figure 190. Schematic of Helium
Seeding Technique for Measuring o Phase IV with increased seal loading plus
Regenerator Leakage. welded seal diaphragms
mixed at selected stations upstream of the HP Each of these seal sets were installed and
inlet. Flow passing through the HP side of the tested in the hot regenerator rig under nom-
core is then ducted overboard. Conditioned inal operating conditions (T3.1 =120Cor250F,
LP flow is then introduced to the slave burner T5.1 --650C or 1200F).
to maintain proper LP and HP flow match.
Fuel is introduced and combustion initiated. The third seal set with welded diaphragms
Regenerator seal leakage from the HP side exhibited such high seal loading resulting in
(helium seeded flow) to the LP side (unseeded) excessive drive torque and prohibited the mea-
can then be determined using a mass spec- surement of useful data. In fact, while
trometer to measure helium concentration attempting to test these seals, the rinK gears
levels at the LP exhaust. Leakage is measured were separated from two regenerator cores as
136
Table 25. Regenerator Rig
Test Results.
Metal
Engine
(871C - Regenerator
1600F) Rig Test
Predieted Test Results
Regenerator HP 42.3 37.4

Inlet Pressure, (6.14) (5.42)
kPa (psig)
HP Side Inlet Flow 4.57 5.03

kg/min (lb/min) (10.07 (11.08)
LP Side Inlet Flow 4.48 5.05

kg/min (lb/min) (9.88) (11.13)
HP Inlet Tempera- 7O 74
NOVEMBER20. 1981 ture C (F) (158) (165)

15:!6:31 HRS
LP Inlet Tempera- 788 746
NREG = 13 RPM
ture C (F) (1451) (1375)
O LP DISCHARGE TEMPERATURE X RP DISCHARGE TEMPERATURE

LP Exhaust Tem- 115 157
perature C (F) (239) (314)
C iF) C (F)
115 239 803 1279 HP Exhaust Tem- 766 718

117 243 894 1281
118 244 894 1282 perature C (F) (1411) (1325)
122 281 706 1302
133 272 707 1354
136 277 711 1311 HP Temperature 0.969 0.959
137 278 712 1314 Effectiveness
139 282 716 1320
152 _ 717 1322
.158 313 718 1324
lnterpath Leakage 0.196 0.261
158 316 721 13,29
171 339 721 1330 kg/min (lb/min) (0.431) (0.576)
178 349 723 1333
178 353 724 1336
179 354 726 1339
181 357 728 1340 seals are shown on Figures 192 and 193,
194 381 728 1342
195 383 729 1344 respectively.
197 387 730 1346
201 393 731 1348
291 394 732 1349 The baseline seal set exhibited lower drive
torque requirements and substantially higher
Gll7-078-374
leakage than the tapered seal set. The dis-

parity in seal performance is probably a result
Figure 191. Steady State Regenerator of the higher seal stiffness of the tapered seal
Diseharge Temperature Distribution. set.
a result of drive torques in the 41 to 62 kg-m Relative to the best of the earlier seal
(300 to 450 ft-lb) range. configurations neither of these seal sets exhib-
ited improvement.
Successful tests were completed on the
baseline and tapered seal sets. The drive Instrumentation was added to this rig to
torque and leakage characteristic for these measure the variations in seal clearance at the
137
PHASE IV -- BASELINE
m_ HX.-021
CX..021
HOT-OYNAMIC
407
(3OOJ
PHASE IV -- TAPERED
8 HX-023
CX-022
117-6l
HOT DYNAMIC
271
NREG ,,_ 12 RPM
'_" {2OOJ
T3.1 = 121C (250FI
6
T5.1 = 649C (1200F)
ll3.2)
_12 E
t IV -- BASELINE
CX-021
(111111
E IV -- TAPEREO
HX-023 CX-022
O COLD BUILD POCKET 9.88 CM {3.890 INCHESI

2
fO) I I 1
(4.41 138 276 414
IO} 12OJ 1401 160J
SEAL PRESSURE,kPa (PS]Ol
GB7_7_379A
o
(Ol
0 138 276 4]4 Figure 193. Regenerator Drive Torque.
{OI 1201 140} 160l
SEAL PRESSURE.kPa IPSIO) the inner peripheral regenerator pocket with

687_7_374A increased rig pressure.
Several tests were run and the cold build
Figure 192. Regenerator Seal Leakage.
seal clearance was varied to determine opti-
mum clearance with respect to seal leakage
regenerator seal inner periphery crossarm pos-
and drive torque. Figures 194 and 195 show
itions. This instrumentation consisted of a set
seal leakage and regenerator drive torque
of Bentley clearance probes mounted on a
plotted versus rig pressure for 1.35 mm (0.053
bracket anchored to the exhaust cover bolt
inch) seal clearance (optimum) and also for a
circle. The Bentley targeted a "finger" fixed
1.55 mm (0.061 inch) seal clearance. Leakage
to the regenerator shield. Another Bentley
at low pressures proved insensitive to seal
targeted the flat land of the exhaust cover.
clearance while, at high p._essures, sensitivity
By comparing these clearance measurements,
to leakage increased dramatically. Seal clear-
taking into account thermal growth of the
ances smaller than 1.35 mm (0.053 inch) were
regenerator shield, and the thermal effects on
tested, but excessive drive torques precluded
the Bentley calibrations the local variations in
gathering data at the higher pressures and the
regenerator pocket can be calculated.
data that was obtained showed no improve-
ment over the leakage with 1.35 mm (0.053
The data from this clearance probe instru-
inch) seal clearance.
mentation was inconclusive due to the lack of
information on the thermal growth of the Unlike earlier generation seals, the drive
regenerator shield. Qualitatively however, torque at the optimum seal clearance is al-
these data seemed to indicate an opening of most constant with respect to rig pressure.
138
4.5
I1O.OI
© 1.32 MM (0.052 IN) SEALCLEARANCE
0 1.32 MM 10.052INI SEALCLEARANCE A 1.55 MM (0.061 IN) SEALCLEARANCE

z_ 1.55 MM (0.061 IN) SEALCLEARANCE
407
HOTSEAL:NX-027
(3001 COLDSEAL:CS-024
CORNING CORE:S/N FM-44-006
SPEED:14 RPM
T3.1 -- 121C {250F)
TS.I -- 649C (1200F)
z 271 TESTDATE:3-16/3-17-84
i 0
I,OOl
136
_-- [I00) COLDSEAl.:CS-024 _ 2
PHASEV
CORNIN6CORE:S/N FM-44-006 ._ i4.4)
..=.
SPEEO:14 RPM
T3.1 -- 121C 1250F) -_
T5.1 -- 649C1120OF)
0 TESTOATE:3-18/3-17-84
IO) I ] I
138 276 414 1

iOl 120J 140] (60j (2.2j
0
GBT-O78-380A SEALPRESSURE,
kPa [PSIG)
0
Figure 194. Phase V Regenerator Seal fOl
0 138 276 4 4
Drive Torque Versus Seal Pressure.
fOl 120J 140) 160}
SEALPRESSURE.
kPa (PSIG)
Relative to the best of the earlier genera- GB7-O78-381A
tion seals, the Phase V seals represent an

improvement in leakage. A comparison is Figure 195. Phase V Regenerator
shown with Phase Illleakage curves in Figure Seals Clearance Test.
196. The seal leakage, measured at two values
for total seal clearance, is shown plotted engine operation at efficient peak cycle tem-
versus rig pressure. peratures for all power levels while operating
over a limited speed range of approximately
Phase V seal drive torque is also shown 50 percent speed at idle to 100-percent speed
compared with earlier seal (Figure 197) con- at fullpower.
figurations. Although improved over Phase IV,
these seals stillhave excessive drive torque. The transmission system selected for the
AGT101 consists of a high-speed variable
5.5 Gearbox/Transmission single-stage differential planetary gearbox in
conjunction with a modified Ford four-speed
Single-shaft gas turbine engine optimiza- automatic overdrive (AOD) transmission. This
tion for automotive applications requires the combination results in a nearly stepless, con-
use of a continuously variable speed trans- tinuously variable speed transmission required
mission, (CVT) with high ratio reduction. Be- for single-shaft gas turbine vehicle system
cause the AGT101 was designed with an auto- optimization.
motive application in mind, early phases of the
project included design of a continuously The planetary gearbox incorporates a dif-
variable transmission. An ideal CVT will allow ferential in which power is split into two
139
KS/Mill
PHASEIV
paths. Power is input at the planetary sun
n_o)'i / ROVEMRER
SS gear and extracted from the planetary carrier.
/ I0 i't CLEARANCE
= 1,3-2.0 mm IO.05Z-O.077tNI In this planetary system, all elements rotate
and the relative speeds are interdependent.
:--,] /-.,
-- CLEARANCE: 1,3 in \ One power path extends directly from the sun
gear to the carrier; the second power path
passes through the planetary ring gear to a
_ IS.el B variable stator torque converter (VSTC) that is
used as a speed changing device. The torque
converter variable output speed is connected
1 ./ SE
to the carrier through a clutch that is open
12,2
2
1 .//.I_RASE
_-
Ill
during vehicle idle operation and closed during
I1
/ 10.053 till normal driving modes. Speed ratio modulation
across the VSTC results in a variable output
O'l Ot , , , , , i
O 138 276 414 speed of the planetary carrier, which in turn is
IOI 1201 1401 1601 geared directly into the AOD transmission.
SEALPRESSUREOIFFEREllTIALILIPI. kPaiPSli Step ratio changes in the AOD are phased with
I I a off-setting speed ratio changes in the variable
0 200 400
speed gearbox to produce a continuously vari-
SEAL PRESSURE.KPa [PSIGI
R74711-3iN
able drive line system.
Figure 196. AGT101 Regenerator Seal Leakage. The AGT101 drive system schematic is
shown in Figure 198.
PHASEIV The AGTI01 transmission system cross

NOVEMBER63
CLEARANCE: 1,70-1.96 am section is shown in Figure 199. The gearbox
542
10.067-0.077INi
14OO1 geartrain has been emphasized by solid black
shading for clarification.
APRIL83
CLEARANCE
= 1,01-1.27 toni
{O.OilG-O050Illl A 23.3 cm (9.18-inch) diameter VSTC of
407 Ford design and manufacture was tested to
13001
generate a performance map covering the
PHASEIII __
extended speed ranges, ratios,and stator blade
L-- angles required for the AGT101 application.
w
The VSTC map is shown in Figure 200.

l__, 271
Efficiency islands as high as 95 percent were
i i_Oi identified and the gearbox differential ratios
__ i_ll4 10.053 INt were selected to operate the VSTC in the most
efficient areas over the CFDC.
CLEARANCE
= 1.35 mm
138
I1001 PHASEV In March 1981, when budget cuts occurred
MARCH84
CLEARANCE
= 1.27 mm and the new Administration dictated that the
iO.O_lllINI Government should sponsor long-range high-
technology programs that industry could not
0 afford, the transmission and vehicular tasks
IOI ' ' ' ' '
138 275 414 were deferred for industry sponsorship, with
IOI Illll 1401 1601 remaining funds Oriented at developing the
SEALPRESSURE
DIFFERENTIAL
ILIPI, kPa IPSll AGT101 high-risk technology elements.
S17_7_313A
The variable stator torque converter for

Figure 197. AGT101 Regenerator Core Torque. the AGTI01 concept had been fabricated,
140
drawings were completed for the Ford AOD 5.6 Controls and Accessories
transmission modifications, and drawings for
the differential gear and housings had been The engine electronic control unit (ECU)
made at the time of curtailment of the trans- and the fuel control unit evolved significantly
mission and vehicle phase of the AGTI01 pro- during the course of the program. Initial
gram. Consider that although the transmission design of these components was made assum-
arrangement illustrated in Figure 199 repre- ing an automotive installation. The ECU soft-
sents new technology, the components includ- ware and hardware provided for many vehicle
ing variable stator torque converter consist of powertrain functions such as transmission op-
proven technology. eration in addition to engine control, and the
KEY
T = GEAR TEETH (NUMBER)

Dp = DIAMETRAL PITCH
1"5 ACCESSORY
N = 2637 - DRIVE
5663
N = RPM RANGE 4T
AO0 = AUTOMATIC OVERDRIVE
JIIZe
X RATIO 12 Op
.2' 31T OVERRUNNINGCLUTCH
=" jg" N = 1716-3684
CARRIER
II
PLANETARY 74T 39T _ ' -- VSTC

N = 0 - 2652
i :GI"7 ;:;,,<, 83T I "9
N = 4069 - 8736
I _ r--I
ZTLJ 35T [____J J MAIN OIL
i
[]72T 0
Z RATIO alw
Dp = 12
75T
AOD
AOD INPUT SHAFT N = 0 - 2564
(SR = 0.85l
REARAXLE RATIO
/
-__ 8RAKE
- _ STARTER
PLANETARYRATIO = il.2727 (28 Dp]

X RATIO _ 1.9176 (12 Op)
Z RATIO = 1.04167 (12 DPl
G87-078-158
Figure 198. AGT101 Drive System.
141
ORIGINAL PAGE IS
011 POOR .QUALITX
41.9 CM
(16.5 IN] = i
I
ACCESSORY
ORIVE
"X" RATIO
\
\
10.80 IN
DIA
AO0
"Z" RATIO
ViEW ROTATED
90 OEGREES
LUBRICATION
PUMP
AOO
INPUT SHAFT
STARTER PAO
G87-078-356
Figure 199. AGT101 Transmissiofi System.
fuel control was designed as a compact unit the transmission as well as the engine. This
for vehicular installation that operated on a 12 unit became obsolete as control technology
volt dc electrical system. Though they rep- progressed during the course of the program
resented the state of 1980 technology, these and the demands of laboratory ceramic engine
components lacked the flexibility required for testing became more and more difficult to
a laboratory test program. As a result, both meet.
components were redesigned Io delete the
vehicular features and enhance the flexibility A new ECU was designed and fabricated
required for laboratory testing. which featured improved flexibility over the
original ECU. The new ECU, shown in Figure
5.6.1 Engine Electronic Control Unit 201, consisted of 3 standard off-the-shelf
boards, 1 iSBX board, 5 custom driver boards,
The initial ECU was based on an Intel 8086 a standard card cage and a RS232 adapter. By
microprocessor and incorporated 10 analog and utilizing the commercial boards the ECU will
2 digital control boards. This unit featured be easily supported or expanded in the future
both static (manual) and dynamic (automatic) to include new functions as they are needed.
control software and monitored all aspects of The commercial boards used in this new design
the vehicle powertrain operation controlling are:
142
1000.
700-
64o
62o
_o
75°
697-078-36 101198-1
Figure 201. Engine Electronic Control

!0 o Unit (ECU).
0o
port and modification of the ECU. First the

entire software filewas written in PLIM 86, a
high-level programming language. The ECU
G| 7_OTil,-35dt
control logic flow is clear and easy to under-
stand from an examination of the program
Figure 200. VSTC Test Results. listing. Second, all setpoint variables and
tables are defined and all RAM variables are
o Systek 8810C CMOS CPU STD board grouped according to their function in the
o Intel iSBX 328 analog output board same programming module. Third, the ECU
o MSI C412 CMOS Analog I/O board control logic is divided into different functions
o Prolog 76C14 CMOS Digital I/O board with each function programmed in a different
o Prolog BR16-T card cage with 7102 STD module consisting of various related sub-
Motherboard routines.
o Systek CA-101 RS232 serial adapter
To enhance the capability of the ECU, up
The software for the new ECU consists of to 24 different sets of setpoints and tables can
background and foreground modules. The be saved, each containing 340 bytes of data.
foreground modules contain the control logic. Different setpoint and table values can be
The background modules in the final product input to the ECU through an external keyboard
are used to monitor and change the various terminal, stored in internal memory and recal-
setpoints and tables. During the ECU develop- led as they are needed during the course of a
ment phase the background modules were also test series. This allows different control par-
used as tools for debugging and testing the ameters to be used from one run to another
foreground control logic. without physical modification to the ECU. A
control software self-test mode was imple-
Three important features were incorpor- mented to test the torque motor drivers and
ated in the "software to faciliate future sup- solenoid functions.
143
The ECU logic provides for full authority range of fuel flows from 0 to 20.4 kg/hr (0 to
control of the engine during testing. Engine 45 Ib/hr). The system was driven by a con-
fuel flow and variable inlet guide vanes(VIGV) stant speed 440 vac motor and mounted in its
are modulated according to operator changes own cart for easy access and transportation.
in engine speed setpoints and dynamometer
load conditions. The ECU will provide for 5.6.3 Variable Inlet Guide Vane Actuator
either manual or automatic operation of the
VIGV and start functions. Engine protection is The variable inletguide vane actuator used
provided for low oil pressure,high oil temper- during AGT101 engine testing was a torque
ature, overspeed, underspeed, and extreme motor actuated hydraulically driven unit. The
engine station temperatures (turbine inlet, actuator design was identical to that used in
turbine discharge, regenerator LP discharge). Garrett's GTCP331 series APUs and no devel-
The ECU logic diagram is shownin Figure 202. opment or modification was required.
5.6.2 Engine Fuel Control 5.6.4 Electrical Accessories
The original AGT101 engine fuel control Engine temperature measurement for con-
was a three piston'pump with a pulse width trol and monitoring are provided by standard
modulated (PWM) fuel metering unit. This Garrett laboratory thermocouples. Type K
system was compact, driven by a 0.09 kw (I/8 (Chromel/Alumel) thermocouples are used for
hp) 12 vdc motor and mounted on the engine all temperature locations except for turbine
test cart. Repeatability, hysteresis,linearity, inlet which used Platinum/Platinum-10 per-
and frequency response were allimproved over cent Rhodium thermocouples. These probes
existing proportional metering systems. Dif- are a shielded open bead design which opti-
ficulty arose when attempts were made to mizes thermocouple response with respect to
measure fuel flow with the PWM system. radiation error.
The fuel flows required for engine testing The hydraulic system low oil pressure
ranged from 0.36 to 20.4 kg/hr (0.8 to 45 switch incorporates a "shunt disc" actuating
Ib/hr). This exceeded the range of the turbine element. A resistive element temperature
flow meters available and necessitated the use sensor is specified to monitor high oil temper-
of a Coreiolis type mass flow measurement ature and ambient air temperature.
unit such as the one provided by Micromotion
R. The pulse frequency of the PWM system Engine speed sensing is accomplished using
with the piston pump, interfered with the mass an eddy current killed oscillator transducer
measurement of Micromotion system causing with a miniature 1.78 mm (0.070 inch) diam-
erroneous readings. Since fuel flow measure- eter probe tip supplied by Bently, Nevada
ment was required for laboratory evaluations, Corp. This probe is mounted in the engine
an equivalent fuel control was designed and inlet housing and targets the stepped surface
fabricated for use in the laboratory. of the compressor shaft OD.
The new fuel control was based on the Connectors selected for development use
design used in Garrett's GTCP36-100 APU. are MIL-C-5015 threaded connectors with 16
The schematic in Figure 203 shows the fuel gauge contacts. The ECU and engine harness
circuit in this design. This unit featured a had two connectors with 48 and 36 contacts
gear pump with an internal bypass circuit and for total of 84 terminations. The 48 and 36
a combination of 70 micron filterscreen and a contact connectors were common with the
cartridge filter. Fuel was metered with a boom cables and interface connectors in the
torque motor specially selected to allow a Garrett test facilities.
144
ORIG_qAL PAGE I8
_OF POOR QUALIT_
I
I
cp
Figure 202. ECU Logic Diagram.
145
HIGH-PRESSUREPUMP
PUMP OISCRARGEPORT
METERING VALVE H
70-MICRON SCREEN MOTOR
_ PLUMBING TO BE
ItM
TESTING
il PROVIOEDFOR
METERED OUTLET PORT II
METEREO
FUEL SUPPLY, E PRESSURE,PNi

INLET
,-"L% HI "_/tO SIMULATOR

PUMP OISCHARGE
GB7-078.37 PRESSUREIeoJ
Figure 203. Fuel Control Test Setup (P/N 3601500).
O_:LIGINAL PAGE IS
OE POOR QUALITY
146
!
6.0 CERAMIC MATERIALS AND COMPONENT DEVELOPMENT
6.1 Ceramic Materials Testing Many silicon nitride materials were tested
to determine suitablity for AGT rotor use.
Many materials such as ASEA, Ford RM-20,
6.1.1 Fast Fracture and Stress Rupture
Kyocera SN 220M, NGK SN-73, and SN-81
Testing
were eliminated for rotor use based on the
inadequate elevated temperature properties.
The fast fracture, four-point flexure Four more advanced materials, NGK SN-82,
strength of ceramic materials used for com- NGK SN-84, Kyocera SN 250M and Kyocera
ponents in the AGT engine have been contin- SN 270M did have adequate properties for
uously monitored over the AGT program. Both rotors based on the testing of fabricated test
fabricated test specimens and specimens ma- bars. Unfortunately, two of the four mater-
chined from components have been tested. ials, NGK SN-82 and Kyocera SN 270M could
not be successfully formed into the complex
rotor configuration. The remaining two ma-
All four-point flexure testing was con-
terials Kyocera SN 250M and NGK SN-84 were
ducted using self-aligning metal (for room
successfully fabricated into the AGT101 rotor
temperature) and hot pressed SiC (for elevated
shape. Rotors of both materials were cut into
temperature) fixtures. The outer span dimen-
test specimens. The specimens were tested
sion is 38.1ram (1.5 inches) and the inner span
for both fast fracture strength and stress
dimenson is 19.1ram (0.75 inch). The load is
rupture. The material cut from the Kyocera
applied using a crosshead speed of 0.5ram/rain
SN 250M rotors has considerably lower
(0.02 inch/minute). Specimen cross sections
strength at elevated temperatures (Figure 204,
were typically 3.2 x 6.4ram (0.125 x 0.250
View G) and two- to three-orders of magni-
inch) except where noted.
tude lower stress rupture life than specimens
fabricated as test specimens (Figure 204, View
Test specimens were machined with 320 H). The poor elevated temperature material
grit diamond wheels longitudinal to the speci- properties of the Kyocera SN 250M rotors
men length unless noted otherwise. The speci- makes them unsuitable for long term, 1371C
men corners were chamfered 0.127-0.254mm
(2500F) TIT engine testing.
(0.005 - 0.010 inch) at a 45 degree angle. Test
data generated during the first five years of Two NGK SN-84 sintered silicon nitride
the AGT program is summarized in Table 26. rotors were cut into test specimens to com-
In general, the strength of fabricated test pare the fast fracture strength and stress
specimens has been very close to the strength rupture life to that of fabricated test speci-
of specimens cut from components. mens. The fast fracture strength of rotor S/N
135 material (Figure 204, View Q) is slightly
During the last two years of the program more lower than the fabricated specimens (Figure
extensive material property testing was initi- 204, View P), at elevated temperature but is
ated to characterize the advanced silicon ni- still>689.5 MPa (i00 ksi) at temperatures
tride and silicon carbide materials. Typically, <1260C (2300F). The fast fracture strength
fast fracture testing was conducted at up to of rotor S/N 136 is lower than measured on
ten temperatures with between five and thirty rotor S/N 135 but the elevated temperature
specimens per temperature. In addition flex- fast fracture strength at temperature
ural stress rupture testing was conducted at <1260C (2300F) is greater than 586MPa (85
typically three temperatures and at four k-si). The flexural stress rupture life of the
stresses chosen to result in stress rupture Hves rotor specimens are approximately one order
of <150 hours. The materials, test and of magnitude less than the stress rupture life
specimen sources are listed in Table 27. of the fabricated test specimens.
147
Table 26. AGT101 Component And Material Summary.
Qualification Bar Ceramic Component
Room Temperature Elevated Temperature Room Temperature Elevated Temperature
Supplier Material Process Condition I1

O'C2 M3 lion O'C2 II M3 F -II
lion O'C2 M3 lion O'C2 II M3 F 1 tion
ACC
Inner DIffuser RBSN Slip As-Fired 49.7 4.5 30 52.1 0.t 2200 10 44.34 8.0 21 ID
Outer Diffuser (RBNI04) Cast I,ongitudineUy 53.2 6.5 10 46.3 5.3 27 OD
Turbine Shroud Ground 50.0 15.4 17 1"5
Oaffle Ileat Treated 54.8 8.5 29 31.1 5.7 5 O
Stator RBSN Injection As-Fired I 40.1 4.8 19 44.5 8.4 2200 23

(RBN124) Molded
Rotor SIntered Slip Longitudinally 123.9 16.4 10 98.6 5.7 1800

Si3N 4 7 Cast Ground
Code 1 71.0 36.8 2200
Sin tered Slip Longitudinally 89.8 7.2 9 50.3 13.0 2200 10 119.0 21.4 10 99.0 10.3 [800
Si3N 4 7 Cast Ground
Code 2 63.7 9.3 2200
Standard Oil
Turbine Shroud, SASC Injection As-Fired 48.6 9.5 30 45.0 5.0 2500 10 55.45 7.1 9
ate for Molded
Combustor SASC SUp LongitudinalP/ 49.4 5.8 30 41.4 6.7 2500 10 53.0 7.3 9
oa rrle Cast Ground 51.6 8.9 4
Transition SASC Isopre_ed I,ongitudinally 57.79 7.7 30 56.2 11.9 2500 10 54.4 ! 6.2 5
Due4 Regen- Ground 34.3 ; 9.1 23 36.9 6.8
erator Shield, 41.8 14.2 12
lluckshroud
Ford
I
Rotor SRBSH Sllp I,Onl_itudinally 109.3 19.B 6 73.1 !16.4 2200 6
(RM-2) Cast Ground
Rotor SR.RSN Slip Longitudinally 98.66 19.0 10 77.66 10.9 1832 5

(RM-3) 8 Cast (;round TI.66 12.9 2192 5
70.86 11.3 2552 5
Rotor SRBSN SKip Longitudinally 139.4 13.3 I0

(RM-20) 8 Cast Ground
Stator RBSN Injection As-Fired 43.1 9.2 39 45.8 7.7 2200 10

Molded
NGK
Baekshroud, SIntered Isopr espied Longitudinally 87.6 10.5 10 47.1 13.6 2000 7
Trmlsl lion Si 3 N4 (;round
Duet, (aN-50)
Diffusers
Turbine Sintered Isopre_ed Longitudinally 109.5 10.3 5 67.9 3.6 1800 2

Shroud Si3N 4 Ground 53.0 6.1 2000 5
(SN-54) 40.7 25.0 2200 5
23.2 7.0 2500 2
Kyocera
Turbine Shroud Sintered lsopre_ed Longitudinally 65.4 5.2 5

SiC Ground
SC201
Ground and 78.5 10.4 5 66.5 19.7 5
Ilent Treated 70.5 12.0 5
Coming
Flow Separatoe LAS Slip longitudinally 14.0 19.8 II 13.5 9.7 2000 12 12.9 5.3 10 13.4 3.4 1800 5
IIousing l5.4 8. l 2000 10
16.0 7.7 2050 9
All test bars 0.250 x 0.125 inch cross section unle.,B noted. Bars tested in 4-point Rexure, 1.50 inch outer span end 0.75 inch inner _pan. Cross head speed, 0.02
inch/minute
ITcst bar cross seetlon 0.31 x 0.15 inch 4"rest bar cross section 0.236 x 0.1 inch 7Code I = 8%-Y203, 4%-AI203
2Characteristic strength, kst GAs machined, longitudinally ground Code 2 = 6%-Y203, 2%-AI203
3M = Weibull Modulus 6Average flexure strength, ksi at)ate supplied by vendor
9Not heat treated
OF POOR QUALITY
148
Table 27. Fast Fracture and Stress Rupture Data was Measured for
Many Advaneed Si3N 4 and SiC Materials.
(Figure 204)
Fast Fracture Stress Rupture
Test Bar Component Test Bar Component

(View Ref) (View Ref) (View Ref) (View Ref)
Silicon Nitride
ASEA A m
Ford RM-20 B m m
Kyocera SN 220M C D
Kyocera SN 250M E G F H
Kyocera SN 270M I J
NGK SN-73 K
NGK SN-81 L M
NGK SN-82 N O
NGK SN-84 P Q R R
Silicon Carbide
Kyocera SC201 S
Standard Oil SASC T mD
Based on this data SN-84 appears to be a contact area. The specimens were heated to
material that can be used for limited 1371C the appropriate temperature while dead
(2500F) TIT engine testing. weight loaded at a minimum simulated engine
assembly load. Once at temperature, addi-
Over the seven years of the AGT101 pro- tional dead weight loads were applied, as
gram large improvements have been made in required, to simulate engine aerodynamic or
high performance ceramic materials. Addi- mechanical loads and held for 20 hours. Un-
tional work must be done to increase the loading and cooldown followed. The specimens
elevated temperature material properties, then were inspected for sticking phenomena.
fracture toughness and shape capability. Three classifications were assigned; no stick
or reaction (NSR); light sticking (LS) upon
6.1.2 Ceramic Compatibility Testing cooldown, ie, came apart during handling; and
hard sticking (HS), not separable by hand.
In conjunction with ceramic component Concern is raised for LS and HS conditions due
tests, an interface compatibility study was to the following:
conducted. All interface conditions for the
AGT101 have been identified, as shown in
o L8 - longer exposure may result in hard
Figure 205, along with respective stress and
sticking
temperature environments for the three
engine operating conditions [871, 1149, and
1371C (1600, 2100, and 2500F) TIT]. Test bars O HS - could result in component fracture if
of as-machined or as-processed material were the parts thermally expand or cool down at
stacked with a 6.4 x 6.4ram (0.25 x 0.25-inch) differing rates
149
MPz IKStl
150-
1000"
14D-
130-
120-
800'
liD-
100-
90-
600"
80-
IlOl
70-
60-
4O0"
50-
1101
40- MATERIAL -- ASEA SI3N4 2*/2% Y203
DATE RECEIVED-- 1966 IlOI _'t
DATE TESTED -- 1986

30- 1101
200" SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM lye X % INCH]
CROSS HEAD SPEED -- 0.5 MM/MINUTE [0.02 INCH/MINUTE)

20- OUTER SPAN -- 38.1 MM (I.50 iNCH)
INNER SPAN -- 19.1 MM )0.75 INCH]
BARS INOICATE OATA RANGE

10-
( J NUMBER OF SPECIMENS TESTED
O" 0 | l I ! I !
0 500 1000 1500 2000 2500 3000
TEMPERATUREIF}
,I
0 ,
200 v
400 i
600 w
800 '
1000 I 2'0
0 I 4'0
0
TEMPERATURE. C
GII7-OTB-BO A
Figure 204. View A. Four-Point Flexure Strength of ASEA Si3N4-2½ Pereent Y203 .
150
MPa (KSII
150-
1000'
140-
130-
120-
800"
IlO.
100-
151
90-
600"
80- 151
16D\
\
70-
\_,,\
60-
400" 151 _._
151
50=
MATERIAL -- FORD RM20 SRBSN
CUT FROM ASEA HIPped ROTORS/N 96

40-
DATE RECEIVED -- 1986

DATE TESTEO -- 1986
30-
SPECIMEN CROSS SECTION -- 3.2 X 8.4 MM I% X % INCH)
200-
CROSS HEAD SPEED -- 0.5 MM/MINUTE (0.02 INCH/MINUTE)
20- OUTER SPAN -- 38.1 MM [1.50 INCH}
INNER SPAN -- 19.1 MM [0.7§ INCH)

BARS INOICATE DATA RANGE
I0-
( ) NUMBER OF SPECIMENS TESTEO
0
I I ! I l I
500 1000 1500 2000 2500 3000
TEMPERATURE (F)
o 20o 400 500 ,o'oo , oo 1;oo

I I I I
TEMPERATURE. C
GB7Q78.61A
Figure 204 (Contd). View B. Four-Point Flexure Strength of Ford RM-20, HIPped by ASEA,
Cut From Rotor S/N 96.
151
MPz tKsq
150-
1000'
140-
130-
120-
800'
IlO-
100-
I101
90-
•- 600' %%
%
8o- {101
70- ItOi 't

fiO-
I10i
MATERIAL -- KYOCERA SN220M SSN

50-
DATE RECEIVED-- APRIL 1985
DATE TESTEO -- APRIL 1986

40- I101 b
SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM 1% X % INCH1 i _
CROSS HEAD SPEEO -- 0.5 MM/MINUTE [0.02 INCH/MINUTE) tlOl\
I
30- OUTER SPAN -- 36,1 MM {1.50 iNCH]
INNER SPAN -- 19.1 MM {0.75 INCH)

DARS INDICATE DATA RANGE
ZO- (1o1
{ ) HUMBER OF SPECIMENS TESTED
I0-
O,
U" ! I ! | ! I
500 1000 1,500 2000 2500 3000
TEMPERATURE {FJ
! ! ! ! I I ! !
O 200 400 600 800 1000 1200 1400
TEMPERATURE, C
GD7,.076-$2 A
Figure 204 (Contd). View C. Four-Point Flexure Strength of Kyoeera SN 220M Sintered
SUieon Nitride.
152
MPa (KSI)
80-
5OO MATERIAL -- SN 220M SSN

70_
OATE RECEIVEO-- APRIL 1986
TEST TEMPERATURE --
e- 1093C (200OF)
•- 1204F (2200FI
60-
400, SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM (% X ',_ INCHI
OUTER SPAN -- 38.1 MM (1.50 INCH)
INNER SPAN -- 19,1 MM [0.75 INCH)
50- INOICATES SPECIMEN DIO NOT FAIL
m= 300"
40-
30-
200 -
20-
100"
I0-
0 I ] I I I
O01 0.1 1.0 10 I00 1000
TIME TO FAILURE. HOURS
GO7-O78-83A
Figure 204 (Contd). View D. Kyocera SN 220M Flexural Stress Rupture.
153
MPa IKSli
150-
1000"
140-
130-
120-
800 • iSi
110-
100-
_m m _ _l_ _ _ m _ _ m _ m ......
_ f301 151 T
80- \ .... 15
f51
I_OI
70-
151 L'lO)
400-
60- IIOI
!
50-
MATERIAL -- KYOCERA SN 250M SSN k
40- OATE RECEIVED -- AUGUST 1985, 1968
DATE TESTED -- AUGUST 1985. 1986

• 1986 VINTAGE
SPECIMEN CROSS SECTION -- 3.2 X 8.4 MM [Vi X % INCH(
30- • 1985 VINTAGE 15i
200- CROSS HEAO SPEEO -- 0.5 MM/MINUTE (0.02 INCH/MINUTE)
OUTER SPAN -- 38.1 MM (I.50 INCH(

20- INNER SPAN -- 19.1 MM (0.75 INCH(
BARS INDICATE DATA RANGE
I J NUMBER OF SPECIMENS TESTED

10-
O
O" J i , i l !
500 I000 1500 2000 2500 3000
TEMPERATURE IFI
I I I i
o 200 4o0 800 ,&o ,4'oo

G87_7_84A
TEMPERATURE. C
Figure 204 (Contd). View E. Four-Point Flexure Strength of Kyocera SN 250M Sintered
Silicon Nitride.
154
MPa IKSII
80-
500 %1 • • _A
70-
"_%% %%
• %.. | %%% %%%

6_
400.
'_%%% %%% %%,%
%%_ • _ • %
50-
300-
40-
MATERIAL -- KYOCERA SN 250M _%
30-
OATE
RECE,"EO-
lg_ "_"
200-
TEST TEMPERATURE %%
.-12O,C)Z200F) %% i_cFi
nn- 1315C)240OF)
2'0- • -- 1370C 12500F)
SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM (% X % INCHI
OUTER SPAN -- 38.1 MM [1.50 INCHI

100-
INNER SPAN -- 19.1 MM )0.75 INCH)
I0- INOICATES SPECIMEN 010 NOT FAIL
0 "
0, I 1 I I I
0.01 0.1 1.0 10 100 1000
GB7-O78-BSA
Figure 204 (Contd). Yiew F. Kyocera SN 250M, Flexural Stress Rupture.
155
MPa IKSll
150-
1000"
140-
130-
12D-
I10,
Z_5I
100-
__ _'_" _ _ SN 250M
k
90-
80-
__,_ _l A
70-
00°
400-
% 15i ROTORS/N 253

MATERIAL -- KYOCERASN 250M SSN %
40.
DATE RECEIVED-- ROTORS 1986 _151 ROTOR S/N 251
OATE TESTEO -- ROTORS i986
SPECIMEN CRDSS SECTION -- 3,2 X 6.4 MM ('/i X % INCH)
30,
CROSS HEAD SPEEO -- 0.5 MM/MINUTE (0.02 INCH/MINUTE)
OUTER SPAN -- 3_i MM (!.50 ]NCHI

20-
INNER SPAN -- 19.1 MM (0.75 INCH]
I0- l] NUMRER OF SPECIMENS TESTED
0" ! I ! ! ! J
0 500 1000 1500 2800 2500 3000
TEMPERATUREIFI
I I I I I ! I I
0 200 400 600 800 1000 1200 1400
6117
o7sM TEMPERATURE C
Figure 204 (Contd). View G. Four-Point Flexure Strength Comparison of Kyoeera SN 250M
Cut From Rotors and Fabricated as Test Bar Specimens.
156
MPa IKSI)
80-
5O0
70- MATERIAL -- KYOCERASN 250M SSN
9ATE RECEIVED-- 1986
TEST TEMPERATURE-- 1370C (2500F)
qb_ • SPECIMEN SOURCE

60- "%_ • FABRICATEDAS TEST BARS
400,
'%% • ROTOR S/N 251
qb'q_ • ROTORS/N 253
_qb SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM I'/e X _/4 INCH)

50-
_qb_- _ OUTER SPAN -- 38.1 MM (1.50 INCH)
"%4b,_ INNER SPAN -- 19.1 MM 10.75 INCH}
300- %
40-
%%%%_ee
30- • • • • %_
200-
20-
100-
10-
0 ! I I I m
0.01 0.1 1.0 I0 lOG 1000
GII7..O78-87A
Figure 204 (Contd). View H. Stress Rupture Lives of Kyocera SN 250M From Rotors is
Considerably Less Than That Fabricated as Test Bars.
157
MPa (KStJ
150-
1000-
140-
130-
120 -
800'
110.
I00._
go.4 tsJ
_=
f5i 15i
804
70.
50.
400-
50-
MATERIAL -- KYOCERASN 270 SSN

40.
OATE RECEIVEO-- MARCH 1986
OATE TESTED -- APRIL 1986
30- SPECIMEN CROSS SECTI0N -- 3.2 X 6.4 MM [¼ X ¼ INCH)
200"
CROSS HEAO SPEEO -- 0.5 MM/MINUTE (0.02 INCH/MINUTE)
20- OUTER SPAN -- 38.1 MM (I.50 INCHI
INNER SPAN -- 19.1 MM 10.75 INClll

OARS INDICATE DATA RANGE
10-
I ] NUMBER OF SPECIMENS TESTED
0 =
0
I I I / I I
500 1000 1500 2000 2500 3000

TEMPERATURE IF)
I I I I I I
/ /
9 zoo 400 ,ooo ,4oo
6117_711-UA TEMPERATURE. C
Figure 204 (Contd). View I. Four Point Flexure Strength of Kyoeera SN 270M
Sintered Silicon Nitride.
158
MPz (KSII
80-
5OO 70-
O0 •
400,
• "... ....,_,.... -.o,, •

50-
300"
40-
"*_L,_,,,_, _,,_ o--

_ 1370C
__ (2500FI
30 I MATERIAL -- KYOCERA SN 270M SSN _ 1427C
DATE RECEIVED-- APRIL 1986 (2800F)

TEST TEMPERATURE
•- 1370C (2500F)
20a
•_ 1427C (26QOFI
SPECIMEN CROSS SECTION -- 3.2 X 5.4 MM I_/I X % INCHI
100- OUTER SPAN -- 38.1 MM 11.50 INCH)
INNER SPAN -- 19.1 MM 10.75 INCH)

10_
INOiCATES SPECIMEN 010 NOT FAIL
O
I I I I t
0.01 0.1 1.0 IO IO0 tO00
TIME TO FAILURE, HOURS
G87_tO-89
Figure 204 (Contd). View J. Kyoeera SN-270M, Flexural Stress Rupture.
159
MPo (KSll
150-
I000'_
140-
130-,
120-
800'
110-
I00" 151
90-
0= 600"
80-
\
70-
60.
41111"
50-
151 \
\
40. MATERIAL-- NGKSN-73SSN
OATERECEIVED
-- 1985 't
gATE TESTED-- 1985 151
30.
200" SPECIMENCROSSSECTION-- 3.2 X 6.4 MM [i/, X _/,,INCHI
CROSSHEADSPEEO-- 0.5 MM/MINUTE(0.02 INCH/MINUTEI
20- OUTERSPAN-- 38.1 MM 11.50INCH)
INNERSPAN-- 19.1 MM 10.75INCHI
10. BARSINDICATEDATARANGE
I J NUMBEROF SPECIMENSTESTED
O" 0 I I I I I '
500 1000 1500 2000 2500 3000

TEMPERATURE
(F|
I ! I ! I I I |
O 200 400 600 800 1000 !200 1400
TEMPERATURE.
C
GBTQ70-90A
Figure 204 (Contd). View K. Four-Point Flexure Strength of NGK SN-73 Sintered
Silicon Nitride.
160
MPa (KSI)
150-
1000"
140-
130-
120-
800"
I10-
I00-
f
90-
600"
80-
I101 i101" f/ " _ I --

70.
60,
400 -
IlOI
50- rio) \
40. MATERIAL -- HGK SN-81

OATE RECEIVED -- MARCH 1986
DATE TESTED -- APRIL 1986

30.
200" SPECIMEN CROSS SECTION -- 3.2 X 8.4 MM {Yz X ¼ INCHI
CROSS HEAO SPEED -- 0.5 MM/MINUTE ]0.02 INCH/MINUTE)
20- OUTER SPAN -- 38.1 MM (I.50 INCH)
INNER SPAN -- 19.1 MM [0.75 INCH) 15i

10- BARS INOICATE OATA RANGE
I ] NUMBER OF SPECIMENS TESTED
O" D,
I I I I I !
0 500 1000 1500 2000 2500 3000
TEMPERATURE IF)
!
0 200
I 400
I 600
I 800
I 1000
I 1200
I I 410O
TEMPERATURE. C
GB7-O76-91A
Figure 204 (Contd). View L. Four-Point Flexure Strength of NGK SN-81 Sintered
Silicon Nitride.
161
MPa IKSII
80-
IP
500 70- 'k

%
%
% io93c
60 m % [2000FI
%
4OO
MATERIAL -- NGK SN 81 SSN %% • • •
DATE RECEIVED -- MARCH 1986 '_b,
TEST TEMPERATURE "_'_ _
50-
0-- 1093C 1200OFI %%_
i-- I_C 12_FI %_I
30O &-- 1315C {240OF) _',_
40- SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM I% X % INCH) %%
OUTER SPAN -- 38.1 MM {1.50 tNCHI %_ -- • • " It=:

INNER SPAN -- 19.1 MM 10.75 INCHI "_'%_
--z,.. INOICATES SPECIMEN 010 NOT FAIL 'rob%

30-
200-
%%%%_ • • • %%% ,_c,

tLL_rl
20-_
_%%% •• •
1
100- !
%%_ • •
%%_ 1315C • •
(2400F)
ol ! I I I I
001 0.1 1.0 I0 I00 1000
087-078-92
A TIME TO FAILURE. HOURS
Figure 204 (Contd). View M. NGK SN-81, Flexured Stress Rupture.
162
MPa IKSII
150 T
ksi " .,_
IOOO-
140
130
120
.I--'"" L'"'"'I
800"
I10
lOO-
11Ol
" [lOl ] t
[101
I10l
'%%.
,,,
"
I
i
904
600-
_z
x
704 I101
.d
llOll
I
I
I
MATERIAL -- NGK SN-82 SSN
4O0" OATE RECEIVED -- JANUARY 1986
5O OATE TESTED -- FEBRUARY 1986
SPECIMEN CROSS SECTION -- 3,2 X 6.4 MM (1/z X V,_INCH)
CROSS HEAD SPEED -- 0.5 MM/MINUTE )0.02 INCH/MINUTE) 1101

40
OUTER SPAN -- 38.1 MM 11.50 INCH)
I
INNER SPAN -- 19.1 MM (0.75 INCHI
3_ BARS INDICATE DATA RANGE
2OO
( I NUMBER OF SPECIMENS TESTED
20-
10-
I I I I I I
500 1000 1500 2000 2500 3000

TEMPERATURE IF)
I | I II I / i
O 200 400 600 800 IDO0 1200 1400
TEMPERATURE. C
G97_7893A
Figure 204 (Contd). View N. Four-Point Flexure Strength of NGK SN-82 Sintered
Silicon Nitride.
163
MPz (KSll
90-
80- • • • • •
%%%%
5oo zo %%
%%
%%
400 %%%% • ,_
%%%%
MATERIAL -- NGK SN 82 SSN _L '_b% 1277C
2OO 30 GATE RECEIVED-- JANUARY 1986 'lrzP_t _ • (2300FJ

TEST TEMPERATURE
• - ,003c
t_FI "....
m- _c 1_00F1 ",,_ _ fl=::
20 .-_.c 12300FI ,3_c
.-13_8c(:,,.00F] (_400F]
lOG SI'EC,MEN
CROSS
SECT,ON
-- -".2
X8.4M,,'_,X'_.,NC,,
OUTER SPAN
- _.1MM,.._o,NCHj
10 tNNER SPAN -- 19.1 MM (0.75 INCH]
o
-,,,. INDICATES SPECIMEN 010 NOT FAIL
0 I I I I I
O.Ol 0.1 1.0 IO 100 lOG0
GII7_I78-G4A
Figure 204 (Contd). View O. NGK SN-82, Flexural Stress Rupture.
164
MPa [KSI]
150.
I000-
140.
800"
130-
120-
110.
!.... I.......
100-
90-
600-
80-
151
704
eo_
400"
50,_ MATERIAL -- NGK SN-84
{
DATE RECEIVED-- AUGUST 1986
40. DATE TESTED -- SEPTEMBER 1986
SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM (% X ¼ INCH)
CROSS HEAD SPEED -- 0.5 MM/MINUTE (0.02 INCH/MINUTEI

30- (101 i
200" OUTER SPAN -- 38.1 MM 11.50 INCHI k
INNER SPAN -- 19.1 MM (0.75 INCH)

20- BARS INDICATE OATA RANGE
J } NUMBER OF SPECIMENS TESTEO

(51
10-
O" 0
I T T 1 "1" 1
500 1000 1500 2000 2500 3000
TEMPERATURE IFI
i I I
o 2oo _oo _ 8_o lO_0 doo 14'oo
G67_79-gSA TEMPERATURE. C
Figure 204 (Contd). View P. Four-Point Flexure Strength of NGK SN-84 HiPped
Silicon Nitride.
165
MPa IKSII
150-
1000"
140-
130-
120-
800"
I10-
100-
(4)
]-
(101
_0-
80- (4i
151
600- 70-
9C 14)
400"
50- 14l
MATERIAL -- NGK SN-84 _ ROT0R S/N 135 • ROTOR S/N 136

40-
OATE RECEIVEO-- FEBRUARY 1987 JUNE 1987
OATE TESTED -- MARCH 1987 JULY 1987

30- SPECIMEN CROSS SECTION -- 32 X 6.4 MM (I/, X % INCHI
200"
CROSS HEAD SPEED -- 0.5 MM/MINUTE (0.02 INCH/MINUTE)
20- OUTER SPAN -- 38.1 MM 11.50 INCXI
INNER SPAN -- 19.1 MM [0.75 INCHI

BARS INOICATE DATA RANGE
10-
I ] NUMBER OF SPECIMENS TESTEO
0 = 0 !
I I ! ! I
0 500 1000 1500 2000 2500 3000
TEMPERATURE IFI
I
0 '
200 '
400 '
600 '
800 '
1000 12'00 I 4'00
TEMPERATURE, C
H7-O78-9_A
Figure 204 (Contd). View Q. Four-Point Flexure Strength of NGK SN-84 HIPped
Silicon Nitride Cut From Rotors S/N 135 and S/N 136.
166
MPa (KSI)
Ai•i 4D
\ \ \ \\
AA • • • _0 •
500
7
\ '_ • • II| •
1277C 1277C 1204C 1204C 1204C

6 (2300FI (2300Fi [2290F) i2200Fi (2200FI
400 ROTOR TEST SPECIMENS ROTOR ROTOR TEST SPECIMENS
S/N 135 S/N 135 S/N 136
51
MATERIAL -- NGK SN 84 SSN
300 TEST 8ARS ROTORSN 135 ROTOR S/N 138

DATE RECEIVED-- AUGUST 1996 FEORUARY1987 JUNE 1987
TEST TEMPERATURE
-- 1204C (2200F) • •
-- 1277C (2300FI • •
SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM I'/, X % INCH)

200, ]1
OUTER SPAN -- 38.1 MM (1.50 INCH)
INNER SPAN -- ]9.1 MM (0.75 INCH)

INDICATES SPECIMEN OlO NOT FAIL
?C
ZO0-
O 0
t I I 1 i
OOI 0.1 1.0 I0 t O0 lO00
G87_7H7A
Figure 204 (Contd). View R. NGK SN-84, Flexural Stress Rupture of Test
Specimens and Rotors.
167
MPa IKSII
150-
1000-
140-
130-
120-
800"
I10-
100-
600-
o!
,mC
400" 1101 [101_OI
I101 (lOll [1
1101
MATERIAL-- KYOCERA
SC-201
(lOl
OATERECEIVED
-- 1985
30- DATETESTEO-- 1965
200'
SPECIMENCROSSSECTION-- 3.2 X 6.4 MM I% X '/, INCH[
20- CROSSHEAOSPEED-- 0.5 MM/MINUTE[0.02 INCH/MINUTEI
OUTERSPAN-- 36.1 MM [1.50 INCH}
INNERSPAN-- 10.1 MM [0.75 iNCH)
10.
BARSINDICATEOATARANGE
[ ) NUMBEROF SPECIMENSTESTEO
0 ! I i ! I !
0 500 1000 150O 2000 2500 3000
TEMPERATURE
IFI
I I I ! I I I !
0 200 400 600 800 1000 1200 1400
TEMPERATURE,
C
GB7-O79.98A
Figure 204 (Contd). View S. Four-Point Flexure Strength of Kyoeera SC201 SSC.
168
MPa (KSll
1000"
800 •
600-
=_, .. ]
..=.
_0,
400 -
50. [lOl
1101 I101 1101
I101
MATERIAL -- STANDARD OIL SASC 1101 (IOI
I101
DATE RECEIVED -- 1985
200- DATE TESTEO -- 1985
SPECIMEN CROSS SECTION -- 3.2 X 6.4 MM (% X % INCH)

CROSS HEAO SPEED -- 0.5 MM/MINUTE 10.02 INCH/MINUTE)
OUTER SPAN -- 38.1 MM )1.50 INCH)
INNER SPAN -- 19.1 MM 10.75 INCH)

10.
{ ) NUMBER OF SPECIMENS TESTED

O = 0
! I ! I I I
0 500 I DO0 1500 2000 2500 3000
TEMPERATURE (F]
! I I I I I l I
0 200 400 500 80Q I OOO 1200 1400
GB7-O78-99A TEMPERATURE. C
Figure 204 (Contd). View T. Four-Point Flexure Strength of Standard Oil SASC.
169
o Detailed inspection after test must be con-
Q Ill 2300F I
12j,O75FATZ.5Ks, ducted to identify problem areas
j ' ',,\I\ 13j,55oo
o Crowned surfaces appear to help in allevi-
ating sticking phenomena
Ill
13) I ,T
Alumina - LAS Sticking Study "
12i 1975F_' AT a.3 KSI / {f--_lJ)-..-_l J .,_
13i 1550F/_ _ " Alumina thermocouple spacers and seals
,1, 2300F j / i I ('_u _ liii, .,nnn_:
have been fabricated for use on the T 4 ther-
12l 1855F _, AT 2.5 KSI "4 1 [ ""_.._'_/ [21_ .....
mocouple. These spacers and seals seat
against the LAS flow separator housing and
the SASC transition duct and are exposed to a
III 1500F AT 350 PSI """"_%'_#_" ENGINE CONFIGURATION
I'l 2500F TIT
peak temperature of approximately 1093C
12) 2100F TIT (2000F). To evaluate the potential of sticking
13) 160OF TIT
at these interfaces, a stack of LAS, A1203,
G67078-100
and SASC materials was exposed to 1093C
(2000F) for approximately 20 hours with a
compressive 15-pound load simulating engine
Figure 205. Interface Conditions for assembly. After completion of exposure and
871, 1149, 1371C (1600, 2100, 2500F) cooldown, no sticking between the LAS and
TIT Engines. alumina or between the SASC and alumina was
observed.
Table 28 summarizes the testing for interface
conditions I, II,and Ill. Testing at condition 6.1.3 Ceramic Interface Evaluation
Ill, depicted in Figure 206, was conducted
under a constant 36.9kN (8.3-ksi)load (engine Studies under the DARPA*/NAVY/Garrett
assembly preload) and results showed no stick- Ceramic Engine Demonstration Program show
ing or reaction. Materials tested include ACC that chipping and cracking of static compo-
RBSN, Standard Oil SASC, Kyocera SC201 nents, which ultimately led to engine failure,
SSC, Standard Oil SiSiC, Corning LAS, and were due to a contact stress mechanism at
NGK SN-50 SSN. interfaces. This mechanism was identified in
the DARPA program and a parallel ONR** pro-
As noted in Table 28, several material gram as localized high tensile stress due to
combinations result in sticking under condi- biaxial loading resulting from a combination of
tions I and II. Additional testing was con- axial aerodynamic loading and radial thermal
ducted for these combinations using flame- expansion relative movement.
sprayed mullite coating on one interface sur-
face. Results indicate that no sticking or Based on the DARPA and ONR program
reaction occurred with flame-sprayed mullite- studies, contact stress appears to be one of
coated surfaces. the most critical considerations of static-
structure component design and material eval-
Test results indicate that the AGT101 cer- uation. In summary, these studies show that
amic engine testing can be performed under for high sliding contact loads a large strength
the following guidelines: reduction resulted when bare surfaces were in
contact, moderate strength reduction resulted
o Attention must be given to material com- with high temperature lubricants present, and
bination selection
o Coated SASC or RBSN can be used for *Defense Advanced Research Projects Agency
selected interface combinations **Office of Naval Research
170
Table 28. Ceramic Interface Test Summary
Interface 1 interface I interface 2

I II m
Material 799C 1010C 1177C 843C 1077C 1260C 843C 1077C 1260C
Combination (1470F) (1850F) (2150F) (1550F) (1970F) (2300F) (1550F) (1970F) '(2300F)
RBSN/RBSN NSR NSR NSR NSR NSR NSR
RBSN/SASC NSR NSR NSR NSR NSR NSR
SASC/SASC HS 3 HS 3 HS 3 NSR NSR NSR
SASC/SC201 LS 4
SASC/SiSiC NSR NSR HS 3
SASC/RBSN NSR NSR NSR
RBSN/LAS NSR NSR NSR
LAS/SASC NSR NSR H_CmP
SASC/SN-50 NSR Dis NSR NSR NSR NSR
SN-50/RBSN NSR Dis NSR
Test Cycle
1 Evaluated at 0.889, 4.89, and ll.lkN Heat at 0.889kN
(0.2, 1.1, and 2.5 ksi) (0.2 ksi) NSR = No stick or reaction
2 Evaluated at 36.9kN
(8.3 ksi), crowned Load HS = Hard stick
3 Mullite coated- NSR Hold 20 Hrs LS = Light stick

Q
4 Evaluated at 0.889kN (0.2 ksi) Unload Dis = Discolored
Cool at 0.889kN
(0.2 ksi)
171
interface evaluation was used to evaluate the
LOAD
materials and estimated contact conditions of
-0"
AGT components. (The test procedure is de-
%_ _.J scribed in the First AGT semiannual report.)
SASC
I
p,s Initial tests on
Oil's SASC were conducted
ACC's RBSN
with point contact
and Standard
RBSN
I 6.4ram crown against a 6.4ram crown (I/4-
inch crown against I/4-inch crown) at loads
ranging from 44.48 to 254.8N (I0 to 57.3
GB7OT_IOi
pounds) normal force (F N) and at temperatures
ranging from room temperature to I093C
Figure 206. Compatibility Test Condition (2000F). For the elevated temperature tests,
HI, Interface Material Combinations. the specimens were held at temperature for 30
minutes prior to application of the tangential
little or no strength reduction resulted with a force (FT). A relative movement of I mm
suitable metal compliant layer between the (0.04 inch) was applied at a rate of 0.5ram/rain
surfaces. The test apparatus developed in the (0.02 in/rain) to allow calculation of the static
DARPA program (shown in Figure 207) for and dynamic coefficients of friction.
'I I
[ CROSS
I LOAD
[ CELL [ "EAO
INSTRON
CROSS HEAO
i I
SUPPORT t !
.!
! !
PNEUMATIC
CYLINOEB
LOAD BAR AND i
LOAD BLOCK PIVOT !
LOAD _:_
Q
PLATE_
CERAMIC
SPECIMENS --"" FURNACE
I
INSTRON BASE
GB7_79-102
Figure 207. Interface Test Apparatus.
172
Each test was repeated six times to verify a) At room temperature, the curves for both
the test rig reproducibility and to provide a materials were continuous with no abrupt
good data base for the RBSNand SASC. Fig- change between static and dynamic coeffi-
ure 208 shows the degree of data scatter for cients of friction. The curve for RBSN
the static coefficient of friction versus the gradually increased and then decreased.
normal load at room temperature. The scatter Examination of the contact surface sug-
is not excessive. The most important observa- gested that surface damage had occurred
tions are that the SASC has a substantially and that particulate debris was being push-
lower room temperature static coefficient of ed ahead of the contact interface, which
friction than does the RBSN, and the static could explain the steady increase in fric-
coefficient of friction does not vary signifi- tion. Once a sufficient amount of debris
cantly over the load range evaluated for the was'present, it could begin acting as a dry
point-contact condition. lubricant, which could explain the later
decrease in friction. Very little surface
Further information is available by com- damage was visible on the SASC stationary
paring the actual FT versus contact traverse specimen, but a roughening had occurred
distance curves for each material at each on the moving specimen. The accumula-
temperature. These curves for RBSN and tion of surface damage on the moving
SASC under 6.8 kg (15 pounds) point-contact specimen could account for the continuous
are shown in Figure 209 with coefficient of increase in friction.
friction values. These curves were reproduc-
ible from specimen to specimen; the following b) At elevated temperatures, the curves for
observationswere noted: both. materials had a discontinuity between
the static and dynamic friction, represent-
07 ative of a breakaway condition. The mech-
anism has not been fully identified, but is
, / i, / //,"///, "///,/.,",,0 suspected to be associated with sticking
06
due to surface oxidation. Contact surface
;. examination suggests that oxidation is en-

hanced in the contact region and that, once
0.5
breakaway occurs, the oxide acts as a
i viscous lubricant.
_ 9.0
0.3 // l lFl il ilt . . .
sAsc.
R,-/. o.8
0.2
-f ,' _ 4.5 0.4_ ,,
"--'---- SASC RT _ " 0.2 .._

_ _
o., _ ,
_,, OATA SCATTER _ I l,.1
0.25 '
0.50 0 5$ 1.OQ
0 TRAVERSEOISTANCE, MM
0 4.5 9 13.6 18 23 27 687-078-1048
I01 [10) 120) 130) (401 1501 160l
Ge7_78-103A NORMAL LOAO, KG ILB)
Figure 209. Friction Versus Relative Motion

Figure 208. Static Coefficients of Friction for Bare SASC and RBSN Under the Point
Versus a Normal Load for RBSN and SASC at Contact Condition With a Normal Load of
Room Temperature. 25 Pounds.
173
c) The dynamic coefficient of friction for but had little effect for elevated temper-
RBSN is lower at I093C (2000F) than at atures.
room temperature and is approximately
constant at 0.45 for normal loads of 4.5 to The point-contact tests at 11.36 kg (25
26 kg (I0 to 57.3 pounds). The dynamic pounds) are considered more severe than the
coefficient of friction for SASC increases static structure components will be exposed to
with traverse distance at 1093C (2000F) in the AGTI01 engine. Current estimates are
and room temperature to the 0.45 range. that the load will not exceed 1.6 kg (3.5
Friction also appears to increase with pounds) over a comparable point-contact area.
normal load at I093C (2000F). The purpose of the exaggerated contact loads
was to identify the mechanisms of contact
Most engine components are pre-oxidized damage and to determine the sensitivities of
prior to rig and engine testing. Figures 210 the candidate ceramic materials to surface
and 211 summarize contact data for pre- damage.
oxidized RBSN and SASC. The RBSN and
SASC were pre-oxidized for 2 hours at 1204C Additional tests were conducted at 11.36
(2200F). Pre-oxidation appeared to decrease kg (25 pounds) or less with line contact. Initial
the room temperature coefficient of friction tests suggest that the data scatter is larger
for line-contact than for point-contact. The
FRICTION
COEFFICIENTS high-temperature friction data are almost
3o _-_3L8 ù,06, conditions.

identical forTheboth
roomRBSN
temperature
and SASCcoefficient
contact
F ùd 0.50 0.52
FRICTION
COEFFICfENTS
i i
2000F 57.3 LB f N ,us 0.40
I0 I0 F. LB F
ùs 0.70 Pd 0.36-0.43
• "d 057 079 . 10 LID F N _'s 060
_'d 0 34 033
ID ==_-o
140(]F, 25 LB F N
ùs 057 10 _ Pd ' 0.30-0.40
ùd 0.70-ID.56 Ps 0.48

Pd 035-0.27
Pd 0.93-0.70 Ps 030
_ ! io ID ' F 'OFN
0
o
I0 ùs 0.15
i'd 0.11 017

ÙS 011
Ps 010
ùd 0.07-0.21
_ Pd 0.050.13 Pd O.Og-O.i O
ùs 0.12
ID _ " _ _d 0.12 0.28 ID _ i aT I0 LB FN " j s,s QIO
0 IDOI 0.02 003 0 04 0 IDOl O.ID2 0.03 004 ;'d 0 liD 018
TRAVERSE DISTANCE. INCH TRAVERSE OISTANCE. INCH
GD/OlO 105 [i81076 }06
Figure 210. Coefficients of Friction Figure 211. Coefficients of Friction

for Oxidized RBSN Point Contact. for Oxidized SASC in Point Contact.
174
of friction appears to be substantially lower After contact testing, each specimen was
for the line-contact condition for both materi- fractured in four-point bending. The strength
als. The line contact friction data for RBSN data are plotted in Figure 213 for point-and
and SASC for the ll.36-kg (25-pound) normal line-contact conditions. Obviously, the RBSN
load are shown in Figure 212. material has lower residual strength than
SASC, and the point-contact results in lower
line-contact conditions. Obviously, the RBSN
_ 90 RBSN. 1100C 0.8
material has lower residual strength than
llOOC SASC, and the point-contact results in lower
0.6
strength than line-contact. More specifically,
4.5 .,,...._ - SASC. RBSN. 1100C the RBSN exposed to point-contact biaxial
0.4 _'-
,, loading, has a residual strength of 121 MPa
v 0,2 __ (17.5 ksi). Strengths of RBSN prior to contact
w,.
z SASC. RT
typically ranges from 275 to 380 MPa (40 to 55
0 r I I
ksi). Strength degradation in the range of 60
Q 0.25 0.50 0.75 1.00
TRAVERSEOISTANCE. MM percent has occurred.
G87-078.107 B
The line-contact strength curve in Figure

Figure 212. Friction Versus Relative Motion 213 for RBSN intersects the point-contact
for Bare SASC and RBSN Under the Line curve, although most of the individual line
Contact Condition With a Normal Load of contact data points are higher. Examination
25 Pounds. of the line-contact surfaces shows that con s
STRESS. MPA
70 80 90 I00 200 300 400 500 500
99
90 /
/ EGEND
70
AVERAGE
/ MATERIAL CONTACT
CONTACT TEST
TEMPERATURE
4 POINT
MOR. MPA
_.. 50
RSSN {RBNI041 POINT 1100C 121
• RBSN IRON104) LINE I100C 186
0 RBSN (NC3501 POINT RT AND IIOOC 132

0 SASC POINT IIOOC 258
20
• SASC LINE IIOOC 433
IO
2
10 20 30 40 50 50 70 80
STRESS. KSI
GB7.078 1088
Figure 213. Room Temperature Strength RBSN and SASC After

Line and Point Contact Tests.
175
tact (and the resulting damage) only occurred contact loading, the contact apparatus data
at high points and varied from specimen to for RBSN and SASC were run through a con-
specimen. Specimens having a low residual tact stress analysis computer program devel-
strength typically had contact only along a oped under an ONR Program. The results for
small segment of the line. As the degree of room temperature and 1100C point- and line-
contact along the line increased, the residual conditions are summarized in Table 29. The
strength increased, suggesting that load distri- II00C results are plotted in Figure 214 and
bution and load sharing are important in mini- compared with baseline four-point flexural
mizing contact damage in RBSN. strength data for the line-contact configura-
tion.*
Figure 213 shows that SASC has higher
residual strength than RBSN. Specifically the The predicted stresses for the point-
average strength of SASC after point-biaxial contact condition considerably exceed the
contact is 258 MPa (37.4 ksi). The average baseline strengths of both RBSN and SASC and
strength prior to contact exposure for the would be expected to produce the degree of
configuration tested is approximately 450 MPa surface damage that would yield the severe
(65 ksi). The resulting strength degradation strength degradation measured. Peak stresses
for the SASC, therefore, is approximately 43 for line contact are substantiallyreduced, only
percent for point-contact and 4 percent for
line-contact.
*Specimen-oriented in the four-point test fix-
To estimate the actual tensile stresses ture with the flat surface in tension and the
applied at the ceramic surfaces during biaxial 0.63-cm radius surface in compression.
Table 29. Predicted Contact Tensile Stresses For Point- and Line-Contact Conditions.
Contact Point-Contact Line-Contact

Normal Test
Load, Temperature, O** a**
Material lb C Ps* ksi Ps* ksi
RBSN 10 Room 0.62 148 0.18 10

25 0.62 219 0.22 20
57.3 0.64 293 0.22 31
10 1100 0.64 152 0.54 32
25 0.60 212 0.68 64
57.3 .52 236 0.53 75
SASC 10 Room .25 67 0.25 20

25 .19 77 0.09 11
57.3 .26 154 0.15 28
10 1100 .66 170 0.62 5
25 .51 200 0.61 75
57.3 .51 290 0.70 131
]JS
static coefficient of friction
Op calculated peak tensile stress
176
ap. KSI
induce spalling has not been quantitifed, how-
50 100 i50 200 250 300
ever a review of Garrett in-house data sug-
,_ RBSN BASELINE STRENGTH
8O gests that spalling of RBSN parts may occur if
35 SASC BASELINE STRENGTH
weight gains, which occur during the nitriding
I 70
3O I
process, are below 59 percent. A weight gain
I RBSN 6O below this figure indicates that a significant
I • LiNE • _ 0
25
z , I CGNTACT/sAscOBS"
! /.,., portion of Si has not been converted to Si3N 4
If /LINE POINT [ / _'n,_r
a
during the reaction process. In addition, high
2o green densities and thick cross sections inhibit
//CONTACT CONTACT// COI_'r_llr-T
x 15 free N 2 passage to the interior, which is

==
necessary for complete reaction. Green densi-
1.g ties and part thickness limitations must typi-
/.//:
I I
10
cally
fabrication
be experimentally
approach
determined
is applied
as this
to larger or
I_1 I I i I
thicker engine components.
4_ 900 1200 1600 2000
CALCULATEOPEAK TENSILE STRESS, ap, MPA

Additional influences on complete nitriding
GDI_I_IOHD
include Si purity. Contaminants such as car-
bon will react with the Si at the elevated
Figure 214. Calculated Peak Tensile nitridation temperatures to form undesirable
Stresses for I10OC Contact Tests of compositions. Thus, any component processing
RBSN and SASC. between green fabrication (ie, casting or
molding), may adversely affect the final Si3N 4
exceeding the baseline strengths of RBSN and product if contamination may occur.
SASC at normal loads above 7.0 kg (15.4
pounds) for the configuration studied. Fabrication processing of the ACC turbine
shroud requires that green machining occur
Correlation between analytically estimated prior to part nitriding. If mounting compounds
peak stresses (see Figure 214) and retained and machining coolants are used, the result
strength are not as good for SASC in line- may be contamination of the porous Si part.
contact. Estimated contact stresses substan-
tially exceeded the base-line strength of the The significance of these and other vari-
material, yet very littlestrength degradation ables on spalling were suspected, but not
occurred. determined with certainty. RBSN fabrication
techniques were nonetheless modified with the
6.1.4 RBSN Spalling Study intent of reducing spalling. Parts were fabri-
cated with lower green densities and green
Spalling damage has occurred on three machined with no coolant, and with minimal or
ACC RBSN (RBN104) turbine shrouds when no mounting compound. In cases where
these parts were exposed to temperatures mounting compounds were used, they were
ranging from 871 to 1204C (1600 to 2200F). In subsequently machined off the Si surfaces
two instances the spallingresulted in complete prior to the final cleaning operation to prevent
destruction of the component. In the third compound penetration into the green part.
instance only limited spallingoccurred and the Turbine shrouds processed using these methods
turbine shroud subsequently passed I0 simu- were thermally cycled to 1204C (2200F) after
lated lightoffcycles to 871C (1600F). nitridingwithout any occurrence of spalling.
Spalling typically has been attributed to 6.1.5 Laser Marking of Ceramic Materials
thermal stress generated within RBSN when
residuals of Si remain un-nitrided during pro- The effect of laser marking on the strength
cessing. The amount of free Si required to of five ceramic materials was investigated to
177
determine whether this technique would be marking of these materials without heat treat-
suitable for permanently marking AGT101 ing showed a reduction in strength and all
parts. Positive identification of engine hard- fractures originated at the laser marking.
ware for the AGT program is necessary to
keep component traceability. Samples of the ACC SN-501 SSN and Norton NC132 HPSN
five ceramic materials (RBSN, SASC, LAS, test specimens failed at the laser marking,
HPSN and SSN) were used to determine the both with and without heat treatment. But
laser parameters that would clearly mark the the flexure strength of the specimens was ade-
ceramic surface without visible damage. Only quate in allcases.
laser marking of the LAS materials was unac-
ceptable, due to visible damage of legibly In sum mary, laser marking parameters
marked materials. Subsequently, test bar have been established for marking four ce-
samples were marked with a series of XHs ramic materials frequently used on the
along the length of the test bars to simulate AGTI01 program. Properly processed laser
worst case marking orientations. One-half of marked materials show littleor no material
the test bars from each material were given strength loss as a result of laser marking.
the established heat treatment for healing
surface flaws. All test bars were flexure 6.1.6 Slip Casting Rheology Studies
tested at room temperature with the laser
marked surface in tension. Table 30 summar- Two subcontracts were initiated during the
izes the flexure strength data of laser marked summer of 1983 to study the rheology of Si
specimens. Standard oilSASC and ACC RBSN and Si3N4-Y203-A120 3 particulate dispersions
laser-marked and heat-treated materials in water. The program objectives are briefly
showed no reduction in strength due to the discussed in subsequent paragraphs and the
laser marking and did not fracture in the laser results are presented in detail in Appendices D
marked area, as long as the materials were and E to the ninth AGT semiannual progress
heat treated. The flexure strength after laser report.
Table 30. Flexure Strength and Fracture Origins of

Laser Marked Test Bars
Laser Marked and

Laser Marked Heat Treated
Mean Mean
Strength Fracture Strength Fracture
MPa (ksi) Origin MPa (ksi) Origin
Standard Oil 225.5 (32.7) At laser 352.3 (51.1) Not at laser

SASC (5 samples) marking (5 samples) marking
ACC RBN104 228.9 (33.2) At laser 263 (38.2) Not at laser

RBSN (5 samples) marking (5 samples) marking
ACC SN-501 676.4 (98.1) At laser 583 (84.6) At laser

SSN (7 samples marking (7 samples) marking
Norton NC-132 615 (89.2) At laser (lO5.O) At laser

HPSN (8 samples) marking (8 samples) marking
178
The Si slip program was subcontracted to - Particle size control by elutriation
Dr. M.D. Sacks of the University of Florida to
support reaction-bonded Si3N 4 processing at o Electrophonetic mobility for each powder
ACC. in aqueous suspension
Objectives/tasks included: o Zeta potential versus pH
o Baseline characterization of ACC-supplied o Preparation of cast specimens for sintering

Si powder and evaluation at Garrett
- Particle size distribution The study report, "Development of Homo-

- Powder surface area measurement geneity in Si3N 4 Ceramics by Colloidal
Filtration", was co-authored by I.A. Aksay
o Electrophonetic mobility versus pH and C.A. Ambarian of UCLA and covers
June through September, 1983. The report
o Zeta potential versus pH is included as Appendix E of the ninth
semiannual AGT progress report.
Shear stress versus shear rate as a function
of solids contact, pH, and ultrasonic agita- 6.1.7 Turbine Diffuser Insulation Evaluation
tion
The use of a fibrous insulation in the
o Viscosity versus shear rate as a function of turbine diffuser was stimulated by the suc-
solids content, pH, and ultrasonic agitation cessful use of Babcock and Wilcox vacuum-
formed 3000 material in rig testing. Results
o Total porosity and median pore radius of of the rig testing indicate that a fibrous
cast samples as a function of slip solids insulation could withstand high temperature,
content, pH, and ultrasonic agitation high gas velocities for a short period of time.
However, surface erosion on the insulation
o Screening study of potential dispersants indicated that a more durable system would be
required for longer term exposure and higher
The study report "Dispersion of Silicon gas velocities. In pursuit of an improved
Powder", covering June through September insulation system, several samples of
1983, is included as Appendix D to the ninth Lockheed High Temperature Protection (HTP)
semiannual AGT progress report. A technical were obtained for evaluation. The sample
paper was subsequently prepared by Dr. Sacks consisted of cylinders of sintered and ma-
and submitted (after NASA approval) to the chined high purity SiO 2 and Al20 3 fibers. The
American Ceramic Society for publication. samples were machined to form cylinders with
a 7.87cm (3.l-inch) bore and a 7.62cm (3-inch)
The Si3N4-Y203-A1203 slip program was length. As-machined cylinders, and cylinders
subcontracted to Dr. I.A. Aksay of UCLA and partially coated with a boro-silicate glass,
subsequently of the University of Washington designated by NASA and Lockheed as reaction
to support sintered Si3N 4 rotor processing cured glass (RCG), were provided by Lockheed
technology. and shown in Figure 215.
Objectives/tasks included: Two samples, one uncoated and one coated,

were exposed to combustor discharge gases as
o Baseline characterization of commercial illustrated in Figure 216. During this test,
Si3N 4, Al20 3 powders vitiated air at 1149C (2100F) was directed
through the cylinder bore at 9 kg/min (20
- Agglomerate content and structure Ib/min) [approximately 76.2 m/see (250 ft/
- Particle size distribution see)] for 46 minutes. This condition exceeded
• 179
minor surface erosion on the uncoated sample.
This erosion was judged a result of local
surface damage resulting from the machining
operation rather than the exposure conditions.
These preliminary results indicate that coated
or uncoated HTP may be suitable for use in
the current AGT101 engine configuration and
that the RCG coated HTP may be suitable for
use in higher gas velocity locations. Addi-
tional testing was directed toward evaluating
insulation capabilities under representative
turbine rotor discharge conditions, simulating
gas velocities and temperatures rather than at
the turbine diffuser dump location• Two test
conditions were selected representative of
engine maximum power [966C, 304.8 m/see
(1770F, I000 ft/sec)] and idle [1093C, 106.7
84636-1
GB7-078.110
m/see (2000F, 350 ft/sec)] flow conditions. To
perform these tests, additional RCG-coated
and uncoated test cylinders were obtained
Figure 215. RCG Coated and Machined HTP from Lockheed. The cylinders were 7.62 cm
Insulation From Lockheed. (3-inches) long with a 6.4 cm (2.75-inch) inner
diameter (needed to attain the higher gas
velocity conditions)• The new cylinders were a
THERMOCOUPLE (3l
higher density HTP-16 (16 Ib/ft3)than the
HTP-12 material previously tested•
_ _ SAMPLE CYLINOERS ./ OISCIIARGE
Testing was performed for 7 cycles pro-

filed in Figure 217. The test rig was dis-
assembled for inspection after completion of
Cycles i, 3, and 7. At completion of Cycles 1
WA
(111 I -II.80RO_SlUCAT I and 3, no material distress was observed.
I GLASS COATING Following Cycle 7, however, disassembly
revealed coated sample shrinkage, uncoated
sample shrinkage, and melting (illustrated in
Figure 218). Based on previous testing at both
Garrett and Lockheed, these results did not
\oo , correlate with 1093C (2000F) exposures.
the two samples were returned
Thus,
to Lockheed
_'_PLATE RETAINER
GBI-O7&I I ] for post-test analysis. Meanwhile, the test
conditions were reviewed at Garrett, including
reealibration of the three control thermo-
Figure 216. AGT101 Insulation Test Rig. couples located downstream of the test cylin-
der discharge. Three conditions were identi-
the 22.9 m/see (75 ft/see) gas velocity pre- fied that may have contributed to the atypical
dicted for the turbine exhaust diffuser dump results• These included the use of a gas
region currently utilizing the Babcock and discharge water quench downstream of the
Wilcox fibrous insulation. test cylinders for rig ducting protection. This
discharge may have provided recirculated con-
Post test visual examination did not reveal tamination or vapor that could conceivably
any damage to the coated sample and only have reached the test articles• Contaminants
180
_, =. _ _i'i_ _Li _i
_:_
107 M/SEC 94 M/SEC
1093 1350 FT/SECI 1300 FT/SECI
""l / I I ,o,,,,,EC I ""1I IJ;;N! ,,O,,NO,ES ,%

oee/ I I .ooo_r/sEc) I I I
,,,,o,
| TEAROOWN| | |TEAROOWN |
I_ -INSPECTION- | INSPECTION |
I l I l II a l
I 2 3 4 5 6 7/
CYCLE
/
TEARDOWN
GII7,.078-112A INSPECTION
Figure 217. Maximum Power and Idle Condition of Lockheed HTP-16 Insulation.
produced local burning and hot spots, which

would have contributed to insulation melting.
The most likely contribution to insulation

damage probably occurred at overtemperature
test conditions. Post-test calibration of the
three Type S thermocouples indicated that one
thermocouple was reading substantially lower
microvolt output than the others. Due to the
averaging of the output of these thermo-
couples by the control and recording systems,
there is a strong possibility that temperatures
500C over the I093C (900F over the 2000F)
GB7-078-113 86821-3 desired test condition occurred in the cycles
subsequent to Cycle 3. Thermoeouple deter-
ioration is believed to have occurred during
Figure 218. Lockheed HTP-16 Cylinders the final phases of testing (Cycles 6 and 7).
After Cyclical Flow Testing.
Continuing discussions with Lockheed
may have contributed to a reduction in the regarding these results confirmed the typical
melting temperature of the insulation materi- occurrence of HTP material shrinkage at
als. Water quenching was used only in test 1593C (2900F), supporting the indications of
cycles subsequent to Cycle 3. an over-temperature condition. Nonetheless,
encouragement is gained in that the test
Another possible contributor was the oc- samples demonstrated short-term durability
currence of wet starts in which unburned fuel during the first 3 cycles, and that even under
may have been sprayed onto, and by, the the excessive conditions of Cycles 4 through 7,
insulators. Subsequent operation would have only limited deterioration occurred.
181
OR!GIN._L PAGE IS
OF POO_ QUALLT_
6.2 Ceramic Turbine Rotor
The AGT101 radial turbine rotor has re-

quired extensive development in design, mate-
rial, and fabrication. Through close involve-
ment with AGT101 ceramic suppliers, Garrett
was able to define the barriers to successful
rotor fabrication and determine possible solu-
tions. The material strength requirements
were determined through detailed analysis.
High-temperature materials with the required
strength were selected for rotor fabrication
development. Initial development concen-
trated on fabrication of bladeless simulated
rotor hubs to demonstrate that thick-section
hubs of the required strength could be fabri-
cated. Suppliers successfully achieving this
task were released to pursue bladed rotor
development.
6.2.1 Initial Evaluations
Garrett selected four ceramic suppliers to

develop the AGT101 turbine rotor:
AiResearch Casting Co. (ACC), Pure Carbon
Co., Standard Oil, and Ford. Difficulties in
fabrication were expected due to the large
volume and thick cross section of the rotor.
For this reason, suppliers were requested to Figure 219. Simulated Rotors Were
provide bladeless simulated rotor hubs for Used to Demonstrate Thick Section
evaluation to demonstrate the capability to Fabrication Capability.
fabricate thick-section rotor hubs of the
required strength. A typical simulated rotor is Oil hot-pressed alpha-SiC material met the
shown in Figure 219. strength requirements and was a possible rotor
candidate material if a suitable fabrication
Rotor materials evaluations indicated that method could be found.
the majority of candidate rotor materials met,
or nearly met, the first-year material strength All four suppliers successfully demon-
goals. Nominally, these goals were a charac- strated simulated rotor shape capability using
teristic strength of 469 MPa (68 ksi) and a a variety of fabrication techniques. ACC
Weibull modulus of 12 at 1093C (2000F). approached the challenge with a slip casting
These goals were based on a rotor stress technique. Ford also concentrated on a slip-
distribution that peaks at approximately 207 cast approach. Standard Oil selected an initial
MPa (30 ksi) within the rotor hub. Final two-piece fabrication approach in which a hub
material goals were a characteristic strength and shell were isopressed, baked-out, and
of 552 MPa (80 ksi) and a Weibull modulus of bonded (hot pressed) together. Standard Oil
14. Suitable candidate rotor materials were selected this approach due to the difficulties
demonstrated for first-yearrotor development in binder bake-out of cross sections greater
for three materials: Ford Sintered RBSN, than one inch. Improved binder removal
ACC Sintered Si3N 4, and Pure Carbon Refel methods allowed Standard Oil to later abandon
Reaction-Sintered SiC. In addition, Standard the two-piece approach in favor of a solid
182
OZ _(] 7,',] '"
OE POem qUALIT_
injection-molded rotor. Pure Carbon, in con- Through changing the slip theology (viscosity
junction with British Nuclear Fuels, Ltd and specific gravity) Ford was able to obtain a
(BNFL), fabricated rotors from reaction sin- more consistent density throughout the rotors,
tered SiC. eliminating the cracking problem. Further
improvements in simulated rotor quality were
Although advancements in rotor fabrica- obtained through controlled drying. These
tion were made by all suppliers, only minimal improvements could be seen in the successful
rotor deliveries were made early in the pro- spin testing of simulated rotors.
gram. Standard Oil produced a "spin quality"
SASC simulated rotor, i.e., free of x-ray 6.2.2 Bladed Rotor Fabrication
indications and significant visual defects.
Oarrett spin-tested the rotor, obtaining good Initial bladed rotor fabrication began in
correlation between analysis and material early 1982, with the suppliers demonstrating
properties. A room temperature burst speed limited success. As discussed in paragraph
of 72,400 rpm was achieved. Pure Carbon suc- 5.2.4, the blade cross section was redesigned
cessfully fabricated a simulated rotor from to improve fabricability (ref. Figure 136).
reaction sintered SiC. One simulated rotor, Nominal rotor thickness at the inducer blade
successfully formed, reacted, and inspected, tip was modified from 0.762 to 0.774 mm
was delivered for evaluation. Evaluation (0.030 to 0.040 inch). These changes reduce
indicated that the Pure Carbon Refel material the difficulty encountered in achieving com-
would require additional development for plete.blade fill. Late in the AGTI01 program,
AGTI01 rotor application; thus it was discon- another design iteration was made, shown
tinued from the program. pictorially in Figure 220. This change resulted
During the first half of 1981, simulated HOLLOW SHAFT:

IMPROVES
rotors were received from ACC and Standard
Oil. Evaluation of the ACC simulated rotors
FAOR,CAO,L,T"
demonstrated rotor hub integrity. Additional
evaluation was required to determine the re-
producibility of the demonstrated rotor integ-
rity. This was achieved late in 1981. _ ACC
was given approval to fabricate bladed
AGTI01 rotors, with initial bladed rotor deliv-
eries received in early 1982. Initial testing of
Standard Oil's SASC material showed it did not
meet the initial required strength goals when
fabricated into simulated rotor shapes and was
deleted from future consideration. Two addi-
tional Standard Oil materials--hot-pressed
alpha SiC and reaction-sintered SiC--were
evaluated. These materials also did not meet
strength goals when fabricated into rotor
shapes and were also deleted from considera-
tion. Efforts by Standard Oil on AGTI01 rotor
development were postponed until Standard
Oil could demonstrate sufficient rotor mate-
rial properties.
GB7-078-041 93402-1
Ford encountered difficulties in forming

simulated rotors, due primarily to density Figure 220. Hollow Shaft Design
gradients, which resulted in internal cracking. Improved Fabrieability.
183
in both an improved bearing surface and investigated to clean both the powder and the
rotor/compressor attachment. Rotor/com- slip prior to casting but the problem was not
pressor attachment methods will be discussed solved. This experience emphasized the re-
later in section 6.2.4. quirement for a clean-room processing envi-
ronment. Incomplete blade fill also remained
The remainder of the AGT101 rotor effort a problem (Figure 221) despite centrifugal
focused on fabrication by ACC and Ford using spinning during" casting. While ACC was not
a slip cast Si3N 4 approach. Problems were able to solve blade breaking during densifica-
encountered in slip preparation, casting, and tion, it was attributed to non-uniform green
densification. density and/or furnace thermal gradient and/
or non-optimized sintering cycle. Despite
To consistently prepare a high quality slip this, ACC did achieve near theoretical densi-
for casting, the starting powder must be con- ties and produced rotors capable of passing
sistent with respect to purity and particle size 115,000 rpm spin testing. The high tempera-
distribution. While the powder may meet the ture properties of the materials developed by
powder supplier's specifications, there is ACC, however, did not meet AGT101 RPD
usually batch-to-batch variations for which temperature goals.
the user must compensate, normally through
additional powder processing. Powder impuri-
ties can degrade the material properties by
affecting grain boundary chemistry and/or in
the formation of strength reducing inclusions.
A controlled optimum particle size distribu-
tion is required to consistently achieve high
strength and minimize shrinkage and porosity.
To optimize strength, a powder that can be
densified quickly with minimal grain growth is
desired.
Casting is affected by the slip properties.

The viscosity must be lowenough to allow
complete mold fill, yet the solids content must
be high enough to achieve a reasonable casting
rate and green density. AGT101 experience
showed that too-slow a casting rate would
result in density variations due to particle
segregation, while too-rapid a casting rate
would result in blockage of narrow passages in
the mold when casting under atmospheric EXAMPLES OF
pressure conditions. INCOMPLETE .......
687-079-42 OLAOE FILL 91678-6
Problems specifically encountered by ACC

included powder purity, partical size distribu- Figure 221. Example of Incomplete
tion, incomplete blade fill and cracking during Blade FilL
densification. The powder particle size distri-
bution and purity was not consistent from Ford encountered similar problems, includ-
batch to batch as received from the powder ing poor tip definition,poor blade resolution,
supplier. Powder impurities resulted in con- and cracks at the blade roots and in the hub
taminant migration to the blade tips, causing region. The tip definition and blade resolution
laminar platelet formation and subsequent problems were improved by using a centrifugal
blade tip cracking. Several approaches were casting technique that forced slip into the
OR2G:NAL PAGE IS
184
_)_]_ P_O.OR QUALITY
blade cavities. Blade root cracking appeared observable interior defects confirm the rotor
to be improved by modifications in the wax hub integrity. This initialbladed rotor testing
removal technique. The hub cracking problem indicated the successful transfer of simulated
was improved by more careful removal of the rotor technology to the bladed rotor. Second,
plaster base. Internal cracks were determined the surface features responsible for the rotor
to be caused by density gradients, similar to failure were easily observable prior to spin
the problem encountered in simulated rotor testing. This fact implied that casting and
fabrication. This was corrected through use of sintering processes which produce good sur-
an organic suspending agent and greater con- face finish rotors will increase spin speed
trol of slip viscosity to eliminate slip settling capability. From these results,both Ford and
during casting. ACC were directed to apply fabrication pro-
cesses to improve visual appearance (surface
In addition, as Ford improved material high finish)to rotors being fabricated.
temperature properties, changes were made to The end.of 1982 saw three additional
the slip. One problem this caused was a low
bladed rotors from ACC judged suitable for
green strength and thus a greater risk of
cold spin testing to assess rotor development
handling damage. However, by adding a pre-
progress. All three rotors were slip cast at
sintering step to the densification process,
ACC and sintered by Ford under a Ford/ACC
handling strength was increased. This allowed
agreement. Table 31 gives the densities and
minor machining operations to be carried out
spin test speeds of the rotors. One rotor
on the rotor in the green state, precluding the
failed due to a surface flaw while the remain-
need for expensive, time-consuming diamond
ing two survived the spin test and were select-
grinding of the fully dense rotor.
ed for hot turbine rig testing.
ACC bladed rotor S/N 10191, east in
Table 31. ACC Bladed Rotor Test Results.
October 1981, was judged suitable for spin test
evaluation based on visual inspection, x-ray
radiography, and density evaluation, despite
S/N I Densi_
_m J
Speed, [
incomplete blade fill. Spin testing was per-
L
formed to verify rotor hub integrity in a

bladed rotor. Initialspin testing resulted in 213 3.17 102.0 Burst due to sur-
the loss of two blades at 83,000 rpm. A face flaw
secondary blade tip fracture occurred in one
blade adiacent to that of the two blades; no 277 3.27 115.0 Successful over-
hub or shaft damage occurred. Visual and 30x speed test
microscopic analyses revealed that both blades
fractured at surface casting defects. The 211 3.09 80.0 Qualified to 80.0
second spin test achieved 95,300 rpm at which krpm hot turbine
time complete rotor destruction occurred. rig testing
Partial reconstruction was possible since sev-
eral large pieces of rotor remained. This Although both Ford and ACC continued
indicated that although the specific origin was rotor fabrication development efforts, neither
missing, the primary fractures initiated at was able to produce rotors with the quality
surface locations with secondary fractures required for engine testing. In order to fulfill
propagating inward to the hub. Examination contractual requirements for engine demon-
of interior fractures revealed no internal de- strations, Garrett began purchasing rotors
fects. from Kyocera and NGK towards the end of the
AGT101 program.
Two encouraging facts came from this
test. First, the fact that the rotor fractures Initialrotors received from NGK were fab-
initiated from surface sites, and the lack of ricated from both SN-73 and SN-81 (sintered
185
silicon nitride materials). The first rotor mance engine testing. The remaining ceramic
failed the 115,000 rpm spin proof test, burst- engine tests utilizing SN 250M rotors were
ing at a speed of 112,000 rpm. The second limited in speed and temperature in an
rotor was proof tested to 100,000 rpm, which attempt to avoid stress rupture failures due to
it successfully passed. The remaining two the lower material properties.
rotors were returned to NGK for further pro-
cessing. This processing included additional Ttie SN 250M rotor (S/N 256) which had
heat treatment to recrystallize the grain been used in the 85-hour, 1204C (2200F)
boundaries in an attempt to improve the high endurance test (engine S/N 004C, Build 6) was
temperature strength. Both were re-spin spin tested to burst, which occurred at 108,600
tested by NGK prior to shipment to Garrett. rpm. The rotor had been spin tested twice
One burst during the spin test, the second was prior to engine operation: once at Kyocera to
spin tested to 95,000 rpm. This rotor was 115,000 rpm before shaft bore machining, and
never used for engine testing due to the avail- again at Garrett to 105,000 rpm after the
ability of improved rotor materials. (Rotor shaft bore was machined.
material properties are discussed more fully in
paragraph 6. I). The failure mode of rotor S/N 256 (spin pit
tested) was similar to that of rotor S/N 259
Initial rotors received from Kyocera were from engine testing in S/N 002C, Build 18.
fabricated from SN 220M (sintered silicon These failures are compared in Figure 222. In
nitride). These wheels were proof spin tested both rotors, the hub centerline remained
to 110,000 rpm at Kyocera. This was lower
than the desired proof spin speed of 115,000
rpm due to concerns of the material strength
of SN 220M. In response, engine testing was
limited in speed and temperature when testing
the SN 22/)M wheels. Later rotors received
from Kyocera were fabricated from SN 250M,
also a sintered silicon nitride material. These
wheels were proof spin tested to 115,000 rpm
at Kyocera. The SN 250M material has a
higher room temperature strength than the SN
220M material and could thus be spin tested to
the higher speed.
Results of ceramic engine testing utilizing

these SN 250M wheels are discussed fully in
paragraph 4.3.3, Ceramic Engine Develop-
ment. Results pertaining directly to rotors ARROWS
S/N 259
INDICATE
will be discussed here. In two engine tests, CRACK
S/N 256
the inducer blade tips were damaged on the PROPOGATION
PATHS
two SN 250M rotors, rendering them unaccept-
able for further engine testing. These rotors • ROTORS/N 259 FAILED iN ENGINE 002C,
were sectioned into test bars for fast fracture BUILD 18. CRACKS PROPAGATEDFROM
SHAFT AREA
and stress rupture property evaluation.
Results indicated that the material in these • ROTORS/N 256 FAILED DURING SPIN PIT
TEST. CRACKS PROPAGATEOFROM SHAFT
excised sections exhibited lower strength AREA.SIMILAR TO ENGINE FAILURE
properties than those measured in bars pro- G87-078-43
vided as test specimens of the same material. Figure 222. Rotor S/N 256 Failed
These lower material properties in SN 250M During Spin Test in Mode Similm- to
rotors were deemed unacceptable for perfor- Previous Engine Failure.
186
intact and the fracture surface propagated with density. Burst occurred at 121,000 rpm
from the exducer hub zone to the rotor back- resulting in a maximum principal stress of
face. A close examination of rotor S/N 256 332.3 MPa (48.2 ksi). As shown in previous
revealed that the fracture originated near the Figure 223, this correlates favorably with pre-
interior surface of the shaft bore in the area vious high density rotor data. Ultrasonic
of some transverse grinding marks. The stress inspection of rotor S/N 688 revealed a large
concentration characteristic of the machining flaw, approximately one inch from the back-
marks could have caused the rotor to fail. face, rendering this rotor unsuitable for engine
testing. Due to the size of the flaw, expecta-
Rotor Density tions were that the rotor would burst in the
70,000 to 80,000 rpm range. The actual burst
Testing showed a definite correlation be- speed was 101,000 rpm, with a maximum
tween rotor density and burst speed (Figure principal stress of 235.8 MPa (34.2 ksi). The
223). Rotors with a density of 98.5 percent failure of rotor S/N 688 reinforced the reli-
ability of ultrasonic inspection in revealing
flaws and the need for additional data corre-
OENSITY AS AN INDICATOR OF STRENGTH
lating flaw size with burst speed and density
THE DATA SHOWS THAT A DENSITY
OF 98.5 PERCENTOF THEORETICAL1OR 3.22g/cm 3
to determine a maximum allowable flaw size.
iN THIS CASE)IS REQUIREDTO ACHIEVE 15 PERCENT
OVERSPEEOi115K rpm] OF AN AGTI01 ROTOR
6.2.3 Hot Turbine Rig
130-
125- "0 SPIN TO BURST ,_/N 425

The hot turbine rig was designed, fabri-
• SPIN TO 115.000 RPM __._S/N 552C
% 12o- cated, and mechanically tested using metal
engine turbine rotors. The objective of the
=115-
E hot turbine rig tests was to subject ceramic
110- turbine rotors to transient thermal and me-
105- .... /. _ chanical

in the AGT101
loads simulating
engine.
startup
These
and shutdown
transients were
100-
I,- designed to induce blade stresses equivalent to
=
E
95- o o/ I engine-like operational stresses, thus providing
=
N
O0-
/0 ,, further
engine
screening
testing.
of the ceramic
Initial rig
rotor prior to
testing using a
/ ',
_ 85-
metal rotor showed that the rotor could be
B0-
controlled at all speeds. Thermal transients
I
75 I I l I I I approximating expected engine conditions
3.05 3.10 3 15 320 3.22 3.25 3.30 could be simulated with the rig.
DENStTY. o/CM 3
GB7_7e-44
After initial rig checkout, the rig was
Figure 223. Increasing Consistency reassembled with an ACC silicon nitride tur-
Evident in Cold Spin Test. bine wheel which had been cold spun to 80,000
rpm (S/N 211). Three cycles of rig operation
theoretical typically have sufficient strength were completed. Each cycle imposed stresses
to pass the cold spin test. Low density rotors in the ceramic wheel that were 2 to 3 times
can be immediately rejected, thus eliminating more severe than would be expected in a
costly inspection and testing. As part of the normal shutdown cycle when operated in the
continuing materials evaluation, rotors S/N 871C (1600F) metal engine. Inspection after
652 and 688 (ACC, Si3N4) were selected as disassembly revealed no damage or signs of
spin-to-burst candidates. Rotor S/N 652, due distress to the ceramic components. As a
to its high density, was selected to provide an result of this testing, the rotor could be used
additional data point to correlate burst speed in a metal engine restricted to 80,000 rpm.
187
A second ACC turbine rotor, cold spun to on silicon nitride shafts (Figure 224, View A).
115,000 rpm was tested in the hot turbine rig. Thermal cycling consisted of 15 minutes of
A cycle was defined to simulate a combustor soak at 538C (1000F). Coupling environment
flameout while operating at i00,000 rpm at temperatures were measured at the coupling
full load. Three cycles were accomplished on surface using a standard thermocouple. Since
the wheel. After teardown inspection showed a limited number of test specimens were
no evidence of any cracking, rubbing, or other available, the ceramic shaft was removed and
distress on the turbine wheel. As a result of then reinserted into the sleeve at different
this testing, the wheel could be used in a 871C circumferential orientation to avoid single-
(1600F) metal engine without speed restric- assembly effects during testing.
tions.
INITIAL SLEEVESWERE STRAIGHT /BRAZEO JOINT
WALLED WITH DIFFERENT INNER
As confidence was gained in rotor hub
_. IWALL THICKNESSES
integrity and material properties, it was
determined that the hot turbine rig was no
longer necessary. Rotors could be successfully
screened for engine testing through visual in-
spection, density determination and cold spin
pit testing.
6.2.4 Turbine Rotor Attachment
Attachment mechanisms of the ceramic /' V,EWA

" •
turbine rotor to the metallic rotating group _ OUTER
were analyzed and a "shrink fit" coupling ar- s_ENEEvRE

_ REF,NEMENT RESULTED IN CONTOUREO _SLEEVE
rangement was selected for initial evaluation.

]ENOS ON THE INNER WALL
This type of arrangement was necessary due to
the initial solid shaft turbine wheel design.
Detailed thermal and stress analyses indicated
that a low expansion, high-strength alloy could
be used with the ceramic rotor to effect the SHAFT
attachment. IN-907 material was selected for
Si3N4 __,
testing based on the strength and thermal
expansion coefficient in the temperature
range of interest 1120 MPa at 0.2 percent f ] •
yield and 1.05 and 1.67/C x 10 -6 coefficient of VIEW 8
expansion (162.5 ksi at 0.2 percent yield and GB7.07B..45

B
4.5 to 5.0 in/F x 10 -6 coefficient of expan- Figure 224. IN907 Ceramic Test Sleeves.
sion).
Initial thermal cyclic testing showed that
Preliminary assembly tests confirmed that thermal "ratcheting" and consequent varia-
diametral interferences between 0.11 and 0.12 tions in length of the coupling assembly were
mm (0.0045 and 0.005 inch) could be made concerns of the sleeve design. Based on these
without subjecting the IN-907 to temperatures results, a coupling was designed, as shown in
in the recrystallization range and further, no Figure 224, View B, to include a straight inner
apparent damage to the ceramic shaft was wall and a contoured open end. Testing was
observed. Therefore, a series of thermal resumed. The -I version, 3.53 mm (0.139 inch)
cycling tests were conducted to determine wall thickness and 22.5 mm (0.886 inch) diam-
dimensional stability after assembly. IN-907 eter ceramic shaft, showed essentially no
sleeves of two different thicknesses, 1.9 and racheting after 350 thermal cycles. The -2
3.0 mm (0.075 and 0.120 inch), were assembled design, 1.9 mm (0.075 inch) wall thickness and
188
25.8 mm (1.015 inch) diameter shaft racheted VACUUM HOLGING
toward the contoured end (shrunk) from 0.058 JAR

\
TOOL
to 0.21 mm (0.0023 to 0.0083 inch) in each 50

cycles out to 350 cycles; thus the -i version \
was selected for further evaluation. Addi-
tional testing was conducted on the foil bear- _\
ing and rotor dynamic test rigs, including
thermal cycling tests, to evaluate the coupl-
ing. It should be noted that since the AGTI01 /COUPLING
""-'-- INOUCTION
utilizes an air lubricated bearing located over
HEATING
the coupling area, a second sleeve is brazed to
the inner sleeve to reduce thermal input to the
bearing and to maintain a uniform journal
diameter. (ref. Figure 224. BASEROTOR CO,L
FIXTURE
Initial rotors were of the stub shaft config- L

uration, shown in Figure 225. The initial
•
L
CLOSECONTROLOF COUPLING TEMPERATUREDURING iNSTALLATION
I
INITIAL ATTACHMENT METHOD INVOLVEO • OXIDATION OF COUPLING (INCONEL 907 -- NO CHROME)
SOLID STUB SHAFT ROTOR • RAOIAL POSITIONING OF COUPLINGRELATIVE TO SHAFT
]'-"_ TITANIUM TIEBOLT • VACUUM PUMP VIBRATION CAUSED ROTOR TO SHIFT
• AXIAL POSITIONING OF COUPLING RELATIVE TO SHAFT
• BLIND ASSEMBLY
GB7@7647A
Figure 226. Installation Difficulties

with Initial Attachment Method.
u -'TURBINE i
cation to the coupling. Electroless nickel was
plated onto the couplings, eliminating the
GII7_78-48A oxidation problem. This in turn eliminated the
positioning problems through removal of the
vacuum chamber requirement. A new attach-
Figure 225. Initial Attachment Method
ment tool (Figure 227) eliminated the radial
Involved Solid Stub Shaft Rotor.
and axial positioning difficulties.
coupling attachment method, shown in Figure

226, was difficult to perform because of the Further foil bearing development work in-
attachment apparatus. Close control of the dicated that sintered silicon nitride, when
coupling temperature during installation was properly polished, made an excellent load
required. Installation had to be performed in a bearing surface. A rotor design modification
vacuum due to oxidation of the coupling at the was made with the resulting rotor/compressor
temperatures required for successful installa- attachment configuration shown in Figure 228.
tion. Radial positioning of the coupling rela- This design has two major benefits over the
tive to the rotor shaft was difficult due to stub shaft design: (I) the hollow shaft reduces
misalignment in the fixturing. Due to the the amount of heat transferred to the journal
necessity of assembling the rotor/coupling in a area, keeping the journal surface cooler, and
vacuum, axial positioning of the coupling rela- (2) ceramic journal/gold plated foils is a very
tive to the rotor shaft had to be done blind. durable foil bearing combination with high
These problems were solved through a modifi- load carrying capabilities.
189
6.2.5 Cold Spin Testing
! Discussion of cold spin testing is incorpor-
ated into the discussion of bladed rotor fabri-
cation and test activities,paragraph 6.2.2.
6.2.6 Rotor Blade Vibration
DUMMY COUPLING HOLDINGTOOL Natural frequencies, mode shapes, and

WITH OVERSIZED INCLUDES AXIAL
strains were predicted and verified for the
I0 ALIGNS TOOL _ / POSITIONING PIN
AND ROTOR I" ON BOTTOM AGTI01 sintered silicon nitride rotor. Finite
element model (FEM) modal analysis was used
to predict natural frequencies, mode shapes,
and normalized stresses and strains. Results
from shaker table, acoustic ring, holography,
and cold motored engine tests were used to
verify room temperature frequencies, mode
shapes, and vibratory strains.
6.2.6.1 Rotor Blade Medal Analysis
Finite element models were generated both

from tooling layouts (TLs), and hardware in-
spection data. Inspection data were used to
generate a second model because Kyocera's SN
• ELECTROLESSNICKEL PLATED COUPLINGS
250M rotor geometries were noticeably dis-
ELIMINATES OXIDATION PROBLEM crepant from TLs.
• NEW ATTACHMENT TOOLING ELIMINATES
RADIAL ANO AXIAL POSITIONING A typical 3-D AGTI01 blade FEM model
DIFFICULTIES
used for the modal analysis is shown in Figure
• OUMMY COUPLING
229.
• LOCATING PIN ON TOOL
GB7-07848 A
A Zeiss series 500 coordinate measuring
Figure 227. Installation Difficulties machine was used to inspect blade surfaces
Solved with New Method. (both pressure and suction sides). Side screen
inspection was used to define the blade outline
in the axial-radial plane. An SN 220M rotor
FINAL ATTACHMENT METHO0 INVOLVED
HOLLOWSHAFT ROTOR (S/N-0005-6), and an SN 250M rotor (S/N 254)
were inspected.
The Zeiss machine was programmed to

measure surface locations along blades in
planes normal to the axis of rotation (Z sec-
tion cuts) and normal to the radial reference
axis (L section cuts). Comparisons of inspec-
tion data to TLs at a Z section for both SN
220M and SN 250M rotors are shown in Figures
230 and 231.
L] COLLET
GB7-07_IgA
Zeiss inspection showed a major geometric

Figure 228. Internal Collet Attachment discrepancy in the SN 250M blade exducers.
Method. The blade exducers leaned toward the pressure
190
A B
S|;.Olfl.t
Figure 229. Typical 3-D Views of the Finite Element Blade Vibration Model.
side, so that the blade cross section is no elements after the problem with skewed ele-
longer radial. This SN 250M exducer discrep- ments was identified.
ancy is apparently caused when exducers droop
The Campbell diagram, shown in Figure
under their own weight during densification.
232, was generated for the SN 250M rotor
The SN 220M rotor did not exhibit this droop.
blade at 1371C (2500F) T 4 and i00,000 rpm.
Inducers of the two rotors appeared very simi-
lar. Operating temperatures and stresses were
included to account for thermal softening and
centrifugal stiffening. Rotor temperatures
Turbine blade inspection data were used to
generate a 3-D finite element blade model. used for the 1371C (2500F) T 4 case are the
same temperatures used in the complete rotor
First, a 2-D blade grid was created from side
stress analysis and are discussed in paragraph
screen inspection data. Then, using a least
6.2.7.2 Rotor Stress Analysis.
squares fit, 3-D polynomial surfaces for the
pressure side and suction side were generated Garrett analysis programs ISOVIB and
from Zeiss inspection data. The 2-D grid was EIGEN were used to set up and solve the eigen
projected onto the pressure and suction side problem for each of the seven analysis models
surfaces to create a 3-D FEM. listed in Table 32. Solutions consisted of
natural frequencies predicted for the first
Finite element models of the SN 220M three modes for the seven models, in Table 32.
rotor blade, SN 250M rotor blade, and nominal
blade were generated with skewed exducer 6.2.6.2 Blade Vibration Testing
elements. Another SN 250M and nominal Rotor vibration tests were performed to
blade model were constructed with normal measure natural frequencies, verify mode
191
SN220M SN250M
ROTOR ROTOR
Z = 2.54 mm = 2.54 mm
I0.10 IN) (0.10 INi
.5.1mm -5.1 mm
I-0.20 IN] I-0.20 Inl
x x
40,§ nlm
40.6 mm
11.60 INI it,60 iN)
Y AXIS ram, (INCHES) Y AXIS ram, )iNCHESI
|IT_P+tA
Figure 230. Comparison of Inspected Z Seetions (heavy lines) to TLs at 2.54 mm

(0.I0 ineh) From Inducer Bacldaee.
shapes, and measure vibrational strains. The Acoustic Ring Testing

type of tests run and rotors tested include:
Due to concerns of possible blade vibration
o Acoustic Ring (SN-73, SN-81, SN 220M, problems during engine operation, acoustic
and SN 250M) ring testing was used toward the end of the
program to determine natural frequencies of
o Holograph (SN 250M - S/N 252) each blade of each rotor. The frequencies can
be converted to equivalent engine speeds
o Bullen (SN-73- S/N=1001, and SN 220M- based on the assumption that the blades are
S/N-l-1000) excited by the wake of the 19 stator segments.
This provides a guideline to avoid dwelling at
o Cold engine (SN-73 - S/N-1001) resonant speeds. In addition, blade frequency
response data could be used as a quality check
Description and results of each test meth- to confirm blade shape and uniformity of mass
od are given in the following discussion. distribution. A rotor with consistent blades
192
-22.9 mm
-22.9 man
(-0.gQ IN)
{-0.90 INi I
SN250M
ROTOR
\\ '\ z : 368,n,, Z = 3.68 mm
11.45 INI
+
x
Y AXIS rnm, IINCHES) Y AXIS ram. (INCHESi

667-OTS-3A
Figure 231. Comparison of Inspected Z Sections (heavy lines) to TLs at 3.68 em

(2.45 inch) From Inducer Backface.
has very little blade-to-blade variation in fre- cited with a piezoelectric transducer glued to
quencies. Rotors from the same supplier fab- the blades. Holography results are compared
ricated from the same material and by the to predicted mode shapes in Figures 234, 235,
same process also have very little rotor-to- and 236. The close match between predicted
rotor variations in frequencies, as illustrated and observed mode shapes provided increased
by Figure 233. confidence in the accuracy of the finite ele-
ment model.
Data collected from acoustic ring tests are
given in Table 33. These data indicate that
Bullen Test
rotors of the same material had nearly the
same blade natural frequencies. However, SN
250M blades and SN 220M blades had signi- A test plan was developed to address con-
ficantly different natural frequencies. Since cerns about blade vibration in engines. Two
these materials have nearly identical densities different tests were conducted: a magneto-
the difference in natural frequencies was restrictive shaker (Bullen machine) test (to
attributed to geometry differences associated determine strain-to-failure), and a (cold) en-
with the different shrink factors of the two gine motoring test (to assess strain during
materials in densification. engine operation).
Holography The Bullen machine is a 1600 watt magnet-

orestrictive shaker capable of frequencies up
Holography was used to verify predicted to 40,000 Hz. It was used to find the vibratory
mode shapes. Natural frequencies were ex- strain magnitude at which blades fracture.
193
O PREDICTED
O TEST fACOLISTIC RINDI
19/REV
ISTATOR
COUNTi
..t
5/REV
4/REV
/REV
T I"
2O 40 80 8O 1DO
ROTOR SPEEO KRPM
a1_711..4
Figure 232. Campbell Diagram for SN 250M Rotors Showing 19/Rev Stator Excitation Line.
Table 32. Analytical Results.
Mode 1 Mode 2 Mode 3

Geometry T4 Freq Freq Freq Elements
rpm
Nominal 0 70 15356 21738 34974 Normal
Nominal 0 70 16167 21796 44700 Skewed
SN 250M 0 70 13189 18265 25785 Normal
SN 250M 0 7O 13999 20581 41012 Skewed
SN 250M 0 2500 12649 18010 36850 Skewed
SN 250M 100,000 2500 12998 18359 37008 Skewed
SN 220M 0 70 16941 23079 44958 Skewed
194
I"1--
21
2-2 ¸
3-4-
_E
4.s-
_., 5-6-
CONSISTENCY OF 8LAOE FREQUENCYRESPONSE
250-
ii°
ALTHOUGH BLAOE FREQUENCYRESPONSE OF THE
I003- ROTORS FROM ONE SUPPLIER )PLOTTED AT THE
TOP OF THE CHART) OIFFERS FROM THE 8LADE
x FREQUENCYRESPONSEOF A SECOND SUPPLIER
1002-
{80TTOMJ. THE CONSISTENCY DEMONSTRATES
=. UNIFORM MASS 81STRIBUTION ANO GOOO
1001- OIMENSIQNAL CONTROL8Y OOTH SUPPLIERS
I I I I I I% i I I I ! I
14 16 18 2o 28 zs 3o 31 3,
FREQUENCY,Hz x 1000
G87-078-29]
Figm'e 233. Consistency of Blade Frequency Response.
Table 33. Acoustic Blade Ring Results.
Freq I, Hz Freq 2, Hz Freq 3, Hz Additional

Material sIN (Std Vev) (Std Dev) (Std Dev) Testing
NGK SN-73 1001 13094 17163 26167 Cold Engine test

(271) (210) (529) and Bullen test
NGK SN-81 1002 12826 16884 25625

(3205 (222) (561)
Kyoeera SN 220M 1-1000 16826 19836 31836 Bullen test

(115) (173) (187)
Kyocera SN 220M 0002-1 17144 19230 34298

(108) (353) (222)
Kyoeera SN 220M 0005-6 16192 19009 31250 Zeiss Inspection

(181) (173) (343)
Kyocera SN 250M 252 13438 16923 27009 Holography

(95) (243) (211)
Kyoeera SN 250M 254 13759 17076 27038 Zeiss Inspection

(166) (78) (165)
Kyoeera SN 250M 255 13798 17163 27385

(167) (133) (292)
195
0P_GD!AE PAGE I8
OF POOR QUALITY
PREDICTEDSN 220M
PREDICTEDSN 220M BLADEOISPLACEMENT
BLADEDISPLACEMENT
Hz = 23,079
HZ = 16,941
PREDICTEOSN 250M
BLADEDISPLACEMENT
PREOICTEDSN 250M
BLAOE DISPLACEMENT Hz = 20,581
Rz = 14,000
HOLOGRAPHY
OF AN SN 250M ROTOR8LADE HOLOGRAPHY
OF AN SN 250M ROTOR8LADE
G87.078-5 G67_078-8
Figure 234. Comparison of Holography Figure 235. Comparison of Holography

Results with Predicted First Exducer Results with Predicted First Inducer
Mode. Mode.
196
ORIGI_L_.L PAGZ ?_
OE POOl{ QUAI_IY
In this test, selected blades on the rotor

were stress mapped and instrumented with
cold strain gages at the point of peak stress.
The rotor was rigidly mounted on the

machine, and the frequency band was swept at
PREOICTEO
SN220M
low amplitude to find the blade natural fre-
t// _ BLAOE
OISPLACEMENTS quencies. The amplitude was then increased
at each natural frequency. Four Bullen tests
were conducted on an SN 220M and an SN-73
rotor.
The SN 220M rotor was mounted axially, as

shown in Figure 237, and tested. This orienta-
tion primarily excited the exducer tips. No
blades broke even at the highest input levels.
The rotor was mounted transversely in the
second test, as shown in Figure 238, "co excite
/___ PREOICTEO SN 250M
the inducer blades more efficiently. One
blade broke at the inducer, but it was not one
/ ] _ 8LAOE OISP[.ACEMENTS
of the gaged blades. This test was repeated
and a gaged blade broke at 2280 micro
inches/inch. The NGK SN-73 rotor was
mounted transversely and tested, but blade
failures could not be induced.
Cold Engine Vibration Tests
Following the Bullen test, a similar rotor

(NGK S/N-1001 of SN-73 material) was strain-
HZ = 26.634
VIBRATION
lo,,,c,,o
HOLOGRAPHY OF AN SN 250 ROTOR BLAOE
G87_7_7
GB7_7_276
Figure 236. Comparison of Holography
Results with Predicted Second Exducer Figure 237. Initial Rotor Orientation
Mode. Excited Exducer Tips (Bullen Test).
197
MPa (PSIAI
20.7, 3
MECHANICAL _'_ /
STATOR WAKE
AT SOKRPM
pNOUCERRESONANCE)
STATOR WAKE
A AT4OKRPM
(EXOUCERRESONANCE)
0.9- 1--
DIRECTION
VIBRATION
OF
/ ' / 50 KRPM
-.--'-' 40 KRPM
O- oJ |
d a
, !
j
_.02 0.00 0.62 {INCH
•'.5 o5 M.
007-070-279 CIRCUMFERENTIALSPACING
607.076-200
Figure 238. Rotor Re-oriented to Excite Figure 239. Cold Motoring Test
Inducer Tips (Bullen Test). Analysis Results.
gaged and installed in a metallic-structured The resulting predicted blade stresses are
engine (S/N-001, Build 40). The engine was shown in Table 34. This indicates a low hot
then motored with ambient temperature air operational cumulative stress which combines
through the blade resonant speeds. The blade the vibratory peak-to-peak stress to the
vibration amplitudes were recorded. centrifugal and thermal stresses in the blade.
A turbine stator wake analysis was perfor- At around 43,000 rpm the exducer reso-
med to establish the pressure profile at the nates at its first natural frequency with a
stator discharge for the cold test case and a peak-to-peak vibratory stress of 33.1 MPa (4.8
similar profile for the predicted hot operating ks]). During operation this stress is super-
conditions. On the assumption that the blade imposed on a steady state stress of 13.1 MPa
response is proportional to the pressure wake, (1.9 ksi). In comparison with the 620.5 MPa
the blade resonant amplitudes measured in the (90 ksi) stress, calculated from the measured
engine motoring test were corrected to pre- 2280 micro inches/inch strain, that caused
dicted engine conditions. Result from the failure in the Bullen test, fracture is highly
wake analysisare illustratedin Figure 239. unlikely.
Table 34. Predicted Ceramic Blade Stresses.
Cold Hot Hot

Measured Vibratory Vibratory Cumulative
Strain, Stress, Stress Range, Tensile Stress,
Mode pin/in P-P MPa (ksi P-P) MPa (ksi P-P) MPa (ksi)
Exducer 460 138 (20) 33 (4.8) 30 (4.3)
Inducer 1730 538 (7S) 101 (14.6) 130 (18.8)
198
At about 53,000 rpm the inducer resonates Blade and hub geometries measured from
at its first natural frequency with a peak to the SN 250M rotor, S/N 254, were used to
peak vibratory stress of 101 MPa (14.6 ksi). develop both models. The models thus in-
This is superimposed on a steady state stress cluded the exducer blade "droop" anomaly
of 79.2 MPa (11.5 ksi). Again vibratory failure shown by the "Z-section" measurements
is not expected at this stress level. (reference paragraph 6.2.6, Figure 23 i).
Because this anomaly introduces non-radial
Analysis indicates that the cold condition blade elements, bending stresses are present
is much more severe than the hot operating at speed which would not be present in a rotor
condition with larger driving forces; however, made to blueprint.
this condition can be avoided in engine testing
Thermal boundary conditions for partial
by limiting the speed of the engine without a
flame in the combustor. power and maximum power conditions were
determined from analyses and engine test re-
sults. Analyses were done for maximum
6.2.7 Rotor Stress Analysis
power, steady-state conditions then corrected
according to rotational speed, turbine inlet
AGT101 rotor stresses were predicted for
temperature, compressor discharge temper-
partial power 70,000 rpm and 1204C (2200F)
ature, and mass flow for partial power condi-
T 4, maximum power 100,000 rpm and 1371C
tions. The maximum and partial power ther-
(2500F) T 4, and 115,000 rpm proof test condi-
tions. mal distributions are shown in Figure 241 and
Figure 242 respectively.
6.2.7.1 Model Development 6.2.7.2 Stress Analyses
An initial model without a fillet between Stresses predicted with these models are a
function of applied rotational speeds and ther-
the hub and blade was developed to help guide
mal boundary conditions. Blade pressure loads
engine test plans specifically for the proposed
were assumed minimal and were not evaluat-
100-hour endurance test. A later model with a
ed.
fillet was developed to generate a more com-
plete stress solution for failure risk assess- Stresses from the first model were deter-
ment as discussed in the following paragraph mined using the Garrett finite element pro-
6.2.8. These models are shown in Figure 240. gram IS03DQ. The principal stress distribution
ROTOR MODEL ROTOR MODEL
WITHOUT FILLET _a,-_-"_. WITH FILLET
a,.
GS/-07Ot
Figure 240. AGT101 Rotor Finite Element Stress Models.
199
410C (770F)
471C [880F}
/ 532C (990F)
593C (1100F)
1181C12158F}MAX
649C {1200F)
710C (1310F)
771C 11420F)
TEMPERATURES
IN 827C (1520F)
C IF)
888C (1630F]
949C (1740F]
IOO4C(1840F)
1066C11950F)
[_ 1127C12060F)
710C
1182C12160F)
{1310FI
/ 471C
(880FJ
827C
1066C11950F] 949C (1740F) 11520F}
I 127C{2060F]
GB7-O78-9A
Figure 241. Maximum Power Thermal Distributions, T 4 = 1371C (2500F).
for the room temperature 115,000 rpm proof (paragraph 6.2.7). Life predictions were made
test, maximum power operating condition, and by integrating the risk of fast fracture and
for partial power conditions are shown in Fig- time dependent failure (slow crack growth)
ures 243 and 244. over the area and volume of the AGT101
rotor. These results have helped define the
Stress analysis showed the largest maxi- capabilities of current ceramic probabilistie
mum principal stress location occurred in the methods and have laid the ground work for
blade exducer, where the off-desig n geometry future verification efforts. These results also
induces bending stress. provide an assessment of current ceramic
materials capabilities and their suitability for
6.2.8 Rotor Life Prediction heat engine applications.
Garrett applied probabilistic design Kyocera SN 250M sintered silicon nitride

methods to evaluate the risk of rotor failure material properties were calculated for rotor
during partial power 70,000 rpm and 1204C failure analyses. Fast fracture and slow crack
(2200F) TIT, and maximum power 100,000 rpm growth properties were obtained from bend
and 1371C (2500F) TIT, conditions for which test results. Fast fracture failures were
stress solutions had previously been generated identified with optical microscopy and cate-
200
432C (810F)
485C 1905F)
B 538C IIO00F)
1117C (2043F) MAX
591C (1095F)
643C (1190F)
B 696C (1285F)
TEMPERATURES 749C {1380F)

IN 802C (1475F)
C IF)
854C {1570F)
907C (1665F)
960C (i 760F)
1013C(1855F)
696C
591C 1066C11950F)
11095F]
/ 1118C(2045F)
\
1013C 802C
(1855F) (1475F) 485C
1066C 1905FI
11950F)
GB7_7_IOA
Figure 242. Partial Power Thermal Distributions, T 4 = 1227C (2240F)
gorized as surface or internal. Details about the curves in Figure 245 based on a hypo-
test procedures are given in paragraph 6.1. thetical surface characteristic strength and
Weibull modulus of 621 MPa (90 ksi) and 20
Independent surface and volume strength respectively, and a volume characteristic
properties were required to predict the com- strength and Weibull modulus of 689 MPa (I00
bined probability of surface and internal ksi) and 10. These material properties were
initiated rotor failure. The probability of used to generate independent surface and
survival when these failure modes are concur- internal failure occurrence curves A and B.
rent is the product of the probabilities of When both failure modes occur in the same
survival of each failure mode. population, curve C is observed. The apparent
surface and volume curves D and E were
Independent surface and volume Weibull generated by separately plotting surface and
characteristic strength and modulus cannot be internal failure data used to plot curve C. The
simply determined from test data that con- reason for different independent and apparent
tains both modes of failure. The interdepend- failure distributions can be explained by
ence of combined failure modes tends to mask examining the volume failure distributions. At
their underlying independent failure distribu- 724 MPa (105 ksi), 20 percent of the popula-
tions. This is demonstrated in Figure 245. tion would survive volume failures. When both
Monte Carlo methods were used to generate surface and volume failures are present the
201
MAXIMUM PRINCIPAL
STRESS. MPa IKSli
MAXIMUM BLADE
STRESS = 398 MPa (57.7 KSI)
TEMP = 21C [7OF)
tSOPLETHS = 5 KSI (35 MPa}
\
MAXIMUM HUB
STRESS 301 MPa (43.6 KSII
GBT-OTO-11k TEMP : 21C 17OF)
Figure 243. Principal Stresses for Room Temperature 115,000 RPM Proof Test
Predicted From No-Fillet Model
entire population fails by 689 MPa (100 ksi). Slow crack growth and fast fracture ma-
Therefore, the apparent volume failure dis- terial properties were determined from speci-
tribution is truncated to zero at 689 MPa (100 mens tested under uniaxial stress conditions.
ksi) which is significantly different than the Batdorf's multiaxial fracture criteria was used
independent volume failure distribution. to predict risk of failure in the AGT101 rotor,
Curves D and E demonstrate the distributions which has largely biaxial stresses. The ratio
of data that would be observed from testing of multiaxial risk to uniaxial risk with respect
this material with both failure modes present. to increasing biaxial stress ratio is plotted in
Data for the curves A and B, however, are Figure 246.
needed to determine the independent surface
and volume Weibull strength properties for A multiaxial correction factor, rex, was
risk analysis. selected from these curves according to the
principal stresses at the location being evalu-
Determination of independent surface and ated with the Weibull modulus of the material.
volume strength distributions was accom-
plished through data censoring. Surface and At elevated temperatures the risk of frac-
volume Weibull strength parameters for SN ture increases with time due to slow crack
250M, determined with censoring, are given in growth. Slow crack growth causes cracks to
Table 34. propagate from existing flaws according to the
202
MAXIMUM PRINCIPAL
STRESS. MPz {KSll
MAXIMUM BLADE
STRESS = 290 MPa [42.0 KSI}
TEMP = 1263C (1823FI
\
iSOPLETHS : 5 KSI [35 MPal
MAXIMUM HUB
STRESS = 219 MPa 131.8 KSII
667-07D-12 A TEMP = 1098C (2008F]
Figure 244. Principal Stresses for Maximum Power, Predicted From No-Fillet Model.
slow crack growth power function. The crack TIT, and maximum power I00,000 rpm and
growth rate power function was combined with 1371C (2500F) TIT, are given in Table 35.
the equation for the critical stress intensity to
derive the time to failure equation. The risk of failure in each of the rotor
regions and over the total rotor vo]ume and
The time to failure relationship was com- area for the partial power condition are shown
bined with the Weibull fast fracture equation in Figure 247. SN 250M (as-fabricated) speci-
to calculate the risk of fast fracture at any men slow crack growth data were used for
IzT_e. these analyses. At part power, the hub does
not have the highest stress in the rotor but it
The risk of AGT101 rotor failure was eval- does have significantly higher initial failure
uated with a Garrett developed program that risk than the other regions. This is due to the
performs Gaussian integration of risk over high relative volume and biaxial nature of the
each element from a finite element stress stress. Only the hub exhibits higher internal
solution. The product of all element survival failure risk than surface failure risk, high
probabilities was calculated to determine the enough to initially give the entire rotor a
total survival probability. greater internal failure risk. Later in the
rotor life, surface failure risk exceeds internal
The predicted risk of fast fracture for failure risk, due to the higher temperature at
partial power 70,000 rpm and 1204C (2200F) the rotor surface.
203
k. B = INDEPENDENTSURFACE AND VOLUME FAILURE
OCCURRENCECURVES
C : BIMOOAL CURVE PROOUCEDWHEN A AND B

OCCUR IN THE SAME POPULATION
m
O, E = APPARENT SURFACE AND VOLUME CURVES

FROM THE BIMODAL POPULATION j zo
0.10 i r
I
A' I
I 7
0.08 I
I
|
L O p /lo
0.06 / /
//
0.04
,' ; - //
|
o.02 //
4O 60 8O I00 120 140

- /iJ
STRESS, KSI
G87,078-t
7A STRESS, MPa
Figure 245. Comparison of Concurrent

and Independent Surface and Internal
Failure Distributions.
Similar risk analyses were performed for

maximum power conditions using specimen I I I I /
slow crack growth data. These results are 0.2 0.4 0.6 0.8 1.0
shown in Figure 248. The same trends

o.2/o 1
observed at partial power conditions were seen
here. However, the effect of these higher
stresses and temperatures is an order of mag- Figure 246. Effect of Biaxial
nitude reduction in rotor life. Tension on Fracture Risk According
to BatdortNs Criteria.
An additional set of analyses were run for
the partial power case using slow crack growth When the rotor data was used a 10 percent
properties determined from cut up rotors probability of failure was reached in just one
shown in Figures 249. These life predictions hour (reference Figure 249). The overall life
were significantly different than predictions was thus several orders of magnitude less than
made using specimen slow crack growth data. predicted from specimen data.
Specimen data predicted a 10 percent failure
probability after more than 10,000 hours oper- While test bar data will continue to be the
ation at part power (reference Figure 249). primary indicator and comparator by which
Q
Table 35. Rotor Fast Fracture Predictions.
Percent Failure Probability

Load Condition surface internal total
risk risk risk
Partial Power 0.006 0.006 0.014
Maximum Power 1.42 3.09 4.47
204
........... 7 ,`¸ /" _'_i _
::i:
"o 'TOTAL
I
.m
--IC_ _,,"r_,'LETR,SK I , ,_:,_

'_ .-1 _ll I I I I i I ! I
I_,_LAnE RSt( i i i I ! !I!
! 1 I I ' 'l ' itl
' I I_iIl l i ! i i ; I
;i; i I t....rSHAFT RSK/ ! I , ! !:
I O: 0s S 6 789] O' 2 3 4 5678g'10 2 5 4 56"_.:_; Z,'
TIME. HOURS
GD7@7_20
Figure 247. Rotor Failure Probability for Partial Power Operating Conditions.
*o
II [ III
I ] IIIll ]_ _ J_ _ : J ,<
[
]I.' _, i !IIli',_
TOTA<R'"\I II
-_ ----\
HSK zI ,tlil
:][ _._ -'l _l.!/
$ ib , ;-._., 8LAOERISK,
, i I,_ ' _
--Iz-_ l Ili .-'T z ' :
.... i ..... ?tFF_ I I L:,,,,'!i' I i ; ii_ii_
..a _.
/ I11111 I iTI]'i'ir., I I i_j':lij
E
F=o /. I "_FILLET RISK : i "
L ' z , |1! _ / I J_//I I I , i ':_l
"-J---'#"i''_'_'i - I _ [I bSHAFT RISK_
'O
(
.....itttt i-I
-
TIME HOURS
ira74711-,z2
Figure 248. Rotor Failure Probability for Maximum Power Operating Condition.
advanced materials are measured, the findings material as fabricated in the rotor, is an
presented here give reason for caution. The indication of the importance of verifying
severe reduction in predicted rotor life, as a properties of components, rather than relying
result of the decreased properties of the exclusively on test bar data.
205
ioo
10-1
10-2
10-3
..I s
_ 10"4
,,_ 10-4 10-3 10-2 10-1 100 I03
._" 100
0
O_
0,,,
111-1 HUB
111-2
10"41-..
10-5,
t_ 10-3 10-2 10-1 100 10i 102 103
TIME, HOURS
687-078-24
Figure 249. Rotor Failure Probability for Partial Power Conditions Based
on Properties of SN 250M Rotors.
L._. .... .,,AL PA<;E ]_

.O_ P_OORQUALITY
206
;' .. _ , <, _,i_. _ ,._
6.3 Ceramic Structures which assumes that a given volume of ceramic

under a uniform stress field will fail at the
The AGT101 ceramic hot section consists most severe flaw. To account for the fracture
of 75 static structural and support system statistics of a polyaxial stress distribution, the
components, and a ceramic radial inflow tur- probabilistic method was modified by an ap-
bine rotor, as illustrated in Figure 250. Static proach used by Batdorf 2.
structural components are predominantly
stressed from start transient induced thermal
gradients with pressure loading imposed as a lWeibull, W.A., "Statistical Distribution Func-
comparatively low stress. tion of Wide Applicability," Journal of Applied
Mechanics, Vol. 18, September 1951, pp. 293-
297.
In the initial design phase, the probabilistic
design prediction method developed by
Weibull 1, commonly used to characterize the 2Batdorf, S.B., "Some Approximate Treat-
flaw distributions in ceramics, was used to ments of Fracture Statistics for Polyaxial
establish an allowable stress level. This meth- Tension," International Journal of Fracture,
od is based upon the weakest link theory, Vol 13, No. 1, February 1977, pp. 5-10.
FLOW
SEPARATOR
HOUSIN|
REGENERATOR
SHIELD
TUOIIIIOE TUIIlll STIITOII BOLT ASSEMBLY

COMBUSTOfl OUTER
MCKIHOUO DIFFUSER
ASSEMBLY
HOUSING
TURBINE
ROTOR
TOAHSITIOH
OUCT ]ii!:]]!i]i
&117,478-79
Figure 250. AGT101 Ceramic Components.
207
At the beginning of this program, correla- trol. Design and fabrication of state-of-the-
tion between probabilistic prediction methods art structural ceramic components of the size
(especially for polyaxial stress distributions) and shape required for the AGT101 had not
and the limited test data had not been estab- heretofore been attempted. Therefore, the
lished. As a result, the realistic design cri- design of simple component configurations
teria had to be established by testing--initial- with uniform wall thicknesses were pursued.
ly by extensive tests in engine simulated me- Whenever a deviation from a uniform wall
chanical and thermal screening rigs, and final- thickness was required, a gradual contour
ly, in actual engine conditions. change was used to minimize the stress con-
centration effects. The large thermal expan-
The overall design/development philosophy, sion differences between ceramic and metallic
therefore, has been to define problem areas structures dictated the use of unique and inno-
and initiate design changes based on feedback vative piloting systems. To accurately predict
from component testing. Continuous feedback component stresses for this program, exten-
has been made to the analytical process and sive engine instrumentation and data reduction
material selection throughout all stages of rig were necessary to develop the proper analyti-
and engine testing to arrive at a viable and cal thermal boundary conditions. By creating
optimized component design. a high degree of analytical confidence, a se-
lection criteria was established for various
6.3.1 Component Design/Analysis component configuration/material options.
For example, a component design utilizing
6.3.1.1 Introduction varying wall thickness to reduce excessive
thermal gradients during transient thermal
Development of AGTI01 ceramic compo- conditions could only be accomplished by ac-
nents was initiated during the engine configu- curately knowing the flow path heat transfer
ration conceptual design study phase. Twenty- coefficients (HTCs). Likewise, only by instru-
three different engine concepts were review- menting the various rigs and components could
ed. confidence be placed in the fact that these
rigs were properly simulating engine condi-
Based on experience in earlier ceramic tions and proof testing the ceramic compo-
programs, component symmetry, both in ge- nents to a predetermined stress level.
ometry and loading, was pursued in the basic
design strategy to eliminate circumferential A wide variety of available ceram'ic mate-
stress variations and to reduce the complexity rials was considered in this program. To
of the finite element analysis. This simplifi- eliminate extensive design optimization on
cation resulted in reduced cost, both in man- each material/component combination, a pre-
power and computer demands, and rapid anal- liminary evaluation was performed for each
ysis turnaround that allowed for extensive component to identify primary and alternate
design iterations. The engine symmetry also materials. Subsequently, design optimization
provided the capability to fabricate and rig- was performed using a prime material candi-
test components early in the program, thereby date for each component. After design opti-
allowing test data feedback into the design mization was achieved, alternate candidate
program. The flow separator housing is the materials again were evaluated, at which time
only exception to the concept of symmetry. all acceptable material/component combina-
This component will be discussed in detail in tions were assessed.
the component design section.
Because of the lack of ductility in ceram-
Individual component geometries were de- ics, localized stress concentrations cannot
fined to minimize stresses and to provide ease plastically redistribute stresses. Consequent-
of component fabrication and engine assembly ly, conducting detailed finite element model-
while maintaining the required piloting con- ing and eliminating known stress concentra-
208
tions from the design wherever possible be- M = 5.4 KG/MIN 112 LB/MIN) M = 8.8 KG/MIN I}917 LB/MIN]
comes of paramount importance. Whencom- C (FI
ponent analysis revealed significant tensile 14_-

2600-
stresses near known stress concentrations, 1200

T4' 12'0
additional analyses using a finer substructure I000 : 4.3 KG/MiN 19.6 LR/MIN
model meshwere performed.
Stresseswere analyzed for three modesof
loading enountered during the normal engine 20O
M = 3,8 KG/MIN (5.5 LB/MIN]
operation; one pressure (maximum power) and O'
0 2 4 8 O I0 29 30 4O 5O 6O 70 SO O0 tO0
two thermal (steady-state operation and light-
IIi/-0711-_ WORST CASE COLO START. SECONOS
off transient)• Shutdown transient conditions
were initially considered but were insignifi-
cant due to the thermal inertia effects of the Figure 251. Worst Case Light-Off
Cycle (1282C).
regenerator core.
Detailed discussion of the design and de-

velopment of the ceramic components follow. figuration are shown in Figure 253. A
Specific components are addressed although predicted maximum principal stress of 391.5
overlapping occurs due to the interrelation of MPa (56.8 ksi) for the worst case engine
one component upon the others, for example: lightoff cycle was reduced 54 percent with the
the finite element model to evaluate the tur- scalloped configuration.
bine shroud included the outer diffuser• Re-
sults from these analyses, both for outer dif-
fuser and turbine shroud are addressed in the Both a slotted and a two-piece configura-
turbine shroud section. tion were found to offer substantial further
reductions in stress but to realize this poten-
6.3.1.2 Diffusers tial a loss in piloting capability would have to
be taken• The design solution finally chosen
was found by eliminating the radial piloting
A thermal stress problem was encountered
during light-off conditions on the inner and slots, and instead pilot the diffusers by sleeves
on the rocker assembly (Figure 254). Anti-
outer turbine diffuser housings (Figure 250).
rotation would be maintained by frictional
Rig testing of the diffuser housings resulted in
loading against the diffuser assembly by an
recurring failures originating from the pilot
axially spring loaded mechanism• The outer
slots in the outer diffuser and the pilot holes
in the inner diffuser. These failures were diffuser thus became a completely axisym-
metric structure. A 41-percent reduction in
identified as being thermal fractures primarily
maximum principal stress from the scalloped
caused by the high radial transient thermal
configuration was estimated for this config-
gradient occurring during light-off condition•
uration.
A 2-D plane stress model of a _ Radian

(180 degree) angular sector of the outer dif- The analytical effort on the inner diffuser
fuser was created to evaluate the localized proceeded on a parallel path to outer diffuser
effect of the rim pilot slot. Two light-off analysis. A 2-D finite element model was
cycles were evaluated; the initial cycle was made of the existing inner diffuser, which
the existing worst case engine light-off cycle extended along the flow path from the turbine
(Figure 251) and the latter was for a 871C rotor exit to the flow separator housing. An
(1600F) screening rig cycle (Figure 252). The additional finite element model was made of a
material evaluated for all analyses was RBSN. truncated inner diffuser, scalloped about the
pilot bolts. All analyses were made using
A comparison between the stress results RBSN as the inner diffuser material. Thermal
obtained for the original and a scalloped con- stresses only were considered•
209
1371
(25ool
1
1093.{
12oo01
I
871C (1600F)
) _
260 .._ = 0.076 KG/SEC
(5o01
] (0.167 LBM/SEC}
- 93C 12OOF)
! | ! I | l I I I
(0) 0 80 160 240 320 400 480 560 640

TIME AFTER LIGHTOFF, SECONDS
687-078-391
Figure 252. Rig Cycle 871C (1600F) Static Structures.
12.4 MPa (1.6 KSI] 391.5 MPa {56.8 KSI) 965 MPi ll4.O KSI{ 181.3 MPa (25,3 KSII
• FLAT • FLAT
• NO DIFFUSER DUMP SECTION • NO OIFFUSER DUMP
• SCALLOPED BOLT SLOTS • NO BOLT HOLES
UNSCALLOPED SCALLOPED • REDUCED O.O.
NOTES: 1l MAXIMUM HOLE RAOIUS USEO

2i TEMPERATURE DISTRIBUTIONUSED FROM
ENGINE WORST LIGHT-OFFCASE 115,55KTIT)
({67,07{I.51
Figure 253. Effect of Scalloping Outer

Diffuser. Bl B2
GB7_71-59
The stress analysis results for the initial Figure 254. Outer Diffuser BI and B2
inner diffuser is shown in Figure 255. To Configurations.
simulate 3-D effects on this 2-D model, the
outer section of the inner diffuser was The stress analysis results for the scalloped
modeled as a thick rim and parametrically inner diffuser was done for both the maximum
plotted with principal stresses at three loca- and minimum tolerances on the pilot slot. As
tions. The results show that the effect of the shown in Figure 256 the singular effect of the
increased rim thickness is to reduce stresses tolerance within the pilot slot radius ac-
at the pilot hole. counted for approximately a I0 percent varia-
210
8 HOLE OUTER RADIUS HOLE INNER 8AOIUS 7 ¸
100.6 MPz
6 114.6 KSII
5.
4.
3'
2'
__ M,N,MUM
HOLE
OUTER EDGE
_ RADIUS USED ,
I.
_,._ , - _ , _ ......2 4 6
{7.9 KSII TENSILE
MINIMUM HOLE RAOIUS USED O"
MPz (KSII
165.4, 24- MAXIMUM PILOT RAOIUS

RIM THICKNESS
T
137.8. 20- ._OUTER RAOIUS
7
16-
_ 110.3-
12-
_J_ _ g,.oM_,
_. 82.7"
2.76"
8-
4
R EDGE =:
o ,
0.4 ,
0.8 ,
1.2 ,
1.6 :
2 0 2 :,, 2:s 3:2
RIM THICKNESS. INCH
'
25 '
5O 7'5
RIM THICKNESS, MM {7.9 KSII TENSILE
GO7-O78-53A
MINIMUM PILOT RADIUS

Figure 255. Inner Diffuser Initial
Configuration Stress Analysis Results. GB7-078-54
tion, the higher stress being predicted for the Figure 256. Inner Diffuser Scalloped
larger radius. The higher stress for the larger Configuration Stress Analysis Results.
radius initially appeared contrary to the com-
mon rules of thumb concerning stress concen- final outer diffuser design. The component
trations. This apparent contradiction was was truncated to inside of the pilot bolts to
rationalized by realizing that the smaller fillet form a completely axisymmetric design. Pi-
radius between the sides and the bottom of the loting is maintained by frictional loading from
slots produced a flat at the bottom, where the an axially spring loaded mechanism. A 60-
peak stress occurred. The larger radius pro- percent reduction in maximum principal stress
duces a full radius at the bottom of the slots. has been estimated by the total culmulative
Compared to the initial configuration, the design changes on the inner diffuser.
scalloped configuration was estimated to offer
a 34-percent improvement. The truncated inner diffuser required the
design and fabrication of an additional com-
The design finally chosen for the inner ponent to provide the outer flow path between
diffuser to meet the program goal had a the outer diffuser to the flow separator
configuration similar to that chosen for the housing. The I093C (2000F) gas temperature,
211
lack of mechanical loading, and the low peak 257 consisted of the shroud, outer diffusers,
gas velocity of 15.2 to 22.9 m/sec (50 to 75 the seal area, and portions of the transition
ft/sec) allowed materials other than mono- duct, stators, and flow separator housing.
lithic ceramic components to form this gas Heat transfer coefficients on the turbine
path surface. Babcock arld Wilcox, Inc., shroud and the outer diffuser were modeled to
vacuum formed A120 3 fiber insulation,coated reflect the absence of the turbine rotor.
with Johns-Manville Cerakote at the surfaces
exposed to the hot turbine exhaust gases, was
TURBINE SHROUD TURBINE
selected for this component and is currently SHROUD
being used. OUTER DIFFUSER
FLOW SEPARATOR
6.3.1.3 Turbine Shroud
S HOUSING
The turbine shroud design (Figure 250) was
the most substantial design challenge of all
the ceramic static structural components.
Functionally, this component must maintain
the outer turbine flow path as energy is con-
verted into power through the turbine rotor.
The turbine shroud design, therefore, must

survive the thermal transient stresses that are
STATOR
induced as local boundary condition change SUPPORT
along the meridional flow path. These non- REGION
uniform boundary conditions result in high

thermal gradients and polyaxial stresses within
the shroud, predominantly during engine light-
off conditions. The greatest influence on the
thermal gradients is the variation in heat
697@7i_55A
transfer coefficients. Although not of primary
concern, it also withstands a significant pres- Figure 257. Turbine Shroud/Outer Diffuser
sure load at maximum power conditions. Ad- Critical Stress Areas (Static Structures Rig
ditionally, stress concentrations are imposed Model).
due to the discontinuities required for turbine
stator vane piloting and pressure sealing. The Early optimization of the turbine shroud
classical solution to the reduction of high configuration used the static structure rig
thermal stresses, the multiple piece approach, model. High thermal stresses during light-off
was not feasible in the AGT101 application were found at the junction of the shroud outer
due to the performance requirement for close cone with the stator support (Figure 257). The
rotor clearance control and static sealing stresses in this area were from high hoop
complications.
stresses coupled with axial-radial bending.
The primary cause was attributable to the
Two-dimensional finite element programs
large variation in heat transfer coefficients
were the primary tools used to design the
along the shroud structure. High heat transfer
turbine element. Three primary models were
coefficients are generated at the turbine
created to represent the static structure rig,
section from the high velocity core flow,
the thermal screening rig, and the engine.
whereas, low heat transfer coefficients are
generated at the diffuser exit.
6.3.1.3.1 Static Structure Rig Shroud
Analyses
Various eonfigurations, as shown in Figure
The static structure rig model used to 258, were investigated and the results are
analyze the turbine shroud, as shown in Figure shown in Table 36. All stresses are for a light-
212
r--
The following results were found from this
SCALLOPED
investigation:
o The Type B3 configuration had the lowest

predicted stress 54.4 MPa (7.9 ksi) for a
28-percent reduction from the baseline
configuration
TYPE A [BASELiNEi
o Compared to the baseline, the B3 configu-
ration increased the time to critical stress
from 175 to 240 seconds, whereas the B4
decreased the time to critical stress from
175 to 120 seconds
o Types B1 and B4 moved the location of

_ _ VAH,AHLE
/-_ critical stress from the seal area (location
A) to behind the stator pocket (location B)
TYPE B1 TYPE B2
o A material change from RBSN to SASC

resulted in a 76 percent higher stress
UNIFORM As a result of this study, a design change

WALL was made on the turbine shroud to reflect the
THICKNESS
Type B3 configuration.
UNIFORM
WALL The next area to be considered was the
THICHNESS
effect of various light-off and operating cy-
TYPE 63 TYPE 84 cles on the shroud stresses in the static struc-
UT,,OTI,,54 ture rig. Six different cycles were considered.
Three of these cycles were light-off cycles to
I149C (2100F), all at idle mass flow conditions
but at different transient rates as shown in
Figure 258. Phase B Turbine Shroud Figure 259. In addition, three other operating
Configurations. cycles were considered, as shown in Figure
260. These included a light-off to steady-
state conditions (cycle 2), a light-off followed
off transient to 871C (1600F) as shown in
by a shutdown after a 1-minute hold at
Figure 252. The comparisons were made
maximum temperature (cycle 3), and a
assuming RBSN as the primary shroud mate-
shutdown from steady-state conditions (cycle
rial. The heat transfer coefficients were
6). Again mass flow was at idle condition
based upon aerodynamic analyses.
during the cycle until the rig was shutdown,
then mass flow was reduced 50 percent.
The Type B1 thru B4 configurations had a
scalloped flange whereas the baseline config- Table 37 shows the results for the effect of
uraton (Type A) did not. In addition, all Type cycle variations. Both the turbine shroud and
B configurations attempted to increase the outer diffuser were assumed to be RBSN. The
structural flexibility to relieve the thermal heat transfer coefficients were based on a
growth mismatch between the outer and inner first set of thermocouple data from a metal
sections of the turbine shroud during transient engine test.
light-off conditions. The Type B4 was a two-
piece configuration with the interface at the The screening rig cycle was selected after
stator support. this study to include a long hold period
213
Table 36. Phase B Shroud Configuration Study.
Shroud Maximum Principal Stress (MPS)
Time
After
Case Magnitude Start,
NO. Type Material Location* MPa (ksi) sec
1 A RBSN A 75.8 (11.0) 175

2 B1 RBSN B 73.1 (1o.6) 240
3 B2 RBSN A 57.9 (8.4) 240
4 B3 RBSN A 54.4 (7.9) 240
5 B4 RBSN B 59.3 (8.6) 120
6 B3 SASC A 95.8 (13.9) 240
B1 Configuration Outer Diffuser (RBSN)
Analytical heat transfer coefficients (HTCs) used
Static Structure Rig Model
*For locations see Figure 257.
IF X 402i
25
42- 4149C 12100F)
40-
15
4-
2-
I_ = 4.5 KG/MIN (40 LR/MINI

O- 0 93C {20OFI
{2OOFi z I
I I ! ! I
0 I0 20 3O 40 50 60 70 80
T4ME AFTER LIGHT-OFF,SECONDS

GBI_7&57
Figure 259. Static Structure Rig Light-off Cycle.
214
C X 102 IF x IO 2)
]4 25
CONOITION A = 4.5 KG/MIN {10 LB/MIN)

CONOITION B = 2.3 KG/MIN [ 5 LB/MIN]
CYCLE 2 1149C (2100FI
20-
I0-
8- 15-
f4 = CONDITION B
6-
10-
4-
5-
2- S/D
_ 93c 1200Fi
O- 0
I I ! I I I I
0 80 | 60 240 320 400 480 560 640
TIME AFTER LIGHT-OFF. SECONOS
687..076-50
Figure 260. Static Structure Rig Operating Cycle.
(6 minutes) to ensure that the maximum prin- the absence of flow around the shroud flange
cipal stress was realized during lightoff. In and cone. The screening rig also had been
addition, the cycle ensured that the cool-down used to screen turbine shrouds initially without
was less severe than the lightoff. The poten- the use of stators and a backshroud. After
tial for high cool-down stresses occur only in passing this initial test, the stator assembly
the rigs since the regenerator system on the and backshroud were added to the assembly
engine negates the possibility of significant and the turbine shroud was retested. Conse-
cool-down rates. quently, the initial turbine shroud screening
rig test did not include the stator exit swirl
6.3.1.3.2 Shroud Screeninff Rift Analyses flow effect encountered in the latter test.
In the fall of 1983 turbine shroud failures

in the screening rig were noted that were Rig testing in January 1984 with a metallic
inconsistent with analysis. Fractography re- turbine shroud instrumented with thermo-!
vealed that the failure origins began in the couples allowed for empirical evaluation of
seal radius rather than radially inboard on the heat transfer coefficients. As expected,
flow path surface predicted by analysis higher heat transfer coefficients occurred in
(Areas A2 and C respectively in Figure 261). the shroud region aft and inboard of the stator
As a result, investigation was directed toward exit when the stator assembly and backshroud
validating the rig heat transfer coefficients was added to the screening rig. But after the
and refining the finite element model. heat transfer coefficients were revised in the
analysis, the results still were unable to con-
The shroud screening rig is similar to the firm the location of the failure origin with
static structures rig as concerns the absence the existing model. Model refinements were
of a turbine rotor but different in regard to necessary to confirm the failure origin.
215
0 0 _ 0 0 0
• , . w
v
v v v v v
• _ _ _ _
0
J
_3
A
A A A _
c/3
v v V v
0
o_
111
0 _
w
III
A
el
V
? ? ? ? o 0
o
0 .__
E_
o
216 tl.
•--4--_ m
m Concurrent with this effort, a revised rig

test was developed. To more accurately simu-
late engine conditions all subsequent turbine
shroud screening tests were conducted with
the backshroud and stator assembly. Insula-
tion was modified within the rig housing to
improve radiation cross connections between
the shroud cone and flange. The screening
cycle was modified to subject the turbine
shroud in the rig to 125 percent of the maxi-
mum predicted principal stress in the engine.
Testing was resumed and the Type B3 shroud
survived.
Table 39 shows the effect of different

ceramic component materials. RBSN, SASC,
and SSN were considered for the turbine
shroud and SASC and SSN were considered for
the backshroud. Following are the results:
o The lowest shroud maximum principal

stress (154.4 MPa) was predicted using an
607-078-(i0 RBSN turbine shroud.
o SSN turbine shroud was next at 86 percent
Figure 261. Turbine Shroud Critieal Stress higher than RBSN.
Areas (Shroud Screening Rig Model).
o SASC turbine shroud was highest at a max-
Since the shroud failures occurred in a rig imum principal stress 145 percent higher
test without the stator assembly and back- than RBSN.
shroud the finite element model was likewise o A SSN backshroud had a maximum princi-
configured. Table 38 shows the results of this pal stress 23 percent higher than a SASC
study. The model light-off (cycle 13, shown in backshroud.
Figure 262) corresponded to the test cycle
that caused shroud failures. Heat transfer 6.3.1.3.3 Shroud Engine Analyses
coefficients and shroud material (RBSN) also
reflected test conditions. Two levels of model The engine model for the turbine shroud
refinements were made, the second is shown in structurally is the same as the static structure
Figure 263. In addition, one design model except the boundary conditions are
modification was considered. The shroud seal modified to include the effect of the turbine
area was to be moved toward the combustor rotor on the heat transfer coefficients.
6.1 mm (0.240 inch) to increase the neck cross
section adjacent to the shroud seal. The The initial use of this model was to eval-
results were as follows: uate various thermal cycles. Table 40 shows
three cycles considered. The first analysis
o The analyses with model refinements pre-
considered the same cycle (cycle 5), used in
dicted the shroud failure origin. Maximum
the static structure rig. Another considered
principal stress increased a total of 68
cycle 8 shown in Figure 264. This cycle
percent (182 percent at location A2)
corresponds to a normal engine lightoff in an
o The configuration change in the shroud seal operational environment. Cycle 11 shown in
area resulted in a 31-percent reduction in Figure 265 was considered to assess the effect
maximum principal stress. of a slow ramp instead of a step increase to
217
Table 38. Turbine Shroud Seal Radius Study.
Maximum Principal Stress (MPS)

Time
After
Case Magnitude Start,
NO. Type Material Location* MPa (ksi) $eo Case Description
14 B3 RBSN C 108.9 (15.8) 6O0 Baseline Model
15 B3 RBSN A2 164.? (23.9) 660 First model refinement

in seal
16 B3 RBSN A2 182.6 (26.5) 660 Second model refinement

in seal
1? B3 RBSN A1 126.1 (18.3) 660 Seal land decreased 6.10 mm

(0.240 inches)
Shroud Only Screening Rig Model

First Revision HTCs Used (Rig Condition)
Cycle 13 Light-off
*For locations see Figure 261.
C X 102 IF X 1021
14=
4 MIN
12- 1149C 1210OF)_
I0-
6 MIN -- 871C
18OOFI
0-
649C
1120OF)--
4 g
: 0.075 KG/SEC 10.167 LB/SECl

21
¢1C 1200FI
0 I
1 1 I 1 1
4 6 8 I0 12 14 16
TIME AFTER LIGHT-OFF,MINUTES
GOT-070-61
Figure 262. Screening Rig Cycle 13.
218
The significance of the last item was that
the maximum stresses could be controlled to
occur during the initial light-off where the
environment is more fully characterized and
the remainder of the cycle would function at a
lower stress level. The first ceramic engine
light-off (engine 002C/Build i) had a similar
cycle to cycle ii except stresses were con-
trolled to a 25-percent lower stress level by
decreasing the initiallight-off step and ramp
rate.
Concerns about the viabilityof the existing

type B3 shroud led to a Phase C turbine shroud
redesign effort. The engine model was up-
G67_78,62A
dated to reflect the refined stresses at the
shroud seal radius and updated heat transfer
Figure 263. Second Model Refinement to coefficients (Figure 268). The outer diffuser
Shroud Fillet Radius. was changed from a type B2 to A3 configura-
tion as shown in Figure 269. Stresses were
operating temperatures. The first engine test calculated for an RBSN turbine shroud and
with all of the ceramic structural components outer diffuser for a normal engine lightoff. In
would use a similar type light-off. The addition, the radiation from the shroud outer
analyses were all done assuming an RBSN cone to rig insulation was considered as shown
turbine shroud and heat transfer coefficients in Figure 270.
(first revision) generated from metal engine
S/N 002/Build 4 test data. The shroud model The various shroud configurations consid-
in these analyses was based on the original ered are shown in Figure 271. The primary
model without the refined seal radius. Criti- objective in the C type configuration was to
cal stress locations for the engine model are increase the heat transfer coefficients in the
shown in Figure 266. The results have been outer portion of the shroud from the seal area
summarized as follows: to the flange. This was to be achieved by
reducing the hydraulic diameter of the flow
The effect of the turbine rotor causes the over the shroud by increasing the diameter of
turbine shroud maximum principal stresses the shroud outer section and decreasing the
in the engine to be 25-percent higher than diameter of the opposing outer duct wall. The
those in the static structure rig for the type D configuration was an attempt to
same light-offcycle. decrease the bending stresses at the seal area
by providing a more direct path from the
The predicted maximum tensile stress for stator support to the flange. All of the
normal engine light-off was 115.8 MPa various cases considered are shown in Table
(16.8 ksi) based on the first revision heat 41. The significantresults are as follows:
transfer coefficients.
o The finite element model refinements in
O The maximum principal stresses predicted the shroud seal area was less significant in
for a ramp type engine cycle occurred two the engine model (increasing maximum
minutes after light-off resulting from the principal stress 36 percent) than in the
initial step change in gas path tempera- thermal screening rig model.
ture, as shown in Figure 267. Maximum
principal stresses during the ramp heatup o Increasing the heat transfer coefficients on
and cooldown were marginally lower. the cone area of the type B3 configuration
219
E-_ m
+°°°
,i,,q
L.
,eu
e-I
o_
_u
E-
"_o
L, ,_J;
m_ 0 0'_
°+1i_ _a
0
m._ _e_ _
220
r13
'0
r;3
.o
m
U3 0
m
r_
E-
E_
0 0 0
! I I '_ .._
_gg
f.
o
_.o
,It
221
1371 _
12_oo)] 222_F
_ ,ogaJ/
" 65,000 RP
=_ 8z6.,J /
,,500, ,,ooo,!/
538.
/
/'.
i
_ (1000) __ / M(0:205125
LBM/SEC)
KG/SEC M = 0.095 KG/SEC
260 / (0.210 LBM/SEC)
II
(OJ0 10
I 15
I 20
| 25
I 30
! 35
!
4O
TIME AFTER LIGHTOFF, SECONDS
GB7-078-390
Figure 264. Normal Engine Light-Off, (Cycle 8).
CX 102 IFX 1021
14 25
3 MIN
12
20-
10-
-_ 6- 15-
z.--
8- 10 MIN
M'J
5-
52C 1125F)
O
0 I I I ] ! I !
0 10 20 30 40 50 60 70 8O
TIME AFTERLIGHT-OFF.
MINUTES
6DT_T&70
Figure 265. Ramp Type Engine Cycle 11.
222
reduced maximum principal stresses by 63
Q TURBINE SHROUD TURBINE
SHROUD percent (without radiation effects from the
/_ OUTER DIFFUSER shroud to the rig insulation considered).
FLOW SEPARATOR
o The type C3 configuration shroud had the
HOUSING lowest predicted maximum principal
stresses of all configurations considered
(29 percent reduction).
o Changing from the first revision to the

second revision set of heat transfer coeffi-
cients resulted in an 18-percent reduction
in MPS.
STATOR o Changing the model boundary conditions to

SUPPORT consider the effect of radiation to the rig
REGION
insulation resulted in an l1-percent reduc-
tion.
o Modeling radiation to the rig insulation

reduced the sensitivity of the MPS from
the HTCs over the shroud cone area.
GB7-O76-55A
o The outer diffuser MPS decreased 15 per-

Figure 266. Turbine Shroud/Outer Diffuser cent by changing from a type B2 to A3
Critical Stress Areas (Engine Model). configuration.
MPa (KSII
I0-
I i I I I I i
B
= LOCATION A (NOOE 1601 = "1 = LOCATION 0 INOOE 403]
• LOCATION 9 (NODE 1281-- DA = LOCATION E INODE 891
• LOCATION C [NODE 541)=

.._...--
_-x = LOCATION F (NODE 419]
40- 'Ill
_1_:35.1 MPa 15.1 KSll
,,,j:
20-
I
J%
/
O--
0 10 20 30 40 50 60 70 80
TIME AFTER LIGHTOFF, MINUTES
6D7_7i1-72
Figure 267. Turbine Shroud Stress Response for Ramp Type Cycle (Engine Cycle 11).
223
D]_ _.'00[{QUALITY
UNITS FOR HTC = WATTS/M2C [BTU/HR-FT2-F}
T4 = 671C IIGQOFI
= 75.5 G/SEC (0.167 LBM/SECI
l_l
131 (23l
17 (5]
17 (31
125 1221
toe(35i 125
199 [35i
148
227
358
482
392
551 i971
267 1471
267 1471
!
312 (55J 306 154l
446 IT91 330 (671
SECOND REVISION
FIRST REVISION
GDT-O7t74A
Figure 268. Engine Heat Transfer Coefficients on Turbine Shroud/Outer Diffuser Model.
, FLAT
o NO DIFFUSER DUMP
. NO DOLT HOLES
, REDUCEDO0
o ORIGINAL SHAPE _ "_

, NO DOLT SLOTS _',j,
,, REDUCED00 TO
DIFFUSER OUMP
62 REGION _A3 ,. _w
GOT_7&75
PREVIOUS CURRENT
GIlT 071i.?ll
Figure 270. Shroud Cone to Insulation

Figure 269. Design Change From Type B2
to Type A3 Outer Diffuser. Radiation.
224
o The RBSN turbine shroud had the lowest
MPS (115.1 MPa) of the three materials
considered.
o The SSN turbine shroud had the next lowest

MPS (104 percent higher than RBSN)
o The SASC turbine shroud had the highest

MPS (188 percent higher than RBSN).
o The outer diffuser MPS was affected by

the choice of shroud material. The MPS
was 38 and 46 percent higher for an SSN
and SASC turbine shroud respectively.
Development engine testing normally oper-

ates with the inlet guide vanes (IGV) fully open
C3 as opposed to being partially closed in a nor-
mal engine lightoff. To achieve the same
stress level as in the normal engine lightoff,
light-offs must be made to lower temperatures
to account for the higher core flow. The
light-off cycle to 982C (1800F) shown in
Figure 272 has been determined to be the
development engine light-off to simulate the
normal engine lightoff. Figure 273 shows the
01
MPS response of the turbine shroud for the
development engine lightoff. For comparative
purposes the MPS response from the screening
G87_7_77
rig and normal engine lightoffs are also shown.
Figure 271. Shroud Configurations

6.3.1.4 Transition Duct/Combustor Baffle/
Investigated in Phase C Study. Turbine Backshroud
o Considering radiation to the rig insulation

The transition duct, combustor baffle, and
and the second revision of heat transfer
the turbine backshroud (reference Figure 250)
coefficients resulted in a total reduction of are addressed collectively in this section.
12 percent in the MPS on the outer dif- Since they are closely interrelated, they were
fuser. analyzed in a common finite element model.
o The change from a type B3 to C3 turbine The transition duct provides the outer flow
shroud results in a 36 percent reduction in path between the combustor and turbine
the outer diffuser MPS. shroud. It axially supports the combustor
baffle with a standoff midway along tkge com-
Various materials were also considered for ponent. Three holes are provided to monitor
the turbine shroud and outer diffuser and are the turbine inlet temperature. Mechanically,
presented in Table 42. The analyses con- the component must carry a radially inward
sidered the latest refinements in the model pressure load and an axially compressive load
and boundary conditions and represented the due to engine pressure loading. Thermally, the
thermal conditions for a normal engine light- component is exposed to both a through-the-
off. The current type B3 turbine shroud and wall and axial thermal gradients. Pressure
type A3 outer diffuser was used in the anal- loading is normally negligible, compared to
yses. The results are as follows: thermal loading.
225
m
m
0
_J _J _J _J _J
=_°omo
_J
226
m _
° ,+ ,< .,_ ,<
I=
+_i,i
p,
o_M
"F_
_ _I. el.
=E
el .,_ .< .<
L •
_+i °-
_._ ,_ ,o
0
0
.M.
227
g82
_'- 782 j
60.000 RPM
I
w.
382 43000. RPM _ -- 0.129 KG/SEC
(720)
i
182 J M : 0.065 KG/SEC
0 21 C
,(70FI , i , , , , ,
(0) 0 60 120 180 240 300 360 420 480
TIME AFTER LIGHTOFF. SECONDS
GB7-078-392
Figure 272. Development Engine Lightoff (Cycle 17).
The combustor baffle diverts the flow from backshroud is significantly influenced by the
the can combustor into an annulus flow circu- turbine rotor. The heat transfer coefficients
lation for the turbine section. It is axially along the backshroud reflect a complex inter-
contained between the transition duct and action between the rotor speed, blade back-
turbine backshroud. The baffle utilizes three face clearance, hub backface clearance, and
struts to interface with the transition duct the resulting hub backface flow distribution.
through the flow path. Mechanically, the Additionally, the large mass of the rotor hub
component must carry a small pressure drop provides a radiation heat sink during light-off
across the component and a spring load. transient conditions.
Normally, mechanical loads are negligible.
The thermal environment consists of radiant A 2-D finite element model as shown in
heat from the combustor and convective heat Figure 274 was created to evaluate transient
transfer along the flow path, peaking at the thermal stresses. The model was initially used
stagnation point on the centerline. to evaluate component stresses in the static
structure rig. Two transient cycles were
The turbine backshroud provides the flow analyzed, the light-off (cycle 9) and the shut-
path boundary at the axial backface of the down (cycle 10) shown in Figures 275 and 276
radial turbine rotor. The backshroud has a respectively. Two different materials, RBSN
circular disk geometry with concentricity to and SASC, were considered for the combustor
the engine centerline provided by a pilot baffle material. The cases considered and
diameter at the turbine stators. The pressure their results for the combustor baffle and
stresses acting on this component are negli- transition duct are shown in Table 43.
gible when compared with the thermal High stresses were found on the combustor
stresses. The temperature distribution in the baffle struts during shutdown. To reduce
228
LOCAT,ON
A, (NODE
_ll
MPz [KSll
°o ,NODE
T.j ----.__to_,o. A_
30 • : LOCAT,DN
C\
NOOE 9331 _ _, ._ COMOUSTOR BAFFLE
24
154.4 (22.4l-
fm
_ _ _ _ _ _ _ • j _, TRANSITION OUCT Z'_
I_] TURBINE OACKSHROUO

_ __© 140'] 18
_ 1111
1Z
o//
80 180 240 320 400 480 580 64O
TIME AFTER LIGHTOFF. SECONDS
LOCATION A LOCATION B LOCATION O
MPI iKSl] • = (NOOE 8211 _."_INOOE 1111 -_ (NODE 454i I
._ _,..o _ ---- . . --
,oo 1B
•=i_= 8o 1o -
i 2o 5/ GI/17I.II4
Figure 274. Transition Duct/Combustor

0 80 160 240 320 400 480 560 840 Baffle/Turbine Backshroud Critical Stress
TIME AFi'ER LIGRTOFF. SECONOS
Areas.
MPI IKSll LOCATION k LOCATION B LOCATION 0
° : [NOOE 8251 "_,_.INOOE 1111 _ : INOOE 4,.ql I respectively for an SASC combustor baffle
it
and transition duct during light-off.
_ -p An RBSN combustor baffle has a MPS 34

percent lower than a SASC combustor
baffle.
I _J
The MPS is higher during shutdown (cycle

10) than during light-off (cycle 9) for the
combustor baffle.
0 80 180 240 320 400 480 580 640
6074t17! TIME AFTER LIGHTOFF. SECONOS
The MPS on the combustor baffle is highest
on the struts during both light-off and
shutdown.
Figure 273. RBSN Shroud Stress for
Various Stages of Development. o The MPS occurs more rapidly during shut-
down than light-off on both the transition
these stresses, two alternate design configura- duct and combustor baffle.
tions for the combustor baffle were consid-
ered. The first had struts with the chord o The combustor baffle MPS during shutdown
reduced 33 percent and the second had struts could be reduced 22 percent by decreasing
with wall thickness doubled. The results are the strut chord by 33 percent.
summarized as follows:
o The combustor baffle MPS during shutdown
o Maximum principal stresses (MPS) of 100.1 could be reduced 33 percent by doubling
and 27.6 MPa (14.6 and 4.0 ksi) were found the strut thickness.
229
C X 102 IF X 1021
14 - 25
12-
1093C [2000Fi
20-
10-
8- 15"
__ 8-
10-
4-
5- : 0.075 KG/SEC 10.167 LB/SEC)

2u
--93C (2OOFI
0 I
! I I J I I I
5 10 15 2O 25 30 35 4O
TiME AFTER LIGHT-OFF,SECONDS
GDT.07&115
Figure 275. Static Structure Rig Light-off Cycle 9.
cxlo 2 IF X 1021
14-
12-
20 C (2OOOF],STEADY-STATE
I0-
25_093
8-
649C (1200FI
__. 5-
I0.
=__
4 m
5- = 0.075 KG/SEC [0.187 LB/SEC)

2_
I l I I I I i
0 5 10 15 20 25 30 35 40
TIME AFTER LIGHT-OFF.SECONDS

GO7_7&II8
Figure 276. Static Structure Rig Cool-down Cycle 10.
230
c_
0 0 0 0 0
o_
0
I¢1
231
Changes to the combustor baffle strut off cycle to 1149C (2100F) maximum power
design were not made since shutdown condi- condition (cycle 15), as shown in Figure 278 to
tions during engine operation have been made evaluate rotor backflow effect with engine
much less severe than the shutdown cycle speed. Following is a summary of the signif-
analyzed. To ensure against failures in the icant results:
static structure rig, the shutdown cycle was
made more gradual. The combustor baffle o High turbine backshroud MPS were found in
alternate design configurations were kept as all cases with the rotor modeled.
backup designs.
The consideration of turbine bore cooling
The transition duct/combustor baffle/tur- significantly increased the MPS (66 to 70
bine backshroud model was then modified to percent increase) on the turbine back-
reflect engine conditions. The primary differ- shroud. MPS occurs after engine speed
ence consists of including radiative and reaches I00,000 rpm.
convective heat transfer to the turbine rotor.
The turbine rotor component has a large O The effect of rotor backflow circulation
thermal inertia and acts as either a heat sink reduced the MPS on the turbine back-
or source during transient operating condi- shroud.
tions, in addition for the 1149C (2100F)
development engine with the dual alloy O The effect of mixing in rotor backflow
metallic turbine wheel, turbine bore cooling circulation increased the MPS on the tur-
would impinge upon the backshroud. bine backshroud.
Table 44 shows the results of a study on o The effect of the rotor on the transition
the dual alloy rotor effect on the transition duct and combustor baffle stresses is
duct, combustor baffle, and turbine back- insignificant.
shroud stresses. For MPS locations refer to
Figure 274. The study considered effects with o The MPS on the transition duct does not
and without turbine bore cooling, backflow occur during the initial light-off, but after
circulation between the rotor and backshroud, rotor speed reaches I00,000 rpm.
and mixing of the circulation. Midway through
the study the analysis was updated to include Table 45 shows the results for various
the change from an A1 to A2 type combustor component materials considered for the all
baffle (Figure 277) and an SSN to SASC ceramic engine with an SSN turbine rotor.
turbine backshroud. All results are for a light- The analyses assumed rotor backflow circula-
tion with mixing, but no rotor bore cooling.
For comparative purposes the operating cycle
(cycle 15) was the same as for the previous
engine analyses. The results for these anal-
_THICK
LOCALLY < '_ yses are as follows:
THINNEO--'-4"
O The transition duct MPS is almost indepen-
dent of the choice of materials for the
combustor baffle or turbine backshroud.
- _ WALL
WALL/__
O An RBSN combustor baffle has the lowest

kl A2 MPS followed by SSN and SASC with the
GII?.O/8.11
?k highest.
O The turbine backshroud MPS decreases 6

Figure 277. A1 and A2 Configurations percent when an SSN combustor baffle is
Combustor Baffle. substituted for an RBSN combustor baffle.
232
_'_ o,1 _1 _ _1
E-.<
0
.m
Q;
.,_ .,_ .< .< .<
:E •D _D 0 0 0
E-.<
a
!-
0
S 0
< < < ,,< ,<

0
:E
r-
.<
0
E-.<
e. °_
o
iT. UO_
_J
0
c-
O
o
._ 0 _
ZI
0,._
m_
a
o _ o o
Z Z _ 0
r_
_J
_
.!
_J
233
cxlo z IeXtOZl
1149C [2100Fi
21, 100
11.5-
M = 0,37 KG/SEC 10.811 LD/SEC)

16.8- -8O
9.16-
60 KRPM
-60
i 6,8?.- 12.6-
-o.
v,
KDPM
E 8.4- .40
4.49"
M = 0,12 KG/SEC (0.294 LD/SEC)
216- 4.2- .2O

L M = 0.66 KG/SEC 10.150 LO/SEC)
-- 21C I70F) 0
Oo O I I I I I I I
O 60 129 180 240 300 360 420 480

TiME AFTER LIGHT-OFF.SECONDS
G|T.OTli.11|A
Figure 278. Engine Cycle 15 (Light-Off to 1149C Max Power).
Fracture of a ceramic turbine backshroud o An A2 type turbine backshroud has a MPS

during a structures rig qualification test in 82-percent lower than for the original (A1)
August 1984 led to a design change which configuration.
incorporates a hole in the center of the tur-
bine backshroud. In this way, the highly o A B1 type turbine backshroud has a MPS
stressed area of the part was, in fact, physi- 23-percent lower than for the original (A1)
cally eliminated from the design. A piece of configuration.
insulation was installed between the back-
shroud and the baffle to avoid thermally o A SASC type B1 turbine backshroud has a
stressing the combustor baffle. This is MPS 8-percent lower than a SSN back-
particularly significant in the 1149C (2100F) shroud. The corresponding MPS in the
configuration engine, where bore cooling is combustor baffle increases 7 percent.
used to cool the dual-alloy rotor.
6.3.1.5 Flow Separator Housing
The results for an analysis of this config-
ured backshroud are shown in Table 46 along The flow separator housing (Reference Fig-
with the original and another proposed con- ure 250) is the largest asymmetrical ceramic
figured backshroud. The original, the trun- component in the AGTI01. This housing pro-
cated center configuration, and the recontour- vides the flow path split between the high
ed turbine backshroud have been designated as pressure and low pressure portions of the
the AI, A2, and B1 type as shown in Figure engine to accommodate the regenerator waste
279 respectively. The contoured backshroud heat recovery system. The low pressure (gas
was analyzed for both an SSN and SASC side) of the flow separator housing is rapidly
material. All analyses are for the engine with heated after lightoff while the high pressure
an SSN turbine rotor. The significant results (air side) heating lags due to the regenerator
are: core thermal mass. Pressure loading of this
234
0
_A
0 O
0 _ 0
¢1
_-' III
e_
C
0
°_
a
0
II
!°
0
-o_E._ e_o,._ _u_ _'-c _

• _._ _
II.
235
,,,q
m. <
_ _ n
_ _ |
a
ii!
236
UNIFORM CENTER THINNED
THICKNESS HOLE SECTION
_Y AI A2
6117-0711-1
Ilia /
Figure 279. Alternate Configurations
of the Turbine Backshroud.
housing further complicates the stress field as

the high pressure (air side) is asymmetrically
loaded at higher engine pressure ratios. Due
to the complex shape and the locally varying
thermal and pressure environment, 3-D finite
element analysis were utilized to evaluate
component
didates were
deflections along sealing surfaces
and associated stresses.
evaluated
Two material can-
for this component,
!
LAS and RBSN, although LAS was the only
material from which actual components were
made. STRESS VALUES AT SIMILAR HOUSINGLOCATIONS
LOCATION STRESS MPo IKSI]

Strain gage testing of a LAS flow separator
GARRETT FORO
housing was done in May 1981 at Corning Glass AT 366.1 kPa (53.1 PSI)
Works. Fourteen uniaxial and seventeen tri-
(_ 1.5 (2.31 13.8 12.01
axial strain gages were attached to S/N-7
(_ 13.1 li 91 15.2-172 12.2-2.51
housing in critical stress areas determined
{_ 6.3-13.8 II .0-2.0l 3.r.7.9 10.45-1.15)
from earlier 3-D finite element analyses.
15,2 12.2l 17.8-19,3 12.55-2.81
Strain gage data were taken up to 414 kPa (60 11.71 2.8 10.41
(_) 11.7
psi) in increments of 27.6 kPa (4 psi) both in
M74711- I L'g
air and water environments in the flow separ-
ator housing mechanical screening rig.
Figure 280. Flow Separator Housing
Figure 280 shows the stress values at Strain Gage Test Correlation.
similar housing locations for a comparison
between the stresses attained during the strain physical characteristics of the test and analyt-
gage tests and the earlier 3-D stress analysis. ical model differ, and secondly, the stress
In both cases, high stresses occur on the pres- locations may not be exactly coincident.
sure half of the housing. The three notable
similarities are along the outer rim (location Due to the considerable difference in
2), in the center of the inner side of the rim material properties, the RBSN component de-
(location 1), and along the fillet radius at the sign was sized to have considerably thinner
corner of the HP slot (location 4). There are walls to take advantage of its higher strength."
two reasons for the discrepancies. First, the
237
A 3-D finite element model was created of gradients were being created in the housing
the thin wall configuration and stresses and from the heat sink effect of the outer flange
deflections were determined. The model was being exposed to compressor discharge air. To
constrained against rigid body motion only for eliminate this heat sink, another configuration
evaluation of thermal stresses. For evaluating (type Xl) was created without this flange.
pressure stresses, the model was axially con- The axial support was thereby removed to the
strained at all of the circumferential row of diffuser end of the flow separator housing as
nodes that interface with the flow separator shown in Figure 282. Insulation would be
housing support ring. In addition, nodes at the
location of the piloting slots were constrained LOCALIZEO
STRUTS
against rigid body radial motion. Pressure
stresses and deflections were found accept-
able, but both thermal stresses and deflections
at the regenerator seal surface were found to [
be excessive. The thermal stresses and
deflections during light-off increased with
-_ I LE
time during the transient interval and were
found to be maximum at steady-state operat-
ing condition. To reduce the deflections at
the regenerator seal surface, a strutted con-
figuration was created. Four struts were SUPPORT PIN
located 60 degrees apart, two in the HP side, FIXED HERE
and two in the LP side. The finite element

model of the strutted configuration is shown in Figure 282. Flangeless Flow
Figure 281 alongside the original configura- Separator Housing Outer Cone
tion. and Attaehment Configuration
(Type Xl).
The strutted housing was analyzed and was
found to have improvements in seal deflection applied at the outer diameter of the flow
but not in thermal stresses. Large, thermal separator housing to shield the component
from compressor discharge air impingement.
Axial pins would extend from the axial support
to provide pilotingfor the component.
Table 47 shows the results of the finite

element analyses for the maximum power
pressure conditions. All three configurations
have been evaluated with RBSN; the firsttwo
configurations have also been evaluated with
LAS material. The location of the critical
stress areas are shown in Figure 283.
Comments regarding pressure stresses on the
flow separator housing follow:
o Pressure stresses are almost independent

of component material.
ORIGINAL 1611 STRUTTEO (B21

CONFIGURATION CONFIGURATION
O The MPS due to pressure loading is 8
percent lower in the strutted (B2) configu-
gll/-O/ll-121 ration compared to the original (BI) con-
Figure 281. The Original and Strutted figuration.
RBSN Flow Separator Housing FEMS.
238
Table 47. Flow Separator Housing Pressure Stresses.
,i
Maximum Principal Stress

(MPS)
Flow Separator
Housing
Case Configuration Magnitude,

NO. Type Material Location MPa (ksi)
,u
01 B1 RBSN A 47.6 (6.9)
02 B2 RBSN B 44.1 (6.4)
03 Xl RBSN C 18.5 (2.7)
04 B1 LAS A 47.6 (6.9)
05 B2 LAS B 43.4 (6.3)
1371C (2500F) Maximum Power Condition

100,000 rpm
*For locations see Figure 283
O PRESSURE STRESS lOCATIONS
] THERMAL STRESS [STEAOY.STATEI

Because the axial loading due to pressures
THERMAL STRESS {TRANSIENT
is reacted out directly behind the HP
cavity in the flangeless (XI) configuration
as opposed to being reacted out at the
flanges in the original (BI) configuration,
pressure _stressesare much lower (61 per-
cent) in the flangeless configuration.
Because of the lower strength of LAS,

pressure stresses are more criticalfor the
LAS material.
An acceptable design for the flow sepa-

rator housing must be able to provide a flat
[] platform for the regenerator seal surface as
well as provide a low MPS. The corresponding
regenerator seal platform deflection for the
© previous pressure analyses are shown in Fig-
ures 284, 285, and 286. These curves trace the
axial deflection of the inner platform cir-
cumferentially 180 degrees from the HP to the
LP side of the component. Figure 286 also
traces deflections along the outer platform
since the type X1 configuration is supported at
the opposite side from the platform. Con-
cerning pressure deflections on the flow sepa-
Figure 283. Flow Separator Housing rator housing, the following conclusions are
Critical Stress Areas. drawn.
239
_m IMILSI _ ; RGSN TYPE B2 {STRUTTEOJ "- : RBSN TYPE 61 [ORIGINALI
_m {MILS) v , HOT I.D. SEAL e , HOT GO. SEAL
1015-
152.4 -
50.8-
101.6-
1a_o.n ' I'='_ o

I RP _ LP
D- .=.
iG
_ -50.6-
",.,... ,,_,_ i'°',
' g
50.8-
=,_.J
O-
"",-- _... '-9O I-,1.9)
-IOl.fl-
-50J-
-152.4_
-1el,G,
-9O -80 -30 0 30 80 9O DEGREES
-9O -80 -6O 0 30 8O 90 OEGREES
ANGULAR POSiTiON
ANGULAR P_SITIGN
G|7_/|-124
||7-o70-1 _6
Figure 284. Pressure Deflections Figure 286. Pressure Deflection

at the Regenerator Seal for Types at the Regenerator Seal for Type
B1 and B2 RBSN Flow Separator Xl RBSN Flow Separator
Housings. Housing.
ym (MILSI v _ LAS TYPE B110RIGINAL} o :LAS TYPE D2 ISTRUTTEDI platform deflection than the B1 configu-
101.0- ration.
] [
[2.7J 8#
HP.._I i _ LP
z.,,_,. The X1 configuration had 95 percent less
5_.G-
11.51 38 _-_,..
variation in the inner regenerator seal

platform deflection than the B1 configu-
o- '-5 I-0,21
ration.
,_ -50.8-
% The low spot on the inner regenerator seal
axial surface is located midway between
-101.6-
'-,,, 1 the crossarms on the HP side of the flow
-152.4 -
_'-,, .137 {-5.41 separator housing.
i 1 h
-9O ..60 -30 0 30 6O 90 OEGREES
The thermal stresses and deflections were

ANGULAR POStTION
also investigated at steady state cruise condi-
Gill,,iDT|.i_ tion [1218C (2225F) T4]. Circumferential
thermal gradients are created by the differ-
Figure 285. Pressure Deflections ence in turbine exit temperature (T5.1) and
at the Regenerator Seal for Types combustor inlet temperature (T3.5) at the LP
B1 and B2 LAS Flow Separator and HP side respectively. Radial gradients are
Housings. created by the combustor inlet temperature
(T3.5) and the inner diameter and the com-
The material with the highest modulus of pressor discharge temperature (T3.1) at the
elasticity (RBSN) had the least variation in outer flange. Radial gradients were found to
deflections at the inner regenerator seal contribute most to the overall thermal gra-
platform for the type B1 and B2 configu- dient. Figure 287 shows the variation in
rations. steady-state temperature along the crossarms
of an LAS flow separator housing. The sen-
o The B2 configuration had 81 percent less sitivityof thermal stresses and deflections to
variation in the inner regenerator seal
240
The MPS is very sensitive to the assumed
heat transfer coefficient between the flow
LASTYPE B1 CONFIGURATION separator housing flange and the compres-
120(] FLANGE
HTC= 31.2 WATTS/CM
2 -K {5.5 BTu/HR-FT
2 -F]
i2192l-
sor discharge temperature. The MPS is
more sensitive for components made from
RBSN than LAS.
"\
o The strutted (B2) configuration MPS is sig-
\ nificantly higher than the MPS for the
\
[1112]-
\ original (BI) configuration.
'%
k-- o The MPS occurs at the outer surface of the
300 L
outer flange except for the single case

with high heat transfer coefficients on LAS
0
ISZ) where the MPS occurs on the inner surface
50.8 101.5 152.4 203.2 250.4 of the inner cone.
121 (41 IOI IOI ItOJ
RADIALLOCATION.
mm (INCHESI The type B1 flow-separator housing regen-
GB7_/_127A erator seal thermal deflections are shown in
Figure 288 for both the RBSN and LAS
Figure 287. Steady-State Cruise Radial material at steady-state cruise conditions.
Temperature Distribution Along the Flow The axial deflections at the regenerator seal
Separator Housing Crossarm. inner surface are plotted against the angular
position for two different sets of heat transfer
the heat transfer coefficients over the outer
coefficients over the outer flange. Figure 289
flange were assessed by analyzing two dif- shows the regenerator seal surface deflection
ferent sets of heat transfer coefficients. The
results for the type B2 RBSN flow separator
results of the stress analyses are shown in
housing. The deflection results for the steady
Table 48. The type Xl flow separator housing state thermal cruise condition are summarized
was not analyzed since the worst case thermal as follows:
stresses and deflections occurs during light-off
for this configuration. The results are sum- o The effect of the heat transfer coefficient
marized as follows: over the outer flange on the platform
Table 48. Flow Separator Housing Steady-State Thermal Stresses.

(MPS)
Flange Flow Separator Housing
Heat Transfer
Coefficient,
Watts Btu Magnitude,
Case Configuration
No. Type Material em2C Hr - ft2-F Location* MPa (ksi)
06 B1 RBSN 31.2 (s.s) D 133.0 (19.3)

0'7 B1 RBSN 5.7 (1.o) D 42.0 (6.1)
08 B1 LAS 31.2 (5.5) E 26.9 (3.9)
09 B1 LAS 5.7 (1.0) D 20.0 (2.9)
10 B2 RBSN 28.4 (5.0) F 192.3 (27.9)
Steady-State Cruise 1218C (2225F)

241
0 At the higher value of heat transfer coeffi-
0 RBSNH = 31.2 WATTS/CM
2 -C 15.59TU/HR-FT
2 -FI cients (31.2 watts/cm2C) the regenerator
* RBSNH = 5.7 WATTS/CM
2 -K 11.0 BTU/HR-FT
2 -F) seal platform deflections are greater for
LAS H = 31.2 WATTS/CM
2 -C [5.5 RTU/HR-FT2-F[ housing. At the lower value of heat trans-
fer coefficients (5.7 watts/cm2C) the op-
pm MILS) Q) LAGH = 5.7 WATTS/CM
2 -C II.O 8TU/HR-FT
2 -F)
posite is true.
254-
I0 t
"" "_
"
; 21618.5)
__
il L. O The strutted (B2) configuration has a signif-
293.?-
icantly lower level of regenerator seal
_ 152+4-
_=
--I_I
tP,
......... -+,,
t2715.+oI
platform
average)
deflections
than
(68 percent
the original
on the
(B1) eonfigura-
-, 101.6- lion but has a higher variation (100 per-
I I _ cent) in-deflections.
50.8'
The transient light-off to cruise condition

II (cycle 8 in Figure 264) was also evaluated for
O"
-_ 450 -30 0 30 60 90 the flow. separator housing. The maximum

thermal gradient for both the B1 and B2 type
667.,076-1269
ANGULAR
POSITION.OEGREES flow separator housings occurred at steady
state conditions. Consequently, the MPS for
Figure 288. Steady-State Cruise Thermal these configurations also occurred at the
Defleetions at the Regenerator Seal for steady state condition. Figure 290 shows the
Type B1 RBSN and LAS Flow Separator
TRANSIENT
L/O CYCLE8
Housing. Mrs (KSIJ
FLANGEHTC= 31.2 WATTS/CM
2 -K )5.5 BTU/HR-FT
2 -P
172.3' 25
I tL
ROSNB2 CONFIGURATION 137.9" 20
&/m [MILS) FLANGE
HTC= 26.4 WATTS/CM
2 -C 15.0BTU/RR-FT
2 -F] [ 133 MPI 119.3K',
($TEADY-STATEiJ_
254-
103.4- 15
_.2"
/t
f
68.9. lO
_ 152.4.
34.6- 5 //
._ 101.6- /
o/
50.8- 0 4 8 !2 I6 20 24 28 32
TIME AFTERLIGHTOFFo
MINUTES
167-071-130
O-
-90 -80 -30 O 30 60 90 Figure 290. Type B1 Flow Separator

Housing Thermal Stress Response
ANGULAR
POSITION,DEGREES
During Lightoff.
11174111-1201
Figure 289. Steady-State Cruise Thermal transient MPS response for the type B1 con-
Deflections at the Regenerator Seal for figuration, which was typical for the con-
the RBSN Type B2 Flow Separator Housing. figurations with the outer flange exposed to
the compressor discharge air. The MPS for
deflection is similar to the MPS. Specific- the flangeless (X1) configuration was found to
ally, the deflections are very sensitive to occur during the transient light-off interval of
the assumed value of heat transfer coeffi- the operating cycle. Table 49 shows the MPS
cients, greater for RBSN than for LAS. for two cases that were considered. The first
model, as with the steady-state analyses
242
Table 49. Flow Separator Housing Transient Thermal Stresses

(MPS)
Flow Separator Housing
With
Inner to Time After
Case Configuration Outer Cone Magnitude, Start,
No. Type Material Radiation Location MPa (ksi) SCC
11 X1 RBSN No G 900
12 X1 RBSN Yes G 104.8 600

(15.2)
Cycle 8 Lightoff
neglected the radiation between the inner and transient temperature responses were com-
outer cones of the flow separator housing. pared to the test results. Initialcomparison
The latter analysis included this radiation. showed a significant lag in the temperature
The results are summarized as follows: response in the analytical model compared to
test results. Inclusion of cross radiation
Without cross radiation from the inner to between the inner and outer cone of the flow
outer cone, the MPS is higher for the separator and increased heat transfer coeffi-
flangeless (Xl) compared to the original cients helped but still were insufficient to
(B1) RBSN flow separator housing at show correlations. The most reasonable
steady-state conditions. hypothesis is that combustor heat reradiating
from the transition duct should have been
With cross radiation, the MPS is lower for modeled since it apparently provides a signifi-
the flangeless compared to the original cant heat input to the flow separator housing.
RBSN flow separator housing at steady- The foregoing conclusion is that the transient
state conditions. response and perhaps the MPS is less than
determined from analyses.
o Including cross radiation reduces the trans-
ient time to MPS. 6.3.1.6 Turbine Stators
In September 1985, the metal engine (S/N The design and development of the turbine
001/Build 34) was tested with 32 thermo- stators (reference Figure 250) is primarily
couples attached to a metallic flow separator controlled by its severe thermal environment.
housing. Additional instrumentation was used High core flow velocities produces high heat
to monitor the inlet guide vane setting, flow transfer coefficients, which coupled with the
path temperature, and engine speed. Data was required nonuniform vane profile,creates high
recorded and reduced for three transient light- thermal stresses at the vane trailing edge
offs to low cruise (60,000 rpm) condition. The during engine light-offs. High reliabilityand
boundary conditions and light-off cycle was survivability is also essential to ensure that
duplicated with an analytical model of a even minor stator damage does not lead to
metallic flow separator housing and resulting secondary impact failure of the turbine rotor.
243
The initial development effort concen-
trated on the segmented stator concept
designed in 1980. Both RBSN and SASC
stators were fabricated. In November 1984 a
stator thermal
the turbine shroud
screening test was run using
screening
objective was to thermally
rig. The test
shock stator seg-
/
ments (RBSN) to normal
tions (plus 25-percent
engine start condi-
stress margin). Follow- /
ing the test, no cracks were found in the
stator trailing edge.
In February 1985, a segmented stator set I

(SASC) was again run to the same
rig cycle used to proof the RBSN
test was unsuccessful;
segments
screening
stators. The
nearly all of the stator
were fractured, and two were fully
\
parted down the vane centerline. Fractures
were seen to initiate from
trailing edge area, which
the stator vane
is the most highly
/
stressed portion of the part. As a result of
this testing, it became apparent that surviv-
ability of stators is a concern under thermal
transient conditions. In order to improve the
survivability of stators a design modification
was made which incorporated a 3 mm (0.120-
inch) radius cutback at the trailing edge. This NEW TRAILtNG EDGE
scallop of the trailing edge represents approxi-

J
mately a l-percent penalty in performance,
but appeared to improve the survivability of
the stators under thermal transient conditions.
This was seen in subsequent testing, as for
example, an RBSN stator set on March 6, 1985
and another SASC stator set on May 3, 1985.
The cutback employed is illustrated in Figure
291. 3MM [0.120 IN) RADIUS
An analytical effort was initiated to assess ORIGINAL VANE CONFIGURATION

the merit of a cutback trailing edge compared GB7-078-131
II
to the original configuration stator vane.

Finite element analysis models as shown in Figure 291. Turbine Stator Cutback
Figure 292 were created for both configura- Vane Trailing Edge Redesign.
tions. The thermal model included adjacent
sections of both the shroud and backshroud as junction with this thermal model an ANSYS
shown in Figure 293 to provide the correct stress model was created.
transient heat fluxes at the stator support
interfaces. Radiation and convective heat Before any analysis was made for the cer-
transfer were considered at the non-contact- amic stators, verification of the stator flow
ing interfaces between the stator and shroud, path heat transfer coefficients was first
and stator and backshroud. Adiabatic wall accomplished. The metal engine, S/N 001/
conditions were assumed at the remaining sur- Build 34, was run in September 1985 with 23
faces of the backshroud and shroud. In con- thermocouples attached to an integral
244
metallic stator assembly as shown in Figure
294. This run occurred in conjunction with a
similarly instrumented flow separator housing.
The transient thermal resl_onse for three light-
STRAIGHT
CUTBACK
G97-071£132
Figure 292. Turbine Stator Finite

GII..OI& 133
Element Stress Models.
Figure 293. Turbine Stator Finite
Element Thermal Model.
•6
8.18
9.19 e 4
e2
FILLET
X
Gmt_7_l_4 ENGINE
Figure 294. MetaUic Stator Assembly Thermocouple Locations.
245
offs were recorded and reduced. Engine light-
off parameters were then simulated in a 3-D
thermal analysis of the metallic turbine stator
model. Temperatures from the analysis were
than compared to the measured test results.
A good correlation was found using the origi-
nally calculated heat transfer coefficients if
the measured thermal response was assumed
to lag due to thermocouple limitations. Two
representative comparisons between the MIN = 239C 1402F)
measured and calculated thermal responses C [FI

are at the mid-span leading and trailing edges TEMP
28a15501[--7
as shown in Figure 295. 343 16501
F--}
3o0(T50{[--1
The analytical comparison between the 454{e501
original straight and the cutback trailing edge 5_0 le50l
566 110501_
stators was then done using the foregoing
621 (115011_
verified heat transfer coefficients. The 677 (1250}
thermal analyses assumed the turbine stator, T32 I!S50} nag
backshroud, and shroud was RBSN. The engine GBT_7_IN|
start cycle (cycle 17) corresponded to the
development engine cycle as shown in Figure Figure 296. Turbine Stator
272. Both the original and cutback vanes were Temperature Distribution 20 Seconds
analyzed. The results were almost identical. After Light-Off.
The temperature distribution at 20 seconds
after lightoff is shown in Figure 296. The stress response. The MPS was found to occur
peak temperature occurs at the mid-span of at the mid-span of the vane trailing edge for
the vane trailing edge and cools off spanwise both stator configurations. As shown in Figure
and chordwise along the vane. 297 the MPS response at this location was very
similar; the MPS for the original configuration
Stress analyses were then conducted at was slightly higher (88.7 MPa compared to
various times after lightoff to determine the 88.3 MPa). No discernible difference was
T15J T12J
1200 1200
,4 . ,,,
(32) D 8 16 24 32 40 48 56 54 132l O O Ie 24 32 40 48 56 64
TIME. SECONDS TIME. SECONOS
6|7_T&i35A MIO-VANE
LEADING
EDGE MID-VANE
TRAILING
EDGE
Figure 295. Typical Correlation Between Test and Analytical Turbine Stator
Thermal Response.
246
MPa IKSII
• STRAIGHT-CUT VANE * CUTBACK VANE
103.4.
(12.9
KSll _._ ..........
68.9-
a.
E
34.4-
/ MAX
: 95.,
M,,,
,,2.o7
K.,,,
MIN = 16.1 MPa [-2.329 KSI]
O-
MPa (KSli
C (F) -0.7 ,0.1
815 ° 13.1 119
I v SIG'I STRESS
zeo 30
40.7 5.9
/ 54.4
68.2
7.9
9.9
/
[,-
82.0 11.9
260-
Z i i 95.8 13.9
MATERIAL: RBSN
CYCLE 17 GD7.076-138
Oq
0 lO 2O 30 4O 5O 6O Figure 298. Original Turbine Stator

TIME SECONOS MPS Distribution 20 Seconds After
Gll7.079-137 II Lightoff.
Figure 297. Turbine Stators Maximum

Temperature and MPS Response (at
Mid-Span Trailing Edge).
88.3 MPa (12.8 KSi) _ _i_'_-;
found in the temperature response at the mid-

span trailingedge. The MPS occurs 17 to 20
seconds after lightoff. The 3-D plots of the
MPS at this time (20 seconds after lightoff)
are shown in Figures 298 and 299 for the
original and cut-back trailing edge turbine
stators respectively.
MAX = 69.3 MPa (12.808 KSI)
To resolve this apparent discrepancy MIN = -11.8 MPa [-i.717 KSI)
between analysis and test results further
MPa (KSII
testing needs to be conducted. For these •0.7 -0.1 _ SI61 STRESS
tests, a significant sampling of stators should 13.1 1.9
be carefully inspected and documented in the 26.9 3.9
criticalstress areas then destructively tested 40.7 5.9

54.4 7.9
under increasingly severe thermal transient
conditions. Only after correlation between
952 9.8
02.011.R
inspection data and test results have been o5.oi39
made can an opinion be formulated as the
667-079-139
whether this discrepancy between analysis and
test is due to analysis error or non-nominal Figure 299. Cutback Turbine Stator
effects of the component geometry and/or MPS Distribution 20 Seconds After
material processing. Lightoff.
247
6.3.1.7 Combustor Assembly The critical design loading for the com-
bustor component is thermal loading. Mechan-
The combustor assembly (Reference Figure ical loading consists of small axially com-
250) was designed to have a low emissions and pressive loading from the combustor spring
specific fuel consumption (SFC) capabilities. and compressive hoop loading from the small
To meet these goals a lean burn concept was radially inward combustor p.ressuredrop. The
utilized in con]unction with a regenerative thermal loading is induced by the flame radia-
combustor inlet air source. The lean burn tion and gas convection of the combustor.
concept creates a uniform lean fuel-air ratio Circumferential and axial hot spots can also
that minimizes NO x. The high inlet temper- be created, the severity of which depends upon
atures cause the CO and HC to be oxidized to the fuel nozzle performance.
an acceptable level.
To evaluate thermal stresses in the com-
Development of the combustor design has bustor assembly, the axisymmetric finite ele-
evolved to the existing combustor assembly ment model shown in Figure 301 was used.
configuration as shown in Figure 300. The The model consisted of the combustor end of
modular main fuel nozzle incorporates an axial
the engine that included all components from
ignitor. It includes a pilot combustor with a the engine housing inward. The finite element
full open radial inflow swirler at the pilot exit.
model was based upon an earlier design that
The absence of holes in the liner walls and the
did not include the pilot combustor compo-
low temperature differential between combus- nent. Convective heat transfer coefficients
tor inlet and discharge temperatures mini-
were analytically derived and applied as
mizes thermal stresses.
boundary conditions. Flame radiation was
considered as well as radiation between engine
NOZZLE ROME components and to the engine external
\ SPACER SWIRLER / environment.
The temperature distribution determined

for steady-state maximum power condition is
shown in Figure 302. The peak temperature
1465C (2669F) was found at the lip of the
nozzle holder. The corresponding MPS for the
combustor assembly components are shown in
Figure 303. The components with the highest
LINER
MPS (61.3 and 66.2 MPa) were found at the
outer edge of the nozzle holder and combustor
6117-EI78-1_
support respectively. Both of these com-
Figure 300. Combustor Assembly Design. ponents were analyzed as RBSN.
GII;..071_I 4|
Figure 301. Combustor Assembly Finite Element Model.
248
982C i1900F1
II 87,c, EI - -------
II
l I i___ r_,] t"
....
#'#-'I@_,_=__:_.,,___._-----1204C
_IIOI_CIIDOOFi
12200F) ALL TEMPERATURES IN C IFI
_......__- ISOPLETH iNTERVAL 111C {2OOF)
1465C i2669FI
GiT-OT&I42A
Figure 302. Combustor Assembly Steady-State 1371C (2500F) Max Power Temperatures.
MAXIMUM PRINCIPAL
STRESS
COMPONENT MATERIAL NOOE MPA IKSll
SPACER SLEEVE ROSN 753 44.1 16.4i

NOZZLE HOLDER ROSN 390 61.3 (6.91
COMBLJSTORSUPPORT RBSN 45 66,2 (9.6i
COMBtiSTOR LINER SASC 592 12.4 t1.81
REGENERATORSHIELD SASC 935 32.4 14.7l
G67-076-|43
Figure 303. Combustor Assembly Steady-State 1371C (2500F) Max Power MPS.
The transient start stresses were evaluated variations of the flame or gas within the
with the lightoff cycle to 1281C (2337F) cruise combustor were not considered in these
condition as shown in Figure 304 (cycle 22). analyses. Consideration of these variations
Shutdown stresses were not analyzed due to would probably make significant increases in
the benign effect of the regenerator system. MPS levels but was not evaluated due to the
The MPS responses are shown in Figures 305, uncertainty of the actual combustive environ-
306, 307, and 308 for the four ceramic comment.
bustor assembly components. The highest MPS
was 32.4 MPa (4.7 ksi), located at the outer 6.3.1.6 Component Assembly/Sealing
edge of the combustor support, and which
occurred 180 seconds after lightoff. The design of ceramic components must
take into consideration the interfacing of high
In summary, the analyses showed very low temperature brittlematerial into an assembly.
thermal stresses for the assumed combustor This section specifically addresses these
burner condition. Axial or circumferential design considerations. They include designing
249
CX 102 (FX 102j
14-
100
1280C 12337F]
-80
35 KRPM
.80
KRPM
I_ = 0.16 KG/SEC [0,357 L8/SEC) -40
s
M = 0.07 KG/SEC
(0.150 Lg/MINi
(70F)
I I I i I I I
5 I0 15 29 25 30 35 40
TIME AFTER LIGHT.OFF, SECONDS
UT.07D-144
Figure 304. Engine Start to Cruise (Cycle 22).
LOCATION C LOCATI0N A LOCATION B

[NOOE824] INOOE803) 7 [NOOEOIli
MPa IKSII
MPa [KSli \ 80.0"
80.9-
NODE 399 ..,,,,,_ _ti_,_ _'_ 1_

55.1-
41.4
-- 41.4"
55.1-
27.6
41 e C i 27.6" f
J
13.8"
13.8 /
/
O- p'A O- O
0 40 90 129 IN) 20O 240 280 329
O 40 60 129 160 290 240 280 320
TIME AFTER LIGHTOFF,SECONDS
TIME AFTER LIGHTOFF,SECONDS
607-076-1450
Figure 305. Spacer Sleeve (RBSN) Start Figure 306. Nozzle Support (RBSN)
Transient MPS. Start Transient MPS. *
the axial and radial support system, sizing and address these design considerations, it is a
controlling the critical dimensions to ensure prerequisite necessity to have a thorough
against interference between ceramic com- understanding of the mechanical and thermal
ponents, and designing the sealing interfaces growth of the overall engine during its duty
to obtain optimum engine performance. To operation. On the AGTI01 program initial
250
MP| KSli
68.9- 3
REGENERATOR COMBUSTOR8AFFLE/
L STACKUP STATOR/BACKSHROUO
, NODE 45 __J__l
5,5.1-
_ 41.4- li" _-I I I1_,",1_1......1 r-r_CHt}Mll))7IItlIUV

eL
27.6
!
(/
0 40 80 120 160 200 240 280 320

TIME AFTER LIGHTOFFoSECONDS
GB7-076-147
Fiture 307. Combustor Support (RBSN)

Start Transient MPS.
ll/Ill-lSth
NODE 818
MPa IKSIJ Figure 309. Ceramic Components

68.9- 0
Stackup Load Paths.
55.1- 3 ,_ I-t b I t I I I I I-I=M t_b-VI il 1-1 i ! I_

bustor baffle/backshroud/stator st_Ickup, and
the diffuser stackup.
--_ 4t.4" _
The inner stackup consists of the rocker

_ 27.6- 4 assembly, turbine shroud, transition duct, and
E
combustor assembly. They are supported by
_ 13.e- z
the inner combustor spring that has been sized
L
to provide stability under a 6-g lateral shock
0
0 40 80 120 160 200 240 268 320 load throughout the engine operating cycle.
TIME AFTER LIGItTOFF, SECONDS The spring was also sized to provide adequate
Gll7-078-148
travel to compensate for the relative thermal
Figure 308. Combustor Liner (SASC) and pressure growths between the inner stack-
Start Transient MPS. up and engine housing.
design effort was accomplished using analyti- A subassembly of the inner stackup is the
cally determined _cowth and was updated as combustor baffle/turbine backshroud, and
engine test data became available. turbine stators stackup. The inital design for
this engine did not have this subassembly
To control the axial positioning of the stackup, but had the inner ceramic loadpath
ceramic components, springs have been used going directly through these three ceramic
to axially load these components against con- subassembly components. The incorporation
trolled surfaces. This thereby eliminates the of this subassembly stackup design was a con-
possibility of component failure due to sequence of seal improvements, as discussed
thermal interference between the ceramic and later in this section, made at the flow separ-
supporting metallic components. Five sets of ator housing, transition duct, and turbine
ceramic component stackup as shown in Figure shroud interface.
4
309 are found on the engine. These are the

inner ceramic stackup, the outer ceramic The outer stackup includes the regenerator
stackup, the regenerator stackup, the corn- shield and the flow separator housing, although
251
the flow separator housingis axially piloted by The components are axially loaded by the two
the regenerator stackup. The regenerator regenerator diaphragm seals. The flow
shield was designed to be loaded by an outer separator housingsupport ring was constructed
combustor spring more lightly loaded than the to, minimize the thermal growth differential
inner stackup. Again, this spring was sized to •between the engine housing and the regener-
meet a 6-g lateral shock load and to allow for ator components. This provided a more uni-
growth differential between the outer stackup form seal performance and regenerator drive
and the enginehousing. torque.
The diffuser stackup consist of the inner The radial support for ceramic components
and outer diffusers and is loaded by a com- must be able to maintain the concentricity of
pression spring against the turbine shroud. As these co.mponents without introducing any
shown in Figure 310 changes have been made mechanical interference. Two types of
to the first generation support system. The piloting methods were used to maintain con-
first generation load spacer in a series of tests centricity of the ceramic components to the
had a tendency to become cocked and wedged engine centerline. One of these, an eccentric
in the guide sleeve. The load spacer was pin in a radial slot, was used to pilot the flow
changed to a ceramic ball supported by a separator housing. This system consists of
conical spring. A stiffer spring was also used three radial slots in the outer rim of the flow
to increase the axial support load from 9.5 kg separator housing through which three
(21 Ibs) to 36.3 kg (80 Ibs) force. On previous metallic eccentric pins, mounted to the main
testing (static structure build 9) failure of the engine housing, maintafn component concen-
outer diffuser was attributed to vibratory tricity. A more innovative method is used to
energy from this spring-mass system. pilot the turbine shroud module. This method
consists of a three support rocker assembly
The regenerator stackup consists of the (Figure 311) that replaces sliding friction with
regenerator core and flow separator housing. rolling friction at a highly loaded interface.
FIRST GENERATIONLOAD SPACER SECONOGENERATIONLOAO SPACER
TURBINE TURBINE
% i
SHROUD SHROUD
J
I
OUTER
DIFFUSER
OUTER
DIFFUSER
I iNNER
DIFFUSER
INNER
OIFFUSER
__ LOAD
SPACER
6|7_7_140
COMPRESSOR
BACKSqROUO
1
Figure 310. First and Second Generation Diffuser Support System.
252
Component concentricity is maintained in this
configuration through interfacial friction at
the rocker contact zone.
Figure 312 shows the change in the rocker

assembly as a result of development testing.
The rocker was modified to eliminate a
thermal interference problem of the outer
diffuser between the shroud and rocker. The
TURBINE SHROUO
rocker assembly was therefore eliminated
from the diffuser stackup in the axial direc-
WASHER
tion. The rocker assembly still remained as
the radial pilot for the diffusers.
ROCKER
In the turbine shroud mechanical screening
rig test the turbine shroud is axially loaded
LOWER CONTACT WASHER and reacted (supported) by the rocker assem-
bly. Repeated failures during simulated en-
gine maximum power pressure condition led to
a design change in the turbine shrouds. The
GB7_T&ISO COMPRESSOR8ACKSHROUrl(METALLIC) locating slots at the turbine shroud mounting
flange were eliminated as shown in Figure 313.
Figure 311. Inner Ceramic Support The turbine shroud mounting system, which
Rocker Assembly. includes the rocker and bolt assemblies, was
SECONOGENERATION
ROCKER/ECCENTRIC
FIRST GENERATION
ROCKER/ECCENTRIC
TURBINE
;RROUO
GB7_TO-151A
Figure 312. First and Second Generation Rocker/Eccentric Design Comparison.
253
clearances between the rotor to shroud inter-
face were determined for transient and
/----'T--V
steady-state operating conditions. Since close
turbine running clearances are critical to opti-
mum engine performance, instrumented engine
PREVIOUS FAILURE ORIGIN testing is normally conducted to verify
analysis. This criticality is compounded with
SLOTTEO CONFIGURATION
the low rub tolerance of ceramics compared to
metallics. On metallic engines, verification is
usually made by using capacitance probes
between the turbine shroud and rotor to mea-
I ,'
I I I sure clearances.
capacitance
On ceramic
probes do not function.
fore, the approach followed
components,
There-
was to measure
SLOTLESSCONFIGURATION
running clearances with the early metallic
version of this engine. These clearances were
G|7_7_152 then adjusted to account for the difference in
materials and higher operating temperatures
Figure 313. Slotted Versus Slotless Turbine of the all ceramic engine. Further discussion
Shroud Configuration. of the turbine clearance evaluation can be
found in engine design and development (Sec-
shown to adequately retain the turbine shroud tion 4).
in position in a shake-table test. The slotless
configuration, which relies upon frictional There are three high temperature leak
loads alone to maintain turbine shroud posi- paths from the HP to the LP cavities in the
tion, withstood a 12-g load. static structure of the engine as shown in
Figure 314 that require the use of ceramics.
The radial clearance between ceramic
These seal areas are:
components was also sized to eliminate poten-
tial thermal interference during engine opera-
tions. The stackup of ceramic components has
to be designed for unconstrained thermal REGENERATORSHIELD/
FLOW SEPARATORHSG. -_
growths without any radial interference _,__ .TURBINE
between adjacent stackups. In addition, /" SHROUO
corner and fillet radii had to be sized to
preclude the possibility of mating components
contacting within the radii. An example of
this occurred during development assembly of
static structure rig, build 7 when the flow
separator housing failed. Disassembly and
inspection of the fractured component
revealed the presence of an interference
between the fillet radius of the flow separator
housing inner lip and the corner radius of the
mating regenerator shield. The corner radius
was increased to a larger value than the fillet
REGENER*TO,
SR,E,.O
EXHAUST ]'
radius in the mating part, thereby eliminating
GDT_T&IS,1
the possibility of localized contact within the
radii.
Axial clearances between the rotor to Figure 314. High Temperature Ceramic
backshroud interface and axial and radial Seal Areas.
254
o Turbine shroud/flow separator housingseal housing area. Because of space limitation, the
o Regenerator shield/flow separator housing wave spring was removed and the piston ring
seal was loaded by the inner combustor spring. To
o Regenerator shield/exhaust housingseal allow the for differential thermhl growth
between this seal area load path and the
Early testing (static structure rig, build 4) combustor baffle, backshroud, and stator load
investigated the usageof a ceramic fiber seal path, a wave spring was inserted into the
between the turbine shroud and flow separator latter load path as shown in Figure 316.
housing. The rig was assembledand a static
leak check performed. The results indicated a The axial interface between the
gross internal leak at the fiber seal area. The regenerator shield and the flow separator
fiber was replaced with ceramic ring seals for housing relies upon the outer combustor spring
all subsequent testing. load to control the seal leakage. Testing of
this seal area indicated a very low leakage
To investigate the performance of ceramic rate that met design criteria.
ring seals, static seal leakage rigs were de-
signed. These rigs were designed to measure The two piston ring seal concept was
seal leakage as a function of seal force and adopted for the regenerator shield/exhaust
pressure differential at the three high housing seal area. The axial load was provided
temperature seal areas. In addition, each rig by a ceramic wave spring in the stackup.
has the capability of establishing the degree of
concentricity between parts being sealed so 6.3.2 Rig Design/Development
that seal performance can be measured and
compared to predetermined amounts of 6.3.2.1 Introduction
eccentricity.
The ceramic structural parts received from
For the baseline configuration the leak AGT101 program suppliers are qualified for
path at the turbine shroud/flow separator engine testing according to the flow chart
housing area was found to be the most critical shown in Figure 317. The iterative design/
in terms of performance loss. The baseline inspection/test/redesign approach in devel-
configuration consisted of a single LAS piston oping engine quality hardware is designed to
ring between the turbine shroud and flow eliminate discrepant parts before they reach
separator housing axially loaded by a ceramic the engine test phase. Each component is,
wave washer. The following results were therefore, subjected to 40X visual and dimen-
noted. sional inspection, non-destructive evaluation
(NDE), and series of rig tests for final qualifi-
o The effect of the axial spring load on the cation prior to engine installation. A list of
seal performance in the baseline seal all of the rig tests used in this qualification
configuration was negligible. process is shown in Table 50.
o The effect of eccentricity between the

Structural components which are subject to
turbine shroud and flow separator housing
high mechanical loads, such as the regenerator
on the seal performance of the baseline
shield, flow separator housing, turbine shroud
seal was significant.
and rocker assembly components, are tested in
o A dual piston ring seal configuration mechanical rigs to simulate the pressure and
reduced seal leakage by 50 percent com- contact loading encountered in engine opera-
pared to the baseline line configuration, as tion. Likewise, components which are sub-
shown in Figure 315. jected to significant thermal stresses are
proof tested in thermal screening rigs to simu-
The dual piston ring configuration was late the maximum principal thermal stresses
adopted for the turbine shroud/flow separator encountered in engine operation. The thermal
255
CONFIGURATION2
.
[---1
LOAO
1
REACTION
CONFIGURATION__ '
IL
]i]. 2/
i O.IR-D.
O"D
I I I I I I l
O 16 32 48 64 (PSIG]
I I I l I
O 11 22.1 33.1 44.1 kPa
RIG PRESSURE
6B7.07i1-154
Figure 315. Dual Piston Ring Seal Performance Benefit.
cycles for the screening rigs are defined so as

to simulate stresses which exceed predicted
DUALSEAL _ "_ engine stresses under normal start conditions
by 25 percent.
CONFIGURATION i .i /
NEWWAVESPRING / i _I |
Thermal screening is intended to eliminate
LOCATION /
components with gross internal defects and/or
smaller defects located in regions of high
bending or tensile stress. These lower quality
components will fracture during thermal
i¸ , screening allowing only the highest quality
parts to continue through qualification testing
and subsequent engine tests. The thermal
screening rigs provide a screening test bed for
various components, configurations, and ma-
terial changes and have an upper temperature
limit of I149C (2100F).
617-070-155 The static structure rig test utilizes all of

the ceramic components in an assembly of the
Figure 316. Combustor Baffle/Backshroud engine structure, without the rotor, and
Wave Spring Location. checks the mechanical functioning of the as-
256
RECEIVING
DESIGN VENDOR
INSPECTION
MECH PASS
FAILURE
SCREENING
ANALYSIS
RIGS
THERMAL I_K{2_F]
SCREENING STATICRIG
RIGS
FAIL STATIC FAIL

STRUCTURE
RIG SECTION
FAIL TEST BED
1422K/I644K
(2100/25001
687478-280
Figure 317. Ceramic Component Qualification Sequence.
sembly. Further, the static structures rig base for evaluating thermal, stress, and risk
evaluates the sealing capability, component analysis tools and methodologies.
capability, axial and radial pilots, contact
loading, and stresses at elevated temperature. 6.3.2.2 Mechanical Screening Rigs
Components that fracture during any phase of
this evaluation undergo intensive fractography Mechanical screening tests have a dual
and analytical reevaluation to determine the purpose; establish overpressure capability and
cause and recommended solutions. The maxi- measure strain, and qualify parts. Initial tests
mum operating condition for this rig was set were performed on the flow separator housing,
at 1093C (2000F), which corresponds to the regenerator shield, and those components that
maximum operating temperature of the regen- provide the attachment between the ceramic
erator in the AGTI01 engine. turbine shroud and metal compressor back-
shroud. These parts have low thermal stresses
To evaluate ceramic structures upstream due to a combination of uniform heating or to
of the turbine exhaust, such as the turbine, a low thermal expansion coefficient (i.e.,lithi-
stator, shroud, backshroud, combustor baffle um aluminum silicate material).
and transition duct at the higher flow path
temperature of the engine, another rig, the The test fixture for the regenerator shield
1371C (2500F) static rig is used. Long-term screening rig is shown in Figure 318. The
testing at 1371C (2500F) temperatures is per- initial regenerator shield test incorporated the
formed in this rig to investigate the high use of strain gages to measure strain distribu-
temperature effects on ceramics, such as tion and magnitude during pressurization. This
static fatigue, contact loading, and endurance. instrumentation was installed as a checkout of
the test procedure. Pressure was applied until
In addition to screening substandard com- a strain level equivalent to 137.8 MPa (20 ksi)
ponents, the component test rig provide a data was achieved. This stress level was selected
257
Table 50. Ceramic Component Rig Tests.
Name Function Description
Regenerator Shield Rig Mechanical screening of Internal pressure proof test

regenerator shields
Flow Separator Housing Rig Mechanieal screening of flow Internal pressure proof test of
separator housings HP cavity
Turbine Shroud/Rocker Mechanical screening of Axial statie load test of the

Assembly Rig turbine shrouds and rocker ceramic inner staekup support
assemblies system
Stator Torch Test Low-cost thermal screening Lightoff thermal shock testing
of turbine stators using oxacetylene torches
Diffusers Rig Thermal screening of inner Lightoff thermal shock testing

and outer diffusers using an external combustor
Transition Duet/Combustor Thermal screening of transi- Lightoff thermal shock testing

Baffle Rig tion duet and combustor using an external eombustor
baffle
Baekshroud/St ator/Shroud Thermal screening of turbine Lightoff thermal shock testing

Rig baekshroud, stators, and using an external eombustor
shroud
Short term maximum temper- Elevated temperature testing

ature interface evaluation up to 1371C (2500F) with a
of turbine section static gentle lightoff and cool-down
strueture components cycle
Static Structure Rig Long term elevated tempera- Elevated temperature testing
ture interface evaluation of up to 1093C (2000F) with a
all static structure compo- gentle lightoff and cool-down
nents cycle. Uses an internal com-
bustor.
1371C Static Rig Long term maximum tem- Elevated temperature testing
perature interface evaluation up to 1371C (2500F) with a
of turbine section static gentle lightoff and cool-down
structure components cycle. Uses an external com-
bustor.
258
because it exceeded the predicted combined
thermal and mechanical stress exposure in
operation and is less than one-half of the
material characteristic strength.
To screen and qualify ceramic components

the proof pressure was reduced to a magnitude
that would simulate 48.2 MPa (125 percent of
the maximum operation stress encountered in
engine operation). Strain gages were not used
in the screening tests.
The flow separator housing test fixture in

the assembled condition is shown in Figure
319. The test fixture was designed to support
the housing, simulating engine mounting condi-
tions while sealing the housing to the fixture
GB7-078-261
with rubber gaskets. The fixture was designed
to use water to apply pressure loads within a
Figure 318. Strain Gaged Regenerator containment vessel. The upper plate is pro-
Shield and Mechanical Screening Rig. vided with fittings for a pressure line and an
G87-078-262
Figure 319. Flow Separator Housing Mechanical Screening Rig.
259
air bleed valve. Using gaskets and o-rings, the shroud attached by three sets of rocker com-
high pressure cavity is sealed so that pressure ponents to an aluminum plate which functions
can be applied to simulate engine operating as the compressor backshroud. The objective
of this test is to subject the three rocker
pressures. The flow separator housings are
screened at pressure levels that represent 125 assemblies and shroud attachment area to a
percent of the maximum mechanical loads 1814 kg (4000 lb.) axial load. This represents a
exerted in the engine at maximum power con- pressure load 25 percent greater than the
dition. maximum power pressure load.
A variation of this rig has been used to

The flow separator housing mechanical test
serve two other functions in this program.
fixture not only serves as a screening appa-
Both used the turbine module assembly with
ratus but also permits the correlation of 3-D
the compressor backshroud replacing the
stress analysis with mechanical stresses en-
aluminum plate. The first function arose early
countered in engine operation. Fourteen uni-
in the program when verification was needed
axial and seventeen triaxial strain gages were
to ensure that the rocker assembly could sus-
attached to S/N-8 flow separator housing in
tain relative radial differential thermal
critical areas for this purpose.
growth motion without compromising the
shroud axial stack or alignment dimensions.
The attachment between the turbine
This test consisted of subjecting the turbine
shroud and the compressor backshroud must
module assembly with compressor bolt loading
survive a large pressure induced axial load
to thermal cycles from room temperature to
during engine operation. To proof these com-
232C (450F). After testing, the contact wash-
ponents, a mechanical rig is used as shown in
ers and rocker interfaces were inspected for
Figure 320. The rig consists of a turbine
damage.
The other use for the turbine module as-

sembly came when demonstration of the rock-
er assembly capability under shock loading was
needed. The module was mounted thru the
compressor backshroud to a shaker table and
submitted to a 10-g lateral load.
6.3.2.3 Thermal Screening Rigs
The thermal screening rigs are designed to

subject critically stressed (thermal) hot gas
flow path components to simulated engine
starting transients to 1149C (2100F). These
rigs duplicate engine flow path geometry, thus
close approximation of actual heat transfer
coefficients is achieved for idle and cruise
conditions. This defined operating envelope
provides a remote test bed for the testing of
individual and/or multiple components in the
desired start transient condition. Components
are thereby screened on a pass/fail basis for
G87-078.253 4ip w,
9_39-1 any given thermal transient.
Figure 320. Turbine Shroud/Rocker An external combustor provides the heated

Assembly Mechanical Screening Test air supply to the rigs which is monitored with
Setup. rig inlet and exit thermocouples.
260 _ _OOR QUALITy

OR_G_?L'_L PAGE IS
DJ_ P_OOR QUALIT_
The ceramic components that have been

designated as critically stressed components
consist of the transition duct, combustor baf-
fle, turbine backshroud, turbine stators, tur-
bine shroud, outer diffuser and inner diffuser.
Three rigs have been designed and dedicated
to qualify (proof) these components. These
are the transition duct/combustor baffle rig,
the backshroud/stator/shroud rig, and the
inner/outer diffuser rig.
An acoustic emission monitoring system is

utilized on all thermal screening rigs in which
667-078-264
high frequency response acoustic transducers
are mounted to a waveguide in direct contact
with the ceramic component being tested. Figure 321. Standard Oil SASC Stator
Component screening is monitored in real time Vanes in the Thermal Shock Test Rig.
so that if a fracture occurs, the event can be
correlated with thermal transient conditions. response of the stator vane trailing edge in the
This equipment allows for diskette recording engine. Initial attempts with one torch did not
of acoustic emission (AE) data and hard copy produce a realistic temperature profile on the
graphics display of the AE data generated by vane or an acceptable thermal response curve.
the test. Additionally, this monitoring ap- The use of two torches yielded a response that
proach provides in-situ component diagnosis so closely simulated the stator llghtoff and was
that unnecessary disassembly/inspection/build used for all torch testing. Although the torch
time and secondary damage due to continued testing proved very simple and quick to use,
rig operation are minimized. its effectiveness to adequately simulate the
engine conditions was questionable. After the
A torch test designed to simulate stator component screening rigs were developed,
lightoff thermal response was used prior to the torch testing was terminated.
development of the thermal screening rigs.
This rig was used to identify potential mate- The inner/outer diffuser screening rig as it
rial deficiencies of the first stator sets from currently exists is shown in Figure 322. The
ACC and Standard Oil. The objective of the rig was designed to screen both the inner and
experimental effort was to duplicate the outer diffuser for short term lightoffs up to
worst-case stator trailing edge heating curve 1149C (2100F). The diffuser assembly is
for RBSN and sintered a-SiC vanes with the mounted within the rig, as in the engine with a
Garrett thermal shock test rig. ceramic rocker assembly, and the flange sec-
tion of a modified ceramic turbine shroud.
The general thermal shock test arrange- Hot gas from a 76.2mm (3 inch) diameter
ment consisted of two castable alumina stator natural gas burner at an engine idle flow rate
vane holders (each designed to hold three provide the heat source. Flow is ducted to the
stator vanes), one or more oxyacetylene diffuser assembly by a metallic nose cone and
torches, two airblast cooling tubes, a timer- outer liner. Purge air enters thru the nose
actuated slide table and a small-focal-spot cone to simulate the foil bearing cooling of
infrared pyrometer. Test setup is illustrated the engine. The inner walls of the housing are
in Figure 321. lined with insulation to keep the housing temp-
erature within acceptable levels. The rig test
A number of iterations were required to is monitored and controlled by two thermo-
identify the setup and torch conditions that couples located at the burner exit. The in-
simulated the analytically-predicted thermal tegrity of the ceramic components is mon-
261
DISCHARGE
The original design for this rig utilized a

metallic strutted standoff for the combustor
baffle. This proved inadequate for the ele-
vated temperatures encountered in this rig.
Consequently, the recent rig configuration
HEATED
AIR
SUPPLY
6117-07D-2115
Figure 322.
Screening Rig.
itored by an acoustic emission probe posi-

tioned off the back side of the inner diffuser.
The existing rig has evolved from the origi-

nal design with only minimal changes. As the
engine design changed the rig was concurrent-
. , ' .
ly updated. These include such changes as
modifying the load spacer design, the rocker
'_ _" CTI-.
assembly eccentric design, and the ceramic
bolt design. In addition, the rig insulation
material was improved. The previous material
Fiber-Frax formable insulation was replaced
with Babcock and Wilcox 3000 alumina-silica
insulation which additionally was coated to
increase strength with Ceracote, where ex-
posed to the turbine discharge gases.
The second thermal screening rig, the tran-

sition duct/combustor baffle rig as shown in
Figure 323 uses spring loaded pushrods to
axially load the transition duct and combustor
baffle assembly against a strutted standoff. It
uses the same laboratory burner as the inner/ _ ,.,k. o
LJ
outer diffuser rig uses. In addition to mon-
GD7-078-266
itoring the combustor discharge temperatures,
two thermocouples monitor the flow temper- Figure 323. Transition Duct and
ature at the T4.1 ports on the transition duct. Baffle Screening Rig With
The acoustic emission probe is positioned onto Original (top) and Final (bottom)
the transition duct at the third T 4 port. Configurations.
/'D
262
uses as a support platform a piece of Lockheed then reconfigured to the previous insulation
HTP-12 insulation which is formed to provide arrangement. Again, tests were run with and
both the support and a portion of the discharge without stators. Finally, the new insulation
flow path. This also has proved inadequate piece, which nests behind the turbine shroud,
due to erosion in the flow path ahead of the was installed with tl_e remainder of the rig
struts. Other changes in this rig include employing the former insulation. This con-
increased insulation along the rig inner walls, figuration was run to provide data on any
change from aviation fuel to natural gas, and effects that this single change may have. The
the addition of a viewport to visually monitor 3
4
the mechanical integrity of the transition
duct. The change was made from a liquid fuel
burner to a natural gas burner for better
lightoff characteristics and more stable opera-
tion under severe transient conditions, leading
to more repeatable screening tests.
The third thermal screening rig, shown in

Figure
shroud
324, is the backshroud/stator/turbine
rig. It uses the same spring axial f i*, • s"
loading mechanism and combustor used on the

transition duct/combustor baffle rig. Temper-
atures were monitored at both the combustor
discharge and at the stator inlet (T4. I) loca- SPRINGLOADTRANSMITTEO BY
tions. The lightoff is controlled from a T4.1 RODS IN HIGH HEAT AREA
P
thermocouple. Since this rig does not closely
simulate the engine (no diffused flow around
the shroud flange) two separate test cycles
must be used to proof the different ceramic TURBINE
components. A shorter and greater transient SHROUO
thermal shock cycle is used to qualify the

stators. A longer and less severe thermal
shock cycle is used to qualify the turbine
shrouds.
In January 1984, the backshroud/stator/ SIATDRS
turbine shroud screening rig was tested to

determine the effect of various rig insulation
configurations on the heat transfer rate into
the turbine shroud. A metal turbine shroud,
instrumented with thermocouples, was used to COMBUSTOR
BAFFLE
model the heat transfer coefficients. The
purposes of the test series were to evaluate TURBINE
the effect of a new more heavily insulated BAGKSHROUD
configuration that was in use when two ROOS
shrouds fractured during screening, and to

PROTECTED
FROM HOT FLOW
1
determine a suitable rig insulation configura-
tion (and thermal transient cycle) for future
GBT_7&267A
ceramic shroud screening.
Figure 324. Turbine Shroud
Tests were run using improved insulation Screening Rig Ori_naJ (top) and
both with and without stators. The rig was Final (bottom) Configuration.
263
results of this test are discussed further in the O Rig modification was made to include a
analysis and design section on the turbine ceramic transition duct, combustor baffle
shroud (paragraph 6.3.1.3.2). and seal rings into the flow path. This was
done tq more closely simulate engine heat
As a result of rig testing with a metallic transfer coefficients on the turbine shroud.
shroud the following changes were made:
The rig discharge section of the insulation
The lightoff cycle required to proof the was replaced with Lockheed HTP-12 type
turbine shroud was made less severe and insulation.
became dependent on the shroud material
being tested [changing from 93 to I149C Two ceramic waveguides are now utilized
(200 to 2100F) to approximately 93 to in this rig test; one in direct contact with
871C (200 to 1600F)]. the turbine shroud mounting flange and the
other in contact with the transition duct.
O The housing cavity insulation was reduced
to allow for more cross-radiation between O A viewport has been added at the forward
sections of the turbine shroud. end of the rig to allow for visual inspection
of the turbine shroud and transition duct
Continued rig improvements were made as interface area.
a result of further rig problems. Some
changes were made only to be later discontin- The thermal cycles for the screening rigs
ued. A chronological order of these problems are defined so as to simulate stresses which
and changes are covered in detail in a previous exceed analytically predicted engine stresses
report*. The design changes made that have under normal start conditions by 25 percent.
been incorporated into the final rig configura- A normal engine start in defined as ambient to
tion include: 1218C (2200F) in five seconds at idle air flow.
These proof cycles are given in tabular form in
*Ninth Semiannual Report - Advanced Gas Tur- Table 51. The test cycles for the turbine
bine (AGT) Technology Development-NASA shroud have been rigorously developed by
CR - 174886 determining heat transfer coefficients for
Table 51. AGTI01 Static Components Proof Test Cycles.
_Jsec
Trig, At_o, ts/D,
Component(s) Material (Ibm/rain) c (F) sec sec
Transition Duct, All 171 21 982

Baffle Materials (22.6) (70) (1800) 10 500
Inner/Outer RBSN 129 21 1093(2000) 10 600

Diffusers SSN (17.1) (70) 1093(2000)
Turbine Shroud RBSN 76.0 93 859(1575) 15 600

(10.0) (200)
SSN 878(1630)
SASC 938(1720)
Stators All 129 294 1218 10

Materials (17.1) (70) (2225) 10 35
264
both the engine and rig from test data and The original and final configurations for
verifying or updating the accuracy of the the ceramic static structure rig are shown in
finite element thermal and stress analyses. Figure 325. The original static structures rig
The cycles for the other components were had two air inlet supplies that simulated com-
developed assuming similar flow conditions pressor discharge (HP), and turbine inlet con-
occur in the engine and rig. ditions (LP) in an open loop configuration. A
6.3.2.4 Static Structure Rig LP
The ceramic structures rig is a high tem-

perature test bed for the complete set of
ceramic structural components. The purposes
of the ceramic structures rig are:
OISCHARGE
o To evaluate the durability of the ceramic
assembly under the following conditions:
l) Normal engine start conditions for the

871C (1600F) TIT configuration and
I093C (2000F) TIT configuration with a
.dual alloy metallic wheel.
HP
2) Steady-state engine conditions for 871C DISCHARGE
(1600F) and I149C (2100F) TIT configu-
rations at idle and cruise
o To evaluate ceramic interface compatibil-

ity at idle and cruise conditions.
o To evaluate the ceramic turbine shroud

support/alignment system under start, DSCHARGE ' J! i _ _',
cruise, and idle conditions.
To demonstrate the ceramic assembly dur-

II j /i _[,,,_.
INLET
ability for ten normal start cycles for the
871C (1600F) and 1149C (2100F) configu-
rations.
The rig utilizes AGT101

outer compressor and exhaust
metallic engine
the hot gas flow path constructed

vidual ceramic
As a ceramic
housings with
of 49 indi-
static structural components.
static structures test bed, the
rig has a maximum low pressure (LP) inlet
24LE
temperature limit of 1093C (2000F) due to the
absence of a turbine rotor and the attendant
temperature drop during simulated engine GB7-O78-268A --'
operation. This 1093C (2000F) limit is a
function of the maximum gas temperature Figure 325. Static Structure Rig
immediately downstream of the turbine rotor Original (top) and Final (bottom)
and is the maximum regenerator temperature. Configurations.
265
three inch lab combustor provided thermal ceramic bolt assembly, the other on the flow
start transient and steady-state conditions separator housing. These locations were se-
under AGTI01 simulated engine idle and cruise lected so that all components could be con-
airflow conditions. tinuously monitored through one of the two
component assembly locations. The acoustic
Air, as hot as I093C (2000F), entered the monitor detects ceramic component fracture
structures rig through the inlet duct and then incidence through a high frequency transducer.
diffused through the ceramic transition duct; The system eliminates rig disassembly be-
thus heating the combustor baffle, turbine tween thermal start transients in which no
shroud piston rings, segmented turbine stators, high energy acoustic events are detected.
turbine backshroud, inner and outer turbine
diffusers, and LP side of the flow separator The last static structure rig test (Build 13)
housing. The air then passed through the was run in March 1985. Sufficient confidence
regenerator. was gained in the ceramic static structures rig
testing so that ceramic structural components
The cold, HP air entered the rig at temper- are now considered qualified for engine use
atures up to 93C (200F), flowed along the without passing through this phase of screen-
compressor backshroud and compressor housing. The static structures rig was therefore
ing, turned in the exhaust cover, and was modified for use as a hot regenerator seal
heated through the regenerator. The air then leakage test rig.
flowed over the HP side of the flow separator
housing, around the transition duct through the 6.3.2.5 1371C (2500F) Static Rig
regenerator shield, and out of the rig.
The intent for this rig was as a vehicle for
Initial testing of the original static struc- performing tests at 1371C (2500F) of major
ture configuration showed a combustor insta- structural components, including transition
bility problem related to an improper pressure duct, combustor baffle, turbine backshroud,
drop across the burner. This can result in a stators, and turbine shroud. During maximum
lack of flame containment within the primary operating conditions, those components being
zone of the burner and, in fact, "allow" the forward of the rotor where work is extracted
flame to extend well into the turbine flowpath from the flow will be exposed to temperatures
region. The ceramic structural components up to 1371C (2500F). Those components aft of
would thereby be locally heated in an atypical the rotor will never be exposed to tempera-
manner. tures higher than I093C (2000F) in the engine.
Therefore, components such as the diffusers,
Attempts were made to alleviate this prob- flow separator housing and regenerator core
lem by modifying the combustor to operate at are fully evaluated in the aforementioned
a higher burner pressure drop, but satisfactory static structures rig.
combustor performance was never fully at-
tained. The solution to this problem was The 1371C (2500F) milestone was initially
achieved by using a singular flow arrangement achieved in a backshroud/stator/turbine shroud
with the engine metallic combustor. The final screening rig in April thru May 1984, although
rig configuration, used since August 1984, only for a short term cycle. To demonstrate
therefore simulates engine conditions more survivability over an extended time period a
completely. separate rig was developed.
An acoustic emissions monitoring system The original configuration as given in Fig-

was used on all ceramic test rigs which pro- ure 326 incorporated a flow separator housing.
vided a "real time" component integrity as- This configuration was used for the first build
sessment. Two acoustic waves guides are of the 1371C static rig. Note that the flow
installed on the structures rig; one on the path of the incoming air to the rig differs
266
THERMOCOUPLE This rig was first used in August 1984, but
TURBINE SHROUD
IT4.1I
ACOUSTIC EMISSIONS PROBE was unable to meet the 1371C (2500F) objec-
TRANSITION OUCT tive because of the fracture of ceramic com-
ACOUSTIC EMISSIONS PROSE ponents. Testing on this rig was terminated
and development of an improved rig was ini-
tiated. In the interim, until a replacement rig
was developed, 1371C (2500F) testing was
done using the backshroud/stator/turbine
shroud screening rig. This was done for two
reasons: to avoid thermal growth damage to
the 1371C static rig inlet housing due to
different heating rates in different areas, and
to avoid risking the flow separator housings,
which are not required for the 1371C (2500F)
testing.
In April 1985, further 1371C (2500F) test-

ing was conducted, this time employing the
new 1371C static rig design similar in config-
THERMOCOUPLE uration to the turbine shroud screening rig, but
FLOW SEPARATORHOUSING ITcAv] oriented in a vertical position. This was done
ACOUSTIC EMISSIONS PROBE
for two reasons: I) to allow a dead weight to
supply the required load to the ceramic stack,
and 2) to preclude the possibility that dis-
_::J OI,
c
charge quench water will back up into the rig.
.-L- In addition, the rig eliminated the flow separ-
ator housing, an expensive and long lead time
item, which would have been unnecessarily
placed at risk in the previous rig design.
THERMOCO_,
The dead weight employed was made of
300 Series stainless steel, insulated for protec-
,_l 4'
ACOUSTIC EMISSION tion. Unfortunately, it reached a high enough
PROBE
temperature at approximately 25 minutes
GBT-O78-269A
after achieving the 1371C (2500F) test tem-
perature, that premature shutdown of the rig
Figure 326. 1371C (2500F) Static was thought prudent.
Turbine Rig Original (top) and Final
(bottom). In May 1985, additional testing at 1371C
(2500F) was attempted. To extend the running
from previous screening rigs in that the in- time at the 1371C (2500F) temperature condi-
coming air enters via the flow separator hous- tion, the dead weight system was changed to
ing "smile" prior to combustion in the rig incorporate ceramics. However, the amount
burner. For this initial test a natural gas pre- of weight that can be physically installed in
heater was not used, as originally intended, the rig envelope was found insufficient. A
due to concerns about rig inlet housing integri- pressure differential at light-off between the
ty when exposed to heated inlet air. This flow path and the cavity surrounding the test
means that incoming air prior to combustion pieces caused the transition duct to move
was in the range of 93-121C (200-250F). The upward (axially) and then bounce back down
heating effect of the regenerator core was not upon the rest of the stack of parts. Further
simulated in this test. modifications to the loading system for this
267
rig were made to allow extended operation at In November 1984, all turbine shrouds were
1371C (2500F) test temperature. These modi- mechanically tested along with the rocker
fications included a component loading system assembly to proof these components against
where a spring load is transmitted through engine maximum pressure loading. A hydraul-
three air-cooled metal bolts. These bolts then ic press was used to incrementally apply a load
load a modified turbine shroud and thus the up to 1814 kg (4000 Ibs). This corresponds to
stack of ceramic parts. This arrangement is 125 percent of the maximum pressure loading
illustrated in Figure 326 (bottom view). of the engine. The first two shrouds to be
tested were an RBSN shroud from ACC and an
6.3.3 Component Rig Testing SN-54 shroud from NGK. The SN-54 shroud
passed the test undamaged while the RBSN
6.3.3.1 Mechanical Screening Rig Testing shroud failed at 1315 kg (2900 Ibs). The origin
of the fracture was located in the radius of
the piloting slot on the shroud flange.
Testing was conducted on three mechanical
screening rigs; namely, the regenerator shield
During December 1984, two additional
rig, the flow separator housing rig, and the shrouds, an RBSN, and SN-54 shroud were
turbine shroud/rocker assembly rig.
subjected to the mechanical proof test.
Again, the SN-54 turbine shroud passed the
The initial testing with the regenerator test while the RBSN shroud failed at approx-
shield rig subjected the ceramic shields to an imately 1315 kg (2900 Ibs). The origin of the
internal pressure loading that would screen out fracture was located in the same area as the
material flaw sizes that had strength less than previous failure.
137.8 MPa (20 ksi). Four regenerator shields
from Standard Oil (SASC) were tested and one Due to the two failures of slotted RBSN
failed at a pressure loading equivalent to 124.1 turbine shrouds during mechanical screening, a
MPa (18 ksi). design modification was implemented to
strengthen the area where these failures initi-
Subsequent regenerator shields were tested ated. This modification resulted in removing
at a lower internal pressure that induced the slots on the flange of the shroud. Mechan-
stresses (48.2 MPa) equivalent to 125 percent ical shock testing of the turbine shroud/sup-
of the maximum stress level encountered in a ports/compressor backshroud assembly had
normal engine operation. Fifteen regenerator shown that these slots, initially designed for
shields made out of SASC from Standard Oil piloting, were not needed to control piloting.
were received and tested and all passed with- The first modified RBSN turbine shroud was
out any damage. mechanically screened during April 1985. The
shroud passed the test, the first RBSN shroud
The flow separator housings were tested in to do so, and indicated that eliminating the
a pressure rig at the vendor (Corning) before stress concentration of the slot increased the
shipping to Garrett. Pressures were specified
load carrying capability to the require_ level.
for this test to simulate the maximum pres-
sure loading in the flow separator housing All turbine shrouds were henceforth fabri-
during engine operations (44.8 MPa). In addi- cated without the slots on the shroud flange.
tion to the testing at the vendor, testing was Of the remaining shrouds tested, only one, S/N
also conducted at Garrett, in another dupli- 726 (RBSN), failed below the required load.
cate rig, to substantiate that no strength de- Failure was attributed to a pre-nitriding crack
gradation had occurred whenever the flow in the shroud flange. Shroud S/N 744 (RBSN)
separator housing was reworked. In all cases did pass the test but was rejected for further
whenever components were reworked and re- use because of the presence of a crack/flow
tested, the components passed the rig test line found on the shroud flange. All the SN-54
without damage. turbine shrouds from NGK and SN 250M tur-
268
bine shroudsfrom Kyocera passedthe mechan- S I
ical screeningtest.
I FAILED
A variation of this rig was used in April
1987 for a series of tests to assessthe struc- PASSED
6m
tural integrity of the lower contact washersin
the rocker assembly. Washers supplied by
NGK (SN-54) and Kyocera (SN 220M) were
tested to load up to 1814 kg (4000Ibs) through E
the turbine shroud, one at each of the three 41 I
rocker supports.
Initially, nine lower contact washers from z
NGK were successfully tested on both sides to 2m 1
a 1814 kg (4000 Ib) load. Following this, a

series of tests was performed with Kyocera i
lower contact washers in which three failures ] [:IN_
were encountered out of 15 test attempts i
(each test setup can proof three contact wash-

ers on one side).
_E
Following this, additional testing was per-

formed to determine if a primary failure of
one contact washer would lead to a secondary TURBINE SHROUD
failure of an adjacent contact washer. To 667-078-353
induce a premature failure of one contact

washer, a washer was scribed on the lower Figure 327. Mechanical Screening
side. The test showed that once the scribed Component Test Results.
washerhad failed, an additional washer failure
would occur in the 1134-1361kg (2500-3000Ib) rate of success has been experienced for all
range. material/component combinations except for
the turbine shroud made out of RBSN. These
It was later discovered that the supporting failures have been traced to material proces-
surfaces in the aluminum test fixture, which sing flaws and are not design'related.
support the lower contact washers and the
rockers did not all have perfectly flat faces. 6.3.3.2 Thermal Screening Test Results
In at least one case, a visible depression was
observed, that was later measured at 38-51 Thermal shock test results were first ob-
microns (0.0015 -0.002 inches). Preliminary tained from the torch test of RBSN and SASC
analysis indicated that the local depressions stators. All stators tested were from the first
below the contact washers would induce high batch made by ACC and Standard Oil and were
bending stresses in the washers. The alumi- tested in the as-sintered condition with no
num test plate was subsequently remachined final machining. In addition all the stators
to eliminate the surface depressions. On June were reject components. The ACC stators
15 and 16, 1987 ten contact washers madeout were rejected due to voids in the leading edge
of NC132 were tested on both sidesto 1814kg while the Standard Oil stators were rejected
(4000 Ibs). In the eight tests required all due to dimensional problems. The objective
contact washerssurvived without any damage. was to identify potential design or material
deficiencies as early as possible. Because of
The results to date for the mechanical the lack of close simulation between the ther-
screening rigs are shown in Figure 327. A high mal response of the test and analysis, this test
269
was intended only as an interim test until the Initial testing of inner and outer diffusers
more refined thermal screening rigs became resulted in fractured components starting
available. from the bolt holes in the inner diffuser and
bolt slots in the outer diffuser. Progressive
Some of the stator vanes from both ACC design changes were made in conjunction with
and Standard Oil survived the worst case start analysis to increase test success. Design
simulation. Those made out of RBSN tended changes on the outer diffuser included scal-
to fail at processing defects that were de- loping the outer ring, slotting from the inner
tected by radiography, others failed at the to outer diameter, and finally eliminating all
mold line at the radius between the trailing of the outer rim and bolt slots by piloting off
edge and the platform. The SASC stators the inside of the rocker assembly (Figure 328).
tended to fail at the radius between the trail- Likewise with the inner diffuser, slotting from
ing edge and the platform. Those stators that the inner to outer diameter, removing the
survived the thermal shock test gave reassur- outer rim and changing to bolt slots, was tried
ance to the adequacy of the existing stator before the final change was made to truncate
design. the outer diameter to below the rocker assem-
INTERMEDIATE
SHAPE- SLOTTED
ORIGINAL
SHAPE
UNSCALLOPED SCALLOPED
INTERMESMTE
SHAPE
FINALSHAPE- TRUNCATEO
Figure 328. Outer Diffuser Housing Design Evolution.
270
Y.O01{ QUALITY
bly, thereby eliminating the bolt slots in the ability, laser marking of all ceramic compo _
component (Figure 329). The initial goal was nents was also initiated. No deterimental
to screen a few components for application in effects from the marking were noted (see also
the static structure rig before a rigorous test paragraph 6.1.5).
cycle was used to proof components for
normal engine use. These components were
The test history for the other screening
qualified and on February 5, 1985 proof testing
rigs also proceeded on a similar and concur-
was initiated. To insure component trace-
rent path, whereby early testing took place in
conjunction with analysis to optimize the com-
ponent design. Numerous design evaluations
were made with the transition duct/combustor
baffle thermal screening rig. Design evalua-
tions on the transition duct centered around
the design of the standoff for the combustor
baffle struts and the T 4 thermocouple holes.
The primary justification for the final design
choice was based on manufacturability and
functionality since, except for one design
choice, no degradation in test results was
evident for any choice. Degradation in com-
ponent strength was noted in testing of two
ORiGiNAL SHAPE (RAO}AL SECTiONI
components with thermocouple bosses brazed
-- WiTH SLOT to the transition duct. The braze material was
-- WfTROUT RAOIAL SLOT
moly-disilicide. Brazing was discontinued

after this trial and all following components
utilized a slip fit design.
The combustor baffle design activity pri-

marily focused on the effect of wall thickness.
A higher rate of test success was achieved
with a uniformly thin wall that was 5.08 mm
(0.200 inch) thick. All subsequent combustor
baffles were changed to this design.
iNTERMEDIATE DESIGN ll/3 SECTIONI
Initial testing of the turbine shroud was

done without any stators in the rig. This was
quickly discontinued when it became apparent
that stators were required to simulate the
therr_al response of the shroud along the flow
path. _ In addition, since no flow occurs on the
outboard of the shroud as in the engine, the
thermal shock was much greater in the rig
than in the engine for the same combustor
discharge temperature rise rate. Consequent-
ly, the proof cycle for the turbine shroud was
FINAL SHAPE IHADIAL SECTION
adjusted to maintain a maximum thermal
GO707B-378 shock stress 25 percent higher than for a
normal engine lightoff. The proof cycle was
Figure 329. Inner Diffuser Housing based on measured transient temperature re-
Design Evolution. sponses in the rig and engine.
271
o
Initial design activity on the turbine shroud data base evolved that allowed for compara-
focused on the effect of wall thickness. When tive evaluation of the performance of various
poor results were obtained for both the thin or ceramic materials in a simulated engine envi-
regular wall thickness components, a com- ronment. Figure 331 shows the test results for
pletely new design concept study was initi- RBSN (AiResearch Casting Company, SN-54
ated. The outcome of that analytical study (NGK-Locke), SASC (Standard Oil) and
was the varied wall thickness turbine shroud SN-250M (Kyocera) materials used on the
design which would provide for a more uniform AGT101 engine components. Both SN-54 and
thermal response (Figure 330). Initial testing SN-250M showed better consistency, although
their data base was somewhat smaller. SASC
showed better success with the combustor
baffle application than with the transition
duct, perhaps, because of fewer stress concen-
trations in the combustor baffle design. RBSN
tended to do better on components aft of the
stator, when combustor pattern factors are
less pronounced.
6.3.3.3 Statie Strueture Rig Testing
NORMAL ENGINE
START 1371C 12500F1
Static structures rig testing was initiated
in July 1982, with primary intent to simulate
At DESIGN 83 DESIGN
173.7 MPa i25,,2 KSII 111.7 MPa (16,.2 KSII steady-state engine idle conditions at a 871C
(1600F) T4.1. The initial build suffered the
N74171_77
fracture of an outer diffuser along a radial
line. Thereafter, the rig was rebuilt sub-
Figure 330. Turbine Shroud Design stituting an updated outer diffuser design,
Evolution. containing a radial stress relief slot (reference
paragrapha 6.3.1.2). Another (partial) tear-
of this design was not encouraging due to the down and rebuild ensued, this time due to a
presence of a large stress concentration at the suspicious acoustic event, before the rig was
seal land radius that was not detected in the successfully run (as Build 03) of the metallic
analysis. Subsequent testing of this design was engine. This successful series was run at
made successful by a combination of a larger turbine inlet temperature of 871C (1600F) at
seal land radius, a better control of the finish airflow rates of 75.6 and 113.4 g/sec (i0 and
within the seal land radius, and improved ma- 15 Ib/min), to simulate engine test bed idle
terial properties. and cruise flow conditions. A total of 15 start
and shutdown thermal transient cycles were
An attempt was made to improve the mar- completed, with an average start transient
ginal capability of the original design by r_a- time of 120 seconds. In addition, shorter
chining a full radius cutback in the stator vane thermal start transients were run in the range
trailing edge. An apparent improvement in of 15 to 25 seconds to 871C (1600F) demon-
strength was noticed in testing. All sub- strating the capability of the ceramic static
sequent stators were made with the cutback structural components to withstand current
trailing edge. Three integral stator assemblies metallic engine test bed start transients.
from Ford were also tested and all resulted in
failures. At test completion, the rig was disas-
sembled and none of the 49 ceramic compon-
As the component designs became finalized ents were damaged. Also, no areas of inter-
face deterioration were detected.
along with the proof cycle requirements, a
272
687-078-354
Figure 331. Thermal Screening Component Test Results.
273
Subsequent builds again encountered set- dimensions of 38.1 microns (0.0015 inches)
backs. In one case, an experiment with a total-indicated-reading (TIR) through the
fibrous seal between the turbine shroud, tran- cyclic testing of Build 08.
sition duct, and flow separator housing gave
unfavorable leakages when compared with the Rig Build 09 incorporated a completely
ceramic seal rings they replaced. In another new set of ceramic hardware and was con-
case, instability of the rig combustor due to ducted to the same test cycle as Build 08. At
the inadequate control of the pressure drop two minutes into the second cycle the acousti-
across it resulted in a typical heating of the cal emission monitor picked up a continuous
ceramic components and consequent fracture. high energy output from a rocker assembly
In still,another case, a stack calculation error sensor. Due to high energy indications, the
resulted in a "thermal jack", or interference test was terminated and the rig disassembled.
between parts as they grew at differing ther-
mal expansion rates during the course of test- Upon disassembly, component inspection
ing, and thus catastrophic failure of several revealed a fractured turbine shroud and a chip
ceramic components. And in the subsequent from the outer diameter of the outer diffuser.
build (Build 07) a radius mismatch between the The turbine shroud fractured at a surface
regenerator shield and the mating pocket or defect approximately 3.18 mm (0.125 inch) in
platform in the flow separator housing was the depth. In addition the outer diffuser housing
cause for failure of the flow separator housing stackup was loose due to apparent sticking at
during assembly. the number two and three bolt locations. It is
believed that the vibratory energy from the
The replacement flow separator housing outer diffuser to the turbine shroud driven by
for Build 08 had a fillet smaller than the the recirculating flow behind the outer dif-
corner radii in the mating regenerator shield. fuser, contained sufficient energy to propa-
In August 1983 a complete set of ceramic gate the existing surface defect and chip the
structures was tested to a cycle 25 percent outer diffuser. To preclude the potential for
more severe than the cycle defined for the this problem to reoccur, the spring load across
first set of I149C (2100F) engine testing with the inner and outer diffuser stackup was in-
ceramic cod_ponents. Seven cycles were com- creased from 9.5 kg (21 Ibs) to 36.3 kg (80 Ibs).
pleted. On the lightoff for the eighth cycle
rig testing was terminated due to excessive The rig was reassembled with a replace-
high-to-low pressure internal leakage. The rig ment turbine shroud as Build 10 and run to the
was disassembled and inspected for component same cycles as with Builds 08 and 09.
damage.
At 19 minutes, 20 seconds into the first
Minor chipping was noted on the following transient cycle (1093C) a 91 db event was
components: combustor baffle strut, eccent- recorded on the flow separator housing acous-
ric spacers, flow separator housing ID seal tic emission monitor. The count rate for the
ring, and T4.1 (TIT) thermocouple load spa- event (approximately 450 counts) indicated a
cers. None of the noted chipping was a result probability of fracture. However, no changes
of material interface sticking. The excessive were noted in rig HP or LP inlet pressures, and
leakage that terminated the test was due to there was no fluctuation in regenerator drive
the collapsed diaphragms on the regenerator torque following the event, therefore, the test
inner seal as a result of overheating. was continued. After the successful comple-
tion of ten cycles the rig was disassembled.
Turbine shroud alignment is of paramount
importance in maintaining proper turbine rotor At disassembly it was noted that the inner
running clearance within close tolerances. diffuser and the flow separator housing had
The ceramic rocker/eccentric assembly suc- fractured. It is believed that the flow separa-
cessfully maintained cold assembly runout tor housing fractured at the time indicated by
274
the acoustic monitor and was caused by a stress in the above listed components. The
radial interference between the flow separator shock wave created by the strain release in
housing and the ring support housing. Inade- the backshroud at fracture is believed to have
quate diametral clearance between the insula- caused the failure in the interfacing compon-
tion preform (in its coated and hardened con- ents.
dition) and the inner diffuser has been deter-
mined to be the primary causeof the fracture The solution to the rig failure was two-
of the inner diffuser. The anti-erosion coating fold. First, the turbine backshroud was par-
on the insulation preform is applied by brush, tially stress-relieved by redesigning, and ma-
and so coating thickness is difficult to control. chining a 3-inch diameter hole through the
Also, the insulation material is subject to center of the component. The void area was
shrinkage in use, another obstacle in control- then filled with Lockheed HTP-12 insulation.
ling its geometry. The 3-inch diameter hole removes the current
peak stress area of the component, effectively
The objective for Builds 11 through 13 was lowering the maximum principal stress by 88-
to subject the ceramic static structure com- percent. Secondly, an improved fuel metering
ponents to a start 25 percent more severethan system was installed to preclude burner/con-
the development engine start, which would be trol instability, thereby minimizing the proba-
equivalent to the normal engine start. The bility of a rig induced failure.
initial test for Build 11 utilized the two flow
path rig configuration used on all previous In September 1984 another set of static
static structure rig test. A three inch labora- structure rig components was assembled as
tory burner is located in-line of the low pres- Build 12 and run with the same test objectives
sure air supply. This burner had never been as for Build 11.
usedat the higher massflows required for this
start cycle and was unable to maintain the Five cycles to 982C (1800F) were accum-
proper burner control to achieve the desired ulated, after which the rig was removed from
lightoff. The rig was consequently recon- the test cell and disassembled. All major
figured to include the engine combustor by ceramic components were intact; only a few
using a single flow configuration as in the minor discrepancies were observed. These
included chips on two of the three spacers
engine.
separating the inner and outer diffusers; two
A successful lightoff was achieved with the very small chips on the inner diffuser, and a
new rig configuration but was terminated prior small chip due to contact stress on the flow
to achieving the desired maximum tempera- separator housing seal ring. Successful com-
ture of 982C (1800F). The test was termi- pletion of this test represented the first time
nated due to the occurrence of coupled acous- that a full set of ceramic static structure
tical emissions from the two probes and a components were proof tested to such severe
simultaneous drop in rig inlet pressure. Disas- thermal shock conditions.
sembly re _"ealed that the transition duct,

stator assembly, shroud, outer diffuser, and Rig Build 13 was assembled from another
backshroud had fractured. set of ceramic components and tested in April,
1985. Three successful test cycles were run to
Fractography results in conjunction with same conditions required for Builds 11 and 12.
thermal and stress analysis for monitored rig
conditions have concluded the cause of failure Late in the third cycle, high level AE was
to have originated from thermally induced noted, indicating a possible fracture of ceram-
stress in the turbine backshroud. The compon- ic components. The test was terminated.
ents are believed to have reached fracture Subsequent rig disassembly revealed only a
stress as a result of a rig burner/control chipped stator segment (reusable) and erosion
instability at the time of maximum (transient) of the high temperature ZAL-45 fiber insula-
275
tion. The high level AE indications may be 1149C (2100F) and 1260C (2300F) respectively.
attributed to the insulation breaking up, or The third cycle reached 1371C (2500F). After
perhaps to the effect of a too rigidly mounted 3.5 minutes at 1371C (2500F) a high acoustical
AE probe sweeping a surface as the rig ex- emission event occurred.
pands and contracts. The test was considered
completed. Disassembly revealed fractures to the tur-
After Build 13 confidence had reached a bine shroud and the integral stator ring. The
stator failure was secondary to the turbine
level that a check screen of the entire ceram-
shroud fracture. The most likely explanation
ic structural assembly was no longer felt nec-
of the fracture was that the pressure differ-
essary as long as components were still quali- ence across the stators created a force suf-
fied on smaller subassemblythermal and me-
ficient to counteract the spring load holding
chanical screening rigs. The static structure
the ceramic stack together. Thus, the as-
rig was therefore modified and used as a hot
sembly was sensitive to any fluctuations in
regenerator seal leakage test rig. pressure. Initiation of fracture was at the
6.3.3.4 1371C (2500F) Static Rig Testing turbine shroud seal land outer diameter (CO)
where contact was made with the transition
This section covers the high temperature
duct. Prior to further testing the rig under-
testing of ceramics to investigate effects such
went modifications to the axial spring loading
as static fatigue, contact loading, sticking,
mechanism.
and endurance. Testing was conducted in
three different rigs throughout the duration of
In mid-May, 1984 another set of ceramic
the program before a specific rig for high
components was assembled in the modified rig
temperature testing was finalized.
and tested to a maximum temperature of
The initial testing was done in the back- 1371C (2500F) according a cycle similar to
shroud/stators/turbine shroud screening rig that used for the previous test. Although
with improved insulation in mid-April, 1984. acoustical emissions indicated a high energy
The objective was to subject the ceramic output after shutdown, rig teardown revealed
transition duct, combustor baffle, turbine all components intact, with the exception of a
backshroud, stator assembly, and turbine spalled wave spring. Wave spring spalling had
shroud to the high temperature cycle defined been traced to the quality of the powder used
in Table 52. The first two cycles only reached in the manufacture.
Table 52. Test Cycle for Initial The design of the backshroud/stator/shroud
1371C (2500F)Testing in the screening rig limited the 1371C (2500F) test-
Stator/Shroud Screening Rig. ing to short term exposure. The hot static rig
was designed for long term exposure of elev-
Time, Temperature
min, C (F) ated temperatures to the ceramic components.
The original configuration for this rig had the
t 6 649 (120Oi flow path of the incoming air different from
4 815 (1500) the flowpath of previous screening rigs in that
4 982 (1800) the incoming air enters via a ceramic flow
4 1149 (2100) separator housing prior to combustion in the
4 1260 (2300) rig burner. Goals for this initial test were to
4 1371 (2500) establish proper rig setup in the test cell,
5 rampdown to 649 (1200) verify rig build procedure and operation, and
4 649 (1200) achieve a 1371C (2500F) temperature. The
Shutdown intended maximum temperature of 1371C
(2500F) was not achieved because of com-
*Light-off to 649C (1200F) in 15 second_ ponent failures during start. Rig disassembly
15 second ramps I revealed a failed turbine shroud, integral
276
stator assembly, seal assembly. The most high stress levels normally would not occur.
likely cause for the fracture was contact load- Also, several stators were noted to have stuck
ing near the fillet area of the shroud where it to the turbine shroud. This phenomenon was
mated with the flow separator housing seal analyzed and it is felt that the sticking was
ring. due to metallic deposits emanating from the
metal-sheathed thermocouples employed in
Further testing at 1371C (2500F) was con- the rig.
ducted in October 1984 using the backshroud/
stator/shroud screening rig as a test bed in In April 1985, further 1371C (2500F) test-
lieu of the previous hot static rig. The ceram- ing was conducted, this time employing a new
ic hardware in the backshroud/stator/screen- rig design similar in configuration to the back-
ing rig Build 06 was subjected to 30 minutes of shroud/stator/shroud screening rig, but orien-
operation at 1371C (2500F) and 75.6 g/sec (I0 ted in a vertical position. This was done for
Ib/min) airflow. No fractures occurred during two reasons: 1) to allow a dead weight to
the operation. supply the required load to the ceramic stack,
and 2) to preclude the possibility that dis-
Two significant observations were made charge quench water would back up into the
during rig disassembly. Insulation which sup- rig. This rig would be used for all subsequent
ports the turbine shroud and the rest of the 1371C (2500F) testing. This rig's first test was
ceramic hardware, consisting of a silicon diox- identified as the 1371C (2500F) static rig
ide insulation with an alumina based rigidizer, Build 01.
yielded during the test, allowing the parts to
drift axially by approximately 1.3 to 2.5 mm The dead weight employed was made of
(0.050 to 0.i00 inch). The insulation also stuck 300 Series stainless steel, insulated for protec-
to the turbine shroud when the latter was tion. Unfortunately, it reached a high enough
removed from the rig. Further, examination temperature at approximately 25 minutes
under magnification on the three struts on the after achieving the 1371C (2500F) test tem-
combustor baffle and the mating areas of the perature that premature shutdown of the rig
transition duct revealed glazing in these areas, was thought prudent. Upon disassembly four
although no component sticking was observed RBSN stators were found lightly stuck to the
(the parts were freely disassembled). Success- RBSN turbine shroud and the LAS spacer ring
ful completion of this test represented the melted. The melting of the LAS spacer ring
first time that these components had been was believed to be caused by a combination of
exposed over a long period (30 minutes plus) to a hot spot in the flame and leakage through
1371C (2500F) temperatures. the seal assembly.
A subsequent attempt to repeat the test, Build 02 incorporated an RBSN spacer ring
(backshroud/stator/shroud screening rig, Build and a stator assembly consisting of both RBSN
07), but substituting a turbine shroud of RBSN and SASC stator segments. To extend the
for the SSN turbine shroud, was aborted ap- running time at the 1371C (2500F) tempera-
proximately 5 minutes after reaching the ture, the dead weight system was changed to
1371C (2500F) temperature. The rig was shut incorporate ceramics. However, the amount
down after a shift in the stack of the ceramic of weight that could be physically installed in
parts was noted. Subsequent disassembly re- the rig envelope was found insufficient. A
vealed that this was due to relaxation of the pressure differential, at the test light-off,
insulation, as had previously occurred. Unfor- between the flow path and the cavity sur-
tunately, quench water used to cool the dis- rounding the test pieces caused the transition
charge ducting backed up into the rig, passed a duct to move upward (axially) and then bounce
dam intended to contain it, and as a result the back down upon the rest of the stack of parts.
turbine shroud and backshroud both were frac- Because of this failure no test time at 1371C
tured. The fractures originated in areas where (2500F) was accumulated for this build.
277
Rig modifications to allow for 1371C this test the coating was not satisfactorily
(2500F) testing were completed in early July evaluated due to failure of the turbine shroud.
1985. An external spring loading mechanism This was caused by inadequate clearance be-
was used to provide the axial force to contain tween the SASC turbine shroud and the RBSN
the ceramic components. This rig configura- transition duct to allow for the differential
tion was tested for the first time in late July thermal growths.
1985 (Build 03) and was successfully operated
at 1371C (2500F) for 2 hours. Stators made of The test objective for Build 08 was to
silicon carbide (SIC) and reaction-bonded sili- evaluate SN 250M material in an assembly
con nitride (RBSN) were alternately used to with SASC material. After 13 hours of testing
evaluate interracial sticking. Teardown in- at 2500F three pairs of SN 250M stator seg-
spection of the rig revealed glass formation, ments were found stuck to each other but not
primarily on the stator and baffle feet sur- at the stator to shroud interface. In addition,
faces; however, there was no evidence of a glassy deposit was found at the center
sticking. portion of the SASC turbine backshroud.
The next build (Build 04) incorporated an Build 09 consisted of a complete set of
all RBSN stator set and RBSN turbine shroud ceramic components made out of SN 250M.
to evaluate interfacial sticking at the shroud/ The rig was run for 14 hours at 1371C (2500F)
stator interface. No results were obtained. A and similar results to build 08 test were found
light-off was attempted using an off-the-shelf on disassembly. The glassy deposit was again
automatic control system. Thermocouple re- found on the center portion of the backshroud,
sponse in this rig (the thermocouples are but this time the backshroud was made out of
shielded in a ceramic sheath) was too slow for SN 250M. This would seem to indicate that
adequate control, and as a result ceramic the glassy deposit was not reacted from the
fracture ensued. backshroud surface. Two pairs of stator seg-
ments were also found stuck to each other and
Builds 05 and 06 each accomplished 12 one stator stuck to the turbine shroud.
hours at 1371C (2500F). Build 05 showed
sticking at the stator/turbine shroud and sta- To date, limited sticking at 1371C (2500F)
tor/stator interfaces. No damage was seen at has been seen. No failure has occurred due to
either interface. Previous builds had showed sticking. Interracial distress seen between the
the transition duct/baffle interface to be a transition duct and combustor baffle can be
problem when using the same material for eliminated by using an alternating material
both parts. This problem can be eliminated by stack in this area. More time at temperature
using an SASC transition duct/RBSN baffle is needed before drawing any conclusions
combination. about possible detrimental effects of the
stator/stator and stator/shroud sticking that
Build 06 showed no sticking. Unlike the was seen.
RBSN stators in Build 05, the SASC stators did
not stick to either the turbine shroud or to 6.3.3.5 Ceramic Seals Testin_
each other. As in Build 05, alternating mate-
rials at the transition duct/baffle interface Performance measurements made during
eliminated problems encountered in earlier engine testing of the ceramic-structured and
builds. all ceramic engines indicated power output
was below design calculations. Excessive
The test objective of stator rig Build 07 leakage through the ceramic sealing areas was
was to determine the effectiveness of A1203 suspected as a primary reason for the lack of
Sol-Gel coating on SASC parts to prevent power. The three sealing areas of interest
sticking. Preliminary results using test bars included the turbine shroud/flow separator
have shown this to be an effective coating. In piston ring seal, the regenerator shield/flow
278
separator housing butt joint seal, and the shroud piston ring seal solid stack arrangement
regenerator shield/exhaust housingpiston ring in the turbine shroud seal pocket. The spring
seal. These seal areas are shown in Figure load was transferred from the turbine shroud
332. In order to isolate and measure leakage stack to the baffle/backshroud/stator stack by
through these areas, three cold static seal rigs inserting a 17.8 cm (7-inch) ceramic wave-
were designed and fabricated. Each of the spring between the baffle and backshroud. Rig
rigs madeextensive use of engine hardware in measurements of the new configuration
order to simplify fabrication. By measuring showed a 50-percent reduction in leakage over
cold seal leakage, the rigs were able to gen- the single LAS configuration as shown in Fig-
erate performance data for new seal designs ure 336.
which could be compared to performance of
existing configurations. The rigs were also Leakage testing with the single LAS seal
successful ]n measuring the effects of seal configuration was done to determine the ef-
loading, concentricity between sealed parts, fects of axial load on the seal performance.
and effects of pressure distortion on seal per- Seal loading was varied from 2.27 to 25 kg (5
formance. to 55 Ibs) over the range of engine pressures.
REGENERATORSHIELD/ TURBINE SHROUD/ As shown in Figure 337 there was no signifi-
FLOW SEPARATORHOUSING FLOW SEPARATOR cant benefit from increased loading.
BUTT JOINT SEAL / PISTON RiNG SEAL
REGEHEBATOR
SH, LO/ / / Single LAS seal testing also indicated that
EXHAUST HOUSING PISTON __ _::_ ./__ /
eccentricity between the turbine shroud and
flow separator has a significant effect on seal
performance. This increased leakage up to 25
percent in a position of 1.52 mm (0.06 inch)
eccentricity over the concentric position.
Overall, a 50-percent reduction in turbine

shroud/flow separator housing piston ring seal
leakage was accomplished by maintaining ec-
centricity below 1.0 mm (0.04 inch) and creat-
ing the dual seal configuration. More re-
cently, a triple seal configuration was de-
signed and fabricated using sintered silicon
nitride. The seals have a split end gap to
account for the difference in thermal growth
667-078-270
between the LAS flow separator housing and
Figure 332. Turbine Shroud Piston Ring the silicon nitride seals. Although testing has
Seal_ not yet begun, the seals will have increased
temperature capability over the LAS design
Since the turbine shroud/flow separator and should exhibit improved performance.
piston ring seal was believed to be most crit-
ical with regard to engine performance, this The regenerator shield/exhaust housing
rig design was fabricated initially and is shown seal is a piston ring type seal shown in Figure
in Figure 333. Baseline leakage measurements 338. The dual seal is axially loaded onto the
using the turbine shroud piston ring seals rig cast iron exhaust housing through a ceramic
were obtained using the single LAS flow sepa- wavespring. The primary purpose of the test
rator seal configuration shown in Figure 334. was to determine the effect of load applied to
This configuration was compared to a new dual the seal on seal performance. The results
seal configuration shown in Figure 335. from this testing indicated that axial load had
a critical effect on seal performance. The
The dual seal configuration utilized two seals are a split end gap configuration which
LAS flow separator housing seals and a turbine
279
PA3612096-1
PA361Z094-1
%
PA3812093.1
PA3612097.1 ,_
MS21043-4
59430-4
47 RI'O
REQ'O
$8990-166
PA3612403.1
PA3612401-1
PA3612091.1
PA3612402.1
PA3610213-1
PA3612095-1
• PA3812092.1
PA3669655-1
-PA3810206-1
PA3809628-4
PA3809621- I
-[
i
PA3609627-1
PA3609627.2
PA3609627-3 \ PA3612098.1
PA3809627..4 \
OF PA3811454
PA3M9657.1 OF PA3811454
$8157.1
]
AN960E516L 2400-1
GD7_7_271 MS9491-24 P
Figure 333. Turbine Shroud Piston Ring Seals.
280
LP
OUAL FLOW SEPARATORSEALS WITH
HP
TURBINE SNROUOSEAL
RP
LP
S SEAL RING
229 CM [9 INCHI
TURBINE 'SHROUO 17.8 CM (7 INCHI
TURBINE DACKSHROUO OUAL SEAL RINGS
WAVESPRING
RP
\
,...,'
2.
l' I
UT@TS-_Z
Figure 334. Single LAS Flow Separator Figure 335. Dual LAS Flow Separator
Seal Configuration. Seal Configuration.
allows for thermal growth of the regenerator activated. Optimum axial load was less than
shield and uses pressure to conform the seal to 2.27 kg (5 pounds). A typical seal leakage
the surface of the shield. Testing showed that curve is shown in Figure 339. Further seal
seal performance needs improvement at lower testing will focus on methods of improving this
pressures where the seal is not pressure- seal's low pressure performance.
281
[LD/MINI
20-
KG/MIN
0._" O ; SINGLE
= OUAL
0,5"
,,t
IPSIGI
O- e
i ! _ I ! f !
54
8 IR 24 32 40 48 56
MPI
,:,, & ,'-.. 2_o & & &
RIG PRESSURE
eJt4;l_74
Figure 336. Single vs. Dual Piston Ring Seals.
KB/MINI
I
KG/MIN
0,75,
1.8-
0 : 2.27 KI 15 LIII LOAO

'0 = 25 K6 i55 LRI LGAO
1.2-
0,5'
025-
0,4"
O- 0 , . , , , , . ,PS,_.,
O IR 24 32 40 48 56
I I I I I a MPa
129 lIK) 240 300 360
RIG PRESSURE
Increased Loading Effect on Single LAS Seal - 5-28-86.
282
(,_JGL_AL PAGE [S
OF. POOR QUALITY_
U7_T&2'Tfl
Figure 338. Regenerator Shield Piston Ring Seal Configuration.
283
ILB/MIN
1.0-
KG/MIN
LEAKAGE
AS A FUNCTION
0.375" OFENGINEFLOW,PERCENT
0.8--
I -- 3.1
2 -- 2.0
3 -- 1.2
0`§ :; I:I e/
0`25-
6 -- K1
0.4" 2
0.125.
0.2 "
O,
I I I I I ! I I IPSIGI
0 8 16 24 32 40 49 56
; I I I I I I MPa
60 120 I DO 240 300 360 420
RIGPRESSURE
G87-0;6-277
Figure 339. Regenerator Shield Seal Leakage over Operating Pressure Range.
284
APPENDIX A
AIRESEARCH CASTING COMPANY

ADVANCED GAS TURBINE (AGT)
TECHNOLOGY DEVELOPMENT PROJECT
FINAL REPORT
1.0 INTRODUCTION 2.2 Pre-Casting Materials Preparation

Development
This report, submitted by AiResearch
Casting Company (ACC), covers the efforts of Proper materials preparation to produce a
ACC in support of the Advanced Gas Turbine - desirable slip is the foundation to the success-
I01 Program (AGT101). This program was ful casting of the rotor. It involves the
conducted as a subcontract to the Garrett optimization of milling procedures (particle
Turbine Engine Company (GTEC). This report size control), slip property adjustments (pH,
covers ACC's materials and fabrication devel- viscosity, solids contents, etc.), and green
opment efforts. These efforts are concerned density. The materials composition selected
with the materials development and fabrica- for investigation at the beginning of the pro-
tion or rotors and ceramic structures by slip gram (1979) was GTE SN502 Si3N 4 + 8 percent
casting and the fabrication of stator vane Y203 + 4 percent AI20 3 (Code 1 composition).
assemblies by injection molding. A rubber lined steel mill jar with Si3N 4 media
was used to blend the powders and control the
2.0 ROTOR - MATERIALS AND particle size distribution. The milling time
FABRICATION PROCESS was 72 hours. The initial slip preparation
DEVELOPMENT procedures resulted in a slip with low solids
contents (31-66 percent) and having a high
2.1 Introduction tendency to gel, and a very high viscosity (500
cps). A subsequent series of iterative experi-
Net shape fabrication of an AGT101 tur- ments resulted in a more desirable slip which
bine rotor was considered to be the most had a solids content of 78 percent and. a
challenging task of the whole AGT program. viscosity of 120 cps. This involved the
No baseline rotor fabrication technology was reducing of milling time from 72 hours to 24
available at the beginning of the program in hours, the introduction of Daxad 30 as a
1979. Every aspect of materials and process- defloculant and some other improvements.
ing procedures had to be developed. Before reaching these optimized slip proper-
ties, many factorial designed experiments
Initially, a rotor fabrication process was were conducted to explore the effects of
developed based on GTE SN502 Si3N 4 powder. calcination temperature, wet milling versus
In the last two years, GTE has not been able to dry milling, mill jar type, type of defloculants,
produce powders with consistent quality. Con- the use of KOH, etc. on slip viscosity, gella-
sequently, new materials and processing pro- tion tendency, casting rate, green density and
cedures were required. sintered density.
The following section describes the pro- In 1983, a new composition (92 percent
gress made in developing the standard process- GTE SN502 + 6 percent Y203 + 2 percent
ing procedures using GTE SNS02 powder, the A120 3) was introduced. This composition,
material properties of the rotors, and the designated as Code 2, was developed in an
more recent efforts of developing alternate AMMRC contract "Low Cost, New Shape
materials and processing procedures. Ceramic Radial Turbine Program" (Contract
285
No. DAAG46-81-C-0006). Because of the 2.3 Mold Fabrication and Casting Procedure
lower Y203 and Al20 3 contents, this new Developments
composition was believed to have superior high
temperature properties. Optimization experi- Net shape casting of a bladed rotor re-
ments were then conducted to establish a quires an organic shell mold with the correct"
procedure to achieve desirable slip properties. configuration, dimensions, and a defect free
The procedure produced well dispersed slip inner surface. A procedure was developed to
with consistent properties when materials fabricate such a mold from either a rubber
were taken from the same lot of GTE SN502. pattern or a water soluble wax pattern. The
development effort was primarily performed
using rubber patterns. This development was
A series of slips was tested using the
aided by the expertise that exists in the Wax
Haake viscometer at 25C, with the CVl00
Department at ACC.
measuring drive, and the AZ30 measuring sys-
tem. Yield points of the slips were then Several shell mold materials were evalu-
determined on the shear stress versus shear
ated with varying degrees of success. A
strain rate curves. Table 53 seems to indicate
paraffin based material, optimized for wax
a correlation between yield point and gelling.
patterns, was initiallyused for mold fabrica-
This investigation led to the use of the Haake
tion with rubber patterns. The paraffin based
viscometer to control the quality of the slip
material was opaque and did not coat the
prior to casting. blade edges well. Ethyl cellulose and butyrate
based dip seal materials provided transparent
Table 53. Yield Point of Slip and
molds helpful in obtaining complete filland
Quality of Casting.
excellent edge coating but were not dimen-
sionally reproducible. Subsequently a trans-
lucent mold material (LN 209-218) was
Slip Yield
developed, similar to the original material,
ID Point,
with improved edge coating capability.
No. Test Date Pa Remarks
All organic shell molds were produced by

06044 06-13-84 0.14 Good casting
one technique. The reusable pattern is dipped
into the advanced shell mold material, LN
06124 06-19-84 0.13 Good casting 209-218. The mold is built up further by
continued dipping into a molten batch of
05314 06-12-84 0.29 Slip gelled; water-soluble wax. The reusable pattern is
rotor cracked then pulled from the mold and the water-
soluble wax is dissolved away.
04274 05-14-84 0.22 Good casting
The wax mold is placed on a plaster base
which provides the configuration for the dome
04184 05-10-84 0.24 Good casting
and also removes the water from the slip.
Initially there were problems with poor dome
04164 04-26-84 0.24 Good casting
surface quality and the presence of pour lines.
A procedure to treat the plaster mold with a
In April 1985, the characteristics of GTE 0.5 percent ammonium alginate solution was
SN502 Si3N 4 were changed significantly due to established which produces a smooth dome
several changes in the powder manufacturing surface.
process implemented by GTE. This required
another series of process modifications before The initial experimental castings experi-
desirable slip properties were once again ob- enced air entrapments, high porosity and high
tained. concentrations of agglomerates. To solve
286
O-Rb_tN_AL PAGff, I_
OF POOR QUALITy
these problems, a screening and de-airing of The as-cast rotor often exhibited cracks on
the slip immediately before casting was incor- the dome because of a fast casting rate in the
porated. dome area where the slip is in direct contact
with the plaster mold. This cracking problem
In addition, a pouring device was developed
was eliminated by soaking the piaster mold in
whereby slip is withdrawn from the bottom of
water prior to use.
the container. This was to eliminate the
contamination from the foaming foreign sub-
2.4 Drying and Pre-Sintering Developments
stance which always appears on the skin layer
of the slip and produced defects in the initial Cracks often developed during drying of a
experimental castings. ceramic part. This is particularly true for
Incomplete fill near the upper blade tips large size rotors with varying thicknesses.
was one of the major problems when bladed Improper drying also would result in cracking
rotor casting efforts were first introduced. To during pre-sintering or sintering. At the
overcome this problem a procedure to spin the beginning of this program even the bladeless
mold during slip pouring was incorporated. simulated rotors cracked consistently. Figures
Another improvement in casting procedure 340 and 341 show two examples. Conse-
was the control of pouring rate by using a quently, a major effort was devoted to solving
separatory funnel. this problem.
Figure 340. Cracked Dome On a Figure 341. Part After Sintering Cracked
Bladeless Rotor Casting After Into Several Pieces Due to Incomplete
Drying. Drying.
287
The initial experiments included the use of Iniection Molding
a desiccant to absorb moisture from the part
and oven drying at 150F as the final step in
A Tempcraft plunger type injection molder
the drying procedure.
was used in this effort. A large number of
This procedure was not successful even silicon nitride simulated rotors were injected
with the simulated rotors. Another experi- to eyaluate the wide variety of molder adjust-
ment involved drying the part at room temper- ments. These included variations in injection
ature for one day followed by vacuum drying pressure, tool clamp pressure, flow control,
for three days. Again the parts cracked into barrels and nozzle temperature, injection and
several pieces. Additional experimental trials cooling time. In spite of the wide choice of
included drying the parts in a plastic box, with parameter variations, a problem of air entrap-
or without venting holes for a period of up to ment within the rotor persisted.
several weeks and the use of manually con-
trolled humidity oven. This was then followed
It was decided to modify the simulated
by oven drying at 90F. None of these experi-
rotor tool configuration. Originally, the base
ments showed any significant improvements.
portion of the tool was made in two halves in
The major breakthrough occurred when a order to permit future simulated vanes at the
programmable humidity oven was introduced. vertical parting lines. However, it was found
This humidity oven enabled more systematic that the tool could not be clamped with suf-
control of the drying environment. A success- ficient pressure to resist flashing. This caused
ful cycle for the rotor was developed. The release of injection pressure resulting in poor
cycle starts at a high humidity (80 percent surface and incomplete mold fill. A new one-
relative) and low temperature (77F) and ends piece base section was built which eliminated
with a low humidity (20 percent relative) and the flashing problem and improved surface
high temperature (185F). Final drying is cosmetics but did not eliminate air entrap-
accomplished in an air oven due to the tem- ment.
perature limitation of the humidity oven. This
drying cycle shortened the previously un-
The next tool modification designed to
successful long cycle to 9 days and produces
solve the problem of entrapped air was to
consistently crack-free bladed rotors.
provide the capability to evacuate the cavity
Prior to the demonstration of ACCts prior to injection. It was found that the
sinter/HIP capability, dried castings were vacuum had to be turned off simultaneously
shipped to Ford or ASEA for densification. To with material injection or the material would
avoid shipping damages, a pre-sintering cycle flow into the vacuum system requiring ex-
was developed. This sintering cycle was to tensive clean-up. However, even with the
impart sufficient strength to the rotor without ability to evacuate the mold cavity, air within
causing cracking or noticeable sintering. The the injection molder barrel enters the mold
cycle was successfully established with the with the batch material when injected.
100 psi furnace. This pre-sintering cycle has
beome a standard procedure prior to sintering
Originally, the entrapped air in injection
even for rotors sinter/HIP'ped at ACC.
molded simulated rotors was assumed to be
the major reason for the inability to produce a
2.5 Configuration Evolution
O satisfactory rotor. In an effort to determine
Rotor efforts at ACC were initially the actual material flowpath during injection,
directed toward fabrication of both injection a clear plastic was substituted for the Si3N 4.
molded and slip cast rotors, a simulated rotor The material flowed from the backface (where
configuration (bladeless rotor) was used in it entered the tool), moved along the tool wall,
both approaches before a bladed rotor config- and on into the shaft end and sealed off the air
uration was introduced. trapped in the center.
288
As a result of this experiment, muchof the Another modification affected blade thick-
problem encountered with voids in the injec- ness at the saddle area, blade to hub filled
tion molded rotors was traced to the fact that radius, transition changes at the saddle area
the rotor was not cooled sufficiently within between the blades and a belled saddle to hub
the tool to solidify the binder bef6re holding line radius definition. This modified design
pressure was lost. Pressure time was thus was named the "B" generation. Another dif-
increasedfrom 6 minutes to 30 minutes. When ference between the "A" & "B" generation is
the resulting rotor was split open, it was solid the "A" generation utilizes reusable rubber
throughout, showing no evidence of voids. patterns while the "B" generation uses injec-
Subsequently more rotors were injected that tion molded water-soluble wax patterns.
did not have any flaws as determined by visual
inspection and X-ray radiography. However, 2.6 Densifieation Process Development
cracks developed in the rotor in the initial
binder removal experiments. Prior to 1983, the only high temperature
furnace available to this program for densi-
To solve the binder removal problem, many
fication of slip cast parts was the i00 psi N 2
studies were conducted. Initially, a study was furnace at ACC. Many cycles were experi-
performed to determine whether rotor hub mented with in this furnace using Code I
size is a critical factor. For this evaluation, composition test plates as well as rotors. The
crack-free molded bladeless rotors were
theoretical density of Code 1 is 3.30 g/cm 3
reduced in size by lathe turning and evaluated (based on the rule of mixtures). Typically the
in the vacuum dewax cycle. Again, dewax test plates achieved densities in the range of
cycles resulted in rotor cracking, indicating 3.18-3.22 g/cm 3, while the rotors could only
that rotor size was not a major controlling achieve 2.78-3.04 g/cm 3. Parameters investi-
variable.
gated included soak temperature and time,
Additional experiments using various heating rate, furnace load, and kiln furniture
thermal cycles in a nitrogen atmosphereat I0- history. In addition to the sample densities
12 psig pressure were tried, but results were (plate versus rotor) effect noted above, it was
inconclusive. The injection molding approach found that kiln furnitures previously exposed
wasdiscontinued in 1981. to an oxidizing environment had a detrimental
effect on densification. It was also found that
Slip Casting "samples with lower weight loss exhibited high
strength at room temperature.
The very first castings were made using
the bladeless rotor or simulated rotor con- Due to the low densities obtained at ACC
figuration. The first casting was made in a using the 100 psi furnace, an arrangement was
metal mold with a plaster base. In subsequent made to densify rotors at Ford and ASEA.
castings, the metal mold was replaced by Ford has a 1500 psi furnace and ASEA has
plaster and then by rubber. glass frit encapsulation plus HIP'ping tech-
nology. Initially Code 1 rotors could be densi-
Bladed rotors were fabricated using rubber
fled only to 3.20 g/cm 3 at Ford. As a result of
patterns based on an actual metal rotor until a
continued materials/process development at
metal pattern with allowance for 14.5 percent
ACC and cooperative efforts at Ford, bladed
sintering shrinkage was obtained. Scale-up
rotors were sintered to 3.27 g/cm 3 in Novem-
was successful. All castings, by November
ber 1982 and further increased to 3.30 g/cm 3
1981, were aimed at producing parts with the
in October 1984. Similar densities were also
14.5 percent shrinkage. obtained in rotors densified at ASEA. In June
Mold design changes were made in 1985. 1983, ACC had demonstrated a capability to
The present rubber pattern was changed from sinter/HIP bladed rotors to 3.26 g/cm 3 (Code 1
14-5 percent (AI generation) to 17.0 percent and Code 2). The Code 2 composition has a
(A2 generation) shrinkage allowance. theoretical density of 3.26 g/cm _. In 1984 the
289
densification processes at ACC, Ford and One of the original slip-cast mold config-
ASEA all could densify rotors to densities urations consisted of a plaster base that con-
higher than 98.8 percent of theoretical. As a tained the backface shape and an all-rubber,
result of the good ACC sinter/HIP densifica- tw.o-piece mold for the hub and shaft. An
tion results, shipments of rotors to Ford and iteration designed to reduce the length of
ASEA were discontinued. casting time involved changing the all rubber
portion of the mold to one-half rubber and
2.7 Alternate Raw Materials Development one-half plaster. Several rotors were cast in
this dual-material mold, but the portion of the
Limited availability, inconsistency in
casting in contact with the rubber face ex-
quality, and the higher cost of GTE SN502 hibited what could be classified as an out-
Si3N 4 powder have been the impetus for the gassing condition with extensive surface
search for alternate sources of Si3N 4. Various
porosity. A thin wax coating was applied to
silicon nitrides were investigated as potential
the rubber face portion of the mold, which
replacements for GTE Si3N 4. Some of the
improved the condition only marginally. The
results are described in the following para-
casting surface in contact with the plaster was
graphs. excellent.
Denka 9S Si3N 4 was milled with additives, Another mold was prepared that was en-
prepared as slip and cast into rotors. Sintered
tirely plaster. The shaft portion was coated to
rotors yielded final densities of 3.23 to 3.26
render it non-absorbent in order to prevent
g/cm 3. However, more optimization experi-
preliminary casting in this area inasmuch as
ments are required before a stable and con-
slip is poured through the shaft end of the
sistent slip can be routinely prepared.
mold. The casting appeared to be free of
UBE SN E02 and SN E05 Si3N 4 were evalu- flaws both visually and by X-ray radiography.
ated as a replacement for GTE. 100 percent The above two experiments were conducted in
SN E02, 100 percent SN E05 and blends of E02 1981 on simulated rotors.
and E05 (25-75, 50-50, and 75-25, percent by
An all plaster mold using plaster inserts
weight) all exhibited lower viscosities and
between the blades was tried. For the purpose
faster casting rates than baseline castings.
of initial evaluation, the plaster inserts were
However, all developed cracks prior to
made individually without proper tooling.
sintering. Most blades were damaged while the inserts
were being removed. It suggested that pre-
Starck H-2 Si3N 4 showed little dis-
coloration on rotor blades but was difficult to cisely fabricated piaster inserts and a con-
make into a consistent slip. A new procedure trolled plaster insert removal mechanism
was recommended by Allied Central Lab. This should be developed before the potential of
this technique could be truly evaluated.
involved preparing the slips of Starck Si3N 4
and the premilled additives (Y203 and AI20 3) An inverted wax shell mold, to form the
separately and then blending them together blades, hub, and shaft was attached to a
after adjusting them to a common pH. How-
plaster base. A wax pattern dome with a
ever, slips continued to be very viscous. No
funnel at the top center was welded to the
improvement in slip viscosity occurred after
upper wax shell that formed the blades (1981).
pH adjustments and lower solids content.
This technique resulted in visually crack free
rotors but it required a much longer casting
2.8 Alternate Casting Procedures
time. In addition, it was difficult to form a
Investigated
precise dome contour. The wax dome in the
Long casting time, poor casting surface above arrangement was replaced by a plaster
quality, incomplete fill and discoloration plate with the dome contour. The slip was fed
problem were addressed by trying a variety of in both from the top and bottom, in separate
alternate casting techniques or mold designs. experiments; all castings were cracked (1985).
290
Evacuation through the plaster mold base I) Analysis
was aimed at reducing casting time, especially
Many organizations have been contributing
for the B generation rotor configuration. This to this effort. Table 54 summarizes the
approach was discontinued due to the lack of results.
proper equipment.
o Garrett - SEM/EDX t 1982
Hydrostatic pressure casting was accom- Green as well as sintered blades were
plished by placing a just poured casting and examined (Figure 342). The discolored
mold into a 1500psi maximum pressurevessel.
area appears denser in the green state
The vessel was then pressurized to various and contains heavier atomic number
pressures and held for different lengths of elements.
time without heating. No obvious improve-
ment in rotor quality wasobserved. EDX (energy dispersive X-ray) was used
to compare the discolored area with the
2.9 Blade Discoloration normal area. EDX shows that the dis-
colored area has higher concentrations
of Al and Y. It should be noted that
As the casting procedures were established EDX at Garrett could not, in 1982, de-
and drying cracks had been routinely elimi- tect elements with atomic numbers less
nated, more attention was paid to other than II (sodium). In the same investi-
aspects of rotor fabrication problems, such as
gation, a flower-like structure was
a blade discoloration. An interesting observa-
found in the discolored region. EDX
tion was first made in 1981 that a darker band found that the flower structure con-
always appeared near the tip of each blade.
tained higher concentrations of Mo and
This discoloration first shows up in the green Y than the normal area. This was the
rotors as a dark grey color. Upon pre-
only time that Mo was found in the
sintering this discoloration turns to black material.
while the base material becomes white. After
sintering, this discoloration remains black o Seal Lab - SEM/EDX
while the base material turns to the normal
dark grey. When examined under a 30X micro- Several green rotor blades were
scope it was found that, after sintering, there analyzed at Seal Lab with a more sensi-
were always mud-like cracks associated with tive EDX (wavelength based) in an
the discolored area. These mud-like cracks attempt to compare the concentrations
were judged to be the cause of failure in the of lower atomic number elements
spin test performed at Garrett. Consequently, (N<II) between the discolored and
the elimination of the discoloration was iden- normal areas. Several energy levels
tified as a major program objective. were used. Table 55 lists the results.
It shows that oxygen, Na (or Zn) and Y
were detected or existed in higher con-
Two parallel approaches were taken to
centrations in the discolored region
combat this discoloration problem, namely I) that in the normal area. Carbon was
analysis and 2) process variation. The first
only found to be higher in the dis-
approach was aimed at determining, chemi-
colored region at 20 KV, thus is con-
cally and microscopically, the difference sidered inconclusive.
between the discolored area and the normal
area. The second approach was empirically to o Oak Ridge National Lab (ORNL) - Super
observe the effect of processing parameters Microprobe
on the degree of discoloration and the move-
ment of the discolored area. Some major A sintered rotor blade was examined at
findings are described in the following: ORNL using their high sensitivity super
291
Table 54. AGT1Ol Rotor Discoloration Analysis.
Discolored Area
Compared with
Analyzed by Method of Analysis Normal Area Sta_e of Process
Garrett ('1982) SEM-EDX Higher AI, Green (AGT)
Higher A1 and Y Sintered (AGT)
ORNL (November 1985) Super Microprobe Higher Y Sintered (T-2)
Allied (March 1986) SAXPS (Small Area X- Lower Y203 Pre-Sintered (AGT)
Ray Photo Electron
Spectroscopy)
SEM-EDS Higher AI and Y Green (T-2)
Higher AI Pre-Sintered (AGT)
Higher A1 and Y Sintered (AGT)
Seal Lab
(September 1984) SEM-EDX O, NaJZn, Y Green (AGT)
NASA (March 1986) SEM-EDX Sintered (AGT)
*No elements were significantly higher in the discolored area.
Figure 342. Green Vane Tip Showing Discolored Area.
ORIGINAL PAGE IS
292
QI_ £OOK QUAL!TM
Table 55. Diseoloration Analysis of a niques. Although there were some dif-
Green Rotor. ferences in results found by different
techniques, the most consistent results
are that the discolored area contains
higher Al and Y concentrations than the
Elements Deteeted
EDX normal area.
Voltage Discolored Normal
(KV) Area Area 2) Process Variation
Several experiments were conducted to

4 C, O, Na/ C, Si determine if the cause of the blade dis-
Zn, Si, Y
coloration was a result of a processing
problem.
7 C, O, Na/ C, Si, A1
Zn, Si, Y, A1 o Rubber Lined Steel Mill Jars Versus
AI2_03 Miller Jar
2O C, O, Na/ Si, A1
Zn, Si, Y, A1 To avoid undesirable oxide contamina-
tions, rubber lined steel mill jars with
39 Si, Y, AI Si, Al, Y Si3N 4 mill media were used to prepare
the powder. Due to the darker color of
microprobe. The mud-like crack fea- the discoloration,it was suggested that
ture was found similar to the one the rubber lining might be the source of
reported by Garrett. The super micro- the black substance. The rubber liner
probe showed that the only difference steel mill jar was replaced by porcelain
between the discolored region versus mills (primarily A1203). The discolora-
the normal region is the Y concentration problem stillexists.
tion being higher in the discolored
region. o Upside Down Castinff
The discoloration showed up on the

o NASA-Lewis Research Center
back face of the blade instead of on the
top face near the tip. This suggested
" Both pre-sintered and fully sintered that the discolored substance is lower
blades were analyzed by several tech-
in density tending to float to the top,
niques. Similar mud-like cracks were
then is trapped on the curved portion of
found in the discolored region. How-
the blade. However, finer particles
ever, river-like cracks were found in
also have a tendency to float. Thus,
the back face of a blade examined. Y
the discolored area might have a higher
concentration was found to be segre-
concentration of finer particles which
gated in small islands which were uni-
formly distributed throughout the may turn out to be Al203. I
blade. No detectable difference o Spin Casting Versus Stationary Casting
between the discolored area and the
No difference was found between spin
normal area was found.
cast rotors and stationary cast rotors.
In both cases, blade discoloration was
o Allied Central Research Lab (ACRL) - evident.
Methods (1986)
o Reversed Spinning
Green, presintered and sintered blades
were analyzed by X-ray diffraction A rotor was spin cast with {he rotation
(XRD), X-ray photoelectron spectros- direction reversed. Again, blade dis-
copy (XPS) and SEM - EMPA tech- coloration was present.
293
This experiment and the above experi- properties through the rotor. The baseline
ment indicated that the discoloration process has produced rotors with very uni-
phenomenon occurred after the mold form PSD in the green rotors. To verify
_as filled, i.e., during the casting this particle size distribution measure-
period. The segregation or migration of ments were made at different locations on
certain species which resulted in dis- several rotors to determine whether there
coloration occurred during the casting is any variation in particle size distribution
time which ranged from one to four within dried green rotors. The results from
days. one randomly selected rotor showed that
no significant difference in PSD was found
o Plaster Patehed Blade Tip Mold within the rotor (Table 56).
One-half of the blade edges of a wax Table 56. Particle Size Variation
mold were cut off and then sealed with Within a Green Rotor.
plaster. This was to provide faster
casting time (minimizing materials
migration/segregation) near the blade Pereentiles of Particle
area where discoloration occurred. Size, l_m
This experiment resulted in a broader
band of darker color parallel to the Loeation 90th 50th 10th
plaster seal on each blade which was Dome 2.62 0.90 0.36
sealed with plaster. The remaining Leading Edge 2.61 0.90 0.36
unmodified blade of the same rotor Blade 2.64 0.90 0.36
showed the usual pattern of discolora- Trailing Edge 2.58 0.90 0.36
tion. Shaft 2.60 0.89 0.36
o Additive Effeets Rotor Deliveries
A rotor was cast with no additives in During the AGT program, 50 turbine rotors
the slip. No discoloration was observed were shipped to Garrett. Many of these rotors
on the blades. were spin tested and exceeded the 115K rpm
proof test requirement. None of the rotors
Another rotor was cast with a slip con- could be considered "engine quality" due to
taining Y203 but no Al203. It also did missing blades, surface porosity, and dimen-
not show any discoloration. sional deviations from the print.
The above two experiments seem to 3.0 CERAMIC STRUCTURES- MATERIALS

suggest that Al203 causes discolora- AND PROCESSES DEVELOPMENT
tion.
3.1 Summary !
Subsequent castings with reduced The objective of this portion of the
AI203 content (1 percent instead of the AGT101 program at ACC was to supply static
baseline 2 percent), no noticeable components to Garrett in support of the
reduction in discoloration was observed. engine test program. At the outset, ACC was
responsible for providing the turbine shroud,
2.10 Rotor Partiele Size Distribution transition duct, turbine stator, turbine baffle,
turbine inner diffuser and the turbine outer
A uniform particle size distribution (PSD) diffuser housing. Since that time, other com-
in a green rotor would normally result in ponents were added which included the turbine
uniform sintering shrinkage (thus no distor- backshroud, flow separator housing, wave
tion), microstructure and mechanical spring and flow separator housing seal rings.
294
The stator vanes were to be injection
molded, and all the other parts were to be slip Other parameters also affect slip
cast reaction-bonded Si3N4. The technology behaviors. Experiments aimed toward density
baseat ACC prior to the start of this program control of silicon slips prepared with various
was the establishment of a silicon slip, desig- fractions of air classified powder were per-
nated as RBN-104, and the demonstration of formed. Limits exist between pH and green
the net shape casting capability of various density such that the controllable range of
small engine components under several green density by pH and adjustment alone
government programs. It was initially appears to be between 1.65 and 1.75 g/cm 3.
assumedthat this baseline technology could be
transferred to the fabrication of large AGT It was also recognized that small amounts
components without any significant develop- of impurities such as calcium, deleteriously
mental efforts. As the program proceeded, it affected slip rheology and subsequent casting
became apparent that improved materials and and nitriding results.
processing procedures would need to be
developed. It was also found that slip casting 3.4 Castin_ Procedure Development
and other processingsteps were very sensitive
to tooling/design changes. A flow chart of the overall slip casting
process is shown in Figure 343.
3.2 Materials Development
One slip casting iteration was tr}ed in
efforts to eliminate center line shrink in the
The original baseline compositions were
backshroud. The backshroud was cast to a
established using silicon powder supplied by
thickness of one to one and a half inch using a
Kawecki Berylco Industries (KBI). Later in the
net shape form one surface of the backshroud.
program KBI discontinued operations necessi-
The backshroud plaster mold to an otherwise
tating evaluation of other sources of supply.
solid ingot, as is shown in Figure 344 and 345.
Silicon was then obtained in lump form as ore
This is an effective method to produce the
directly from the mine which reportedly sup-
backshroud avoiding flow lines, centerline
plied KBI's material. The ore was pulverized
shrink and air entrapment, but requires con-
by an outside source and returned in powder
siderable green machining.
form. Analysis revealed high impurity levels
and slip preparation was difficult and not
3.5 Drying and Prenitriding
reproducible. Other sources of high purity
silicon powder were evaluated and Kemanord,
Drying of large, slip cast components,
Sweden, was finally selected as most reliable
especially those with non-uniform wall thick-
with respect to purity, consistency of product
ness, has proven to be one of the more crucial
and availability of supply.
steps in processing. Extensive evaluation of
drying and prenitriding cycles were conducted.
3.3 Pre-Casting Materials Preparation It is apparent that components such as the
Development turbine shroud and transition duct must remain
in the plaster mold after casting for a length
Materials preparation prior to actual cast- of time to achieve good green strength. Sub-
ing underwent continuous evaluation. Particle sequent to removal from the mold, it is essen-
size distribution (PSD) and surface area (SA) tial that the casting be permitted to dry
of silicon powder are important parameters in slowly and completely before attempting to
controlling slip rheology. Experiments were prenitride. Several techniques have been used
conducted using the air classifier to separate which include controlled humidity drying, dry-
the powder into different PSD fractions and ing in an enclosure and subsequent exposure to
observing slip behavior as a function of warm dry air, slowly heating in an oven over a
particle size. long period of time to the boiling point of
295
HI-PURITY Si
I
PATTERN
PREPARAT ON I MILLGRIND [ I CHEMICAL
ANALYSIS I CASTING I
I I I I
MOLD
[PREPARATONI
[
I
PLASTER
MIXING I PREPARATIONI
I
i PARTICLE
SIZE DISTRI8
I
I MOLD
REMOVAL
i
I
I
I REMOVAL
PATTERN I I SURFACE
AREA I DRY I
I SLIP AGING
I
DRYING SINTER _ NDE 1

I
MEASURE GREEN
VISCOSITY, DENSITY
NITRIDED
pH, SOLIDS AND MACHINE : I DENSITY
CONTENT WEIGHT GREEN I AND
MEASURE.
MOUN?MOLD/
I % WEIGHT
NIT_IDE GAIN
TO PLASTER
I
I FILTER
EVACUATE
AND
I _D_I
I ! I
I SHIP I
Figure 343. Flow Chart Showing Slip Casting Procedure.
296
GlhlGINAL PACE IS
OF POOR Q[_,.,_,.I.1TY
Figure 344. Plaster Base and Mylar Ring

for Casting Solid Ingot. Figure 345. Ingot After Casting.
water and, allowing the part to remain in Plates were cast to determine the effect
vacuum for several days prior to prenitriding. of nitriding aid. Compositions investigated
Several backshrouds and some transition ducts include the baseline 3 percent Fe203, 1.5
have survived prenitriding and machining but percent Cr203 and 1.5 percent Fe203 plus 1.5
the turbine shroud continues to crack during percent Cr203.
prenitriding. The nitriding aid billets were machined
into test bars. One-half of each of the speci-
3.6 Nitriding Development
mens was nitrided in staddard nitriding cycle
Early work investigation nitriding capa- and the remaining half was nitrided in an
bility included nitriding thick sections, atmosphere containing a 4 percent helium
nitriding aids, and the effect of a 4-percent addition.
helium addition to the normal H2-N 2 nitriding The results are tabulated below.
atmosphere.
H2-N 2 H2-N2-He
Test plates of RBN-104 material were cast
Nitriding Aid MOR MOR
in thickness ranging from 1/4 inch to 1 3/4 Chemistries (ksi) (ksi)
inches to evaluate the nitriding capability of
the standard cycle. These plates were nitrided Baseline (RBN-104, 47.0 47.0
and were evaluated for the presence of 3 percent FeO 2)
unreacted silicon. The nitriding of thick sec- 1.5 percent Ye203 39.7 41.7
tions indicated that sections in excess of 1/2 1.5 percent CR203 38.8 38.8
inch thickness contained mostly unreacted sili- 1.5 percent Fe203 +
con in the center. However, these evaluations 1.5 percent Cr20 3 41.8 41.0
were conducted with little mass of material in
the furnace and a relatively short cycle. Since The new nitriding aids and nitriding cycle
then it has ben found that nitriding is far more showed no improvement over the standard
efficient when a large load mass and longer nitriding cycle and the baseline material, RBN
cycles are used. 104.
297
This PageIntentionally Left Blank
298
APPENDIX B
FORD MOTOR COMPANY

FINAL REPORT
INTRODUCTION ment of the ceramic turbine rotor, the one-

piece ceramic stator and the ceramic flow
Prior to initiation of the AGT program in separator housing. As the AGT Program
1979, Ford was involved in the early develop- evolved, the Ford efforts on the 9ne-piece
ment of the various aspects of ceramic turbine stator and the flow separator housing tasks
technology. Much of this work was conducted were terminated, because the work had
under an ARPA-sponsored program entitled reached a satisfactory plateau and the overall
"Brittle Materials Design, High Temperature program was faced with funding reductions. In
Gas Turbine ''(I).The broad objective of this the critical ceramic rotor task, Fo_'d retained
program was to encourage wider use of cer- its interest throughout the program, though
amic materials in engineering systems; the there were, from time to time, budgetary
more specific objective was to develop cer- reductions in all aspects of the program.
amic materials, designs, processes and test
methods for a high temperature, automotive In this report, the ceramic materials and
gas turbine and demonstrate 200 hours of fabrication processes activities are covered
ceramic component durability. Even though according to the task breakdown in the sub-
these goals were met on each of the ceramic contract, as shown below:
flow path components, including the rotor, it
became evident that development of the Task 2.3 - Ceramic Rotor
highly stressed, ceramic turbine rotor for
widespread application was a most difficult Task 2.7- Ceramic Stator and Flow Sepa-
and formidable task. A recommendation was rator Housing
made for a focussed research program that
would investigate and develop materials, Task 2.3 - Ceramic Rotor
processes and designs and define reliability
limits to establish"realizable potential" in this Introduction
criticalcomponent.
Pressureless sintered silicon nitride was
In the AGT101 program, appreciable re- the material selected for the fabrication of
sources were allocated to the development of monolithic rotors for the AGT101 program.
this critical component. Realizing the impor- Two processes were proposed: the primary
tance of ceramic materials, fabrication approach utilized sintered reaction bonded
processes and components, as well as develop- siliconnitride (SRBSN), with the more conven-
ment of the regenerator system, Ford teamed tional sintered silicon nitride powder (SSN)
up with the Garrett Turbine Engine Company serving as a backup material.
as a subcontractor on the DOE/NASA/AGT101
Program. This report appendix covers activ- In both approaches, various developmental
ities in the ceramic materials and processes tasks had to be completed prior to producing a
area. turbine rotor. These tasks included material
and process improvement efforts. The fugi-
Initially, the ceramic materials/processes tive wax slip casting method was employed to
tasks included materials and process develop- fabricate rotors from both materials.
299
PRECEDING PAGE BLANK NOT FILMED
SRBSN Material Development and for the highly stressed turbine rotor. Yttrium
Characterization oxide was selected as the sintering aid for
these materials as good strength at high tem-
Sintered Reaction Bonded Silicon Nitride perature was retained by early hot pressed
(SRBSN) describes a class of material combin- silicon nitrides containing this additive(3).
ing reaction bonded silicon nitride and sintered However, the Y203 - Si2N 4 system is sus-
silicon nitride technologies. ceptible to oxidation, particularly with applied
stress at intermediate temperature regimes.
The consolidation process, outlined in Fig- Consequently, time-dependent property deter-
ure 346, begins with the addition of a sintering mination also became an important aspect of
aid to a fine-grained, high purity silicon the material characterization. These two
powder. Components are fabricated from this characteristics, strength (flexural stress
mixture using standard green forming rupture) and oxidation resistance, dominated
processes including slip casting and injection the evaluation program and determined the
molding. The part is then nitrided and sin- developmental iterationsof this material.
tered to near-theoretical density.
Four grades of slip cast, yttria-doped
SRBSNs were developed at Ford, code
numbered RM-0 RM-I(4), RM-2(5) and RM-3.
FORD SRBSN PROCESS The properties of each of these grades are
summarized in Table 57.
SILICON POWDER + SlNTERING AID[
! RM-0 Material
[ FABR,CATECOMPONENT] The first generation material (RM-0 had

I potentially high strength and Weibull modulus,
despite significant batch-to-batch inconsis-
tencies. Elevated temperature strength was
I acceptable. Oxidation behavior was evaluated
at 700 and 1000C, the temperatures described
in the literature as most critical in yttrium-
containing silicon nitrides. Oxidation tests
Figure 346. Ford SRBSN Process. were performed for 300 hours in a static air
environment. Visual inspection revealed sur-
The inherent advantages of this process are face discoloration, although no cracking had
availabilityand low cost of the raw material occurred. This was a significantimprovement
(siliconrather than silicon nitride)and lower over the earlier Y203/Si3N 4 which cracked
sintering shrinkage. severely after short exposure (<29 hours) at
these temperatures (6).
Sintered reaction bonded silicon nitride has
been under development at Ford since the late RM-1 Material
1970s(2). In the interim, SRBSN has undergone
many iterative improvements in both material The objective of the development of this
and fabrication processes, resulting in signifi- material was reduction of the batch-to-batch
cantly improved microstructural and mechan- strength variabilityexhibited by RM-0. This
ical properites. This report will describe the was accomplished through modification of the
material characterization techniques em- sintering cycle. The RM-I material had high
ployed to identify the problem areas which room temperature strength (91.5 ksi) with a
prompted each iteration. Weibull modulus of 13.3. These characteristic
values are a compilation of a number of sin-
Slip cast SRBSN was developed for use in tering runs. Batch-to-batch variability was
the advanced gas turbine engine, specifically significantlyreduced.
300
Table 57. Slip Cast SRBSN Material Properties.
i,
1979 1980 1981 1982

RM-0 RM-1 RM-2 RM-3
Density (g/cm_ 3) 3.05 - 3.31 3.20 - 3.25 3.26 -3.31 3.26 - 3.34
MOR (ksi) and Weibull

Room Temperature 91.5 (13.3) lO7.6 (21.1) 98.6 (19.0)
1000C 77.6 (10.9)

1200C 81.9 (14.9) 91.7 (15.4) 71.6 (12.9)
1400C 70.8 (11.3)
Oxidation MOR
700C No cracking -58% (700 hr) 0% (700 hr) +7% (300h r)

1000C No cracking -4% (300 hr) -15% (700 hr) -5% (300h r)
1200C -7% (300 hr) -10% (364hr)
1400C -8% (300 hr) -16% (300h r)
Stress Rupture
1000 - 1200C Ul; <40 ksi at

to 90 ksi: No 8000 - 1000C:
time dependent No time de-
failures (600 pendent fail-
ures
1000C, 20 ksi
time dependent
failure (57 hrs
avg)
Oxidation testing was again performed at ments. Again, no visible cracking was ob-
700 and 1000C. Exposure times were in- served under any test condition; however,
discoloration was noted. These data suggested
creased, up to 700 hours, and two new tests
wei-e introduced. Oxidation kinetics were that an oxidation problem stillexisted.
determined by sequential weight gain mea-
surements and thestrength after oxidation
was ascertained. This testing revealed that Residual strength after oxidation exposure
the reaction kinetics were different for each was measured. Figure 348 shows the strength
test temperature (Figure 347). At 700C, the decreasing substantially when samples were
kinetics were linear, and influenced by the oxidized at 700C, but deteriorating only
testing sequence, while at 1000C, parabolic slightlyafter the 1000C exposure. Addition-
behavior was observed. Lower weight gains ally, the strength decrease at 700C was
were recorded for samples exposed for 700 dependent on the testing sequence, as evi-
hours without removal for weight measure- denced by the uninterrupted data point. A
301
120
RM-I Uninterrupted
12
z
O
N 700°C
E 80
o IO ¢:3
X
E 0
n-
8 UJ
__ 60
W t,.
-1-
6 I000 °C I-
Lu (.9
(.9 z 40
Z tU
OC
"t- k- m
O .4 (/3
I-
2O
700°C Uninterrupted
O I 1 I I I I
0 0 200 400 600 800 I000 12OO 140C
0 IOO 200 300 400 500 6OO 700
TEMPERATURE (°C)
TIME (hr.) Figure 348. Strength After Oxidation
Figure 347. Oxidation Kinetics in RM-20. of RM-I.
correlation was obtained between the oxi- attributed to a superior microstructure con-
dation kinetic and the strength after exposure. sisting of a matrix of interlocking needle-like
The strength dependence on test conditions grains (Figure 349), a structure found to result
also suggested a potential thermal shock prob- in optimum properties in hot pressed silicon
lem. nitride(7). Increased process control also con-
tributed significantly to these improved
RM-2 Material results.
The objective of this development iteration The oxidation kinetics at 700C are para-
bolic, as are the 1000, 1200 and 1400C data
was improvement of the low temperature
properties of the material as a function of (Figure 350). However, the 1000C data con-
tinue to reflect anomolous behavior with
both the oxidation kinetics and factors con-
noticeable weight gains. The strength- after-
trolling the strength reduction after oxidation.
oxiation data (Figure 348) show that the
This was accomplished through control of the
secondary Y-Si-O-N phases present in the severe 700C problem of RM-I was solved; only
a minor strength reduction was measured at
material via further refinements in sintering
1000C.
conditions and enhanced quality control.
Testing to this point revealed no serious
This material itertion yielded a higher problems with the RM-2 material. Additional
characteristic strength (107 ksi) and Weibull testing was undertaken to examine the mate-
modulus (21.I). These improvements were rial'sthermal shock resistance (in response to
302
l
!
1.5
I
B
"-. 1.0
1200 %
Figure 349. Microstructure of RM-2. 700°C
1-
test results obtained for RM-I) and stress __
I,i 1000 "C
ol
rupture behavior.
The thermal shock resistance of the mate-

rial was measured by rapid cycling from room
temperature to 1000 and 1200C using the I I I I I
procedure by Johnson and Hartsock(8). 0 I00 200 300
Although heating and cooling rates exceeding TIME (hr.)
2400C/min were employed, no visible damage
was observed. Fracture evaluation after 1,000
Figure 350. Oxidation Kinetics in RM-2.
and I0,000 cycles of thermal shock testing
showed no strength degradation. SRBSN is subject to a stress-enhanced degra-
dation due to oxidation similar to stress cor-
Stress rupture tests were conducted in rosion cracking, a finding not previously
four-point flexure at temperatures between reported. Again a correlation was found
i000 and 1200C. Between 1100 and 1200C, no relating this strength reduction to another
time-dependent failures were observed for oxidation behavior. Complete details regard-
stress up to 90 ksi (specimens tested in ing strength characterization of RM-2 were
stepped stress rupture fashion) and times published recently by Govila, Mangels and
exceeding 600 hours. However, the testing at Baer (9).
1000C revealed another problem. Failures
occurred in an average period of 357 hours RM-3 Material
(based on an average of 10 tests; data ranged
from 19 hours to 93 hours) at a stress level of
only 20 ksi. While a slight strength reduction The goal in developing this material was
was observed after static oxidation (15 per- improvement in the stress rupture character-
cent), stress rupture testing in air revealed a istics at intermediate temperatures, without
reduction in strength of approximately 80 per- degrading the strength (Weibull modulus or
cent. In vacuum, there were no failures at microstructure of RM-2 SRBSN. This was
1000C and 40 ksi (twice the stress levels fo accomplished through the use of a post-sin-
the air tests). These results revealed that tering heat treatment'of the MR-2 SRBSN.
303
The strength and Weibull modulus are 100
ksi and 19.0 respectively, essentiallythe same I000°C
as in RM-2. Again, these data consolidate the
results of numerous batches. Figure 351 com-
pares the strength/temperature data for
RM-I, RM-2 and RM-3, showing slightlylower
strengths for the RM-3. A
E
120 o 1.0
tU
I00 n-
LU
Z
1400 =C
"I"
... 75 0
m
Or) RM -3 "I"
n- LU 1200°C
O
=E
50-
700°C
0 I I I
0 I00 200 50,
25- TIME (hr.)
Figure 352. Oxidation Kinetics in RM-3.
ments, oxidation kinetics and strength after

0 I I I I I I oxidation, reflect improvements over allother
0 200 400 600 800 I000 1200 1400 grades of SRBSN's tested.
TEMPERATURE (=C)
The stress rupture behavior of RM-3 was
Figures 351. Strength at Temperature evaluated between temperature of 800 and
Data for RM-1, 2 and 3. 1000C and stress levels between 40 and 62 ksi.
Limited testing (800C at 62 Ksi for 100 hrs.
The oxidation kinetics for RM-3 are shown and 1000C at up to 40 ksi for 500 hrs) pro-
in Figure 352. The data at all temperatures duced no time-dependent failures. Detailed
describe parabolic curves, with the weight strength characterization of the RM-3 mate-
gains generally inereaisng with temperature, rialwas recently reported by Govila (10).
as expected, for a dense material. The initial
weight gain is a result of the post-sintering The properties of RM-3 exceed the criteria
heat treatment given to this material. The for the AGT101 turbine rotor as defined by the
strength after oxidation exposure is shown in Garrett Engineering Project, who are respon-
Figure 348 as a function of temperature. The sible for overall engine development. It has
data show no degradation in strenght for tem- good strength and Weibull modulus, consistent
peratures up to 1000C with only modest losses oxidation behavior with low weight gains and
at 1200 and 1400C. Both of these measure- experience& no degradation with oxidation
304
exposure. Additionally, it presents no time- stant, and the primary objectives of the pro-
dependent stress rupture problems for stress gram were:
levels up to 40 ksi and temperatures betwen
800 and 1000C. i) Development of process techniques for
casting crack-free turbine rotors using the
While RM-3 exhibits significantly improved fugitive-wax process.
properties over earlier SRBSN materials,
further advances are still possible and appear 2) Fabrication of rotors of the RM-20 mate
desirable. Post-sintering heat treatment rial with identical physical properties to

those of test bars.
results in a coating and the material becomes
subject to the limitations imposed by coatings.
The principal drawback is nonuniformity of h) RM-20 Silicon Nitride Process
coating thicknes which could lead to nonuni-
RM-20 is a derivative of a silicon nitride-
form oxidation of the base material and pre-
mature time-dependent failures. Non-uniform based material originally developed by Ford
coatings are also a potential source of defects for cutting tools and other wear type appli-
resulting in strength reduction. This may cations. This material is made by the pres-
account for the slightly lower room temper- sureless sintering of a silicon nitride chemical
ature strength and the reduced high temper- composition. The process concept involves a
ature strength of RM-3 relative to RM-2, the form of SRBSN, and the use of yttria as a
matrix material. major sintering aid. Property data based on
test bar specimens (Table 58), and the inherent
process flexibility of RM-20 indicate that this
SSN Material Development_ Characterization_ material is suitable for slip casting of the AGT
Fabrication and Testing rotor.
a) RM-20 Introduction Table 58. Physical Properties RM-20

Material.*
Prior to 1985, SRBSN was the prime can-
didate for AGT101 turbine rotor fabrication at
Ford. Inherently, this material was preferred Flexural Strength (RT) 126 ksi
because its low sintering shrinkage would Weibull Modulus 13
permit the fabrication of rotors to near net
shape. Efforts to fabricate crack-free rotors Density 3.29 g/era 3
using the silicon-water slip cast system, how-
Strength (2200F) 82 ksi
ever, were mostly unsuccessful. Problems
were encountered with stability of the silicon-
water base slip, and as a result of these *Strength data are reported here for a cold
problems, and the inherent weakness of the pressed material nitrided and sintered.
silicon castings, SRBSN components could not Complete processing details are given else-
be made free of blade cracks. where (11).
Material iteration RM-20, a sinterable sil-

icon nitride (SSN), provided a solution to some Figure 353 shows the process for fabrica-
of the problems encountered with SRBSN. ting RM-20 SSN gas turbine rotors. This
process consists of dry ball milling selected
The following sections characterize the silicon powders and additives to produce an
RM-20 SSN material, and describe the Ford intimate mix and cold pressing round plates
effort associated with fabricating a crack-free under unilaxial pressure. These plates are
AGT rotor. During this phase of the program nitrided to yield silicon nitride contents
the rotor material composition was held con- greater than 89 percent by weight and
305
II
SILICON POWDER i_:

ADDITIVES
PREFORM
i,
NITRIDE
fill l
COMMINUTION
FABRICATION
(SLIP CASTING)
BISQUE FIRE
GREEN MACHIME
IZ,, ,ECC I ' CThlT

F'RESSURELESS S I N TER _,T_
SIN]ER £NG HOT ISOSTATIC (HIP)
_ t_ i_
INSPECT
ii
SPIN TEST
Figure 353. Sintered Silicon Nitride Process for Fabrication

of Advanced Gas Turbine Rotors.
second phase oxynitride crystallites of mechanisms. These deflocculant systems are

Y10Si6024N2 composition. Thereafter, the covered in the following section of RM-20 slip
plates are broken into coarse particles (14 preparation.
mesh) suitable for primary reduction by roll
crushing. As in the case of SRBSN, slip casting -
combined with the fugitive wax process - was
Ball milling is sued to further reduce the used to fabricate RM-20 rotors. After slip
particle size after roll crushing. By most casting, rotors are dried to remove the resid-
standards the resultant powder is relatively ual water, bisque fired for green strength, pre-
coarse for slip castings, but a stable slip can machined and sintered. Both pressureless sin-
be prepared using deflocculant systems that tering and hot isostatic pressure were em-
provide both physical and chemical suspension ployed.
306
b) RM-20 Rotor Fabrication suspension, would break down over a long
period of time, resulting in poor viscosity
Slip Cast Material Preparation control. Wet milling and screening experi-
ments were designed to eliminate agglom-
Early studies (12) examined the production erates in the RM-20 casting slip. Screening
of stable slipsfor the casting of RM-20 rotors. was only partially effective, and wet milling
The objective of this work was the develop- drastically changed the powder particle size
ment of a slip,free of outgassing, with good distribution. The best result was achieved
flow characteristics,stabilitywithin the cast- with a combination of dry ball milling and
ing time period and demonstrating little limited amount of wet milling.
thixotropic behavior. Slip deflocculants
evaluated were a) Keltex, b) Dispex A-40, c) Table 59 summarizes this procedure which
Dispex N-40, and d) California 226/35. calls for 48 hours of dry milling,followed by
removal of the balls and the addition of water.
Of this group, Dispex A-40, in concentra- This suspension is then tumbled 72 hours. The
tions of 0.26 percent, produced the best mill is recharged with grinding media and the
results. Figure 354 shows the viscosity rela- suspension is wet milled for 1.25 hours. The
tionship with spindle rpm for various levels of grinding media are removed, the viscosity of
Dispex A-40. Initially, a problem was the slip is adjusted to 125-150 cps via water
encountered in casting the shaft (the last addition and the slipis tumbled 16 hours.
portion of the rotor cast), wherein the slip
became thixotropic, causing slumping and dis- Table 59. Improved RM-20 Slip
tortion. This problem was eliminated by sub- Preparation Procedure.
stituting Displex N-40 deflocculant for
approximately 50 percent of the A-40 defloc-
culant. - 48 hours dry milling
One problem associated with slip con- - 72 hours tumbling, no milling media
sistency is the presence of agglomerates in the
slip. These agglomerates are formed during - 1.25 hours wet milling with grinding media
dry milling of the RM-20 powder. The ag-
glomerated powder, when placed into slip 16 hours tumbling, no milling media
& .34%
200 0.43% DISPEX A-40
CONCENTRATION
[3.26 % BY WEIGHT
I00
>
RPM
0 ! I I I
I0 20 50 I00
Figure 354. Relationship of Viscosity Vs. Spindle Speed for
Various Concentrations of Dispex A-40 Deflocculant.
307
This procedure produced a casting slip of caused large deflections in thin sections of the
good stability and excellent consistency. casting molds. The molds had the form of a
Several rotors were cast with this improved rotor sitting on a flat disc and clamped to a
technique. plastic base which loaded the wax mold on the
webs formed between blade tips. Mismatches
Rotor Slip Casting between the plaster and mold caused distor-
tion in the wax. This type of mold, and the
Historically, blade cracks and inferior technique used to clamp it, were changed to
microstructures in blade tip regions resulted in eliminate distortions in the wax.
low rotor yield. Also, a high proportion of spin
pit failures originated in blades that appeared A new method of dipping the wax mold,
sound, which suggested that subsurface using an aluminum contoured plate, was devel-
defects were controlling quality,, even in oped. The plate matched the base contour of
rotors that passed visual and Zyglo inspec- the water soluble positive wax and had the
tions. To overcome these problems, a sys- web sections between blades machined away,
tematic review of all the steps in the casting leaving a ledge of aluminum 0.060 inches wide.
process was undertaken, starting with the The positive wax was held down loosely with a
composition of the waxes used to make the bolt and a cap. Any stress on the positive wax
molds. could cause a crack which would translate into
a web inside the wax mold and a corresponding
Thermal Expansion of Negative Fugitive Wax crack in the cast part. The contour of the
aluminum dipping plate wax was produced in a
The wax used for casting molds at Ford
computer-controlled mill. The same program
was a product of the Freeman Company, was used to mill this exact contour into the
called "Dip Seal". In 1985 Freeman ceased
surface of the plaster on which the wax mold
marketing this product and an analysis of the sits, thus assuring a perfectly matched inter-
wax was made to determine whether it could
face. The first molds, produced by three dips
be duplicated "in-house". By analysis the wax
in a 155F McCaughin dip wax, were heavy
was primarily paraffin (65 percent) and 35
enough to allow filling of the mold without
percent methyl stearate. The later was
mechanical clamping. This arrangement was
deemed undesirable due to its' insolubility in slightly modified when the thickness and
toluene, the solvent used to remove the wax
weight of the molds were changed; however,
from the casting. A large change in the
the unstressed mold feature (except for stress
thermal expansion rate of paraffin occurs at caused by the slip itself) was retained.
rather modest temperatures (28 - 34F). This is
caused by a solid-phase change in the crystal Wax Removal
structure of the wax and results in a five-fold
increase in an already high thermal expansion
While the triple-dip molds produced a high
rate. Therefore, casting and wax removal is percentage of crack-free rotors, 24 hours were
now done at constant temperature to avoid required to dissolve the wax. Removal was
any expansion damage to the casting. accomplished by immersing the wax-covered
To replace the Freeman Dip Seal wax, a casting in toluene without agitation and at
casting wax (LN-209-218) was purchased from constant temperature. Thinner molds could be
J.F. McCaughin Co. This material is a bees- expected to dissolve quicker and less time in
wax-paraffin-carnauba mixture with a melting toluene would be a desirable result. It was
temperature of 150F. also suspected that soaking the wax caused it
to swell slightly, stressing the blades of the
Mechanical Strength of Negative Fugitive Wax casting. Hence, thicker molds could be
expected to stress the castings more than thin
Because of the low elastic moduli and ones. Initial attempts at a single dip, thin
strength of waxes, slight forces or loads mold produced problems in the wax positives.
308
.- _. ._ <,:_b_r;__
Positive Wax Preparation ring about 6 inches in diameter, 1 inch high and
0.060 inch thick was fabricated. This ring
Polishing the wax with 4-0 steel wool elim- weighed 70 grams and, when placed on the wax
inated surface defects and improved wax re- mold at the tip of the blades, held the mold
moval from the casting. Two new problems down (Figure 355). Because of the perfect fit
occurred in the thin wax molds; splitting of of the wax to plaster, there was no deflection
the mold down the blade edges and surface of the wax.
contamination by traces of lubricant used in
Wax removal was accomplilshed in 6 to 8
the injection molding of the wax positive. The
hours (versus 24 hours for the triple-dip wax
lubricant was elimiinated by two consecutive
mold).
cleanings in an organic solvent.
Material Quality In Rotor Blades
The problem of mold splitting was caused
by the uneven coating of wax around the edge Spin pit tests of visually sound rotors pro-
of the blades. At the corners the wax deposit's duced a large percentage of blade failures.
are very thin. Chamfering this edge on all This suggested that the blades had subsurface
blades caused a more even layer of wax and cracks initiating failure(s) at speeds well
reduced this tendency to split. below those at which hub failures occurred.
To assess the qualify of the subsurface blade
Single Dip Molds material, the sound blades were lightly glass-
beaded after the rotors had been bisque fired.
Initial attempts to use single dip molds Removal of the surface disclosed porous mate-
proved difficult. The wax molds were 70-80 rial at the blade tips. To overcome this, the
grams lighter than previous molds. Unless the castings were spun at 100 rpm while pouring.
slip was poured into the mold slowly and This application of centrifugal force enhanced
carefully, a leak would start at the mold/ the migration of the material to the tip and
plaster interface, quickly draining the slip and significantly reduced the number of blade fail-
ruining the casting. To overcome this, a steel ures.
Figure 355. Single Dip Wax, Plaster Base and Steel Retaining Ring.
309
Remaining Problems With Wax Molds and Wax inspection and subsequent sintering. This
Positives problem is most severe during the placement
or "packing" of the rotor in the power blanket
Two problems remain with the wax sys- for sintering. Therefore, a presintering or
tems. First, the positive has a tendency to "bisque" firing treatment was developed to
distort with shelf time. This tendency to partially consolidate the casting and provide
stress-relieve and hence distort is well under- adequate handling strength.
stood in the dental profession where critical
dimensions in wax patterns must be main- The bisque firing cycle consists of slowly
tained. This problem has not been addressed, ramping in a nitrogen atmosphere to 1500C
but it will be a deterrent to accurate blade and holding at temperature for 2 hours.
forming. This deformation also causes the During this cycle the casting undergoes partial
shaft diameter of the wax positive to ovalize consolidation, as indicated by a 2.5 percent
by 0.010 to 0.020 inch over a few weeks. linear shrinkage, which is sufficient to impact
Blade deformation has not yet been defined. suitable handling strength. An additional
Possible solutions for this problem include advantage of bisque firing is that it permits
refrigerating the wax positives until they are machining of the casting with standard high
useds or using them immediately after they speed tool steel, and thus reduces costly
are molded. diamond grinding of the final component. This
pre-sintering cycle does not impede the final
Secondly, the inside surface of the wax sintering kinetics. This process is now used on
mold develops a rough surface even though the all castings.
wax positive is polished smooth before drip-
ping. Apparently, the 150-160F melting tem- Sinterinff of Rotor
perature of the water soluble wax is too close
to the 150F melting temperature of the dip The sintering time/temperature profile
wax. Lowering the temperature of the dip initially developed was a single-step cycle.
wax to its lowest usable point improves the This cycle was adequate for the densification
surface texture, whereas raising the tempera- of the original platformed rotors but yielded
ture or holding the positive in the dip wax unpredictable blade distortion in the rede-
longer than required heightens the effect. signed rotor castings. In addition, consider-
These bumps on the casting not only destroy able fusion of the packing powder occurred,
the smoothness of the component surface, but increasing the difficulty of removal, particu-
give the mold wax a rough surface, making larly between the blades. Packing powder
removal more difficult. The surfaces can be compositional changes reduced but did not
smoothed after bisque firing by sanding, but eliminate the problem.
this is a tedious job and should be avoided.
A two-step sintering cycle was developed
In summary, the process modifications to reduce blade distortion and facilitate re-
described have produced a marked improve- moval of the packing powder. The first step
ment in rotor fabrication. Approximately 70 of the dual cycle incorporated a packing pow-
percent of slip east rotors have been crack der (boron nitride, silicon nitride, yttria and
free. There is, however, a need to improve siliea) and a 12 hour temperature ramping
the quality of the material in the blade tip schedule that had three discrete temperature
region of the rotor. holds: 0.5 hours at 1520C, 0.5 hours at 1620C,
and 4 hours at 1720C. Neglilgible fusion
Rotor Heat Treatment and Equipment occurred within the packing powder.
Pre-Sintering of Rotor
The second step of the sintering cycle
The dried RM-20 rotor castings are ex- utilized a compositional change in the packing
tremely fragile and difficult to handle through powder and a single temperature hold of 12
310
hours at 1800C. This compositional change was generally dark in color and has a well-
and increasedtemperature and hold time pro- crystallized secondary phase. The interior
vided greater availability of Si0(g) to maintain region of the rotor is lighter in color and
the required oxynitride composition within the shows (by x-ray diffraction) a predominantly
rotor, resulting in full densification. This glassy second phase. Scanning electron micro-
packing powder was easily removed. scopy indicates that the densities and chem-
ical compositions of both regions are identical.
At the completion of the two-step cycle, a Cold spin testing of several rotors produced
nominal 18 percent sintering shrinkage and blade separation failures at the intersection of
final densities of 3.24 to 3.28 g/cma were the two regions. Fracture origins could not be
obtained. Observed blade distortion was min- identified but separation occurred near the
imal. interface of the duplex zones in all of the
blades• This zoning interface may have acted
The two-step cycle approach, although as a stress riser and produced premature fail-
greatly improved over the earlier one-step ures.
cycle, considerably increased the process
complexity and total time neededto complete An investigation of heat treatment tech-
the sintering cycle. Each step required a niques to crystallize the interior region of the
complete temperature cycle with re-packing rotor, and the elimination of the duplex inter-
of the rotor between cycles. In addition, an face, was undertaken. A half-fraction of a
undesirable extra handling of the fragile rotor 24-1 statistically-designed experiment was
was incurred. used to study four variables (nucleation tem-
perature and time, growth temperature and
Consequently, a new cycle was developed time) at two levels each, requiring 8 heat
for the final rotor iteration. Its main features treat cycles.
were:
Test bars cut from sintered rotor #40 were
a) A one-step sintering cycle heat treated• A 0.025 inch layer was ground
from each test bar. X-ray diffraction analysis
b) A 36-hour hold (compared to the 12- hour
revealed that all eight heat treatment cycles
hold) at the maximum temperature
resulted in a crystallized second phase. How-
c) Two discrete temperature holds: 1 hour at ever, when the 0.063 inch-level (maximum
1500C and 36 hours at 1800C. depth of test bar) was examined, four of the
eight bars retained some glassy phase.
d) Two discrete temperature ramp rates:
300C/hr up to 1500C, followed by 50C/hr
Table 60 summarizes the eight heat treat
to 1800C.
cycles and the x-ray diffraction results. These
e) The packing powder, which included 24 initial test results indicate that devitrification
percent boron nitride, was packed loosely. of the glassy second phase throughout the
rotor (over 1 inch depth) would require further
In this one-step sintering cycle, shrinkage was cycle modifications.
nominally 18 and a final density of 3.28 to 3.29
g/cm 3 was obtained. The packing powder was Full size rotors were heat treated, sec-
readily removed and blade distortion was min- tioned and examined by x-ray diffraction to
imal. determine phase composition. One rotor (#64)
was heat treated using a cycle similar to cycle
Post Sinterlnff Heat Treatment of RM-20 #HT4; however, the hold time at the "growth"
Rotor temperature of 1450C was increased from 4
hours to 96 hours. Rotor #63 was heat treated
All of the pressureless sintered rotors had using a cycle similar to cycle #HT7 with an
a duplex zoning structure• The exterior casing identical increase in hold time. The relevant
311
Table 60. Summary of Heat Treat Cycles and X-Ray Diffraction Results.
Cycle Bar Nucleation Growth Glassy

# # Temp C Hours Temp C Hours 2nd Phase
HTI 8A 1050 1 1300 1 Some
HT2 9B 1200 1 1300 4 Some
HT3 6B 1050 1 1450 4 No
HT4 6A 1200 1 1450 4 Some
HT5 5A 1050 4 1300 4 No
HT6 8B 1200 4 1300 4 Some
HT7 7B 1050 4 1450 4 No
HT8 7A 1200 4 1450 4 No
differences in their cycles, as shown in the The secondary phase composition of rotor
table, were in their "nucleation" temperatures #64 was identified as:
and duration times.
The results of these tests show that the YIoSi6024N2 ("H") + Y2Si2 07
two heat treatment cycles did affect the sec-
ond phase throughout the interior of the two + Y2SiO5 + a small amount of glass.
rotors, although not to the same degree. The
original #HT4 cycle, used on the test bars, After the heat treatment, MOR test bars
resulted in some retained glass (see Table 60). were cut from rotors #63 and #64 and from
Similarly, the modified #HT4 cycle (rotor #64) rotor #65 which did not undergo post-sinterlng
resulted in more glass than the modified heat treatment.
#HT7s cycle (rotor #63).
The secondary phase composition of rotor Room temperature four-point MOR bend-
#63 was identifiedas: ing tests were then carried out on five test
bars from each rotor. The results are sum-
Y10Si6024N2 ("H") + Y2Si20 7. marized in Table 61.
Table 61. Mean MOR Results for Heat-Treated Rotors.
Heat Cycle
Rotor # Mean MOR Std Dev (See Table 60)
(ksi) (ksi)
63 79.5 3.4 HT4 (96 hrs)

65 87.6 3.4 None
64 79.5 4.8 HT4 (96 hrs)

63 74.0 7.2 HT7 (9 hrs)
312
These results show some deterioration of reduce the faster deterioration of the floor
room-temperature strength with decreasing relative to the rest of the hot zone, caused by
levels of glass in the secondary phases. It is spilled packing powder reacting with the floor.
expected, however, that high temperature To minimize the down-time needed for a hot
strength will be greatest in the rotor with the zone rebuild, two additional hot zones were
least amount of the glassy secondary phase. built, Table 62 shows the marked improve-
ments achieved with in-house furnace modifi-
The overall strength level of these rotors is cations.
approximately 30-35 percent lower than the
test bar strengths of 125 ks{ previously re- d) Spin Testing RM-20 Rotors
ported. A chronic problem in any material/ General Review
component development program is that of
Extensive changes were made in the meth-
obtaining material properties similar to those
ods used to mount spin pit arbors to AGT
of the test bar material in a large component, rotors and the manner in which the rotors are
such as the AGT rotor. In this case, the
balanced. Previously, the final spin pit arbor
difference in strength is attributed to poros-
was mounted using "Loctite" and a 0.001 inch
ity. Such porosity is believed to occur during to 0.002 inch loose diametrical fit between the
the long sintering times required to achieve
arbor and the rotor. This technique did not
high density. An effort was made to eliminate
produce perfect concentricity between rotor
porosity in the RM-20 rotors with hot isostatic shaft and the outer diameter (O.D.) of the
pressing (HIP).
steel arbor on which the assembly is mounted
and spun during balancing. Tests revealed that
Sintering Equipment this method introduced unbalanced forces
which then had to be removed by grinding
The pressureless sinterlng process utilized
away rotor surfaces, usually the blades.
a cold-walled vacuum furnace and a refractory
Obviously, this was unacceptable for rotors
metal hot zone. A major drawback of this
which might be engine candidates. To over-
system was reactivity of the SiC(g) with the come this problem, a press fit arbor was
shot refractory metal. The hot zone was
designed and the balancing process was split
limited to approximately 15 to 18 cycles
into two parts; an initial balance check and a
before a complete rebuild was required. Sig-
final balance operation.
nificant in-house modifications increased the
furnace life to approximately 25 cycles, but Initial Balance Check
further improvement was necessary. There-
fore, a furnace manufacturer was contracted To reduce the extensive grinding previously
to redesign the hot zone and other deficient required to balance rotors, a method of re-
areas to improve the sintering enviornment. locating the geometric center of the rotor
shaft through the mass center of the part was
The new design provided for a smaller, adopted. The mass center was located and the
more efficient hot zone faced with 0.005 inch shaft portion of the rotor was ground as close
thick tungsten sheet. After several cycles, to the center of the blades as possible. A
deterioration was noted which rapidly acceler- steel sleeve, 3.50 inch long with 0.070 inch
ated with subsequent use. The hot zone be- wall thickness, was press-fitted to the shaft.
came unusable after 23 cycles. An in-house The press was 0.0015 inch on the diameter and
redesign and rebuild effort resulted in an the sleeve was fitted by heating before instal-
improved hot zone. Reconstruction included lation. The sleeve was installed while the
the replacement of the 0.005 inch tungsten rotor was still mounted on the grinding ma-
face with 0.020 inch molybdenum sheet and chine. When cooled, the O.D. of the sleeve
molybdenum rods added to the ceiling struc- was ground to fit the half-bearings of the
ture for support. A sacrificial molybdenum balance machine. The initial check of unbal-
sheet was placed over the floor in order to ance was then made. Results revealed from
313
Table 62. Life Improvements Resulting from Furnace Redesign.
Previous Latest
Category Design Design
Total number of cycles 23 23

(still usable)
Number of cycles with max temperature >1750C 8 12
Number of cycles with max temperature between 0 20

1510C and 1750C
Number of cycles with max temperature between 15

1300C and 1510C
Number of hours at temperature above 1750C 96 316
40 _ercent to 60 percent less unbalance than encountered to 0.010 - 0.040 oz-in, an amount
rotor: mounted with Loctite. With the direc- which can be corrected by grinding without
tion and size of the unbalance known, the destroying the rotor blades.
offset of the rotor shaft could be calculated.
The geometrical center of the shaft could be
e) Spin Testing of RM-20 Rotors
relocated in greater proximity to the mass
center, further reducing the unbalance and
Spin tests of RM-20 rotors were conducted
allowing finalbalancing with minimal grinding.
in a vacuum pit using a Barber Stockwell
Final Balancing 150,000 rpm air turbine. A summary of test
results,along with failure analysis comments,
isshown in Table 63.
After the initial check, the sleeve was
removed by heating it to 300F to 400F. The
rotor was returned to the machine shop, This series of tests revealed an increasing
mounted and adjusted until the shaft portion uniformity in the speed and type of component
had zero runout. The machine check was then failure. All rotors in this group were crack
adjusted to reposition the mass center at the free, based on visual and Zyglo inspection, and
rotational center and the shaft was reground fully bladed. All were cast with the same
to run true to the mass center. The steel spin- techniques, by the same people, from RM-20
pit arbor was installed with a 0.003 to 0.004 SSN material. Although previously tested
inch diametrical press fitand itsO.D. ground rotors (RM material) failed at higher speeds,
to fit the balance machine half bearing. The they were not fully bladed and blade failures
quill shaft hole was also ground true to the predominated. Most of the earlier rotors
mass center during this stage. These steps tested were made of SRBSN material.
were allaccomplished without dismounting the
rotor from the original machine position used In the RM-20 rotors, there was a general
to regrind the shaft to the new location over trend toward improved blade quality, as re-
the mass center. The rotor was then final flected by the number of hub failures in the
balanced by grinding on the back and on the early stages of RM-20 material development.
shaft just forward of the blade root on the This suggested that the large flaw size of the
exducer end. These steps reduced the unbal- cast rotor material (as compared to test bars)
ance from the 0.160 - 0.200 oz-in, previously was controlling the ultimate failure speed of
314
Table 63. AGT RM-20 Rotor Spin Test Results.
Rotor HIP Densit_ RPM at Hub

No. Y/N g/cm J Failure Corn ments
O65 N 3.289 95,620 No blade failure prior to hub failure
066 N 3.290 95,000 No blade failure prior to hub failure
069 N 3.289 88,480 One blade failed causing quill failure,

followed by hub failure.
078 Y 3.303 85,000 No blade failure prior to hub failure
080 Y 3.293 101,600 No blade failure prior to hub failure
083 Y 3.294 98,500 • Lost a blade at 96,470 rpm
094 Y 3.304 112,980" Was partial sintered prior to HIPping;

heat treated following HIP
Lost 1st blade at 104,060 rpm

Lost 3rd blade at 109,360 rpm
Lost 3rd blade at 112,980
097 Y 100,780 No blade failure prior to hub failure
105 Y 103.780 Lost 1st blade at 93,000 rpm

Followed by hub failure.
r
*Did not fail, reached max. rig rig speed.
**In several rotors (as noted), blade failure at lower speeds preceeded the final
hub failure.
these rotors. All of these rotors were pres- firing by ASEA Cerama using their proprietary
sureless sintered with a 36-hour hold at tem- glass encapsulation process. HIPping condi-
perature. This long hold time was believed to tions were a maximum temperature of 1700C
result in increased porosity and correspond- and a soak time of two hours at a pressure of
ingly lower strength, having a pronounced 160 MPa.
affect on the highly stressed hub during spin
testing. One noteworthy aspect of the failure of
the HIPped rotors was the recurrence of blade
In an effort to reduce porosity, rotors No• failuresat speeds below the utlimate hub burst
75 through No. 105 were HIPped after bisque speed• Rotor No. 94 failed in three blades but
• 315
reached the limiting speed of the spin pit air It was observed that even when the percent of
turbine without hub failure and was retired total porosity was small, large isolated pores
with the hub intact. This rotor was within 940 existed throughout the sample. The largest
rpm of the 105,000 rpm qualification speed pore of each sample was measured and re-
when a blade failed and it sustained a 112,980 corded. The relationship between the rotor's
rpm ultimate speed without hub failure;indi- burst speed and the largest pore size of its
cating that HIPping raised the hub strength to corresponding fragment was plotted and is
adequate levels, but blades again became the shown in Figure 356. This plot shows the data
limiting factor. Blade quality and strength fallinginto two groups: one group had smaller
adequate for a 95,000 rpm hub capability were pores, corresponding to rotors with the higher
now the criticalparameters determining rotor burst speeds. The second group had larger
durability. This is also illustratedby rotor No. pores, corresponding to the rotors with the
105 which failed in three blades before hub lower burst speed. In addition, when the
failure occurred in excess of 103,700 rpm. individual data points in Figure 356 were eval-
The overall results confirmed that a measura- uated relative to sintering process, it was
ble increase in hub strength resulted from determined that the smaller pores (high burst
HIPping. speed group), corresponded to the HIPped
rotors; whereas the larger pores (low burst
Analysis of RM-20 Spin Pit Tested Rotors speed group), were found in the pressureless
sintered rotors.
A porosity analysis was conducted on both
pressureless sintered and HIPped RM-20 rotors The flexural strength data obtained from
as a means of explaining the higher burst test bars cut from pressureless sintered and
speeds of HIPped rotors. HIPped rotors further substantiate these re-
sults. As shown in Table 64, the average
The spin-to-failure tests resulted in the strength of the HIPped rotor material was 111
rotors bursting into many small fragments. ksi versus 75 ksi for pressureless sintered
These pieces were too small to be accurately RM-20 material. Clearly, the overall results
traced to their original position in the pre- of porosity size determination, fracture
failed rotors. Consequently, nine fragments strength evaluation and spin testing demon-
(3 x 20 mm), one from each rotor, were chosen strate the need to reduce pore size in the
randomly. The nine pieces were cut and RM-20 rotor material.
polished to permit microscopic examination of
the porosity. The porosity evaluation was f) The Effect of Heat Treatment on the
performed using DIMA (Digital Imaging Analy- Flexural Strength of RM-20 Rotor Material
sis). Both the percent of overall porosity and
the largest pore dimension and its area were The effects of heat treatment on the
measured on five consecutive 'fields' (0.62 x HIPped RM-20 rotor material were studied in
0.96 mm each) per sample at 180X magnifica- terms of flexural strength and secondary
tion. phases. Modulus of rupture was measured at
room temperature and in some cases at 1200C.
Noticeable variations in porosity were seen Speciman size was 1.25 x 0.25 x 0.125 inch,
when scanning across the rotor samples. In upper span 0.375 inch, lower span 0.75 inch
one sample, total porosity ranged from 0.5 and the loading rate was 0.02 in/min. Table 65
percent to 9.6 percent. The smallest variation summarizes these results,indicating the type
per sample was 0.2 to 0.7 percent. No corre- of heat treatment, specimen (rotor versus test
lation could be established between total bar), secondary phases present and strength
porosity of the pressureless sintered versus the resultsat room temperature and 1200C.
HIPped rotor material. There was, however,
considerable range (1 to 200 microns) in the The flexural strength of the as-HIPped
size of pores found in the nine rotor samples. RM-20 material cut from a rotor was 111.3
316
RPM @ FAILURE VS. MAX. PORE SIZE
105,000
[]
100,000
A
95,000 V
=Z
o,,,
1
90,O00
am A
85,0O0 I W
80,000
PORE SIZE (MICRONS)
HIPped RM-20 ROTORS
e PRESSURELESS SINTERED RM-20 ROTORS
Figure 356. Rotor Pore Size Versus Spin Test Rotor Speed.
Table 64. Flexural Strength Of Pressureless Sintered and HIPped RM-20 Rotor Material.
Pressureless Sintered RM-20 HIPped RM-20
Sample MOR MOR

No. (MPa) (ksii (MPa) (ksi)
m
63-1 508.4 73.8 79A-1 810.0 117.6

63-2 537.4 78.0 79A-2 767.6 111.4
63.3 502.8 73.0 79A-3 619.0 89.8
63.4 446.9 64.9 79A-4 838.0 121.6
63.5 581.0 84.3 79A-5 798.9 116.0
Average 515.3 74.8 Average 766.6
317
ksi. X-ray diffraction analysis indicated a B- this slightly diminished high temperature
silicon nitride crystalline primary phase with strength due to the small sample size.
an amorphous secondary phase. When heat
treated as a half-section rotor (Heat Schedule An interesting phenomena associated with
C), the RT strength was reduced by approxi- these heat treat studies was the pronounced
mately I0 percent- not totally unexpected in affect of heat treatment on the smaller speci-
view of the reduced secondary phase glass men (in this case, the MOR test bar) versus
content and formation of H-phase the previously discussed rotor subsequently cut
(Y5Si3OI2N) and Y2Si207. The 1200C MOR into MOR bars. This can be seen in the last
was 71.5 ksi for this heat treated rotor three tests reported in Table 65. In heat treat
material. schedule B + D, MOR test bars were heat
treated for 96 hours at 1450C. This treatment
Heat Treatment B, with a complete rotor resulted in a severe room temperature
and a shorter total heat treat time, resulted in strength reduction to 73 ksi. Secondary phases
very little difference in room temperature were H-phase and glass. Schedule B + F, using
flexural strength when compared to non-heat a shorter time but higher heat treat tempera-
treated material. X-ray diffraction of this ture of 1700C, resulted in a further strength
material revealed the presence of H-phase but reduction. As expected, this higher tempera-
no glassy secondary phase. The 1200C MOR ture produced more glassy secondary phase
was 59 ksi. No significance was attached to material. On the other hand, when small
Table 65. The Effect of Heat Treatment On Flexural Strength and

Secondary Phases In RM-20 Rotor Material.*
Heat Treatments Seeondary Phases 4-Point MOR (ksi)

Performed Observed at RT at 1200C
A G 111.3(5)
A+C H_Yi g 99.3(5) 71.5(I)
B H lO4.9(3) 58.7(2)
B+D H+g 73.2(3)
B+E Yi+K 117.1(3) 69.1(1)
B+F G 66.o(3)
* Notations:
1. Heat Treatments
A = As HIPped rotor (No H.T.)
B = Complete rotor, 96 hours at 1450C
C = One half rotor, 16 hours at 1450C followed by 36 hours at 1450C, followed
by 96 hours at 1450C
D = Test bar, 96 hours at 1450C
E = Test bar, 96 hours at 1450C with packing powder
F = Test bar, 36 hours at 1700C
2. Secondary Phases
G = Glass (major); g = glass (minor)
H : YsSi3012N; K : YSiO2N; Yi = Y2Si207
3. Strength
Loading Rate: 0.5 ram/rain
Sample size is shown in ( ) following the strength value (average)
318
specimens were packed in powder and heat the glass recrystallized as H-phase. With
treated using schedule B + E, reduction in either heat treatment, however, stress rupture
strength did not occur• In fact, the highest was not improved at the 1200C temperature
strength of 117 ksi was obtained• Phases and 20-30 ksi stress level. In all probability
produced by this schedule were B-silicon sufficient residual glass, although not detected
nitride, K phase and Y2Si20 7. by x-ray analysis, was present to cause sample
deformation and oxidation.
g) Stress Rupture of RM-2O Rotor Material
In the previously described MOR experi-
The stress rupture characteristics of ments, heat treatment of the test specimens
RM-20 rotor material are shown in Table 66. in packing powder resulted in conversion of
Stress rupture, as in previously described MOR the glassy phase to K-phase (YSiO2N). Unfor-
tests, was measured in 4-point bending• Stress tunately, program timing did not permit these
levels evaluated were 30, 40 and 50 ksi and experiments to be extended to stress rupture
temperature levels were 800, 1000 and 1200C. evaluation. Future testing should address this
Measurements were made on RM-20 material form of heat treatment as a possible means of
prepared as follows: extending the capability of RM-20 silicon
nitride.
Group 1 - Test specimens cut from "as-
HIPped" rotor h) Development of Improved RM-20 Rotor
Material
Group 2 -Same as above but test specimens
heat treated 96 hours at 1450C Failure of the HIPped RM-20 material to
adequately address the AGT rotor stress rup-
Group 3 -Test specimens cut from a rotor
ture requirements initiated a search for other
that was previously heat treated 96
hours at 1450C possible compositions. The short time remain-
ing on the contract, however, discouraged any
As shown in Table 66, the as-HIPped mate- drastic modifications of our present approach
rial sustained over 200 hours without failure at or material. Recent information cited in the
a stress of 50 ksi and test temperature of literature (13, 14) pointed to ZrO 2 additions as
800C. At 1000C under the same load, speci- sintering aids capable of improving the
men failure occurred in 1.1 hours, however, at strength of silicon nitride. Zirconia seemed an
the lower load of 40 ksi, this RM-20 material ideal candidate for investigation--it has the
again sustained over 200 hours of testing at potential for overcoming RM-20's limitations
1000C without failure• At 1200C, the mate- while being relatively easy to test and imple-
rial failed in 1.3 hours at the 40 ksi load. ment.
Failure analysis of the specimens showed evi-
Hot pressing was chosen as the initial
dence of slow crack growth, as well as oxida- means of consolidating the powder mixtures of
tion. At 1200C and 30 ksi, the material did
partially-stabilized zirconia (PSZ) and RM-20.
not fail up to 140 hours of testing• However,
It permitted more rapid assessment of the
visual inspection of the specimen during the
composition's potential properties and, to
test revealed bending• X-ray analysis of this
some extent, the sinterability of the material•
as-HIPped material confirmed the presence of
Hot pressing time can be a measure of the
glass in the secondary phases, accounting for
amount of liquid phase present- the more
the deformation.
liquid, the shorter the time required to hot
The Group 2 and 3 heat treatment experi- press- and the amount present generally cor-
ments were designed to eliminate this sec- relates well with sinterability. However, if a
ondary glass phase. X-ray analysis in both particular composition emerged as a likely
cases disclosed recrystallization of the glass. candidate, a parallel study, undertaken to
In the Group 2 heat treatment of the test evaluate its slip-casting potential, would also
specimen, Y2Si207 was formed. In Group 3, be used to evaluate sinterability.
319
Table 66. Stress Rupture Test Results RM-20 Rotor Material.
GROUP 1. "AS-HIPped" Rotor
* X-RAY DATA: B + G
Specimen Stress Stress Temp. Results

No. (MPa) (ksi) (c) And Observations
79A-7 344.7 5O 800 214 Hrs. sustained
79A-6 344.7 5O 1000 1.1 Hrs. failed
79A-8 275.8 40 1000 212 Hrs. sustained. Slight

discoloration
79A-9 275.8 4O 1200 1.3 Hrs. failed. Slow erack

growth and oxidation
79A-I0 206.8 30 1200 140 Hrs. sustained specimen

shows bending and heavy
discoloration
GROUP 2. After Test Bar Heat Treatment: 96 Hrs. at 1450C
* X-RAY DATA: B + Yi
79A-11 275.8 4O 1200 12 Min. failed. Slow crack

growth and oxidation.
79A-12 206.8 3O 1200 44 Hrs. sustained. Specimen

shows bending and discoloration
* B = Beta Si3N4; Yi = Y2Si207

G = Glass (major); g = glass (minor)
GROUP 3. After Rotor Heat Treatment: 96 Hrs. at 1450C
* X-RAY DATA: B + H


* B = Beta Si3N4; H = Y5Si3012N
.320
Various PSZ and RM-20 powder batches of transformation. The presence of both silicon
200 grams each were mixed for 2 hours in nitride phases in these samples may account
half-liter polyethylene bottles with absolute for their total deterioration during oxidation.
alcohol as a vehicle. A dozen hot pressed In the other samples, oxidation was directly
silicon nitride milling media (1/2 in long x 1/2 proportional to the amount of PSZ present.
in diameter) were included to facilitate mix- The only exception occurred between samples
ing. The powder mixtures were ready for hot IV and V, where the additional alumina in V
pressing after removal of the milling media appeared to alter the effects of the additional
and evaporation of the alcohol. PSZ.
The mixtures were hot pressed in a graph- Although additions of PSZ to the RM-20
ite die lined with molybedenum foil at 4000 material might ultimately have improved room
psi. The cycle time required to hot press was temperature rotor performance, the oxidation
defined to be from the time the die punches test clearly revealed its "Achilles heel" and
started to move until the motion was less than further testing was discontinued.
0.001 in./15 min. The resulting hot pressed
TASK 2.7 - STATOR - FORD
billets were 6.67 mm in diameter by approxi-
mately 1.75 mm in thickness.
Introduction
The compositions and the initial results of The stator configuration selected for the
these experiments are summarized in Table AGTI01 engine was of a 19-vane, radial flow
67. Also included in this table are three
design. Nominal overall dimensions were:
samples of the same composition which were
slip cast and sintered. All of the hot pressed Outside diameter 7.0 inches
samples achieved greater than 99 percent of Inside diameter 5.2 inches
their theoretical density. It should also be Total thickness 0.6 inch
noted that the PSZ containing samples were
hot pressed in much shorter time than the base Design alternatives considered were (a) a
RM-20 material. This suggested the likelihood stator assembled from individual vane seg-
of good sinterability. Indeed, the first slip ments and (b) a monolithic configuration hav-
casting and sintering attempt achieved densi- ing one solid and one segmented shroud. The
ties as high as.98.5 percent of theoretical. latter configuration was designed to reduce
thermal gradients which could result in exces-
Both the hot pressed and sintered samples sive stresses during transients.
were sectioned for further study. Test speci-
mens were cut from each of the billets for use Several potential advantages were inherent
in a static oxidation test and other evalua- in the monolithic construction:
tions, such as MOR, creep, thermal expansion,
o Ease of Handlin[_- One large part versus 19
etc. The static oxidation results (Figure 357),
discouraged any hopes that PSZ additions pieces would be processed through all
could overcome RM-20's limitations. All of stages of fabrication and assembly.
the ZrO 2 containing samples exhibited poor
O Reduced Machinin_ Costs- There would be
oxidation resistance, as evidenced by the dis-
a maximum of four surfaces and no critical
coloration of their initial dark appearance.
circumferential stack-up tolerances requir-
The dramatically different results obtained in
ing machining.
samples of similar composition warrants fur-
ther comment. The abbreviated hot pressing o Reduced Leakage- Leak paths through the
cycle used to achieve full density in composi- shrouds would be reduced by one half.
tions Ai, Bi and Ci (where unreacted raw
materials were used in place of RM-20), was o Experience- Ford had already demon-
too short to complete the alpha-to-beta phase strated the capability of monolithic axial
321
Table 67. The Effect of PSZ Addition to RM-20 Si3N 4 Material.
Hot Pressed
Weight Percent
Temp. Time Density X-Ray

ldent. RM-20 PSZ Si3N 4 KI20 3 Deg. C Min. g/em3 Data (*)
Base 100 0 0 0 1800 271 3.308 B,K
I 99.5 0.5 0 0 1675 94 3.315 B,G
II 99 1 0 0 1675 74 3.323 B,Y,g
[] 98 2 0 0 1675 70 3.336 B,H,g
IV 96 4 0 0 1675 68 3.375 B,Z,G
V 90 8 0 2 1675 17 3.438 B,Z,G
A 60 10 30 0 1675 115 3.436 B,Z,g
B 70.6 5.9 23.5 0 1675 145 3.388 B,Z,g
C 83.3 2.8 13.9 0 1675 107 3.342 B,Z,g
Si3N4 Y203 A120 3
Ai 54 4.8 1.2 10 30 0 1675 44 3.420 A,B,Z,G
Bi 63.5 5.6 1.4 5.9 23.5 0 1675 29 3.353 A,B,Z,G
Ci 75 6.7 1.7 2.8 13.9 0 1675 32 3.322 A,B,Z,G
Slip Cast and Sintered
As 60 10 30 1800 36hr 3.306 B,Z,G
Bs 70.6 5.9 23.5 1800 36hr 3.311 B,Z,G
Cs 83.3 2.8 13.9 1800 36hr 3.262 B,Z,G
(*) (A) Alpha Si3N4, (B) Beta Si3N4, (Z) ZrO2, (Y) Y2Si207, (H) Y5Si3012N ,
(K) YSiO2N , (G) Glass (major), (g) glass (minor)
322
1000°C - 100 hrs.
Figure 357. Static Oxidation Test.
stator molding for the 820 engine program. location. Both can influence melt front weld
(1,15) Molding and forming this component lines and significantly affect component
of individual vane segments was readily strength. For the AGT stator, a diaphragm
supported by existing inhouse technology gate along the inside diameter of the solid
and that of other suppliers. shroud would prevent weld lines in the shroud.
However, the vanes and segmented shrouds
Based on these advantages and seeking to
would be filled in parallel, constricting the
advance the technology of complex shape
flow through each vane to 5 percent (1/19) of
forming, Ford elected the challenge of a mon-
the machine maximum flow rate. Conversely,
olithic stator and worked on this design until
a single point gate (at any location) would
February, 1984 when lack of funding caused
produce a shroud weld ]ine but permit a flow
elimination of this effort.
rate through each vane of 50 percent machine
Flow Modelinl_ maximum by filling paired vanes sequentially.
Among the many important parameters for Cornell University was contracted to con-
injection molding tool design are gate type and duct computer flow modeling studies to ex-
323
amine the influence of flow rate, material lation and operational check out. Several
temperature and die temperature on the vane deficiencies were related to installation and
fill pattern (16). Of primary concern was the construction. The platens of the vertical
possibility of inadequate fill or the formation molding machine were bowing, transmitting
of weld lines in the thin trailing edge of the the clamping forces at the tool perimeter,
vane. leaving the central section completely unsup-
ported. The platen deflection pattern was
Their studies indicated that, for the ideal-
measured and circular shims installed at the
ized isothermal case and projecting flow rates
tool/platen interfaces to correct the problem.
<50 percent of those anticipated through a
Dimensional measurements of molded stators,
diaphragm gate, good vane fillwould occur as
however, suggested that internal tool compo-
shown in Figure 358. In this case, the melt
nents were deflecting. Specific ares of de-
front at the trailing edge advances to the
flection were measured during molding trials
segmented shroud before the shroud fills.
and locations which required internal stiffen-
However, projecting an unheated tool and the
ing were identified. Also, cold weld lines were
expected flow rates for a diaphragm gate, the
occurring in the trailing edge of the vane. The
model predicted complete freezing of the melt tool was returned to the vendor for correction
front along the trailing edge before reaching
of the internal deflection problems and en-
the shroud. Only back flow from the shroud
largement of the cavity to increase the vane
would complete the vane trailing edge, result-
trailing edge thickness from 0.030 to 0.040
ing in a cold weld line. As the flow rate was inch.
increased to that predicted for a point gate,
the melt front approached that of the isother-
Molding Development - Diaphragm Gate
mal case. Thus the study emphasized the need
for precise die temperature control and flex-
Stator molding was initiated using the
ibility in design of the gate region to facilitate
0.010 inch thick gate; 12 full and six short
changes in gate configuration.
shots were made. The fillpattern observed on
Tooling Design the short shots was non-uniform as shown in
Figure 360, Views (a) and (b). Complete cavity
The stator tool design (Figure 359) incorpo-
fill was achieved as indicated in Figure 360
rates three independent water temperature
View (c).
control circuits. Five cavity pressure trans-
ducers and four thermocouples are included to
Measurements of the gate material at the
monitor and control molding parameters.
inside diameter of the solid shroud revealed
Removable die inserts enable rapid changes in
that the actual gate size varied from 0.006 to
gat ing.
0.014 inch. This variation was the result of
Previous experience in the injection mold- eccentricity between the tool details which
ing of complex-shaped components related the form the gate. The direction of eccentricity
occurrence of cracks in the molded part to the matched the fill pattern seen in the short
sequencing of insert removal and parting of shots. This is, the cavity area first to fill was
the die. Consequently, considerable effort at the larger gate dimension. Full shots had a
was spent in designing flexibility into the tool. weld line in the solid shroud opposite the
This included independent sequencing of the larger gate area.
two sets of 19 inserts forming the vane con-
tour. In addition, the die was designed to Seven of the full shots were visually in-
independently sequence the gate parting line spected for defects in the vane trailing edges.
before opening the die and ejecting the stator. In general, the trailing edges were sound with
over half the vanes in each stator having no
Tool DeveloPment
observable defect. Trailing edge knit lines, or
Initial development work on molding the signs of cold packing, were found in the re-
AGT stator was directed toward tool instal- mainder of the vanes. Three types of typical
324
Vi EW xx
VANE
SEGMENTED
SHROUD
! I
VIEW zz
Figure 358. Cornel] Melt Front Advance.
325
g
I
Figure 359. Stator Tool.
trailing edge conditions are shown in Figure thick and an extensive parametric study was
361. The relative frequency of each type of conducted. Stators were molded at material
vane was: temperatures of 200, 250 and 275F and tool
temperatures of 90 and IIOF. Several shots
Vane Average Vanes
were also made at varied injection pressures.
Appearance per Stator
No defect I0.7
The first series of shots was made at
Knit lines 1.3 material and tool temperatures of 200 and
90F, respectively. These stators were visually
Cold packing _. 1
inspected and compared with stators molded
The tool insert forming the gate was re- using the eccentric gate and identical temper-
worked to make a concentric gate 0.015-inch atures. The data are summarized below:
ORIGINAL PAC!,_ Ig
326
10]_ _OOR QUALITY
_GOl_ QUALITZ
(a) (b) (c)
Figure 360. Fill Pattern With 0.010 Inch Eccentric, 0.0O6-0.014-inch

Diaphragm Gate Short Shorts (a) and (b), and Full Shot (c).
(a) (b) (c)

Figure 361. Vane Trailing Edges TypicaLly Formed With Eccentric Gate
Good Fill (a), Knot Line (b) and Cold Packing (c).
327
Gate Knit Cold No
t_
Configuration Lines Packing Defect 2O
t_ 18
Eccentric 1.3 7.0 10.7 ae
w
(0.006 - 0.014 n 16
inch) 14 _
Z __" 90*F
_
12
Concentric 6.0 17.0 0.6
u.
0
(0.015 inch) I0
0_
8
The overall condition of the stators had z 6 _
obviously degenerated significantly and vanes w
4
appeared which had both knit lines and cold ¢Y Z _
packing flaws. u.I
I I I I
0
The effects of material and tool temper- 2oo 225 250 275 300
atures on vane trailingedge flaws are shown in MATERIAL TEMPERATURE (°F)
Figures 362 and 363. The data for these

Figure 363. Effect of Tool
x 20- Temperature on Vane Trailing Edge
0
Cold Packing Flaws.
until the cavity packing pressure was reached.

14 The same factors would be expected to pro-
duce fewer knit lines and, in fact, knit lines
were observed to decrease with increasing tool
io temperature. However, at lower tool temper-
e atures, knit l_es increased with increasing
material temperature. (Figure 362).
z
w 4 The "short shots" shown in Figure 364 dem-
----0
2("
onstrate the effect of material temperature
w on the fillpattern for a constant tool temper-
I I I I
0
ature and flow rate. At low material temper-
200 225 250 275 300
MATERIAL TEMPERATURE (°F)

atures, the fillpattern moves primarily in a
circumferential direction similar to the ec-
Figure 362. Effect of Material centric gate condition which produced fewer
Temperature on Vane Trailing Edge Knit knit lines. As the material temperature is
Line Flaws. increased, the fill pattern changes to more
uniform radial filling, reducing the flow rate
figures were generated from visual inspection through each vane and producing weld lines as
of all the vanes on 20 stators fabricated during predicted in flow modeling.
a parametric study. At least three stators
were formed at each set of conditions. During the parametric study, a more
criticalflaw type was observed under certain
Predictably, the curves show that cold conditions. Vane cracks occurred either in the
packing diminished with higher tool and mate- trailing edge or in the vane-to-shroud fillet.
rialtemperatures. An increase in either tem- The relative frequency of cracks as a function
perature tended to keep the material softer of material and tool temperature is shown in
328
,0_, i:O.i)i :, QJAI,YI'_
(a) (b)
Figure 364. Changing Fill Pattern With Material Temperature Concentric

Diaphragm Gate (0.015 in) Material Temperatures of 200F (a) and 275F (b).
Figure 365. Raising either temperature

n-
tended to produce cracks. o
The effect of the diaphragm gate size was 2O
investigated by increasing the gate thickness n..

i,m 18
from 0.015 to 0.065 inch. Ten shots were
made (nine full and one short) with material 16 m
uJ
temperatures ranging from 225 to 275F at a z
14
constant tool temperature of 80F. The results
Q IlO°F
indicated that with the thick gate: W 12 (
I0
o Uniform radial fill occurred at all temper-
atures. 8
0
6 90°F
o Virtually t no cold packing appeared in the TOOL TEMP.
/
vane trailing edges. 4 -
Z
2 -
o Knit lines were evident in all vanes. W
0
W 2O0 225 250 275 500
Knit lines became "tighter" and less ob-
vious with increasing materifil tempera- MATERIAL TEMPERATURE (OF)
ture.
Figure 365. Relative Frequency of
o Vane or vane/shroud fillet cracks were Vane Cracks as a Function of Material
present at all temperatures. and Tool Temperature.
329
Parametric studies with the diaphragm original gate at the solid shroud inside diam-
gate configuration revealed that no combi- eter.
nation of conditions was likely to produce
flaw-free parts. However, it was clearly evi- Short shots disclosed that a circumferen-
dent that all three defects (knit lines, cold tial fill pattern developed with a uniformly
packing and cracks) could be significantly thick melt front (Figure 366). When full shots
altered. were made, there were no signs of trailing
edge knit lines or cold packing. The fullshots,
Cracks in the vane trailing edge or the however, revealed a localized "hot spot" in the
vane/shroud fillets were attributed primarily shroud near the gate. Changing the injection
to tensile loads on the stator during die open- hold times before tool opening and ejection
ing. The entire conical runner section usuall_ resulted in stators with either "pull-out" of the
adhered to the upper half of the tool, breaking shroud material or shroud cracks (Figure 367).
off at the gate during opening. The most
frequent cracks occurred when using the Fourteen full shots were made during a
thicker gate which developed higher loads. study of the effect of gate size, flow rate and
The thin gate shots at lower tool and material tool temperature on the pull-out and cracking.
temperatures were not cracked as shown The material temperature and injection pres-
earlier in Figure 361. At higher temperatures sure were held constant at 250F and 1800 psi
cracks appeared, believed to result from ad- (9000 psi equivalent cavity pressure). The
herence of the part to the tool. Crack preven- gate size, flow rates and tool temperatures
tion appeared to be a function of the loads investigated were:
borne by the stator during tool opening.
Gate size - 0.015 x 0.5 - 0.015 x 1.0
Molding Development - Single Point Gate (inch)
Flow rate - 4.4 - 8.8
Improvements in tool gating could resolve (in3/sec)
all three problem areas. Switching from the Tool temp (F) - 70 - 90
diaphragm gate to a single point gate had the
advantage of: All 14 stators had either pull-out or cracks,
but no strong correlation between any of the
O Sequentially fillingvanes; effectively in- above parameters and flaw type was evident.
creasing the flow rate and eliminating Since the machine was being operated man-
trailingedge knit lines. ually during the study, significant variations
occurred in the injection molding sequence
o Developing a thick melt front while filling times (injection hold, tool open and ejection).
the vanes and eliminating the cold packing. The timing of these events was available from
analog instrumentation recordings of cavity
o Reducing the gate cross-sectional area, pressures for each shot.
I minimizing the load required to break off
the gate runner during tool opening. The flaw types were quantified and plotted
as a function of time as shown in Figures 368
One potential disadvantage was expected. and 369. Pull-out was quantified by measuring
The circumferential fill pattern would result the length of chord of the affected area.
in one solid shroud weld line 180 degrees from Cracks were quantified as the number of op-
the gate. ticallydetectable cracks in the stator shroud.
A cone-shaped metal insert was fabricated At a tool temperature of 70F, pull-out was
to fillthe original runner cavity. A one-half reduced to zero when the tool opening time
inch wide radial slot was then cut into the exceeded 14 seconds from the start of injec-
insert, running from the central sprue to the tion. Cracks appeared only when ejection
330
_ ii_ _ i i_ _ _ _
Figure 366. Short Hot Fill Pattern When Using Single Point Gate.
(Far Left Shot lneludes Sprne and Runner)
(a) (b)
Figure 367. Full Shots Using Point Gate Material Pull-Out in

Gate Area (a) and Solid Shroud Cracks (b).
331
Seventeen additional stators were modeled
z
o
F
0 I0 20
i
30
I
4O
utilizing
the
tures,
size,
time
an
the point
higher
window,
overflow
injeetion
at the
gate
lower
to
pressures,
solid
further
material
shroud
larger
weld
investigate
tempera-
gate
line
TIME BEFORE TOOL OPENING (sac.) and use of a hot sprue bushing. Preeise
control of hold times for tool opening and part
• GATE O.OI5 x 0.5 IN.
ejeetion. Effeetively prevented pull-out and
o GATE O.OI5 x I.O IN.
solid shroud eracks. None of the other vari-
°iF f
ables had a significant effect on molded part
=5 quality.
OO
trio I I I The best tooling configuration and molding

0 IO 20 30 40 conditions for overall good quality stators
TIME BEFORE PART EJECTION (see.) were:
o Gate size- 0.015 x 3.0 inches

Figure 368. Effect of Injection
o Use of a hot sprue bushing
Sequence Timing on Material PuU-Out
o Tool temperature - 90F (top and bottom)
and Shroud Cracks - Tool Temperature,
o Material temperature - 225F
7OF.
o Injection flow rate - 8.8 in3/see
o Injection pressure - 1950 psi (9750 psi cav-
ity)
20 ° o Injection hold time - 5-7 seconds
o Tool open time - 15-16 seconds
,o o Ejection time- 18-20 seconds
0 I0 20 30 40 A production run of 24 consecutive shots

TIME BEFORE TOOL OPENING (sec.) was made over a three-day period. These
stators were intended for complete processing
• FLO_/ RATE 8.8 IN,3/SEC. through machining, as candidates for possible
o FLOW RATE 4.4 IN.3/SEC. engine testing. Of the 24 stators, four were
rejected for molding flaws during the post
molding visual inspection. All 24 passed x-ray
oo 2
oE_ inspection.
"rOE
trio 0 t I _ :
0 I0 20 30 40 Subsequent inspections during processing
TIME BEFORE PART EJECTION (sac.) disclosed fine fillet cracks at the vane-to-
shroud junction on both leading and trailing
Figure 369. Effect of Injection edges. Re-inspection of "as molded" stators
Sequence Timing on Material Pull-Out revealed that the cracks could be seen in the
and Shroud Cracks - Tool Temperature molded state.
90F.
Fillet Crack Investil_ation
times were greater than 22 seeonds. At a tool
temperature of 90F, the "no pull-out time" The discovery of fillet cracks in the
increased to 19 seconds and creeks did not "shipper candidate" stators mandated a re-
appear at ejection times <_34 seeonds. Clear- inspection of all available stators, with partic-
ly, a time "window" existing in both eases ular attention to the leading edge (LE) and
where no pull-out or cracks would be ex- trailing edge (TE) fillet areas. All available
pected. stators molded from the 2.7 density material
332
were re-inspected: 23 of the batch intended Leading edge cracks were found in all of
for further testing and 10 others molded the test candidate stators,averaging 7.5 vanes
during development studies. A summary of per stator and ranging from 1 to 13. Trailing
the inspection findings relative to leading and edge cracks appeared in an average of 2.8
trailingedge cracks ispresented in Table 68. vanes per stator; and ranged from 0 to 12 in
Table 68. Fillet Craek Inspeetion Results.
Stator
" Serial LE TE Total
Number Cracks Cracks Craeks
S 1206201 12 4 16
H 02 7 1 8
I 02 8 0 8
P 04 3 0 3
P 05 4 1 5
E 06 6 16 18
R 07 u_
08 4 8 12
1208201 9 2 11
C 02 11 0 11
A 03 11 0 11
N 04 5 0 5
D 05 5 7 12
I 06 6 0 6
D 07 9 6 • 15
A 08 1 0 1
T 1209201 13 4 17
E 02 4 8 12
S 03 9 1 10
04 8 0 8
05 11 1 12
06 7 3 10
07 13 1 14
08 7 5 12
D 0113201 0 0
E 027206 0 0
V S 1129203 18 18
E T 04 17 17
L U 1130203 4 4
O D 1202202 5 5
P I 1203201 5 5
M E 04 5 5
E S 05 3 5
N 06 4 4
T
333
number. Six stators had no trailing edge
cracks. The ten experimental stators had a
B
significantly different crack distribution; two 24
had fillet cracks and none had the typical _20
trailing edge cracks. (These ten stators, how-
w[ 8 m
ever, had other defects).
z
-;16
Although 23 stators were molded under
14
nominally identical conditions, some process _-
variability occurred. Size parameters were _n 12
identified which showed significant shot-to- _ I0 m
shot variation: <

m
. o 8
IA.
o Cracked vane number o
a: 6
o Injection hold time (t I) -,
m 4
o Time to die open (t o )
o Time to part ejection (t E) z 2
o Cavity pressure at t o (Pcav)
o Cavity pressure difference, (APcav) O, w _v vv
17 191 2 4 6 8 I0 12 14 16 18191
front to rear, at to
VANE NUMBER
Figures 370 and 371 show the frequency of Figure 371. Frequency of Cracks Relative
cracks relative to vane numbers. Leading to Vane Number (Trailing Edge).
edge cracks occurred I00 percent of the time
where no cracks occurred (vanes 14 through
24 - 17).
-- Z2
LJ
Plots of vane cracks versus injection times
zo
144 (tI, to and tE) are shown in Figures 372 and
373. No clear pattern or trend appears for
leading edge "cracks. For the trailing edge
..I
cracks, however, die open time (to) appears to
(n influence crack frequency where increasing to
•." 12
U reduces cracks. No cracks appear when this
variable is greater than 16.1 seconds. Cavity
U.
o
e_
u_
i1o
z
8
0 I
_7 le
I I
2 4 6 _ i_ i_
/I
14 is
I I
_igl
I I
pressure did not correlate with either type of
crack.
Significant differences
frequency
batch and the 10 experimental
were found
of fillet cracks between
in the
24-stator
stators which
were inspected (Table 68). Two of the latter
I VANE NUMBER
were crack-free. All ten had no trailing edge
Figure 3?0. Frequency of Cracks Relative cracks. In addition, the frequency of leading
to Vane Numbers (Leading Edge). edge cracks in these stators could be sum-
marized in three categories: no vanes cracked,
at vane number 19 in the center of the gate all vanes cracked or approximately 25 percent
region. Vanes 8, I0 and 12- opposite the of the vanes had cracks.
gate- had no cracks in any of these stators.
Trailing edge cracks were more randomly dis- Ten variables were identified which could
tributed, but a four-vane segment existed be used to compare the two batches of stators
334
16
14
I Q
Q•
z
I0
$ m
= 8
_0
6 QO
OOQ
• QO
4
I 1 I I I I I I I I I I I I I I _ I I I t i 1
2 4 6 8 I0 12 14 16 18 20 :)2 z4
0
SHOTTIMES(SECONDS)
Figure 372. Plot of Leading Edge Cracks vs. Injection Time.

16
_ 12 $ • •
_ 10
00 • •
= 8
w
z
6
_ 4
z 2 Q •
L ! I 1 _ --__ _ 1 I I 1 1 1 I .I. *L-_l c I c___ _
2 4 6 8 10 12 M 16 i8 20 22 24
SHOTTIMES(SECONDS)
Figure 373. Plot of Trailing Edge Cracks Vs. Injection Time.
335
relative to leading edge cracks. The variables Table 69 shows the value or range for these
were: variables for the best and worst shots of the
24-batch stators and the three groupings for
o Gate type (full 360 degrees versus arc the developmental shots. Three variables
segment) stand out as being unique to the best develop-
mental shots: gate type, sprue type and injec-
o Sprue type (standard tapered versus hot tion pressure. The eccentric, full 360 degree
sprue) gate had previously produced a fill pattern
essentially the same as an arc-segment gate.
o Injection pressure (hydraulic pressure set- Therefore this variable was considered insig-
ting) nifieant in the development of leading edge
cracks. However, sprue type and injection
o Melt temperature setting pressures were potentially important param-
eters influencing leading edge cracks.
Mold temperature
top half In summary, a review of all available data
bottom half regarding the leading and trailingedge cracks
in the vane-to-segmented-shroud fillet sug-
Injection times gested that four variables had influenced the
cracks:
inject and hold
time to die open o Gate location - allpoint gate (arc segment)
time to ejection shots were made with the gate in the same
location which may have masked another
o Cavity pressures (residualat to) tool problem.
Table 69. Fillet Crack Inspection Results - Variables.
No. of Variables
Leading
Edge Gate Sprue lnj. Melt Mold, Temp. Inieetion Times PCAV at
Cracks Type Type Press.
TemP. (Top) [ (Bot.) (lnj.) I (Open)I (Ejet.) topen
0.015
Shipper X
600 -
Best (1) 3.0 Hot 1980 225 90 90 5.5 16.5 21.5 800
0.015
Shipper X
800 -
Worst (1) 13 3.0 Hot 1980 225 90 90 4.5 15.3 21.3 1600
Full
Eccentric
Other 0.015/ 900 - 225 - 80 - 80 - 5.5
Best (2) 0.025 Taper 1500 240 95 95 8.1
0.015
Other X 1800 - 200 - 70 - 70 - 5.2 12.4 15.6 500 -
Fair (6) 3-5 3.0 Hot 1980 250 90 90 8.1 20.8 30.6 2500
0.015
Other X 5.9 - 14.1 - 20.0 - 600 -
Worst (2) 17 - 18 3.0 Hot 1800 250 75 75 8.6 14.5 21.6 1000
336
O Sprue type- the only stators without lead- Each of the three other variables showed
ing edge cracks were made with a standard an improvement over the repeat shots in terms
tapered sprue. of filletcracks. However, at lower injection
pressure, the parts experienced solid-shroud
o Injeetion pressure- developmental shots pull-out and cracking at the edges of the gate,
made with reduced pressures had no lead- due to localized overheating. This problem
ing edge cracks. persisted when all three variables were com-
bined. A similar pattern was observed to a
o Die open time- trailingedge cracks were lesser degree in the 24-batch stators. Again,
reduced with longer times and none ap- the difference might have been a function of
peared for to >16.1 seconds. material batch inconsistency. This investiga-
tion indicated that the variables identified
Molding Investigations and Development significantly influenced the development of
leading and trailingedge filletcracks.
Molding studies investigated each signif-
icant variable. Stators were molded repeating In order to reduce localized overheating in
the same conditions as used for the 24-stator the gate region, and to continue stator mold-
batch. Several shots were made, changing ing development, the gate size was increased.
only one variable from the baseline conditions. An insert was installed which increased the
Finally, shots were made combining all three gate width to 0.065 inch, maintaining an arc
variable changes. All of the stators were then length of 3.0 inches. Preliminary shots showed
inspected with particular attention given to that the overheating problem was corrected.
the leading and trailing edge fillets at the
segmented shrouds. The inspection results are A parametric study was initiated to ex-
summarized in Table 70, along with compara- amine the effects of material melt temper-
tive data from the baseline. ature. Ten shots were made over the temper-
ature range of 225 to 270F. These stators
The six shots replicating the 24-batch con- were visually inspected for all types of flaws.
ditions generally repeated the baseline results. The inspection results are summarized in
The variance in the number of leading edge Table 71. The total number of flaws should be
cracks may reflect material batch incon- compared to 10.3 flaws per stator observed in
sistency. When the gate was rotated during the batch of 24.
these shots, the crack pattern followed the
gate, indicating a flow-related phenomenon Further molding process improvements and
rather than a localized tool problem. tool temperature measurements indicated that
Table 70. Variable Investigation using 0.015 X 3.0 Gate.
Variable Changed Combined Sprue

Repeat
Shipper Shipper Taper T o = 17
Base Line Conditions Sprue T O = 17 see IP = 1300 IP = 1300
Number of Shots 24 6 4 4 4 3
Average Number 7.5 3.0 1.8 1.3 0.8 0.3

LE Cracks
Average Number 2.8 2.7 0.3 0.0 0.3 0.0

TE Cracks
337
Table 71. Material Temperature Investigation 0.065 X 3.0 Gate.
225F 250F 270F
Number of Shots 2
LE Crack SS 3.5 1.2 1.5

S 0.0 0.0 0.0
other 0.0 0.3 0.0
TE Crack SS 1.5 0.3 0.0

S 0.0 0.3 0.5
other 1.5 1.0 2.0
Shroud Cracks 0.0 0.7 0.5
Total Flaws 6.5 3.8 4.5
Notes:
SS = Vane To Segmented Shroud Fillet

S = Vane to Solid Shroud Fillet
* = Location in Shipper Cracks
trailing edge cracks could be related to the Table 72. Stators Delivered to Garrett.
temperature of the inner inserts which form
the vane trailing edge. As no water temper- Number of Cracks
ature control circuits were available in that Stator
region of the tool,-auxiliary electrical heaters _N LE TE
were installed. However, because of budget
cut-back, no funding was allocated to continue 1215202 1 0
the investigation. 1221306 0 2
1110308 1 4
Lack of funding for stator molding devel- 1130306 1 0
opment and the concurrent availability of 0125408 0 3
single stator vanes, caused this effort on a 0221406 1 2
monolithic, 19-vane stator to be dropped in 1206203 9 0
February 1984. As of that time, twelve 1206206 8 13
stators were delivered to Garrett for evalu- 1020307 1 4
ation. Two of those delivered had only one 1206308 2 1
crack each. Table 72 summarizes leading and 1208205 5 7
trailingedge cracks of these stators. 1209206 7 3
TASK 2.7 - FLOW SEPARATOR HOUSING ture for hot flow path components and pro-
vides a sealing surface for the rotating heat
Introduction exchanger.
The flow separator housing is a ceramic The material selected for this component
component that channels the gas flow through is a low thermal expansion lithium aluminum
the heat exchanger, acts as a support struc- silicate(LAS) ceramic, made by Corning Glass
338
Works as their Code 9458 LAS. The process
for making the flow separator housing consists
of slip casting an aqueous suspension of
crushed glass into the desired shape in Plaster
of Paris molds. The component then is re-
moved from the mold, dried, sintered, and
diamond-machined, as required.
Physical Properties of 9458 LAS
Physical property data on Code 9458 mate-

rial is shown in Table 73 and Figures 374, 375
and 376.
Table 73. Physical Properties of

Coming 9458 LAS.
Bulk Density 2.3 g/cm 3 - 2OO

0
Apparent Porosity Zero

Figure 374. Strength and Thermal
Thermal Instability - Expansion of Coming Code 9458 Lithium
Linear (500 hours Aluminum Silicate.
at 2100F) • 100 ppm* max.
The bulk density of 9458 LAS measured
(500 hours at 2200F • 200 ppm* max. approximately 2.3 g/cm 3 or-92 percent of
theoretical density. The apparent or open
*Particles per million porosity is nominally zero.
.S
N_ 14 .0209 _
1.0196 .,_
Z= 13
.o,87
_.0181 _.
"_.0173
-".0164
o
...1
_ 9 <¢
n,-
_,J
I 1 I I I I I I I "I"
w I--
z 0 400 800 I?,.00 1600 2000
I..-
TEMPERATURE ('F)
Figure 375. Thermal Conductivity of Corning Code 9458 Lithium Aluminum Silicate.
339
The thermal expansion of 9458 shown in subjecting them to temperatures varying from
Figure 374 is characteristic of beta spodu- 1800F to 2200F (in increments of 100F) for
mene-type ceramics. The material undergoes 500 hours at each temperature. Measurements
negative expansion from room temperature to were made on cube faces prior to the test and
approximately 300C, followed by a much at 250-hour intervals during exposure, with an
sharper positive expansion up to 900C. optical comparator and a series gage block to
measure dimensional changes. This technique
The flexural strength data in Figure 374 was accurate to _10 ppm. Results of this test
shows a slight decrease in strength over the are shown in Figure 377. The Corning 9458
range from room temperature to 500C. Above LAS material retained 100 ppm stability up to
500C the strength increases, indicating the 210OF. Above 2100F, additional contraction
possibilityof microcrack healing. occurred, and the material fell below the
lower limit of- 100 ppm.
Thermal conductivity, based on measure-
ments of specific heat and thermal diffusivity, Meehanieal Testing
was averaged using two different techniques
(Figure 375). Unlike most ceramics, there is The flow separator housing is subjected to
an increase in conductivity with temperature. thermal and mechanical loads in the engine.
In order to provide some assurance that this
Figure 376 shows the modulus of elasticity component will sustain the mechanical loads
of 9458 LAS. The MOE is approximately and to eliminate defective housings early in
processing, a test fixture was designed and
fabricated to duplicate the engine static oper-
TEMPERATURE ('F) ating pressures on the housing.
- 592 752 1112 1472 1832
x 11.5 The test fixture, assembled with a wooden
model of the flow separator housing is shown
in Figure 378. The fixture is designed to
__ II.O support the housing, simulating engine mount-
ing conditions and sealing the housing to the
fixture with rubber gaskets. Hydrostatic
_. 10.5 pressure loads could be applied within a con-
tainment vessel available at Corning. Lifting
ua I0.0 eyes assist in moving the assembly in and out
of the vessel. The upper plate is provided with
fittings for a pressure line and an air bleed
valve.
_ 9.5
/
The flow separator housing is mounted on a

,", 9.0 silicone rubber gasket which has the same
0 0 200 400 600 800 i000
=E configuration as the regenerator cold seal as-
TEMPERATURE ('C) sembly. The gasket is molded to the base
plate. On the underside of the clamp ring is
Figure 3_6. Modulus of Elasticity of another silicone rubber gasket which functions
Coming Code 9458 LAS. as the engine's flow divider regenerator seal.
As the clamp ring is bolted to the base plate,
10,000,000 psi at room temperature and ap- the housing flanges are sandwiched between
proaching 11,000,000 psi at 1000C. the two gaskets, and the gaskets are com-
pressed to 80 percent of their free height. In
Dimensional stability of the LAS material the upper plate an "O" ring provides the seal-
was measured by machining one inch cubes and ing normally provided between the transition
340
"'--'-1800
OF 1900°F =I= 2000OF'--"_ 2100'F--_ 2200°F-'"
, _mi .
100
5O
A
E
0
e_
v
,- -50
._J
-I00
z
-150
-200 - O0
I I I I I I I I I I I I I I I I I
I I I I I I I I I I I
0 0 5OO 0 5O0
TIME (HOURS)
Figure 377. Dimensional Stability of Coming 9458 LAS Material at Various Times and
Temperatures.
Figure 378. Flow Separator Housing Pressure Test Fixture Shown Assembled to Wooden
Model of the Housing.
341
duct, the turbine shroud and the flow sepa- rejected due to very large casting voids in
rator housing. Thus the high pressure cavity is areas where mechanical and thermal stresses
sealed and loads can be applied to simulate were highest. During this period, a number of
engineoperating pressures. process modifications and mold changes were
effected. These improvements, coupled with
Each flow separator housing was tested at an acceptance specification that permitted
a pressure 25 percent greater than the maxi- casting voids in non-critical areas of the hous-
mum mechanical load exerted on the compo- ing, resulted in additional acceptable compo-
nent at full engine speed. All components met nents for delivery to Garrett. In summary, for
the 65 psi mechanical load specification. a period of five years up to April 1985, the
accept/reject ratio for flow separator housing
was approximately 1:2. At this point in the
Strain Gage Measurements
AGTI01 program, the decision was made that
housing quality permitted housings to be
The flow separator housing mechanical test
shipped directly from Corning, and the Ford
ratus but also permitted the correlation of 3D
inspection technology was transferred to
stress analysis with mechanical stresses en- Garrett.
countered in engine operation. Fourteen uni-
axial and 17 triaxial strain gages were
CONCLUSIONS AND RECOMMENDATIONS
attached to SN-8 in critical areas, as shown in
Figures 379 and 380. Strain gage data were
This report covers material and process
recorded up to 60 psi in increments of 4 psi in
developments by the Ford Motor Company on
both air and water environments.
three key components of the AGTI01 gas
turbine. These components are the i) silicon
Component Fabrication nitride turbine rotor 2) reaction-bonded silicon
nitride one-piece stator, and 3) lithium alum-
The initial phase of component fabrication inum silicate flow separator housing. The
consisted of wooden model machining, pattern following are conclusive remarks and recom-
design, pattern fabrication and Piaster of mendations for future work on each of these
Paris mold construction. A series of slip activities.
casting experiments was designed to test the
feasibility of casting a relatively flaw-free Rotor Development
flow separator housing. These experiments
revealed the need for a number of design o Two materials - sintered, reaction-bonded
modifications, particularly a double bulkhead silicon nitride (SRBSN) and sintered silicon
wall (Figure 381). With this double wall, nitride (SSN) - were investigated using the
casting could be achieved without the inci- fugitive wax slip casting process to form
dence of very large voids at the junction of AGT rotors.
the outer wall/bulkhead and bore/bulkhead.
o Investigating experiments shows that the

Table 74 lists the flow separator housings instability of silicon and water prevented
delivered by Coming Glass, inspected by Ford the formation of stable slip systems using
and approved for shipment to Garrett for the SRBSN process.
engine or test rig evaluation. At Ford these
housings were visually inspected using dye o However, pre-reacting a compact of silicon
penetrant and radiographic techniques. and sintering additives produced a silicon
nitride powder stable in the presence of
Most of the early flow separator housings, water. This material, designated RM-20,
although by no means flaw-free, were of was used in conjunction with the fugitive
reasonable quality. A period of poor quality wax slip casting process to make prototype
followed, in which several components were gas turbine rotors.
342
Figure 379. Ceramic Flow Separator Housing Showing Strain Gages In Place.
o Test bars cut from RM-20 rotors did not ondary phase which affected the stress
have the same material quallity as mater- rupture properties of RM-20.
ial cut from a billet. These findings
o The glassy secondary phase was reduced,
emphasized the importance of developing
but not eliminated, through heat treat-
and fabricating components of test bar
ment. This heat treatment, when done in
quality.
the absence of packing powder, also re-
sulted in reduced flexural strength. It is
O HIPping of RM-20 rotors versus pressure- recommended that future heat treatment
less sintering resulted in material closest
by accomplished in packing powder.
to billet/test bar quality.
o Crack-free rotors were produced and cold
o Both pressureless sintered and HIPped sil- spin tested to failure speeds in excess of
icon nitride rotors showed a glassy sec- 1009000 rpm.
343
Figure 380. Ceramic Flow Separator Housing Mounted to Mechanical Test Fixture.
O Pore size in the rotor material was most wax process. Successes in this area will
important as related to ultimate failure require improved injection molding binder
speed. Future work should focus on reduc- systems permitting binder burnout of thick
ing pore size as well as total porosity. section ceramic components.
o Slip casting using the fugitive wax proce- Reaction Bonded Silicon Nitride Stator
dure is a viable process for making proto-
type rotor components. Future work using O Injection molding of the complex shaped,
this process should address: monolithic, radial stator was shown to be
feasible.
1) Shorter casting time
2) Increased slip suspension time Lack of funding during th AGT program
3) Reduction of pore size in the finished and concurrent availabilityof single stator
rotor vanes caused the effort on a monolithic,
19-vane stator to be dropped. At that
o For the long range, and particularlyfor low time, twelve stators were delivered for
cost, high volume production, the injection evaluation.
molding process and SRBSN is recom-
mended as a rotor material and forming o The molding development efforts revealed
porcess over slipcast SSN and the fugitive complex interaction of many molding
344
ORIGINAL PAGE IS
_01_ P.OOR QUALITY.
0RIG!NAU PA-GNIg
OE POORQUALITY
Figure 381. Ceramic Flow Separator Housing with Double Wall Bulkhead.
Table 74. AGT Flow Separator Housing Quality (Period 1980 - 1985).
FSH FSH
(No.) Quality/Co m ments (No.) Quality/Com ments
SN-8 Aeeepted Rejeetect/Large Voids

SN-12 Aeeepted Accepted via Deviation
SN-14 Accepted (Large Voids)
SN-15 Rejected/Large Voids SN-30 Accepted
SN-I? Aceepted SN-33 Rejected/Machined
SN-21 Rejected/Large Voids Incorrectly
SN-23 Rejected/Large Voids SN-37 Rejected/Large Voids
SN-24 Rejected/Large Voids SN-38 Rejected/Oversize Bore
SN-25 Rejected/Large Voids SN-41 Rejected/Large Voids
345
variables influencing several different complex components such as the flow sepa-
fabrication defects. As such, more exten- rator housing used in the AGT101 gas tur-
sive use of statistical experimental design bine program.
should be applied to resolve the relation-
ship to, and significance of each of the o The double bulkhead configuration of the
variables to specific flaw types. flow separator housing demonstrated the
importance of component design, not only
Computer flow modeling was shown to be a in meeting performance criteria, but criti-
very useful tool for providing direction in cal fabrication criteria as well. This de-
injection molding tool design and in estab- sign modification eliminated the incidence
lishing material flow rate requirements. of very large voids at the junction of the
Efforts should be continued to explore outer wall/bulkhead as well as the bore/
modeling techniques with specific emphasis bulkhead.
on the nonisothermal modeling of simple
shapes and experimental verification. o Although the components produced were
not totally flaw-free, nondestructive test-
Tool deflection and temperature control ing and proof test procedures that permit-
were shown to be critical considerations in ted casting voids in non-critical areas of
injection molding die design. These should the housing are feasible.
be evaluated analytically during die design
to reduce time consuming delays during This program demonstrated that slip cast-
tool tryout. ing is the correct process choice for fabri-
cating large complex components such as
Lithium Aluminum Silicate Flow Separator the flow separator housing in prototype
Housinff form. On a long range basis, however, it is
felt that a glass ceramic process will be
o Lithium aluminum silicate is a suitable required to attain cost competitiveness in
ceramic material for fabricating large high volume production.
346
REFERENCES
1) McLean, A.F. and Fisher, E.A., "Brittle 9)Govila, R.K., Mangels, J.A., Baer, J.R.,
Materials Design, High Temperature Gas "Fracture of Yttria-Doped, Sintered Reac-
Turbine," Final Report AMMRC Contract tion Bonded Silicon Nltride," J. Am.
No. DAAG 46-71-C-0162, (AMMRC TR Ceram. Soc., 68 (7),413-41 (1985).
81-14), March, 1981.
2) Mangels, J.A., Tennenhouse, G.J., "Densifi- i0) Govila, R.K., "Fracture of Flash Oxidized,
cation of Reaction Bonded Silicon Nitride," Yttria-Doped Sintered Reaction-Bonded
Bull. Amer. Ceram. Sot., Vol. 59, No. 12, Silicon Nitride," accepted for publication
(1980), p. 1216. in J. Mater. Sci. (1986).
3) Gazza, G.E., "Hot-Pressed Si3N4," J.

Amer. Ceram. Soc., Vol. 56, No. 12, 1973, 11)Govila, R.K., "Strength Characterization
p. 662. of Yttria-Doped, Sintered, Reaction
Bonded Silicon Nitride," Tech. Report
4) Mangels, J.A., "Sintered Reaction Bonded
SR-86-36, Ford Motor Co., 03-21-86.
Silicon Nitride for the AGT101 Turbine
Rotor," DOE Automotive Technology
Development Contractors' Coordination 12)Tenth Annual Report- AGT101 Program,
Meeting, Dearborn, MI, Nov. 11-13, 1980. (NASA CR-179485) June 1984-85.
5) Mangels, J.A., "Sintered Reaction Bonded

Silicon Nitride for the AGTI01 Turbine 13) Sanders, W.A. and Mieskowski, D.M.,
Rotor - An Update," DOE Automotive "Strength and Microstructure of Si3N 4
Technology Development Contractors" Co- Sintered with ZrO 2 Additions," Adv.
ordination Meeting, Dearborn, MI, Oct. Ceram. Marls., 1, [2],66-173 (1986).
26-30, 1981.
6)Benn, K.W., Carruthers, W.D., "3500 Hour 14) Dutta, S. and Buzek, B., "Microstructure,
Durability Testing of Commercial Ceramic Strength and Oxidation of a I0 wt. Percent
Materials," Third Quarterly Progress Re- Zyttrite-Si3N 4 Ceramic," 67 [2], 89-92
port, DOE/NASA Contract DEN 3-27, (1984).
1978.
7) Hinsolt, G., Knoch, H., Huebner, H., 15)Arnon, N. and Trela, W., "Evaluation of
Kleinlein, F.W., "Mechanical Properties of Ceramics for Stator Applications- Gas
Hot Pressed Silicon Nitride with Different Turbine Engines," Final Report DOE/NASA
Grain Structures," J. Amer. Ceram. Soc., Contract No. 0019-83/1 (NASA
Vol. 62, No. 1-2, 1979, p. 29-32. CR-168140), March, 1983.
8)Johnson, C.F., Hartsock, D.L., "Thermal

Response of Ceramic Turbine Stators," 16) Hieber, C.A., Isayev, A.I. and Socha, L.S.,
Ceramics for High Performance Applica- "An Explorator Investigation of Injection
tions, Burke, J.J., Gorum, A.E., Katz, Molding a Highly Ceramic-Filled Material,"
R.N., editors, Brook Hill Pub. Company, Cornell University, Ithaca, N.Y., May,
1974, p. 549-562. 1981.
347
348
APPENDIX C
PURE CARBON COMPANY

FINAL REPORT
Pure Carbon Based on these known limitations, BNFL

performed a study to demonstrate binder
Early in the AGT101 project, Pure Carbon bake-out and carbon conversion in the thick
Co., acting as British Nuclear Fuel Ltd rotor sections prior to the initiation of a rotor
(BNFL) U.S. distributor provided simulated fabrication development program. This study
rotors and test bars of Refel reaction-sintered was performed using isopressed and green
SiC for evaluation. Reaction-sintered SiC machined simulated rotors.
products are fabricated by a process in which
a porous SiC/carbon body is exposed to silicon Four of these simulated rotors were
at elevated temperatures and the carbon is received from Pure Carbon. Dimensional
converted to bonding SiC. No material shrink- inspection and surface characteristics were
age occurs during the conversion process. favorable. Density of one rotor was measured
Green materials to be reaction sintered may as 3.07 g/cm 3. Room temperature strength of
be formed by a variety of routes, e.g., slip one cut-up sample was (49.1 ksi), with a Weibull
modulus of i0.
casting, injection molding, extrusion, isopress-
ing and warm molding, and the green body may Two Pure Carbon simulated rotors were
be machined before firing. Requirements of spin tested to failure. The best result was a
binder removal prior to firing and silicon pene- failure at 92,000 rpm.
tration during firing, however, impose thick
section limitations on component geometries. Pure Carbon also provided several seal
These limitations are determined experi- rings for the regenerator hot shield O.D. of
mentally, although penetration up to 2.54 cm reaction sintered SiC. Deliveries and test
(I inch) has been demonstrated. activities were curtailed in mid-1981.
PRECEDING PAGE i_L,a_NK NOT FILMED
349
This PageIntentionally Left Blank
350
APPENDIX D
LIST OF SYMBOLS, ABBREVIATIONS, AND ACRONYMS
Acronym Definition
ACC AiResearch Casting Company

AE Acoustic emissions
AGT Advanced gas turbine
AGT101 AGT model being developed by Garrett/Ford
A1203 Aluminum oxide, brand name Sol Gel
AS Aluminum silicate composition, regenerator core
ASEA ASEA Pressure Systems, Inc., Los Angeles
ATS Air turbine starter
Astroloy High temperature superalloy
CBO Standard Oil, formerly Carborundum Company, Niagara Falls,NY
CFDC Combined Federal Driving Cycle
cfm cubic feet per minute, ft'_/min
Corning Corning Glass Works
dB Decibel
DES-1000 Chrome-oxide coating from Kaman Sciences
DF Diffusion flame (relatesto combustor/nozzle)
DF-2 Diesel fuel grade 2
DOE Department of Energy
ECU Electronic control unit
EDX Energy dispersive x-ray analysis technique
EPA Environmental Protection Agency
F Fahrenheit (degrees of)
FPI Fluorescent-penetrant inspection
FSH Flow separator housing
Ford Ford Motor Company
Fuller'searth fine, dust-like material for detecting flow paths
g Gravity, 1-g = force equal to one gravity
g/cm 3 Grams per cubic centimeter
GE-Cordierite Coating material made by General Electric
Garrett Garrett Turbine Engine Company, Division of The Garrett Corporation
He Helium (used for leak path detection)
HIP Hot isostatic pressing
HP High pressure
HPSN Hot pressed silicon nitride
Hz, CPS Hertz, cycles per second
IBM International Business Machine, Inc.
ID Inner diameter
IDH Inner diffuserhousing
inch U.S. customary linear unit
in.HgA Inches of mercury, absolute
in-lb Inch-pounds (work)
1-85 Copper base coating material (regenerator seals)
I-I12 Zinc-oxide coating material, regenerator seals
351
APPENDIX D
LIST OF SYMBOLS, ABBREVIATIONS, AND ACRONYMS (Contd)
Acronym Definition
JP-4 Kerosene base aviation jet fuel

k Thousands of a unit, e.g., rpm, XI000
ksi One-thousand pounds per square inch
Kyocera Kyoeera International, Inc., Kyoto, Japan
LAS Lithium-alu minu m-silicate composition-regenerator core
Ib-in Pound-inch (torque)
ib/min Pounds per minute (flow)
LP Low pressure
MAS Magnesium-aluminum-silicate composition, regenerator core
MAR-M 247 Nickel based alloy, dual alloy turbine rotor
mil One-thousandth of one inch
MOR Modulus of rupture
N Speed, rpm
NASA National Aeronautics and Space Administration
NDE Nondestructive evaluation
NGK NGK - Locke, Inc.
NOx Nitric oxides
OD Outer diameter
ODH Outer diffuser housing
pH Relative acidity to alkalinity balance
psia pounds per square inch, absolute
psid AP Differential pressure, pounds per square inch
psig pounds per square inch, gage
pound, Ib U.S. customary weight measure 0.373 kilograms
p-p Peak-to-peak
Prig Rig pressure psi
PV Pressure-velocity value
RBSN Reaction bonded silicon nitride
rpm revolutions per minute
RM-20 Sintered reaction bonded silicon nitride ceramic material
SASC Sintered alpha silicon carbide ceramic material
SC201 (SIC) Silicon carbide ceramic material
Si3N4 Silicon nitride
S/N Serial Number
S-77 Zinc-oxide based coating, regenerator seals
SN-71 NGK ceramic materials, sintered silicon nitride
SN 220M Kyocera ceramic materials, sintered silicon nitride
SN 250M Kyoeera ceramic materials, sintered silicon nitride
SN 270M Kyocera ceramic materials, sintered silicon nitride
SSN Sintered silicon nitride ceramic material
352
APPENDIX D
LIST OF SYMBOLS, ABBREVIATIONS, AND ACRONYMS (Contd)
Acronym Definition
shp Shaft horsepower

T Temperature
Ti Titanium
TIT Turbine inlettemperature
T3.5 Rating point - combustor inlet temperature
T4.1 Rating point - turbine inlet temperature
UT Ultrasonic techniques, inspection processes
W Flow, rate as specified
Wayne-Kerr Rotor dynamic measuring devices
watt Electrical power - one volt x one ampere
Waspaloy Heat resistant,high temperature superalloy
Y203 Yttrium oxide
Zyglo Surface defect inspection process, brand name
2-D Two-dimensional
3-D Three-dimensional
Micro, one millionth
Microstrain
353
354
APPENDIX E
REFERENCES
I) Garrett Turbine Engine Company, "Ad- 7) Garrett Turbine Engine Company, "Ad-
vanced Gas Turbine (AGT) Powertrain Sys- vanced Gas Turbine (AGT) Powertrain Sys-
tem Development for Automotive Applica- tem Development for Automotive Applica-
tions," Semiannual Progress Report tions," Semiannual Progress Report Num-
Number 1 (October 1979 through June ber 7 (January 1983 through June 1983),
1980), NASA Report CR-165175, November NASA Report CR-174694, December 1983,
1980, Contract DEN3-167. Contract DEN3-167.
2) Garrett Turbine Engine Company, "Ad-

vanced Gas Turbine (AGT) Powertrain Sys- 8) Garrett Turbine Engine Company, "Ad-
tem Development for Automotive Applica- vanced Gas Turbine (AGT) Powertrain Sys-
tions," Semiannual Progress Report tem for Automotive Applications," Semi-
Number 2 (July 1980 through December annual Progress Report 8 (July 1983
1980), NASA Report CR-165329, June through December 1983), NASA Report
1981, Contract DEN3-167. CR-174809, June 1984, Contract DEN3-
167.
3) Garrett Turbine Engine Company, "Ad-
vanced Gas Turbine (AGT) Powertrain Sys-
tem Development for Automotive Applica- 9) Garrett Turbine Engine Company, "Ad-
tions," Semiannual Progress Report vanced Gas Turbine (AGT) Powertrain Sys-
Number 3 (January 1981 through December tem for Automotive Applications," Semi-
1981), NASA Report CR-167983, June annual Progress Report 9 (January 1984
1982, Contract DEN3-167. through June 1984), NASA Report CR-
174886 December 1984, Contract DEN3-
4) Garrett Turbine Engine Company, "Ad- 167.
tem Development for Automotive Applica-
tions," Semiannual Progress Report 10) Garrett Turbine Engine Company "Ad-
Number 4 (July 1981 through December vanced Gas Turbine (AGT) Technology De-
1981), NASA Report CR-167983, June velopment Project" Technical Summary
1982, Contract DEN3-167. Report I0 (July 1984 through June 1985),
NASA Report CR-179485 July 1986, Con-
5) Garrett Turbine Engine Company, "Ad- tract DEN3-167.
tions," Semiannual Progress Report 11) Garrett Turbine Engine Company "Brayton
Number 5 (January through June 1982). Cycle Solarized Advanced Gas Turbine"
Contract DEN3-167. Final Report (February 1980 through March
1986) NASA Report CR-179559, December
6) Garrett Turbine Engine Company, "Ad- 1986, Contract DEN3-181.
tions," Semiannual Progress Report Num- 12) Govila, R.K. "Strength Characterization of
bet 6 (July 1982 through December 1982), Yttria-Doped, Sintered, Reaction Bonded
NASA Report CR-168246, June 1983, Con- Silicon Nitride", Tech. Report SR-86-36,
tract DEN3-167. Ford Motor Co., 03-21-86.
355
1 Report No 2 Government Acc_sion No 3 Recipient's Catalog No
CR-180891
4 Title and Subtitle 5 Report Date
Advanced Gas Turbine (AGT) December 1987

Technology Development Project 6. Performing Organization Code
7 Author(s) 8 Performing Orgamzation Report No
Engineering Staff of Garrett Auxiliary Power Division 31-3725 (12)

A Unit of Allied-Signal Aerospace Company
10 Work Unit No
9 Perf_mingOr_nizationNameand Addre=
Garrett Auxiliary Power Division 11 Contract or Grant No

2739 E. Washington Street
Phoenix, Arizona 85034 DEN3-167
13 Ty_ of Report and Period Covered

12 S_nsoring Agency Name and Address Final October 1979-
June 1987
U.S. Department of Energy
14+ Sponsoring Agency Code
Office of Transportation Systems, Heat Engine Prop-
ulsion Division, Washington, D.C., 20585 DOE/NASA/0167-12
15 _pptementary Notes
Final Report under Interagency Agreement

Project Manager T.N. Strom, Propulsion Systems Division
NASA-Lewis Research Center, Cleveland, Ohio, 44135
16 Abstract
This report is the final in a series of Technical Summary Reports for the Ad-
vanced Gas Turbine (AGT) Technology Development Project, authorized under NASA
Contract DEN3-167 and sponsored by the DOE. The project was administered by Mr.
Thomas N. Strom, Project Manager, NASA-Lewis Research Center, Cleveland, Ohio. This
report summarizes plans and progress from October 1979 through June 1987.
Project effort conducted under this contract is part of the DOE Gas Turbine
Highway Vehicle Systems Program. This program is directed at providing the United
States automotive industry the high-risk, long range technology necessary to produce
gas turbine engines for automobiles that will have reduced fuel consumption and red-
uced environmental impact. The intent is that this technology will be capable of
reaching the marketplace by the 1990s.
The Garrett/Ford automotive Advanced Gas Turbine has been designated the AGTI01
The AGTI01 nominally is a 74.5 kw (100 shp) engine r capable of speeds to 100,000 rpm
and operating at turbine inlet temperatures to 1370C(2500F) with a specific fuel
consumption level of 0.18 kg/kw-hr(0.3 ibs/hp-hr) over most of the operating range.
This final report summarizes the powertrain design, power section development,
and component/ceramic technology development. Appendices include summary reports of
progress from Ford Motor Company, AiResearch Casting Company and Pure Carbon, acting
as major subcontractors and suppliers of ceramic materials and components. Standard
Oil Company's Unique and Common work accomplished under this program is published
in topical report (TBD).
17 Key Words (Suggested by Author(s)) 18 Distribution Statement
Advanced Gas Turbine Unclassified - Unlimited

Single Shaft Engine Star Category 85
Ceramic Turbine DOE Category UC-96
Turbine Transmission
19 Security Cla_if (of this report) 20 SecurityClassif (of this _) 21 No of Pages 22 Price"
Unclassified Unclassified 384 A.]7
+ For sale by the National Techmcal Information Serwce, Springfield, V=lg=m,'_ 22161
NASA-C-168(Rev I0-75)

Advanced Gasturbine (Agt) Technology Development Project: DOE/NASA/0167-12 Nasa CR-180891 Garrett NO. 31-3725

Uploaded by

Copyright:

Available Formats

Advanced Gasturbine (Agt) Technology Development Project: DOE/NASA/0167-12 Nasa CR-180891 Garrett NO. 31-3725

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Advanced Gasturbine (Agt) Technology Development Project: DOE/NASA/0167-12 Nasa CR-180891 Garrett NO. 31-3725

Uploaded by

Copyright:

Available Formats

DOE/NASA/0167-12

1.1 Powertrain Development 1

3.0 REFERENCE POWERTRAIN DESIGN

4.0 POWER SECTION DEVELOPMENT 13

4.1 Engine Mechanical Development 13

4.1.1 Foil Bearing 13

4.2 Engine Rotor Dynamic Development 26

4.2.1 AGTI01 Rotor Dynamic History 26

4.3 Engine Development Evaluations 34

4.3.1 Metallic Engine Testing 36

5.0 COMPONENT/SUBSYSTEM DEVELOPMENT 47

5.1 Compressor Development 47

5.1.1 Compressor System Requirements 47

5.2 Turbine Development 65

5.2.1 Aerodynamic Design 65

5.3 Combustor Development 88

5.4 Regenerator System 93

Ford Regenerator Development 94

5.5 Gearbox/Transmission 139

5.6.1 Engine Electronic Control Unit 142

6.0 CERAMIC MATERIALS AND COMPONENT DEVELOPMENT 147

6.1 Ceramic Materials Testing 147

6.1.1 Fast Fracture and Stress Rupture 147

6.2 Ceramic Turbine Rotor 182

6.2.1 Initial Evaluations 182

6.3.1 Component Design/Analysis 208

APPENDIX A AIRESEARCH CASTING COMPANY ADVANCED GAS TURBINE

APPENDIX B FORD MOTOR COMPANY ADVANCED GAS TURBINE (AGT)

APPENDIX C PURE CARBON COMPANY ADVANCED GAS TURBINE (AGT)

APPENDIX D LIST OF SYMBOLS, ABBREVIATIONS, AND ACRONYMS

NASA C-168 FORM

1 AGT101 Program Schedule 7

2 AGT101 RPD Powertrain 8

3 AGTI01 Power Section 9

4 AGTI01 Drive System Schematic i0

5 AGT101 Installation- Side View ii

6 AGTI01 Installation- Front View 11

7 AGT101 Estimated Performance Curves 12

8 AGTI01 Specific Fuel Consumption 12

9 Basic Bearing Configuration 14

10 Foil Bearing Deflected State at 100,000 RPM Operating Speed 14

11 Film Pressure Distribution at Operating Speed 15

12 Foil Bearing Designs Compared 16

13 Foil Bearing Surface Coating Development 17

14 Foil Bearing Dynamic Properties Wear Rig 17

15 TiN Coated Ring (Top) Versus Au Coated Disk (Bottom) After

16 Typical Results of Dynamic Friction Test 20

17 AGT101 Hydrodynamic Thrust Bearing Configuration 21

18 Thrust Bearing Design Detail 21

19 Thrust Bearing/Ball Bearing Load Share Characteristic 22

20 Total Analytical Heat Generation for AGTI01 Fluid Film

22 Thrust Bearing Power Consumption Decrease with Oil Flow 23

23 Floating Ring Seal 23

24 Typical Turbine Axial Clearance Transient for Cold Start 24

25 Compressor and Turbine Clearance Loss Due to Thermal Growths 25

26 AGT101 Metal Structured Power Section 27

27 AGTI01 Engine Rotor System Configurations 28

28 AGTI01 Engine InitialRotor System Configuration Response 29

29 AGTI01 Engine Motoring Rotating Group Configurations Tested 30